Datasets:

markhneedham

/

blog

like 0

2015-02-24 22:39:49

Python/nltk: Naive vs Naive Bayes vs Decision Tree

Last week I wrote a blog post describing http://www.markhneedham.com/blog/2015/02/20/pythonscikit-learn-detecting-which-sentences-in-a-transcript-contain-a-speaker/[a decision tree I'd trained to detect the speakers in a How I met your mother transcript] and after writing the post I wondered whether a simple classifier would do the job. The simple classifier will work on the assumption that any word followed by a ":" is a speaker and anything else isn't. Here's the definition of a +++<cite>+++NaiveClassifier+++</cite>+++: [source,python] ---- import nltk from nltk import ClassifierI class NaiveClassifier(ClassifierI): def classify(self, featureset): if featureset['next-word'] == ":": return True else: return False ---- As you can see it only implements the classify method and executes a static check. While reading about ways to evaluate the effectiveness of text classifiers I came across http://streamhacker.com/2010/05/17/text-classification-sentiment-analysis-precision-recall/[Jacob Perkins blog] which suggests that we should measure two things: precision and recall. * Higher precision means less false positives, while lower precision means more false positives. * Higher recall means less false negatives, while lower recall means more false negatives. If (like me) you often get confused between false positives and negatives the following photo should help fix that: </p> image::{{<siteurl>}}/uploads/2015/02/false_positive_negative.jpg[False positive negative,400] I wrote the following function (adapted from Jacob's blog post) to calculate precision and recall values for a given classifier: [source,python] ---- import nltk import collections def assess_classifier(classifier, test_data, text): refsets = collections.defaultdict(set) testsets = collections.defaultdict(set) for i, (feats, label) in enumerate(test_data): refsets[label].add(i) observed = classifier.classify(feats) testsets[observed].add(i) speaker_precision = nltk.metrics.precision(refsets[True], testsets[True]) speaker_recall = nltk.metrics.recall(refsets[True], testsets[True]) non_speaker_precision = nltk.metrics.precision(refsets[False], testsets[False]) non_speaker_recall = nltk.metrics.recall(refsets[False], testsets[False]) return [text, speaker_precision, speaker_recall, non_speaker_precision, non_speaker_recall] ---- Now let's call that function with each of our classifiers: [source,python] ---- import json from sklearn.cross_validation import train_test_split from himymutil.ml import pos_features from himymutil.naive import NaiveClassifier from tabulate import tabulate with open("data/import/trained_sentences.json", "r") as json_file: json_data = json.load(json_file) tagged_sents = [] for sentence in json_data: tagged_sents.append([(word["word"], word["speaker"]) for word in sentence["words"]]) featuresets = [] for tagged_sent in tagged_sents: untagged_sent = nltk.tag.untag(tagged_sent) for i, (word, tag) in enumerate(tagged_sent): featuresets.append( (pos_features(untagged_sent, i), tag) ) train_data,test_data = train_test_split(featuresets, test_size=0.20, train_size=0.80) table = [] table.append(assess_classifier(NaiveClassifier(), test_data, "Naive")) table.append(assess_classifier(nltk.NaiveBayesClassifier.train(train_data), test_data, "Naive Bayes")) table.append(assess_classifier(nltk.DecisionTreeClassifier.train(train_data), test_data, "Decision Tree")) print(tabulate(table, headers=["Classifier","speaker precision", "speaker recall", "non-speaker precision", "non-speaker recall"])) ---- I'm using the https://pypi.python.org/pypi/tabulate[tabulate] library to print out a table showing each of the classifiers and their associated value for precision and recall. If we execute this file we'll see the following output: [source,bash] ---- $ python scripts/detect_speaker.py Classifier speaker precision speaker recall non-speaker precision non-speaker recall ------------- ------------------- ---------------- ----------------------- -------------------- Naive 0.9625 0.846154 0.994453 0.998806 Naive Bayes 0.674603 0.934066 0.997579 0.983685 Decision Tree 0.965517 0.923077 0.997219 0.998806 ---- The naive classifier is good on most measures but makes some mistakes on speaker recall - we have 16% false negatives i.e. 16% of words that should be classified as speaker aren't. Naive Bayes does poorly in terms of speaker false positives - 1/3 of the time when we say a word is a speaker it actually isn't. The decision tree performs best but has 8% speaker false negatives - 8% of words that should be classified as speakers aren't. The https://github.com/mneedham/neo4j-himym/blob/master/scripts/detect_speaker.py[code is on github] if you want to play around with it.

pythonnltk-naive-vs-naive-bayes-vs-decision-tree

https://markhneedham.com/blog/2015/02/24/pythonnltk-naive-vs-naive-bayes-vs-decision-tree

2015-02-15 15:56:09

Python/scikit-learn: Calculating TF/IDF on How I met your mother transcripts

Over the past few weeks I've been playing around with various NLP techniques to find interesting insights into How I met your mother from its transcripts and one technique that kept coming up is TF/IDF. The http://en.wikipedia.org/wiki/Tf%E2%80%93idf[Wikipedia definition] reads like this: ____ *tf--idf*, short for *term frequency--inverse document frequency*, is a numerical statistic that is intended to reflect how important a word is to a document in a collection or corpus. It is often used as a weighting factor in information retrieval and text mining. The tf-idf value increases proportionally to the number of times a word appears in the document, but is offset by the frequency of the word in the corpus, which helps to adjust for the fact that some words appear more frequently in general. ____ I wanted to generate a http://en.wikipedia.org/wiki/Tf%E2%80%93idf[TF/IDF] representation of phrases used in the hope that it would reveal some common themes used in the show. Python's http://scikit-learn.org/stable/[scikit-learn] library gives you two ways to generate the TF/IDF representation: . Generate a matrix of token/phrase counts from a collection of text documents using +++<cite>+++http://scikit-learn.org/stable/modules/generated/sklearn.feature_extraction.text.CountVectorizer.html[CountVectorizer]+++</cite>+++ and feed it to +++<cite>+++http://scikit-learn.org/stable/modules/generated/sklearn.feature_extraction.text.TfidfTransformer.html#sklearn.feature_extraction.text.TfidfTransformer.fit_transform[TfidfTransformer]+++</cite>+++ to generate the TF/IDF representation. . Feed the collection of text documents directly to +++<cite>+++http://scikit-learn.org/stable/modules/generated/sklearn.feature_extraction.text.TfidfVectorizer.html[TfidfVectorizer]+++</cite>+++ and go straight to the TF/IDF representation skipping the middle man. I started out using the first approach and hadn't quite got it working when I realised there was a much easier way! I have a collection of https://github.com/mneedham/neo4j-himym/blob/master/data/import/sentences.csv[sentences in a CSV file] so the first step is to convert those into a list of documents: [source,python] ---- from collections import defaultdict import csv episodes = defaultdict(list) with open("data/import/sentences.csv", "r") as sentences_file: reader = csv.reader(sentences_file, delimiter=',') reader.next() for row in reader: episodes[row[1]].append(row[4]) for episode_id, text in episodes.iteritems(): episodes[episode_id] = "".join(text) corpus = [] for id, episode in sorted(episodes.iteritems(), key=lambda t: int(t[0])): corpus.append(episode) ---- +++<cite>+++corpus+++</cite>+++ contains 208 entries (1 per episode), each of which is a string containing the transcript of that episode. Next it's time to train our TF/IDF model which is only a few lines of code: [source,python] ---- from sklearn.feature_extraction.text import TfidfVectorizer tf = TfidfVectorizer(analyzer='word', ngram_range=(1,3), min_df = 0, stop_words = 'english') ---- The most interesting parameter here is +++<cite>+++ngram_range+++</cite>+++ - we're telling it to generate 2 and 3 word phrases along with the single words from the corpus. e.g. if we had the sentence "Python is cool" we'd end up with 6 phrases - 'Python', 'is', 'cool', 'Python is', 'Python is cool' and 'is cool'. Let's execute the model against our corpus: ~~~python tfidf_matrix = tf.fit_transform(corpus) feature_names = tf.get_feature_names() >>> len(feature_names) 498254 >>> feature_names[50:70] [u'00 does sound', u'00 don', u'00 don buy', u'00 dressed', u'00 dressed blond', u'00 drunkenly', u'00 drunkenly slurred', u'00 fair', u'00 fair tonight', u'00 fall', u'00 fall foliage', u'00 far', u'00 far impossible', u'00 fart', u'00 fart sure', u'00 friends', u'00 friends singing', u'00 getting', u'00 getting guys', u'00 god'] ~~~ So we're got nearly 500,000 phrases and if we look at +++<cite>+++tfidf_matrix+++</cite>+++ we'd expect it to be a 208 x 498254 matrix - one row per episode, one column per phrase: ~~~python >>> tfidf_matrix <208x498254 sparse matrix of type '<type 'numpy.float64'>' with 740396 stored elements in Compressed Sparse Row format> ~~~ This is what we've got although under the covers it's using a sparse representation to save space. Let's convert the matrix to dense format to explore further and find out why: ~~~python dense = tfidf_matrix.todense() >>> len(dense[0].tolist()[0]) 498254 ~~~ What I've printed out here is the size of one row of the matrix which contains the TF/IDF score for every phrase in our corpus for the 1st episode of How I met your mother. A lot of those phrases won't have happened in the 1st episode so let's filter those out: ~~~python episode = dense[0].tolist()[0] phrase_scores = [pair for pair in zip(range(0, len(episode)), episode) if pair[1] > 0] >>> len(phrase_scores) 4823 ~~~ There are just under 5000 phrases used in this episode, roughly 1% of the phrases in the whole corpus. The sparse matrix makes a bit more sense - if scipy used a dense matrix representation there'd be 493,000 entries with no score which becomes more significant as the number of documents increases. Next we'll sort the phrases by score in descending order to find the most interesting phrases for the first episode of How I met your mother: ~~~python >>> sorted(phrase_scores, key=lambda t: t[1] * -1)[:5] [(419207, 0.2625177493269755), (312591, 0.19571419072701732), (267538, 0.15551468983363487), (490429, 0.15227880637176266), (356632, 0.1304175242341549)] ~~~ The first value in each tuple is the phrase's position in our initial vector and also corresponds to the phrase's position in +++<cite>+++feature_names+++</cite>+++ which allows us to map the scores back to phrases. Let's look up a couple of phrases: ~~~python >>> feature_names[419207] u'ted' >>> feature_names[312591] u'olives' >>> feature_names[356632] u'robin' ~~~ Let's automate that lookup: ~~~python sorted_phrase_scores = sorted(phrase_scores, key=lambda t: t[1] * -1) for phrase, score in [(feature_names[word_id], score) for (word_id, score) in sorted_phrase_scores][:20]: print('{0: <20} \{1}'.format(phrase, score)) ted 0.262517749327 olives 0.195714190727 marshall 0.155514689834 yasmine 0.152278806372 robin 0.130417524234 barney 0.124411751867 lily 0.122924977859 signal 0.103793246466 goanna 0.0981379875009 scene 0.0953423604123 cut 0.0917336653574 narrator 0.0864622981985 flashback 0.078295921554 flashback date 0.0702825260177 ranjit 0.0693927691559 flashback date robin 0.0585687716814 ted yasmine 0.0585687716814 carl 0.0582101172888 eye patch 0.0543650529797 lebanese 0.0543650529797 ~~~ We see all the main characters names which aren't that interested - perhaps they should be part of the stop list - but 'olives' which is where the olive theory is first mentioned. I thought olives came up more often but a quick search for the term suggests it isn't mentioned again until Episode 9 in Season 9: ~~~bash $ grep -rni --color "olives" data/import/sentences.csv | cut -d, -f 2,3,4 | sort | uniq -c 16 1,1,1 3 193,9,9 ~~~ 'yasmine' is also an interesting phrase in this episode but she's never mentioned again: ~~~bash $ grep -h -rni --color "yasmine" data/import/sentences.csv 49:48,1,1,1,"Barney: (Taps a woman names Yasmine) Hi, have you met Ted? (Leaves and watches from a distance)." 50:49,1,1,1,"Ted: (To Yasmine) Hi, I'm Ted." 51:50,1,1,1,Yasmine: Yasmine. 53:52,1,1,1,"Yasmine: Thanks, It's Lebanese." 65:64,1,1,1,"[Cut to the bar, Ted is chatting with Yasmine]" 67:66,1,1,1,Yasmine: So do you think you'll ever get married? 68:67,1,1,1,"Ted: Well maybe eventually. Some fall day. Possibly in Central Park. Simple ceremony, we'll write our own vows. But--eh--no DJ, people will dance. I'm not going to worry about it! Damn it, why did Marshall have to get engaged? (Yasmine laughs) Yeah, nothing hotter than a guy planning out his own imaginary wedding, huh?" 69:68,1,1,1,"Yasmine: Actually, I think it's cute." 79:78,1,1,1,"Lily: You are unbelievable, Marshall. No-(Scene splits in half and shows both Lily and Marshall on top arguing and Ted and Yasmine on the bottom mingling)" 82:81,1,1,1,Ted: (To Yasmine) you wanna go out sometime? 85:84,1,1,1,[Cut to Scene with Ted and Yasmine at bar] 86:85,1,1,1,Yasmine: I'm sorry; Carl's my boyfriend (points to bartender) ~~~ It would be interesting to filter out the phrases which don't occur in any other episode and see what insights we get from doing that. For now though we'll extract phrases for all episodes and write to CSV so we can explore more easily: ~~~python with open("data/import/tfidf_scikit.csv", "w") as file: writer = csv.writer(file, delimiter=",") writer.writerow(["EpisodeId", "Phrase", "Score"]) doc_id = 0 for doc in tfidf_matrix.todense(): print "Document %d" %(doc_id) word_id = 0 for score in doc.tolist()[0]: if score > 0: word = feature_names[word_id] writer.writerow([doc_id+1, word.encode("utf-8"), score]) word_id +=1 doc_id +=1 ~~~ And finally a quick look at the contents of the CSV: ~~~bash $ tail -n 10 data/import/tfidf_scikit.csv 208,york apparently laughs,0.012174304095213192 208,york aren,0.012174304095213192 208,york aren supposed,0.012174304095213192 208,young,0.013397275854758335 208,young ladies,0.012174304095213192 208,young ladies need,0.012174304095213192 208,young man,0.008437685963000223 208,young man game,0.012174304095213192 208,young stupid,0.011506395106658192 208,young stupid sighs,0.012174304095213192 ~~~

pythonscikit-learn-calculating-tfidf-on-how-i-met-your-mother-transcripts

https://markhneedham.com/blog/2015/02/15/pythonscikit-learn-calculating-tfidf-on-how-i-met-your-mother-transcripts

2015-02-12 23:45:03

Python/gensim: Creating bigrams over How I met your mother transcripts

As part of my continued playing around with How I met your mother transcripts I wanted to identify plot arcs and as a first step I wrote some code using the https://radimrehurek.com/gensim/[gensim] and http://www.nltk.org/[nltk] libraries to identify bigrams (two word phrases). There's an http://radimrehurek.com/gensim/models/phrases.html#module-gensim.models.phrases[easy to follow tutorial in the gensim docs] showing how to go about this but I needed to do a couple of extra steps to get my text data from a CSV file into the structure gensim expects. Let's first remind ourselves what the https://github.com/mneedham/neo4j-himym/blob/master/data/import/sentences.csv[sentences CSV file] looks like: [source,bash] ---- $ head -n 15 data/import/sentences.csv | tail 5,1,1,1,Son: Are we being punished for something? 6,1,1,1,Narrator: No 7,1,1,1,"Daughter: Yeah, is this going to take a while?" 8,1,1,1,"Narrator: Yes. (Kids are annoyed) Twenty-five years ago, before I was dad, I had this whole other life." 9,1,1,1,"(Music Plays, Title ""How I Met Your Mother"" appears)" 10,1,1,1,"Narrator: It was way back in 2005. I was twenty-seven just starting to make it as an architect and living in New York with my friend Marshall, my best friend from college. My life was good and then Uncle Marshall went and screwed the whole thing up." 11,1,1,1,Marshall: (Opens ring) Will you marry me. 12,1,1,1,"Ted: Yes, perfect! And then you're engaged, you pop the champagne! You drink a toast! You have s*x on the kitchen floor... Don't have s*x on our kitchen floor." 13,1,1,1,"Marshall: Got it. Thanks for helping me plan this out, Ted." 14,1,1,1,"Ted: Dude, are you kidding? It's you and Lily! I've been there for all the big moments of you and Lily. The night you met. Your first date... other first things." ---- We need to transform those sentences into an array of words for each line and feed it into gensim's +++<cite>+++models.Phrase+++</cite>+++ object: [source,python] ---- import nltk import csv import string from gensim.models import Phrases from gensim.models import Word2Vec from nltk.corpus import stopwords sentences = [] bigram = Phrases() with open("data/import/sentences.csv", "r") as sentencesfile: reader = csv.reader(sentencesfile, delimiter = ",") reader.next() for row in reader: sentence = [word.decode("utf-8") for word in nltk.word_tokenize(row[4].lower()) if word not in string.punctuation] sentences.append(sentence) bigram.add_vocab([sentence]) ---- We're used nltk's +++<cite>+++word_tokezine+++</cite>+++ function to create our array of words and then we've got a clause to make sure we remove any words which are punctuation otherwise they will dominate our phrases. We can take a quick peek at some of the phrases that have been created like so: [source,python] ---- >>> list(bigram[sentences])[:5] [[u'pilot'], [u'scene', u'one'], [u'title', u'the', u'year_2030'], [u'narrator_kids', u'i', u"'m", u'going', u'to', u'tell', u'you', u'an_incredible', u'story.', u'the', u'story', u'of', u'how', u'i', u'met_your', u'mother'], [u'son', u'are', u'we', u'being', u'punished', u'for', u'something']] ---- gensim uses an underscore character to indicate when it's joined two words together and in this sample we've got three phrases - 'narrator_kids', 'met_you' and 'an_incredible'. We can now populate a Counter with our phrases and their counts and find out the most common phrases. One thing to note is that I've chosen to get rid of stopwords at this point rather than earlier because I didn't want to generate 'false bigrams' where there was actually a stop word sitting in between. [source,python] ---- bigram_counter = Counter() for key in bigram.vocab.keys(): if key not in stopwords.words("english"): if len(key.split("_")) > 1: bigram_counter[key] += bigram.vocab[key] for key, counts in bigram_counter.most_common(20): print '{0: <20} {1}'.format(key.encode("utf-8"), counts) i_'m 4607 it_'s 4288 you_'re 2659 do_n't 2436 that_'s 2044 in_the 1696 gon_na 1576 you_know 1497 i_do 1464 this_is 1407 and_i 1389 want_to 1071 it_was 1053 on_the 1052 at_the 1035 we_'re 1033 i_was 1018 of_the 1014 ca_n't 1010 are_you 994 ---- Most of the phrases aren't really that interesting and I had better luck feeding the phrases into a Word2Vec model and repeating the exercise: [source,python] ---- bigram_model = Word2Vec(bigram[sentences], size=100) bigram_model_counter = Counter() for key in bigram_model.vocab.keys(): if key not in stopwords.words("english"): if len(key.split("_")) > 1: bigram_model_counter[key] += bigram_model.vocab[key].count for key, counts in bigram_model_counter.most_common(50): print '{0: <20} {1}'.format(key.encode("utf-8"), counts) do_n't 2436 gon_na 1576 ca_n't 1010 did_n't 704 come_on 499 end_of 460 kind_of 396 from_2030 394 my_god 360 they_'re 351 'm_sorry 349 does_n't 341 end_flashback 327 all_right 308 've_been 303 'll_be 301 of_course 289 a_lot 284 right_now 279 new_york 270 look_at 265 trying_to 238 tell_me 196 a_few 195 've_got 189 wo_n't 174 so_much 172 got_ta 168 each_other 166 my_life 157 talking_about 157 talk_about 154 what_happened 151 at_least 141 oh_god 138 wan_na 129 supposed_to 126 give_me 124 last_night 121 my_dad 120 more_than 119 met_your 115 excuse_me 112 part_of 110 phone_rings 109 get_married 107 looks_like 105 'm_sorry. 104 said_`` 101 ---- The first 20 phrases or so aren't particularly interesting although we do have 'new_york' in there which is good as that's where the show is set. If we go further we'll notice phrases like 'my_dad', 'get_married' and 'last_night' which may all explain interesting parts of the plot. Having the data in the Word2Vec model allows us to do some other fun queries too. e.g. [source,python] ---- >>> bigram_model.most_similar(['marshall', 'lily'], ['ted'], topn=10) [(u'robin', 0.5474381446838379), (u'go_ahead', 0.5138797760009766), (u'zoey', 0.505358874797821), (u'karen', 0.48617005348205566), (u'cootes', 0.4757827818393707), (u'then', 0.45426881313323975), (u'lewis', 0.4510520100593567), (u'natalie.', 0.45070385932922363), (u'vo', 0.4189065098762512), (u'players', 0.4149518311023712)] >>> bigram_model.similarity("ted", "robin") 0.51928683064927905 >>> bigram_model.similarity("barney", "robin") 0.62980405583219112 >>> bigram_model.most_similar(positive=['getting_married']) [(u'so_glad', 0.780311107635498), (u'kidding', 0.7683225274085999), (u'awake', 0.7682262659072876), (u'lunch.', 0.7591195702552795), (u'ready.', 0.7372316718101501), (u'single.', 0.7350872755050659), (u'excited_about', 0.725479006767273), (u'swamped', 0.7252731323242188), (u'boyfriends', 0.7127221822738647), (u'believe_this.', 0.71015864610672)] >>> bigram_model.most_similar(positive=['my_dad']) [(u'my_mom', 0.7994954586029053), (u'somebody', 0.7758427262306213), (u'easier', 0.7305313944816589), (u'hot.', 0.7282992601394653), (u'pregnant.', 0.7103987336158752), (u'nobody', 0.7059557437896729), (u'himself.', 0.7046393156051636), (u'physically', 0.7044381499290466), (u'young_lady', 0.69412761926651), (u'at_bernie', 0.682607889175415)] ---- I'm not quite at the stage where I can automatically pull out the results of a gensim model and do something with it but it is helping me to see some of the main themes in the show. Next up I'll try out trigrams and then TF/IDF over the bigrams to see which are the most important on a per episode basis. I also need to dig into Word2Vec to figure out why it comes up with different top phrases than the Phrases model.

pythongensim-creating-bigrams-over-how-i-met-your-mother-transcripts

https://markhneedham.com/blog/2015/02/12/pythongensim-creating-bigrams-over-how-i-met-your-mother-transcripts

2015-02-13 23:38:43

Neo4j: Building a topic graph with Prismatic Interest Graph API

Over the last few weeks I've been using various NLP libraries to derive topics for my corpus of How I met your mother episodes without success and was therefore enthused to see the release of Prismatic's http://blog.getprismatic.com/interest-graph-api/[Interest Graph API] The Interest Graph API exposes a web service to which you feed a block of text and get back a set of topics and associated score. It has been trained over the last few years with millions of articles that people share on their social media accounts and in my experience using Prismatic the topics have been very useful for finding new material to read. The first step is to head to http://interest-graph.getprismatic.com/[interest-graph.getprismatic.com] and get an API key which will be emailed to you. Having done that we're ready to make some calls to the API and get back some topics. I'm going to use Python to call the API and I've found the http://docs.python-requests.org/en/latest/[requests] library the easiest library to use for this type of work. Our call to the API looks like this: [source,python] ---- import requests payload = { 'title': "insert title of article here", 'body': "insert body of text here"), 'api-token': "insert token sent by email here"} r = requests.post("http://interest-graph.getprismatic.com/text/topic", data=payload) ---- One thing to keep in mind is that the API is rate limited to 20 requests a second so we need to restrict our requests or we're going to receive error response codes. Luckily I came across an excellent http://blog.gregburek.com/2011/12/05/Rate-limiting-with-decorators/[blog post showing how to write a decorator around a function] and only allow it to execute at a certain frequency. To rate limit our calls to the Interest Graph we need to pull the above code into a function and annotate it appropriately: [source,python] ---- import time def RateLimited(maxPerSecond): minInterval = 1.0 / float(maxPerSecond) def decorate(func): lastTimeCalled = [0.0] def rateLimitedFunction(*args,**kargs): elapsed = time.clock() - lastTimeCalled[0] leftToWait = minInterval - elapsed if leftToWait>0: time.sleep(leftToWait) ret = func(*args,**kargs) lastTimeCalled[0] = time.clock() return ret return rateLimitedFunction return decorate @RateLimited(0.3) def topics(title, body): payload = { 'title': title, 'body': body, 'api-token': "insert token sent by email here"} r = requests.post("http://interest-graph.getprismatic.com/text/topic", data=payload) return r ---- The https://github.com/mneedham/neo4j-himym/blob/master/data/import/sentences.csv[text I want to classify] is stored in a CSV file - one sentence per line. Here's a sample: [source,bash] ---- $ head -n 10 data/import/sentences.csv SentenceId,EpisodeId,Season,Episode,Sentence 1,1,1,1,Pilot 2,1,1,1,Scene One 3,1,1,1,[Title: The Year 2030] 4,1,1,1,"Narrator: Kids, I'm going to tell you an incredible story. The story of how I met your mother" 5,1,1,1,Son: Are we being punished for something? 6,1,1,1,Narrator: No 7,1,1,1,"Daughter: Yeah, is this going to take a while?" 8,1,1,1,"Narrator: Yes. (Kids are annoyed) Twenty-five years ago, before I was dad, I had this whole other life." 9,1,1,1,"(Music Plays, Title ""How I Met Your Mother"" appears)" ---- We'll also need to refer to https://github.com/mneedham/neo4j-himym/blob/master/data/import/episodes_full.csv[another CSV file] to get the title of each episode since it isn't being stored with the sentence: [source,bash] ---- $ head -n 10 data/import/episodes_full.csv NumberOverall,NumberInSeason,Episode,Season,DateAired,Timestamp,Title,Director,Viewers,Writers,Rating 1,1,/wiki/Pilot,1,"September 19, 2005",1127084400,Pilot,Pamela Fryman,10.94,"Carter Bays,Craig Thomas",68 2,2,/wiki/Purple_Giraffe,1,"September 26, 2005",1127689200,Purple Giraffe,Pamela Fryman,10.40,"Carter Bays,Craig Thomas",63 3,3,/wiki/Sweet_Taste_of_Liberty,1,"October 3, 2005",1128294000,Sweet Taste of Liberty,Pamela Fryman,10.44,"Phil Lord,Chris Miller",67 4,4,/wiki/Return_of_the_Shirt,1,"October 10, 2005",1128898800,Return of the Shirt,Pamela Fryman,9.84,Kourtney Kang,59 5,5,/wiki/Okay_Awesome,1,"October 17, 2005",1129503600,Okay Awesome,Pamela Fryman,10.14,Chris Harris,53 6,6,/wiki/Slutty_Pumpkin,1,"October 24, 2005",1130108400,Slutty Pumpkin,Pamela Fryman,10.89,Brenda Hsueh,62 7,7,/wiki/Matchmaker,1,"November 7, 2005",1131321600,Matchmaker,Pamela Fryman,10.55,"Sam Johnson,Chris Marcil",57 8,8,/wiki/The_Duel,1,"November 14, 2005",1131926400,The Duel,Pamela Fryman,10.35,Gloria Calderon Kellett,46 9,9,/wiki/Belly_Full_of_Turkey,1,"November 21, 2005",1132531200,Belly Full of Turkey,Pamela Fryman,10.29,"Phil Lord,Chris Miller",60 ---- Now we need to get our episode titles and transcripts ready to pass to the +++<cite>+++topics+++</cite>+++ function. Since we've only got ~ 200 episodes we can create a dictionary to store that data: [source,python] ---- episodes = {} with open("data/import/episodes_full.csv", "r") as episodesfile: episodes_reader = csv.reader(episodesfile, delimiter=",") episodes_reader.next() for episode in episodes_reader: episodes[int(episode[0])] = {"title": episode[6], "sentences" : [] } with open("data/import/sentences.csv", "r") as sentencesfile: sentences_reader = csv.reader(sentencesfile, delimiter=",") sentences_reader.next() for sentence in sentences_reader: episodes[int(sentence[1])]["sentences"].append(sentence[4]) >>> episodes[1]["title"] 'Pilot' >>> episodes[1]["sentences"][:5] ['Pilot', 'Scene One', '[Title: The Year 2030]', "Narrator: Kids, I'm going to tell you an incredible story. The story of how I met your mother", 'Son: Are we being punished for something?'] ---- Now we're going to loop through each of the episodes, call +++<cite>+++topics+++</cite>+++ and write the result into a CSV file so we can load it into Neo4j afterwards to explore the data: [source,python] ---- import json with open("data/import/topics.csv", "w") as topicsfile: topics_writer = csv.writer(topicsfile, delimiter=",") topics_writer.writerow(["EpisodeId", "TopicId", "Topic", "Score"]) for episode_id, episode in episodes.iteritems(): tmp = topics(episode["title"], "".join(episode["sentences"]).json() print episode_id, tmp for topic in tmp['topics']: topics_writer.writerow([episode_id, topic["id"], topic["topic"], topic["score"]]) ---- It takes about 10 minutes to run and this is a sample of the output: [source,bash] ---- $ head -n 10 data/import/topics.csv EpisodeId,TopicId,Topic,Score 1,1519,Fiction,0.5798245566455255 1,2015,Humour,0.565154963605359 1,24031,Laughing,0.5587120401021765 1,16693,Flirting,0.5514098189505282 1,1163,Dating and Courtship,0.5487490108554022 1,2386,Kissing,0.5476185929151934 1,31929,Puns,0.5375100569837977 2,24031,Laughing,0.5670926949850333 2,1519,Fiction,0.5396488295397263 ---- We'll use Neo4j's http://neo4j.com/docs/stable/query-load-csv.html[LOAD CSV] command to load the data in: [source,cypher] ---- // make sure the topics exist LOAD CSV WITH HEADERS FROM "file:///Users/markneedham/projects/neo4j-himym/data/import/topics.csv" AS row MERGE (topic:Topic {id: TOINT(row.TopicId)}) ON CREATE SET topic.value = row.Topic ---- [source,cypher] ---- // make sure the topics exist LOAD CSV WITH HEADERS FROM "file:///Users/markneedham/projects/neo4j-himym/data/import/topics.csv" AS row MERGE (topic:Topic {id: TOINT(row.TopicId)}) ON CREATE SET topic.value = row.Topic ---- [source,cypher] ---- // now link the episodes and topics LOAD CSV WITH HEADERS FROM "file:///Users/markneedham/projects/neo4j-himym/data/import/topics.csv" AS row MATCH (topic:Topic {id: TOINT(row.TopicId)}) MATCH (episode:Episode {id: TOINT(row.EpisodeId)}) MERGE (episode)-[:TOPIC {score: TOFLOAT(row.Score)}]->(topic) ---- We'll assume that the episodes and seasons are already loaded - the https://github.com/mneedham/neo4j-himym/blob/master/data/import/himym.cql[commands to load those in are on github]. We can now write some queries against our topic graph. We'll start simple - show me the topics for an episode: [source,cypher] ---- MATCH (episode:Episode {id: 1})-[r:TOPIC]->(topic) RETURN topic, r ---- image::{{<siteurl>}}/uploads/2015/02/graph.png[Graph,333] Let's say we liked the 'Puns' aspect of the Pilot episode and want to find out which other episodes had puns. The following query would let us find those: [source,cypher] ---- MATCH (episode:Episode {id: 1})-[r:TOPIC]->(topic {value: "Puns"})<-[:TOPIC]-(other) RETURN episode, topic, other ---- image::{{<siteurl>}}/uploads/2015/02/graph-1.png[Graph 1,425] Or maybe we want to find the episode which has the most topics in common: [source,cypher] ---- MATCH (episode:Episode {id: 1})-[:TOPIC]->(topic), (topic)<-[r:TOPIC]-(otherEpisode) RETURN otherEpisode.title as episode, COUNT(r) AS topicsInCommon ORDER BY topicsInCommon DESC LIMIT 10 ==> +------------------------------------------------+ ==> | episode | topicsInCommon | ==> +------------------------------------------------+ ==> | "Purple Giraffe" | 6 | ==> | "Ten Sessions" | 5 | ==> | "Farhampton" | 4 | ==> | "The Three Days Rule" | 4 | ==> | "How I Met Everyone Else" | 4 | ==> | "The Time Travelers" | 4 | ==> | "Mary the Paralegal" | 4 | ==> | "Lobster Crawl" | 4 | ==> | "The Magician's Code, Part 2" | 4 | ==> | "Slutty Pumpkin" | 4 | ==> +------------------------------------------------+ ==> 10 rows ---- We could then tweak that query to get the names of those topics: [source,cypher] ---- MATCH (episode:Episode {id: 1})-[:TOPIC]->(topic), (topic)<-[r:TOPIC]-(otherEpisode)-[:IN_SEASON]->(season) RETURN otherEpisode.title as episode, season.number AS season, COUNT(r) AS topicsInCommon, COLLECT(topic.value) ORDER BY topicsInCommon DESC LIMIT 10 ==> +-----------------------------------------------------------------------------------------------------------------------------------+ ==> | episode | season | topicsInCommon | COLLECT(topic.value) | ==> +-----------------------------------------------------------------------------------------------------------------------------------+ ==> | "Purple Giraffe" | "1" | 6 | ["Humour","Fiction","Kissing","Dating and Courtship","Flirting","Laughing"] | ==> | "Ten Sessions" | "3" | 5 | ["Humour","Puns","Dating and Courtship","Flirting","Laughing"] | ==> | "How I Met Everyone Else" | "3" | 4 | ["Humour","Fiction","Dating and Courtship","Laughing"] | ==> | "Farhampton" | "8" | 4 | ["Humour","Fiction","Kissing","Dating and Courtship"] | ==> | "Bedtime Stories" | "9" | 4 | ["Humour","Puns","Dating and Courtship","Laughing"] | ==> | "Definitions" | "5" | 4 | ["Kissing","Dating and Courtship","Flirting","Laughing"] | ==> | "Lobster Crawl" | "8" | 4 | ["Humour","Dating and Courtship","Flirting","Laughing"] | ==> | "Little Boys" | "3" | 4 | ["Humour","Puns","Dating and Courtship","Laughing"] | ==> | "Wait for It" | "3" | 4 | ["Fiction","Puns","Flirting","Laughing"] | ==> | "Mary the Paralegal" | "1" | 4 | ["Humour","Dating and Courtship","Flirting","Laughing"] | ==> +-----------------------------------------------------------------------------------------------------------------------------------+ ---- Overall 168 (out of 208) of the other episodes have a topic in common with the first episode so perhaps just having a topic in common isn't the best indication of similarity. An interesting next step would be to calculate http://en.wikipedia.org/wiki/Cosine_similarity[cosine] or http://en.wikipedia.org/wiki/Jaccard_index[jaccard] similarity between the episodes and store that value in the graph for querying later on. I've also calculated http://www.markhneedham.com/blog/2015/02/12/pythongensim-creating-bigrams-over-how-i-met-your-mother-transcripts/[the most common bigrams across all the transcripts] so it would be interesting to see if there are any interesting insights at the intersection of episodes, topics and phrases.

neo4j-building-a-topic-graph-with-prismatic-interest-graph-api

https://markhneedham.com/blog/2015/02/13/neo4j-building-a-topic-graph-with-prismatic-interest-graph-api

2015-02-22 08:58:57

R/dplyr: Extracting data frame column value for filtering with %in%

I've been playing around with http://cran.rstudio.com/web/packages/dplyr/vignettes/introduction.html[dplyr] over the weekend and wanted to extract the values from a data frame column to use in a later filtering step. I had a data frame: [source,r] ---- library(dplyr) df = data.frame(userId = c(1,2,3,4,5), score = c(2,3,4,5,5)) ---- And wanted to extract the userIds of those people who have a score greater than 3. I started with: [source,r] ---- highScoringPeople = df %>% filter(score > 3) %>% select(userId) > highScoringPeople userId 1 3 2 4 3 5 ---- And then filtered the data frame expecting to get back those 3 people: [source,r] ---- > df %>% filter(userId %in% highScoringPeople) [1] userId score <0 rows> (or 0-length row.names) ---- No rows! I created vector with the numbers 3-5 to make sure that worked: [source,r] ---- > df %>% filter(userId %in% c(3,4,5)) userId score 1 3 4 2 4 5 3 5 5 ---- That works as expected so +++<cite>+++highScoringPeople+++</cite>+++ obviously isn't in the right format to facilitate an 'in lookup'. Let's explore: [source,r] ---- > str(c(3,4,5)) num [1:3] 3 4 5 > str(highScoringPeople) 'data.frame': 3 obs. of 1 variable: $ userId: num 3 4 5 ---- Now it's even more obvious why it doesn't work - +++<cite>+++highScoringPeople+++</cite>+++ is still a data frame when we need it to be a vector/list. One way to fix this is to extract the userIds using the $ syntax instead of the select function: [source,r] ---- highScoringPeople = (df %>% filter(score > 3))$userId > str(highScoringPeople) num [1:3] 3 4 5 > df %>% filter(userId %in% highScoringPeople) userId score 1 3 4 2 4 5 3 5 5 ---- Or if we want to do the column selection using dplyr we can http://stackoverflow.com/questions/21618423/extract-a-dplyr-tbl-column-as-a-vector[extract the values for the column like this]: [source,r] ---- highScoringPeople = (df %>% filter(score > 3) %>% select(userId))[[1]] > str(highScoringPeople) num [1:3] 3 4 5 ---- Not so difficult after all.

rdplyr-extracting-data-frame-column-value-for-filtering-with-in

https://markhneedham.com/blog/2015/02/22/rdplyr-extracting-data-frame-column-value-for-filtering-with-in

2015-11-28 13:56:59

Python: Parsing a JSON HTTP chunking stream

I've been playing around with meetup.com's API again and this time wanted to consume the http://www.meetup.com/meetup_api/docs/stream/2/rsvps/#http[chunked HTTP RSVP stream] and filter RSVPs for events I'm interested in. I use Python for most of my hacking these days and if HTTP requests are required the http://docs.python-requests.org/en/latest/[requests library] is my first port of call. I started out with the following script [source,python] ---- import requests import json def stream_meetup_initial(): uri = "http://stream.meetup.com/2/rsvps" response = requests.get(uri, stream = True) for chunk in response.iter_content(chunk_size = None): yield chunk for raw_rsvp in stream_meetup_initial(): print raw_rsvp try: rsvp = json.loads(raw_rsvp) except ValueError as e: print e continue ---- This mostly worked but I also noticed the following error from time to time: [source,text] ---- No JSON object could be decoded ---- Although less frequent, I also saw errors suggesting I was trying to parse an incomplete JSON object. I tweaked the function to keep a local buffer and only yield that if the chunk ended in a new line character: [source,python] ---- def stream_meetup_newline(): uri = "http://stream.meetup.com/2/rsvps" response = requests.get(uri, stream = True) buffer = "" for chunk in response.iter_content(chunk_size = 1): if chunk.endswith("\n"): buffer += chunk yield buffer buffer = "" else: buffer += chunk ---- This mostly works although I'm sure I've seen some occasions where two JSON objects were being yielded and then the call to 'json.loads' failed. I haven't been able to reproduce that though. A second read through the http://docs.python-requests.org/en/latest/user/advanced/#streaming-requests[requests documentation] made me realise I hadn't read it very carefully the first time since we can make our lives much easier by using 'iter_lines' rather than 'iter_content': [source,python] ---- r = requests.get('http://stream.meetup.com/2/rsvps', stream=True) for raw_rsvp in r.iter_lines(): if raw_rsvp: rsvp = json.loads(raw_rsvp) print rsvp ---- We can then process 'rsvp', filtering out the ones we're interested in.

python-parsing-a-json-http-chunking-stream

https://markhneedham.com/blog/2015/11/28/python-parsing-a-json-http-chunking-stream

2015-11-08 20:58:42

Docker 1.9: Port forwarding on Mac OS X

Since the http://www.infoq.com/news/2015/10/neo4j-2.3-release[Neo4j 2.3.0 release] there's been an https://hub.docker.com/r/neo4j/neo4j/[official docker image] which I thought I'd give a try this afternoon. The last time I used docker about a year ago I had to install http://boot2docker.io/[boot2docker] which has now been deprecated in place of http://docs.docker.com/engine/installation/mac/[Docker Machine and the Docker Toolbox]. I created a container with the following command: [source,bash] ---- docker run --detach --publish=7474:7474 neo4j/neo4j ---- And then tried to access the Neo4j server locally: [source,bash] ---- $ curl http://localhost:7474 curl: (7) Failed to connect to localhost port 7474: Connection refused ---- I quickly checked that docker had started up Neo4j correctly: [source,bash] ---- $ docker ps CONTAINER ID IMAGE COMMAND CREATED STATUS PORTS NAMES 1f7c48e267f0 neo4j/neo4j "/docker-entrypoint.s" 10 minutes ago Up 10 minutes 7473/tcp, 0.0.0.0:7474->7474/tcp kickass_easley ---- Looks good. Amusingly I then came across http://www.markhneedham.com/blog/2014/11/27/dockerneo4j-port-forwarding-on-mac-os-x-not-working/[my own blog post from a year ago] where I'd run into the same problem - the problem being that we need to access the Neo4j server via the VM's IP address rather than localhost. Instead of using boot2docker we now need to use docker-machine to find the VM's IP address: [source,bash] ---- $ docker-machine ls NAME ACTIVE DRIVER STATE URL SWARM default * virtualbox Running tcp://192.168.99.100:2376 ---- [source,bash] ---- $ curl http://192.168.99.100:7474 { "management" : "http://192.168.99.100:7474/db/manage/", "data" : "http://192.168.99.100:7474/db/data/" } ---- And we're back in business.

docker-1-9-port-forwarding-on-mac-os-x

https://markhneedham.com/blog/2015/11/08/docker-1-9-port-forwarding-on-mac-os-x

2015-11-08 11:47:36

IntelliJ 'java: cannot find JDK 1.8'

I upgraded to https://www.jetbrains.com/idea/download/[IntelliJ 15.0] a few days ago and was initially seeing the following exception when trying to compile: [source,text] ---- module-name java: cannot find JDK 1.8 ---- I've been compiling against JDK 1.8 for a while now using IntelliJ 14 so I wasn't sure what was going on. I checked my project settings and they seemed fine: image::{{<siteurl>}}/uploads/2015/11/2015-11-08_11-39-16.png[2015 11 08 11 39 16,500] The error message suggested I look in the logs to find more information but I wasn't sure where those live! I eventually found out the answer via the https://intellij-support.jetbrains.com/hc/en-us/articles/206827517-Locating-IDE-log-files[comments of this support ticket] although I later found https://intellij-support.jetbrains.com/hc/en-us/articles/206827437-Directories-used-by-the-IDE-to-store-settings-caches-plugins-and-logs[a post describing it in more detail]. Looking into the logs revealed the following error message: [source,bash] ---- $ less /Users/markneedham/Library/Logs/IntelliJIdea15/idea.log 2015-11-05 16:31:28,429 [ 428129] INFO - figurations.GeneralCommandLine - Cannot run program "/Library/Java/JavaVirtualMachines/jdk1.7.0_71.jdk/Contents/Home/bin/java" (in directory "/Applications/IntelliJ IDEA 15.app/Contents/bin"): error=2, No such file or directory java.io.IOException: Cannot run program "/Library/Java/JavaVirtualMachines/jdk1.7.0_71.jdk/Contents/Home/bin/java" (in directory "/Applications/IntelliJ IDEA 15.app/Contents/bin"): error=2, No such file or directory at java.lang.ProcessBuilder.start(ProcessBuilder.java:1048) at com.intellij.execution.configurations.GeneralCommandLine.startProcess(GeneralCommandLine.java:368) at com.intellij.execution.configurations.GeneralCommandLine.createProcess(GeneralCommandLine.java:354) at com.intellij.execution.process.OSProcessHandler.<init>(OSProcessHandler.java:38) at org.jetbrains.idea.maven.server.MavenServerManager$3.startProcess(MavenServerManager.java:359) at org.jetbrains.idea.maven.server.MavenServerManager$3.execute(MavenServerManager.java:345) at com.intellij.execution.rmi.RemoteProcessSupport.a(RemoteProcessSupport.java:206) at com.intellij.execution.rmi.RemoteProcessSupport.acquire(RemoteProcessSupport.java:139) at org.jetbrains.idea.maven.server.MavenServerManager.create(MavenServerManager.java:163) at org.jetbrains.idea.maven.server.MavenServerManager.create(MavenServerManager.java:71) at org.jetbrains.idea.maven.server.RemoteObjectWrapper.getOrCreateWrappee(RemoteObjectWrapper.java:41) at org.jetbrains.idea.maven.server.MavenServerManager$9.execute(MavenServerManager.java:525) at org.jetbrains.idea.maven.server.MavenServerManager$9.execute(MavenServerManager.java:522) at org.jetbrains.idea.maven.server.RemoteObjectWrapper.perform(RemoteObjectWrapper.java:76) at org.jetbrains.idea.maven.server.MavenServerManager.applyProfiles(MavenServerManager.java:522) at org.jetbrains.idea.maven.project.MavenProjectReader.applyProfiles(MavenProjectReader.java:369) at org.jetbrains.idea.maven.project.MavenProjectReader.doReadProjectModel(MavenProjectReader.java:98) at org.jetbrains.idea.maven.project.MavenProjectReader.access$300(MavenProjectReader.java:42) at org.jetbrains.idea.maven.project.MavenProjectReader$1.doProcessParent(MavenProjectReader.java:422) at org.jetbrains.idea.maven.project.MavenProjectReader$1.doProcessParent(MavenProjectReader.java:399) at org.jetbrains.idea.maven.project.MavenParentProjectFileProcessor.processRepositoryParent(MavenParentProjectFileProcessor.java:84) at org.jetbrains.idea.maven.project.MavenParentProjectFileProcessor.process(MavenParentProjectFileProcessor.java:62) at org.jetbrains.idea.maven.project.MavenProjectReader.resolveInheritance(MavenProjectReader.java:425) at org.jetbrains.idea.maven.project.MavenProjectReader.doReadProjectModel(MavenProjectReader.java:95) at org.jetbrains.idea.maven.project.MavenProjectReader.access$300(MavenProjectReader.java:42) at org.jetbrains.idea.maven.project.MavenProjectReader$1.doProcessParent(MavenProjectReader.java:422) at org.jetbrains.idea.maven.project.MavenProjectReader$1.doProcessParent(MavenProjectReader.java:399) at org.jetbrains.idea.maven.project.MavenParentProjectFileProcessor.processRepositoryParent(MavenParentProjectFileProcessor.java:84) at org.jetbrains.idea.maven.project.MavenParentProjectFileProcessor.process(MavenParentProjectFileProcessor.java:62) at org.jetbrains.idea.maven.project.MavenProjectReader.resolveInheritance(MavenProjectReader.java:425) at org.jetbrains.idea.maven.project.MavenProjectReader.doReadProjectModel(MavenProjectReader.java:95) at org.jetbrains.idea.maven.project.MavenProjectReader.access$300(MavenProjectReader.java:42) at org.jetbrains.idea.maven.project.MavenProjectReader$1.doProcessParent(MavenProjectReader.java:422) at org.jetbrains.idea.maven.project.MavenProjectReader$1.doProcessParent(MavenProjectReader.java:399) at org.jetbrains.idea.maven.project.MavenParentProjectFileProcessor.processRepositoryParent(MavenParentProjectFileProcessor.java:84) at org.jetbrains.idea.maven.project.MavenParentProjectFileProcessor.process(MavenParentProjectFileProcessor.java:62) at org.jetbrains.idea.maven.project.MavenProjectReader.resolveInheritance(MavenProjectReader.java:425) at org.jetbrains.idea.maven.project.MavenProjectReader.doReadProjectModel(MavenProjectReader.java:95) at org.jetbrains.idea.maven.project.MavenProjectReader.readProject(MavenProjectReader.java:53) at org.jetbrains.idea.maven.project.MavenProject.read(MavenProject.java:626) at org.jetbrains.idea.maven.project.MavenProjectsTree.doUpdate(MavenProjectsTree.java:564) at org.jetbrains.idea.maven.project.MavenProjectsTree.doAdd(MavenProjectsTree.java:509) at org.jetbrains.idea.maven.project.MavenProjectsTree.update(MavenProjectsTree.java:470) at org.jetbrains.idea.maven.project.MavenProjectsTree.updateAll(MavenProjectsTree.java:441) at org.jetbrains.idea.maven.project.MavenProjectsProcessorReadingTask.perform(MavenProjectsProcessorReadingTask.java:60) at org.jetbrains.idea.maven.project.MavenProjectsProcessor.doProcessPendingTasks(MavenProjectsProcessor.java:134) at org.jetbrains.idea.maven.project.MavenProjectsProcessor.access$100(MavenProjectsProcessor.java:30) at org.jetbrains.idea.maven.project.MavenProjectsProcessor$2.run(MavenProjectsProcessor.java:109) at org.jetbrains.idea.maven.utils.MavenUtil$7.run(MavenUtil.java:464) at com.intellij.openapi.application.impl.ApplicationImpl$8.run(ApplicationImpl.java:365) at java.util.concurrent.Executors$RunnableAdapter.call(Executors.java:511) at java.util.concurrent.FutureTask.run(FutureTask.java:266) at java.util.concurrent.ThreadPoolExecutor.runWorker(ThreadPoolExecutor.java:1142) at java.util.concurrent.ThreadPoolExecutor$Worker.run(ThreadPoolExecutor.java:617) at java.lang.Thread.run(Thread.java:745) at org.jetbrains.ide.PooledThreadExecutor$1$1.run(PooledThreadExecutor.java:55) Caused by: java.io.IOException: error=2, No such file or directory at java.lang.UNIXProcess.forkAndExec(Native Method) at java.lang.UNIXProcess.<init>(UNIXProcess.java:248) at java.lang.ProcessImpl.start(ProcessImpl.java:134) at java.lang.ProcessBuilder.start(ProcessBuilder.java:1029) ... 55 more ---- Somewhere I had a JDK 1.7 defined which no longer existed on my machine. I actually only have one JDK installed at the moment: [source,bash] ---- $ /usr/libexec/java_home -V Matching Java Virtual Machines (1): 1.8.0_51, x86_64: "Java SE 8" /Library/Java/JavaVirtualMachines/jdk1.8.0_51.jdk/Contents/Home /Library/Java/JavaVirtualMachines/jdk1.8.0_51.jdk/Contents/Home ---- A bit of exploring led me to 'Platform Settings' which is where the culprit was: image::{{<siteurl>}}/uploads/2015/11/2015-11-08_11-45-00.png[2015 11 08 11 45 00,500] That setting lives actually lives in +++<cite>+++/Users/markneedham/Library/Preferences/IntelliJIdea15/options/jdk.table.xml+++</cite>+++ and once I removed it IntelliJ resumed normal service.

intellij-java-cannot-find-jdk-1-8

https://markhneedham.com/blog/2015/11/08/intellij-java-cannot-find-jdk-1-8

2015-11-24 00:12:59

jq: Cannot iterate over number / string and number cannot be added

In my continued parsing of http://www.meetup.com/meetup_api/[meetup.com's JSON API] I wanted to extract some information from the following JSON file: [source,bash] ---- $ head -n40 data/members/18313232.json [ { "status": "active", "city": "London", "name": ". .", "other_services": {}, "country": "gb", "topics": [], "lon": -0.13, "joined": 1438866605000, "id": 92951932, "state": "17", "link": "http://www.meetup.com/members/92951932", "photo": { "thumb_link": "http://photos1.meetupstatic.com/photos/member/8/d/6/b/thumb_250896203.jpeg", "photo_id": 250896203, "highres_link": "http://photos1.meetupstatic.com/photos/member/8/d/6/b/highres_250896203.jpeg", "photo_link": "http://photos1.meetupstatic.com/photos/member/8/d/6/b/member_250896203.jpeg" }, "lat": 51.49, "visited": 1446745707000, "self": { "common": {} } }, { "status": "active", "city": "London", "name": "Abdelkader Idryssy", "other_services": {}, "country": "gb", "topics": [ { "name": "Weekend Adventures", "urlkey": "weekend-adventures", "id": 16438 }, { "name": "Community Building", "urlkey": "community-building", ---- In particular I want to extract the member's id, name, join date and the ids of topics they're interested in. I started with the following jq query to try and extract those attributes: [source,bash] ---- $ jq -r '.[] | [.id, .name, .joined, (.topics[] | .id | join(";"))] | @csv' data/members/18313232.json Cannot iterate over number (16438) ---- Annoyingly this treats topic ids on an individual basis rather than as an array as I wanted. I tweaked the query to the following with no luck: [source,bash] ---- $ jq -r '.[] | [.id, .name, .joined, (.topics[].id | join(";"))] | @csv' data/members/18313232.json Cannot iterate over number (16438) ---- As a guess I decided to wrap '.topics[].id' in an array literal to see if it had any impact: [source,bash] ---- $ jq -r '.[] | [.id, .name, .joined, ([.topics[].id] | join(";"))] | @csv' data/members/18313232.json 92951932,". .",1438866605000,"" jq: error (at data/members/18313232.json:60013): string ("") and number (16438) cannot be added ---- Woot! A different error message at least and this one seems to be due to a type mismatch between the string we want to end up with and the array of numbers that we currently have. We can cast our way to victory with the 'tostring' function: [source,r] ---- $ jq -r '.[] | [.id, .name, .joined, ([.topics[].id | tostring] | join(";"))] | @csv' data/members/18313232.json ... 92951932,". .",1438866605000,"" 193866304,"Abdelkader Idryssy",1445195325000,"16438;20727;15401;9760;20246;20923;3336;2767;242;58259;4417;1789;10454;20274;10232;563;25375;16433;15187;17635;26273;21808;933;7789;23884;16212;144477;15322;21067;3833;108403;20221;1201;182;15083;9696;4377;15360;18296;15121;17703;10161;1322;3880;18333;3485;15585;44584;18692;21681" 28643052,"Abhishek Chanda",1439688955000,"646052;520302;15167;563;65735;537492;646072;537502;24959;1025832;8599;31197;24410;26118;10579;1064;189;48471;16216;18062;33089;107633;46831;20479;1423042;86258;21441;3833;21681;188;9696;58162;20398;113032;18060;29971;55324;30928;15261;58259;638;16475;27591;10107;242;109595;10470;26384;72514;1461192" 39523062,"Adam Kinder-Jones",1438677474000,"70576;21549;3833;42277;164111;21522;93380;48471;15167;189;563;25435;87614;9696;18062;58162;10579;21681;19882;108403;128595;15582;7029" 194119823,"Adam Lewis",1444867994000,"10209" 14847001,"Adam Rogers",1422917313000,"" 87709042,"Adele Green",1436867381000,"15167;18062;102811;9696;30928;18060;78565;189;7029;48471;127567;10579;58162;563;3833;16216;21441;37708;209821;15401;59325;31792;21836;21900;984862;15720;17703;96823;4422;85951;87614;37428;2260;827;121802;19672;38660;84325;118991;135612;10464;1454542;17936;21549;21520;17628;148303;20398;66339;29661" 11497937,"Adrian Bridgett",1421067940000,"30372;15046;25375;638;498;933;374;27591;18062;18060;15167;10581;16438;15672;1998;1273;713;26333;15099;15117;4422;15892;242;142180;563;31197;20479;1502;131178;15018;43256;58259;1201;7319;15940;223;8652;66493;15029;18528;23274;9696;128595;21681;17558;50709;113737" 14151190,"adrian lawrence",1437142198000,"7029;78565;659;85951;15582;48471;9696;128595;563;10579;3833;101960;16137;1973;78566;206;223;21441;16216;108403;21681;186;1998;15731;17703;15043;16613;17885;53531;48375;16615;19646;62326;49954;933;22268;19243;37381;102811;30928;455;10358;73511;127567;106382;16573;36229;781;23981;1954" 183557824,"Adrien Pujol",1421882756000,"108403;563;9696;21681;188;24410;1064;32743;124668;15472;21123;1486432;1500742;87614;46831;1454542;46810;166000;126177;110474" ... ----

jq-cannot-iterate-over-number-string-and-number-cannot-be-added

https://markhneedham.com/blog/2015/11/24/jq-cannot-iterate-over-number-string-and-number-cannot-be-added

2015-11-14 22:51:38

jq: Filtering missing keys

I've been playing around with the http://www.meetup.com/meetup_api/docs/2/events/[meetup.com API] again over the last few days and having saved a set of events to disk I wanted to extract the venues using https://stedolan.github.io/jq/[jq]. This is what a single event record looks like: [source,bash] ---- $ jq -r ".[0]" data/events/0.json { "status": "past", "rating": { "count": 1, "average": 1 }, "utc_offset": 3600000, "event_url": "http://www.meetup.com/londonweb/events/3261890/", "group": { "who": "Web Peeps", "name": "London Web", "group_lat": 51.52000045776367, "created": 1034097743000, "join_mode": "approval", "group_lon": -0.12999999523162842, "urlname": "londonweb", "id": 163876 }, "name": "London Web Design October Meetup", "created": 1094756756000, "venue": { "city": "London", "name": "Roadhouse Live Music Restaurant , Bar & Club", "country": "GB", "lon": -0.1, "phone": "44-020-7240-6001", "address_1": "The Piazza", "address_2": "Covent Garden", "repinned": false, "lat": 51.52, "id": 11725 }, "updated": 1273536337000, "visibility": "public", "yes_rsvp_count": 2, "time": 1097776800000, "waitlist_count": 0, "headcount": 0, "maybe_rsvp_count": 5, "id": "3261890" } ---- We want to extract the keys underneath 'venue'. I started with the following: [source,bash] ---- $ jq -r ".[] | .venue" data/events/0.json ... { "city": "London", "name": "Counting House Pub", "country": "gb", "lon": -0.085022, "phone": "020 7283 7123", "address_1": "50 Cornhill Rd", "address_2": "EC3V 3PD", "repinned": false, "lat": 51.513407, "id": 835790 } null { "city": "Paris", "name": "Mozilla Paris", "country": "fr", "lon": 2.341002, "address_1": "16 Bis Boulevard Montmartre", "repinned": false, "lat": 48.871834, "id": 23591845 } ... ---- This is close to what I want but it includes 'null' values which means when you extract the keys inside 'venue' they are all empty as well: [source,bash] ---- jq -r ".[] | .venue | [.id, .name, .city, .address_1, .address_2, .lat, .lon] | @csv" data/events/0.json ... 101958,"The Green Man and French Horn, -","London","54, St. Martins Lane - Covent Garden","WC2N 4EA",51.52,-0.1 ,,,,,, 107295,"The Yorkshire Grey Pub","London","46 Langham Street","W1W 7AX",51.52,-0.1 ... ,,,,,, ---- If functional programming lingo we want to filter out any JSON documents which don't have the 'venue' key. 'filter' has a different meaning in jq so it took me a while to realise that the 'select' function was what I needed to get rid of the null values: [source,bash] ---- $ jq -r ".[] | select(.venue != null) | .venue | [.id, .name, .city, .address_1, .address_2, .lat, .lon] | @csv" data/events/0.json | head 11725,"Roadhouse Live Music Restaurant , Bar & Club","London","The Piazza","Covent Garden",51.52,-0.1 11725,"Roadhouse Live Music Restaurant , Bar & Club","London","The Piazza","Covent Garden",51.52,-0.1 11725,"Roadhouse Live Music Restaurant , Bar & Club","London","The Piazza","Covent Garden",51.52,-0.1 11725,"Roadhouse Live Music Restaurant , Bar & Club","London","The Piazza","Covent Garden",51.52,-0.1 76192,"Pied Bull Court","London","Galen Place, London, WC1A 2JR",,51.516747,-0.12719 76192,"Pied Bull Court","London","Galen Place, London, WC1A 2JR",,51.516747,-0.12719 85217,"Earl's Court Exhibition Centre","London","Warwick Road","SW5 9TA",51.49233,-0.199735 96579,"Olympia 2","London","Near Olympia tube station",,51.52,-0.1 76192,"Pied Bull Court","London","Galen Place, London, WC1A 2JR",,51.516747,-0.12719 101958,"The Green Man and French Horn, -","London","54, St. Martins Lane - Covent Garden","WC2N 4EA",51.52,-0.1 ---- And we're done.

jq-filtering-missing-keys

https://markhneedham.com/blog/2015/11/14/jq-filtering-missing-keys

2015-10-05 22:41:10

Mac OS X: Installing the PROJ.4 - Cartographic Projections Library

I've been following http://scottbarnham.com/blog/2010/08/27/uk-postcodes-to-latitudelongitude/[Scott Barnham's guide to transforming UK postcodes into (lat, long) coordinates] and needed to install the https://github.com/OSGeo/proj.4[PROJ.4 Cartographic Projections library] which I initially struggled with. The first step is to download a tar.gz version which is linked from the https://github.com/OSGeo/proj.4/wiki[wiki page]: [source,bash] ---- $ wget http://download.osgeo.org/proj/proj-4.9.1.tar.gz ---- Next we'll unpack the file and then build the binaries: [source,bash] ---- $ tar -xvf proj-4.9.1.tar.gz $ cd proj-4.9.1 $ ./configure --prefix ~/projects/land-registry/proj-4.9.1 $ make $ make install ---- The files we need are in the bin directory\... [source,bash] ---- $ ls -alh bin/ total 184 drwxr-xr-x 8 markneedham staff 272B 5 Oct 23:07 . drwxr-xr-x@ 41 markneedham staff 1.4K 5 Oct 20:46 .. -rwxr-xr-x 1 markneedham staff 20K 5 Oct 23:07 cs2cs -rwxr-xr-x 1 markneedham staff 16K 5 Oct 23:07 geod lrwxr-xr-x 1 markneedham staff 4B 5 Oct 23:07 invgeod -> geod lrwxr-xr-x 1 markneedham staff 4B 5 Oct 23:07 invproj -> proj -rwxr-xr-x 1 markneedham staff 13K 5 Oct 23:07 nad2bin -rwxr-xr-x 1 markneedham staff 21K 5 Oct 23:07 proj ---- \...now let's give it a try. We need to feed in OSGB36 grid reference values and then we'll get back WGS84 Lat/Lng values. We can grab some grid reference values from the http://www.ordnancesurvey.co.uk/oswebsite/products/code-point-open/index.html[Ordnance Survey website]. e.g. the Neo4j London office has the post code SE1 0NZ which translates to coordinates 531950,180195. Let's try those out with PROJ.4: [source,bash] ---- $ ./proj-4.9.1/bin/cs2cs -f '%.7f' +proj=tmerc +lat_0=49 +lon_0=-2 +k=0.9996012717 +x_0=400000 +y_0=-100000 +ellps=airy +towgs84=446.448,-125.157,542.060,0.1502,0.2470,0.8421,-20.4894 +units=m +no_defs +to +proj=latlong +ellps=WGS84 +towgs84=0,0,0 +nodefs 531950 180195 -0.1002020 51.5052917 46.0810195 ---- So it's suggested a (lat, long) pairing of (51.5052917, -0.1002020). And if we plug that into https://www.google.co.uk/maps/place/51%C2%B030'19.1%22N+0%C2%B006'00.7%22W/@51.5056056,-0.1011247,17z/data=!4m2!3m1!1s0x0:0x0[Google maps]\... image::{{<siteurl>}}/uploads/2015/10/2015-10-05_23-29-09.png[2015 10 05 23 29 09,400] \...it's pretty much spot on!

mac-os-x-installing-the-proj-4-cartographic-projections-library

https://markhneedham.com/blog/2015/10/05/mac-os-x-installing-the-proj-4-cartographic-projections-library

2015-10-02 18:42:47

R: data.table - Finding the maximum row

In my continued playing around with the R https://cran.r-project.org/web/packages/data.table/index.html[data.table] package I wanted to find the maximum row based on one of the columns, grouped by another column, and then return back the whole row. We'll use the following data table to illustrate: [source,r] ---- > blogDT = data.table(name = c("Property 1","Property 1","Property 1","Property 2","Property 2","Property 2"), price = c(10000, 12500, 18000, 245000, 512000, 1000000), date = c("Day 1", "Day 7", "Day 10", "Day 3", "Day 5", "Day 12")) > blogDT[, lag.price := c(NA, price[-.N]), by = name] > blogDT[, diff := price - lag.price] > blogDT name price date lag.price diff 1: Property 1 10000 Day 1 NA NA 2: Property 1 12500 Day 7 10000 2500 3: Property 1 18000 Day 10 12500 5500 4: Property 2 245000 Day 3 NA NA 5: Property 2 512000 Day 5 245000 267000 6: Property 2 1000000 Day 12 512000 488000 ---- I wanted to find the biggest difference in 'price' and 'lag.price' grouped by the 'name' column. If we just want to get the max 'diff' grouped by 'name' it's quite easy: [source,r] ---- > blogDT[!is.na(diff), .(max = max(diff)), keyby = name] name max 1: Property 1 5500 2: Property 2 488000 ---- However now we've lost the information about which 'date' that was on and what the 'price' and 'lag.price' were which is a bit annoying. If we only want to keep the rows which have the highest 'diff' grouped by 'name', one way to go about this is to add a 'max.diff' column to each row and then filter appropriately. e.g. [source,r] ---- > maxDT = blogDT[!is.na(diff)][, max := max(diff), by = name] > maxDT name price date lag.price diff max 1: Property 1 12500 Day 7 10000 2500 5500 2: Property 1 18000 Day 10 12500 5500 5500 3: Property 2 512000 Day 5 245000 267000 488000 4: Property 2 1000000 Day 12 512000 488000 488000 > maxDT[diff == max] name price date lag.price diff max 1: Property 1 18000 Day 10 12500 5500 5500 2: Property 2 1000000 Day 12 512000 488000 488000 ---- I've only been playing with data.table for a few days so if there's a better way do let me know in the comments.

r-data-table-finding-the-maximum-row

https://markhneedham.com/blog/2015/10/02/r-data-table-finding-the-maximum-row

2015-10-18 10:03:57

Exploring (potential) data entry errors in the Land Registry data set

I've previously written a couple of blog posts describing the mechanics of analysing the https://data.gov.uk/dataset/land-registry-monthly-price-paid-data[Land Registry data set] and I thought it was about time I described some of the queries I've been running the discoveries I've made. To recap, the land registry provides a 3GB, 20 million line CSV file containing all the property sales in the UK since 1995. We'll be loading and query the data in R using the https://cran.r-project.org/web/packages/data.table/index.html[data.table] package: [source,r] ---- > library(data.table) > dt = fread("pp-complete.csv", header = FALSE) > dt[1:5] V1 V2 V3 V4 V5 1: {0C7ADEF5-878D-4066-B785-0000003ED74A} 163000 2003-02-21 00:00 UB5 4PJ T 2: {35F67271-ABD4-40DA-AB09-00000085B9D3} 247500 2005-07-15 00:00 TA19 9DD D 3: {B20B1C74-E8E1-4137-AB3E-0000011DF342} 320000 2010-09-10 00:00 W4 1DZ F 4: {7D6B0915-C56B-4275-AF9B-00000156BCE7} 104000 1997-08-27 00:00 NE61 2BH D 5: {47B60101-B64C-413D-8F60-000002F1692D} 147995 2003-05-02 00:00 PE33 0RU D V6 V7 V8 V9 V10 V11 V12 1: N F 106 READING ROAD NORTHOLT NORTHOLT 2: N F 58 ADAMS MEADOW ILMINSTER ILMINSTER 3: N L 58 WHELLOCK ROAD LONDON 4: N F 17 WESTGATE MORPETH MORPETH 5: N F 4 MASON GARDENS WEST WINCH KING'S LYNN V13 V14 V15 1: EALING GREATER LONDON A 2: SOUTH SOMERSET SOMERSET A 3: EALING GREATER LONDON A 4: CASTLE MORPETH NORTHUMBERLAND A 5: KING'S LYNN AND WEST NORFOLK NORFOLK A ---- For our first query we're going to find the most expensive query sold for each year from 1995 - 2015. The first thing we'll need to do is make column 'V2' (price) numeric and convert column 'V3' (sale date) to data format so we can do date arithmetic on it: [source,r] ---- > dt = dt[, V2:= as.numeric(V2)] > dt = dt[, V3:= as.Date(V3)] ---- Now let's write the query: [source,r] ---- > dt[, .SD[which.max(V2)], by=year(V3)][order(year)][, .(year,V9,V8,V10,V12,V14,V4,V2)] year V9 V8 V10 V12 V14 V4 V2 1: 1995 THORNETS HOUSE BUILDER GARDENS LEATHERHEAD SURREY KT22 7DE 5610000 2: 1996 24 MAIN ROAD MELTON MOWBRAY LEICESTERSHIRE LE14 3SP 17250000 3: 1997 42 HYDE PARK GATE LONDON GREATER LONDON SW7 5DU 7500000 4: 1998 19 NEW BRIDGE STREET LONDON GREATER LONDON EC4V 6DB 11250000 5: 1999 TERMINAL HOUSE LOWER BELGRAVE STREET LONDON GREATER LONDON SW1W 0NH 32477000 6: 2000 UNIT 3 JUNIPER PARK FENTON WAY BASILDON ESSEX SS15 6RZ 12600000 7: 2001 19 BABMAES STREET LONDON GREATER LONDON SW1Y 6HD 24750000 8: 2002 72 VINCENT SQUARE LONDON GREATER LONDON SW1P 2PA 8300000 9: 2003 81 ADDISON ROAD LONDON GREATER LONDON W14 8ED 9250000 10: 2004 29 HOLLAND VILLAS ROAD LONDON GREATER LONDON W14 8DH 7950000 11: 2005 APARTMENT 1102 199 KNIGHTSBRIDGE LONDON GREATER LONDON SW7 1RH 15193950 12: 2006 1 THORNWOOD GARDENS LONDON GREATER LONDON W8 7EA 12400000 13: 2007 36 CADOGAN PLACE LONDON GREATER LONDON SW1X 9RX 17000000 14: 2008 50 CHESTER SQUARE LONDON GREATER LONDON SW1W 9EA 19750000 15: 2009 CASA SARA HEATHERSIDE DRIVE VIRGINIA WATER SURREY GU25 4JU 13800000 16: 2010 10 HOLLAND VILLAS ROAD LONDON GREATER LONDON W14 8BP 16200000 17: 2011 WHITESTONE HOUSE WHITESTONE LANE LONDON GREATER LONDON NW3 1EA 19250000 18: 2012 20 THE BOLTONS LONDON GREATER LONDON SW10 9SU 54959000 19: 2013 APARTMENT 7F 171 KNIGHTSBRIDGE LONDON GREATER LONDON SW7 1DW 39000000 20: 2014 APARTMENT 6, 5 PRINCES GATE LONDON GREATER LONDON SW7 1QJ 50000000 21: 2015 37 BURNSALL STREET LONDON GREATER LONDON SW3 3SR 27750000 year V9 V8 V10 V12 V14 V4 V2 ---- The results mostly make sense - the majority of the highest priced properties are around Hyde Park and often somewhere near Knightsbridge which is one of the most expensive places in the country. There are some odd odds though. e.g. in 1996 the top priced property is in Leicester and sold for just over £17m. I looked it up on the Land Registry site to quickly see what it was http://houseprices.landregistry.gov.uk/price-paid-record/1943130/24+main+road+asfordby+valley+melton+mowbray+melton+leicestershire+le14+3sp[subsequently sold for]: image::{{<siteurl>}}/uploads/2015/10/2015-10-17_22-06-03.png[2015 10 17 22 06 03,500] Based on the subsequent prices I think we can safely assume that the initial price is incorrect and should actually have been £17,250. We can also say the same about our 2000 winner in Juniper Park in Basildon which sold for £12.6 million. If we look at the http://houseprices.landregistry.gov.uk/sold-prices/juniper%20park%20ss15%206rz[next sale price after that] it's £172,500 in 2003 so most likely it was sold for £126,000 - only 100 times out! I wanted to follow this observation and see if I could find other anomalies by comparing adjacent sale prices of properties. First we'll create a 'fullAddress' field which we'll use as an identifier for each property. It's not completely unique but it's not far away: [source,r] ---- > dt = dt[, fullAddress := paste(dt$V8, dt$V9, dt$V10, dt$V11, dt$V12, dt$V13, dt$V4, sep=", ")] > setkey(dt, fullAddress) > dt[, .(fullAddress, V2)][1:5] fullAddress V2 1: ''NUTSHELL COTTAGE, 72, , KIRKLAND, KENDAL, KENDAL, SOUTH LAKELAND, LA9 5AP 89000 2: 'FARRIERS', , FARRIERS CLOSE, WOODLEY, READING, WOKINGHAM, RG5 3DD 790000 3: 'HOLMCROFT', 40, , BRIDGNORTH ROAD, WOMBOURNE, WOLVERHAMPTON, SOUTH STAFFORDSHIRE, WV5 0AA 305000 4: (AKERS), , CHAPEL STREET, EASINGWOLD, YORK, HAMBLETON, YO61 3AE 118000 5: (ANNINGS), , , FARWAY, COLYTON, EAST DEVON, EX24 6DF 150000 ---- Next we'll add a column to the data table which contains the previous sale price and another column which calculate the difference between the two prices: [source,R] ---- > dt[, lag.V2:=c(NA, V2[-.N]), by = fullAddress] > dt[, V2.diff := V2 - lag.V2] > dt[!is.na(lag.V2),][1:10][, .(fullAddress, lag.V2, V2, V2.diff)] fullAddress lag.V2 V2 V2.diff 1: (ANNINGS), , , FARWAY, COLYTON, EAST DEVON, EX24 6DF 150000 385000 235000 2: (BARBER), , PEACOCK CORNER, MOULTON ST MARY, NORWICH, BROADLAND, NR13 3NF 115500 136000 20500 3: (BELL), , BAWBURGH ROAD, MARLINGFORD, NORWICH, SOUTH NORFOLK, NR9 5AG 128000 300000 172000 4: (BEVERLEY), , DAWNS LANE, ASLOCKTON, NOTTINGHAM, RUSHCLIFFE, NG13 9AD 95000 210000 115000 5: (BLACKMORE), , GREAT STREET, NORTON SUB HAMDON, STOKE-SUB-HAMDON, SOUTH SOMERSET, TA14 6SJ 53000 118000 65000 6: (BOWDERY), , HIGH STREET, MARKINGTON, HARROGATE, HARROGATE, HG3 3NR 140000 198000 58000 7: (BULLOCK), , MOORLAND ROAD, INDIAN QUEENS, ST. COLUMB, RESTORMEL, TR9 6HN 50000 50000 0 8: (CAWTHRAY), , CAWOOD ROAD, WISTOW, SELBY, SELBY, YO8 3XB 130000 120000 -10000 9: (CAWTHRAY), , CAWOOD ROAD, WISTOW, SELBY, SELBY, YO8 3XB 120000 155000 35000 10: (COATES), , , BARDSEA, ULVERSTON, SOUTH LAKELAND, LA12 9QT 26000 36000 10000 ---- Let's find the properties which have the biggest £ value difference in adjacent sales: [source,r] ---- > dt[!is.na(V2.diff)][order(-abs(V2.diff))][, .(fullAddress, lag.V2, V2, V2.diff)][1:20] fullAddress lag.V2 V2 V2.diff 1: , 50, CHESTER SQUARE, LONDON, LONDON, CITY OF WESTMINSTER, SW1W 9EA 1135000 19750000 18615000 2: 44, , LANSDOWNE ROAD, , LONDON, KENSINGTON AND CHELSEA, W11 2LU 3675000 22000000 18325000 3: 24, , MAIN ROAD, ASFORDBY VALLEY, MELTON MOWBRAY, MELTON, LE14 3SP 17250000 32500 -17217500 4: 11, , ORMONDE GATE, , LONDON, KENSINGTON AND CHELSEA, SW3 4EU 250000 16000000 15750000 5: 2, , HOLLAND VILLAS ROAD, , LONDON, KENSINGTON AND CHELSEA, W14 8BP 8675000 24000000 15325000 6: 1, , PEMBRIDGE PLACE, , LONDON, KENSINGTON AND CHELSEA, W2 4XB 2340250 17000000 14659750 7: 10, , CHESTER SQUARE, LONDON, LONDON, CITY OF WESTMINSTER, SW1W 9HH 680000 15000000 14320000 8: 12, , SOUTH EATON PLACE, , LONDON, CITY OF WESTMINSTER, SW1W 9JA 4250000 18550000 14300000 9: 32, FLAT 1, HOLLAND PARK, , LONDON, KENSINGTON AND CHELSEA, W11 3TA 420000 14100000 13680000 10: 42, , EGERTON CRESCENT, , LONDON, KENSINGTON AND CHELSEA, SW3 2EB 1125000 14650000 13525000 11: 36, , CADOGAN PLACE, LONDON, LONDON, KENSINGTON AND CHELSEA, SW1X 9RX 3670000 17000000 13330000 12: 22, , ILCHESTER PLACE, , LONDON, KENSINGTON AND CHELSEA, W14 8AA 3350000 16250000 12900000 13: 3, , BOLNEY GATE, , LONDON, CITY OF WESTMINSTER, SW7 1QW 5650000 18250000 12600000 14: JUNIPER PARK, UNIT 3, FENTON WAY, , BASILDON, BASILDON, SS15 6RZ 12600000 172500 -12427500 15: 10, , WALTON PLACE, , LONDON, KENSINGTON AND CHELSEA, SW3 1RJ 356000 12750000 12394000 16: 84, MAISONETTE C, EATON SQUARE, , LONDON, CITY OF WESTMINSTER, SW1W 9AG 1500000 13400000 11900000 17: 3, , CHESTERFIELD HILL, , LONDON, CITY OF WESTMINSTER, W1J 5BJ 955000 12600000 11645000 18: 39, , ENNISMORE GARDENS, LONDON, LONDON, CITY OF WESTMINSTER, SW7 1AG 3650000 15250000 11600000 19: 76, FLAT 2, EATON SQUARE, , LONDON, CITY OF WESTMINSTER, SW1W 9AW 3500000 15000000 11500000 20: 85, , AVENUE ROAD, , LONDON, CAMDEN, NW8 6JD 519000 12000000 11481000 ---- Most of the entries here are in Westminster or Hyde Park and don't look particularly dodgy at first glance. We'd have to drill into the sale dates to confirm. What you might also have noticed is that our Melton Mowbray and Juniper Park properties both show up and although they don't have the biggest £ value difference they would probably rank top if calculated the multiplier instead. Let's give that a try: [source,r] ---- > dt[, V2.multiplier := ifelse(V2 > lag.V2, V2 / lag.V2, lag.V2 / V2)] > dt[!is.na(V2.multiplier)][order(-V2.multiplier)][, .(fullAddress, lag.V2, V2, V2.multiplier)][1:20] fullAddress lag.V2 V2 V2.multiplier 1: 24, , MAIN ROAD, ASFORDBY VALLEY, MELTON MOWBRAY, MELTON, LE14 3SP 17250000 32500 530.76923 2: LEA HAVEN, FLAT 1, CASTLE LANE, , TORQUAY, TORBAY, TQ1 3BE 38000 7537694 198.36037 3: NIGHTINGALE HOUSE, , BURLEIGH ROAD, ASCOT, ASCOT, WINDSOR AND MAIDENHEAD, SL5 7LD 9500 1100000 115.78947 4: JUNIPER PARK, UNIT 3, FENTON WAY, , BASILDON, BASILDON, SS15 6RZ 12600000 172500 73.04348 5: 9, , ROTHSAY GARDENS, BEDFORD, BEDFORD, BEDFORD, MK40 3QA 21000 1490000 70.95238 6: 22, GROUND FLOOR FLAT, SEA VIEW AVENUE, , PLYMOUTH, CITY OF PLYMOUTH, PL4 8RU 27950 1980000 70.84079 7: 91A, , TINTERN AVENUE, WESTCLIFF-ON-SEA, WESTCLIFF-ON-SEA, SOUTHEND-ON-SEA, SS0 9QQ 17000 1190000 70.00000 8: 204C, , SUTTON ROAD, SOUTHEND-ON-SEA, SOUTHEND-ON-SEA, SOUTHEND-ON-SEA, SS2 5ES 18000 1190000 66.11111 9: PRIORY COURT, FLAT 3, PRIORY AVENUE, TOTNES, TOTNES, SOUTH HAMS, TQ9 5HS 2226500 34000 65.48529 10: 59, , ST ANNS ROAD, SOUTHEND-ON-SEA, SOUTHEND-ON-SEA, SOUTHEND-ON-SEA, SS2 5AT 18250 1190000 65.20548 11: 15, , BREWERY LANE, LEIGH, LEIGH, WIGAN, WN7 2RJ 13500 880000 65.18519 12: 11, , ORMONDE GATE, , LONDON, KENSINGTON AND CHELSEA, SW3 4EU 250000 16000000 64.00000 13: WOODEND, , CANNONGATE ROAD, HYTHE, HYTHE, SHEPWAY, CT21 5PX 19261 1200000 62.30206 14: DODLESTON OAKS, , CHURCH ROAD, DODLESTON, CHESTER, CHESTER, CH4 9NG 10000 620000 62.00000 15: CREEKSIDE, , CURLEW DRIVE, WEST CHARLETON, KINGSBRIDGE, SOUTH HAMS, TQ7 2AA 28000 1700000 60.71429 16: 20, , BRANCH ROAD, BURNLEY, BURNLEY, BURNLEY, BB11 3AT 9000 540000 60.00000 17: THE BARN, , LEE WICK LANE, ST OSYTH, CLACTON-ON-SEA, TENDRING, CO16 8ES 10000 600000 60.00000 18: 11, , OAKWOOD GARDENS, KNAPHILL, WOKING, WOKING, GU21 2RX 6000 357000 59.50000 19: 23, , OLDHAM ROAD, GRASSCROFT, OLDHAM, OLDHAM, OL4 4HY 8000 475000 59.37500 20: THE SUNDAY HOUSE, , WATER LANE, GOLANT, FOWEY, RESTORMEL, PL23 1LF 8000 475000 59.37500 ---- This is much better! Our Melton Mowbray property comes in first by miles and Juniper Park is there in 4th. The rest of the price increases look implausible as well but let's drill into a couple of them: [source,r] ---- > dt[fullAddress == "15, , BREWERY LANE, LEIGH, LEIGH, WIGAN, WN7 2RJ"][, .(fullAddress, V3, V2)] fullAddress V3 V2 1: 15, , BREWERY LANE, LEIGH, LEIGH, WIGAN, WN7 2RJ 1995-06-29 13500 2: 15, , BREWERY LANE, LEIGH, LEIGH, WIGAN, WN7 2RJ 2008-03-28 880000 ---- If we look at some other properties on the http://www.rightmove.co.uk/house-prices/WN7/Brewery-Lane.html[same road] and look at the http://www.rightmove.co.uk/house-prices/detailMatching.html?prop=18993944&sale=35286620&country=england[property's features] it seems more likely that's meant to say £88,000. I noticed a similar trend when looking at some of the others on this list but I also realised that the data needs a bit of cleaning up as the 'fullAddress' column isn't uniquely identifying properties e.g. sometimes a property might have a Town/City of 'London' and a District of 'London' but on another transaction the District could be blank. On top of that, my strategy of looking for subsequent prices to spot anomalies falls down when trying to explore properties which only have one sale. So I have a couple of things to look into for now but once I've done those it'd be interesting to write an algorithm/program that could predict which transactions are likely to be anomalies. I can imagine how that might work if I had a labelled training set but I'm not sure if I could do it with an unsupervised algorithm so if you have any pointers let me know.

exploring-potential-data-entry-errors-in-the-land-registry-data-set

https://markhneedham.com/blog/2015/10/18/exploring-potential-data-entry-errors-in-the-land-registry-data-set

2015-10-27 23:10:47

Spark: MatchError (of class org.apache.spark.sql.catalyst.expressions.GenericRow) spark

I've been using http://spark.apache.org/[Spark] again lately to do some pre-processing on the https://data.gov.uk/dataset/land-registry-monthly-price-paid-data[Land Registry data set] and ran into an initially confusing problem when trying to parse the CSV file. I'm using the https://github.com/databricks/spark-csv[Databricks CSV parsing library] and wrote the following script to go over each row, collect up the address components and then derive a 'fullAddress' field. To refresh, this is what the CSV file looks like: [source,bash] ---- $ head -n5 pp-complete.csv "{0C7ADEF5-878D-4066-B785-0000003ED74A}","163000","2003-02-21 00:00","UB5 4PJ","T","N","F","106","","READING ROAD","NORTHOLT","NORTHOLT","EALING","GREATER LONDON","A" "{35F67271-ABD4-40DA-AB09-00000085B9D3}","247500","2005-07-15 00:00","TA19 9DD","D","N","F","58","","ADAMS MEADOW","ILMINSTER","ILMINSTER","SOUTH SOMERSET","SOMERSET","A" "{B20B1C74-E8E1-4137-AB3E-0000011DF342}","320000","2010-09-10 00:00","W4 1DZ","F","N","L","58","","WHELLOCK ROAD","","LONDON","EALING","GREATER LONDON","A" "{7D6B0915-C56B-4275-AF9B-00000156BCE7}","104000","1997-08-27 00:00","NE61 2BH","D","N","F","17","","WESTGATE","MORPETH","MORPETH","CASTLE MORPETH","NORTHUMBERLAND","A" "{47B60101-B64C-413D-8F60-000002F1692D}","147995","2003-05-02 00:00","PE33 0RU","D","N","F","4","","MASON GARDENS","WEST WINCH","KING'S LYNN","KING'S LYNN AND WEST NORFOLK","NORFOLK","A" ---- [source,scala] ---- import org.apache.spark.sql.{SQLContext, _} import org.apache.spark.{SparkConf, SparkContext} case class BlogTransaction(price: Integer, date: String, postCode:String, paon:String, saon:String, street:String, locality:String, city:String, district:String, county:String) object BlogApp { def main(args: Array[String]) { val conf = new SparkConf().setAppName("Simple Application") val sc = new SparkContext(conf) val sqlContext = new SQLContext(sc) import sqlContext.implicits._ sqlContext.read.format("com.databricks.spark.csv").load("/Users/markneedham/projects/land-registry/pp-complete.csv").registerTempTable("transactions") val rows = sqlContext.sql("select C1,C2,C3,C7,C8,C9,C10,C11,C12,C13 from transactions where transactions.C3 = 'SW3 4EU'").map(x => Row.fromSeq(x.toSeq ++ Array(Array(x.get(4), x.get(3), x.get(5), x.get(6), x.get(7), x.get(8), x.get(9), x.get(2)) .map(x => x.toString) .filter(x => !x.isEmpty) .distinct .mkString(" / ")))) val path: String = "/tmp/tx-" + System.currentTimeMillis() + ".csv" rows.map { case Row(price: Integer, date: String, postCode:String, paon:String, saon:String, street:String, locality:String, city:String, district:String, county:String) => BlogTransaction(price, date, postCode, paon, saon, street, locality, city, district, county) } .toDF() .write .format("com.databricks.spark.csv") .save(path) } } ---- Let's execute that job against a local Spark worker: [source,bash] ---- ./spark-1.5.0-bin-hadoop2.6/bin/spark-submit --class BlogApp --master local[8] --packages com.databricks:spark-csv_2.10:1.2.0 target/scala-2.10/simple-project_2.10-1.0.jar 15/10/27 22:56:41 INFO Executor: Executor killed task 7.0 in stage 1.0 (TID 8) Exception in thread "main" org.apache.spark.SparkException: Job aborted due to stage failure: Task 2 in stage 1.0 failed 1 times, most recent failure: Lost task 2.0 in stage 1.0 (TID 3, localhost): scala.MatchError: [14850000,2013-11-13 00:00,SW3 4EU,9,,ORMONDE GATE,,LONDON,KENSINGTON AND CHELSEA,GREATER LONDON,9 / ORMONDE GATE / LONDON / KENSINGTON AND CHELSEA / GREATER LONDON / SW3 4EU] (of class org.apache.spark.sql.catalyst.expressions.GenericRow) at BlogApp$$anonfun$main$1.apply(BlogApp.scala:24) at BlogApp$$anonfun$main$1.apply(BlogApp.scala:24) at scala.collection.Iterator$$anon$11.next(Iterator.scala:328) at scala.collection.Iterator$$anon$11.next(Iterator.scala:328) at scala.collection.Iterator$$anon$11.next(Iterator.scala:328) at com.databricks.spark.csv.package$CsvSchemaRDD$$anonfun$5$$anon$1.next(package.scala:154) at com.databricks.spark.csv.package$CsvSchemaRDD$$anonfun$5$$anon$1.next(package.scala:147) at scala.collection.Iterator$$anon$11.next(Iterator.scala:328) at org.apache.spark.rdd.PairRDDFunctions$$anonfun$saveAsHadoopDataset$1$$anonfun$13$$anonfun$apply$6.apply$mcV$sp(PairRDDFunctions.scala:1109) at org.apache.spark.rdd.PairRDDFunctions$$anonfun$saveAsHadoopDataset$1$$anonfun$13$$anonfun$apply$6.apply(PairRDDFunctions.scala:1108) at org.apache.spark.rdd.PairRDDFunctions$$anonfun$saveAsHadoopDataset$1$$anonfun$13$$anonfun$apply$6.apply(PairRDDFunctions.scala:1108) at org.apache.spark.util.Utils$.tryWithSafeFinally(Utils.scala:1206) at org.apache.spark.rdd.PairRDDFunctions$$anonfun$saveAsHadoopDataset$1$$anonfun$13.apply(PairRDDFunctions.scala:1116) at org.apache.spark.rdd.PairRDDFunctions$$anonfun$saveAsHadoopDataset$1$$anonfun$13.apply(PairRDDFunctions.scala:1095) at org.apache.spark.scheduler.ResultTask.runTask(ResultTask.scala:66) at org.apache.spark.scheduler.Task.run(Task.scala:88) at org.apache.spark.executor.Executor$TaskRunner.run(Executor.scala:214) at java.util.concurrent.ThreadPoolExecutor.runWorker(ThreadPoolExecutor.java:1142) at java.util.concurrent.ThreadPoolExecutor$Worker.run(ThreadPoolExecutor.java:617) at java.lang.Thread.run(Thread.java:745) Driver stacktrace: at org.apache.spark.scheduler.DAGScheduler.org$apache$spark$scheduler$DAGScheduler$$failJobAndIndependentStages(DAGScheduler.scala:1280) at org.apache.spark.scheduler.DAGScheduler$$anonfun$abortStage$1.apply(DAGScheduler.scala:1268) at org.apache.spark.scheduler.DAGScheduler$$anonfun$abortStage$1.apply(DAGScheduler.scala:1267) at scala.collection.mutable.ResizableArray$class.foreach(ResizableArray.scala:59) at scala.collection.mutable.ArrayBuffer.foreach(ArrayBuffer.scala:47) at org.apache.spark.scheduler.DAGScheduler.abortStage(DAGScheduler.scala:1267) at org.apache.spark.scheduler.DAGScheduler$$anonfun$handleTaskSetFailed$1.apply(DAGScheduler.scala:697) at org.apache.spark.scheduler.DAGScheduler$$anonfun$handleTaskSetFailed$1.apply(DAGScheduler.scala:697) at scala.Option.foreach(Option.scala:236) at org.apache.spark.scheduler.DAGScheduler.handleTaskSetFailed(DAGScheduler.scala:697) at org.apache.spark.scheduler.DAGSchedulerEventProcessLoop.doOnReceive(DAGScheduler.scala:1493) at org.apache.spark.scheduler.DAGSchedulerEventProcessLoop.onReceive(DAGScheduler.scala:1455) at org.apache.spark.scheduler.DAGSchedulerEventProcessLoop.onReceive(DAGScheduler.scala:1444) at org.apache.spark.util.EventLoop$$anon$1.run(EventLoop.scala:48) at org.apache.spark.scheduler.DAGScheduler.runJob(DAGScheduler.scala:567) at org.apache.spark.SparkContext.runJob(SparkContext.scala:1813) at org.apache.spark.SparkContext.runJob(SparkContext.scala:1826) at org.apache.spark.SparkContext.runJob(SparkContext.scala:1903) at org.apache.spark.rdd.PairRDDFunctions$$anonfun$saveAsHadoopDataset$1.apply$mcV$sp(PairRDDFunctions.scala:1124) at org.apache.spark.rdd.PairRDDFunctions$$anonfun$saveAsHadoopDataset$1.apply(PairRDDFunctions.scala:1065) at org.apache.spark.rdd.PairRDDFunctions$$anonfun$saveAsHadoopDataset$1.apply(PairRDDFunctions.scala:1065) at org.apache.spark.rdd.RDDOperationScope$.withScope(RDDOperationScope.scala:147) at org.apache.spark.rdd.RDDOperationScope$.withScope(RDDOperationScope.scala:108) at org.apache.spark.rdd.RDD.withScope(RDD.scala:306) at org.apache.spark.rdd.PairRDDFunctions.saveAsHadoopDataset(PairRDDFunctions.scala:1065) at org.apache.spark.rdd.PairRDDFunctions$$anonfun$saveAsHadoopFile$4.apply$mcV$sp(PairRDDFunctions.scala:989) at org.apache.spark.rdd.PairRDDFunctions$$anonfun$saveAsHadoopFile$4.apply(PairRDDFunctions.scala:965) at org.apache.spark.rdd.PairRDDFunctions$$anonfun$saveAsHadoopFile$4.apply(PairRDDFunctions.scala:965) at org.apache.spark.rdd.RDDOperationScope$.withScope(RDDOperationScope.scala:147) at org.apache.spark.rdd.RDDOperationScope$.withScope(RDDOperationScope.scala:108) at org.apache.spark.rdd.RDD.withScope(RDD.scala:306) at org.apache.spark.rdd.PairRDDFunctions.saveAsHadoopFile(PairRDDFunctions.scala:965) at org.apache.spark.rdd.PairRDDFunctions$$anonfun$saveAsHadoopFile$1.apply$mcV$sp(PairRDDFunctions.scala:897) at org.apache.spark.rdd.PairRDDFunctions$$anonfun$saveAsHadoopFile$1.apply(PairRDDFunctions.scala:897) at org.apache.spark.rdd.PairRDDFunctions$$anonfun$saveAsHadoopFile$1.apply(PairRDDFunctions.scala:897) at org.apache.spark.rdd.RDDOperationScope$.withScope(RDDOperationScope.scala:147) at org.apache.spark.rdd.RDDOperationScope$.withScope(RDDOperationScope.scala:108) at org.apache.spark.rdd.RDD.withScope(RDD.scala:306) at org.apache.spark.rdd.PairRDDFunctions.saveAsHadoopFile(PairRDDFunctions.scala:896) at org.apache.spark.rdd.RDD$$anonfun$saveAsTextFile$1.apply$mcV$sp(RDD.scala:1426) at org.apache.spark.rdd.RDD$$anonfun$saveAsTextFile$1.apply(RDD.scala:1405) at org.apache.spark.rdd.RDD$$anonfun$saveAsTextFile$1.apply(RDD.scala:1405) at org.apache.spark.rdd.RDDOperationScope$.withScope(RDDOperationScope.scala:147) at org.apache.spark.rdd.RDDOperationScope$.withScope(RDDOperationScope.scala:108) at org.apache.spark.rdd.RDD.withScope(RDD.scala:306) at org.apache.spark.rdd.RDD.saveAsTextFile(RDD.scala:1405) at com.databricks.spark.csv.package$CsvSchemaRDD.saveAsCsvFile(package.scala:169) at com.databricks.spark.csv.DefaultSource.createRelation(DefaultSource.scala:165) at org.apache.spark.sql.execution.datasources.ResolvedDataSource$.apply(ResolvedDataSource.scala:170) at org.apache.spark.sql.DataFrameWriter.save(DataFrameWriter.scala:146) at org.apache.spark.sql.DataFrameWriter.save(DataFrameWriter.scala:137) at BlogApp$.main(BlogApp.scala:30) at BlogApp.main(BlogApp.scala) at sun.reflect.NativeMethodAccessorImpl.invoke0(Native Method) at sun.reflect.NativeMethodAccessorImpl.invoke(NativeMethodAccessorImpl.java:62) at sun.reflect.DelegatingMethodAccessorImpl.invoke(DelegatingMethodAccessorImpl.java:43) at java.lang.reflect.Method.invoke(Method.java:497) at org.apache.spark.deploy.SparkSubmit$.org$apache$spark$deploy$SparkSubmit$$runMain(SparkSubmit.scala:672) at org.apache.spark.deploy.SparkSubmit$.doRunMain$1(SparkSubmit.scala:180) at org.apache.spark.deploy.SparkSubmit$.submit(SparkSubmit.scala:205) at org.apache.spark.deploy.SparkSubmit$.main(SparkSubmit.scala:120) at org.apache.spark.deploy.SparkSubmit.main(SparkSubmit.scala) Caused by: scala.MatchError: [14850000,2013-11-13 00:00,SW3 4EU,9,,ORMONDE GATE,,LONDON,KENSINGTON AND CHELSEA,GREATER LONDON,9 / ORMONDE GATE / LONDON / KENSINGTON AND CHELSEA / GREATER LONDON / SW3 4EU] (of class org.apache.spark.sql.catalyst.expressions.GenericRow) at BlogApp$$anonfun$main$1.apply(BlogApp.scala:24) at BlogApp$$anonfun$main$1.apply(BlogApp.scala:24) at scala.collection.Iterator$$anon$11.next(Iterator.scala:328) at scala.collection.Iterator$$anon$11.next(Iterator.scala:328) at scala.collection.Iterator$$anon$11.next(Iterator.scala:328) at com.databricks.spark.csv.package$CsvSchemaRDD$$anonfun$5$$anon$1.next(package.scala:154) at com.databricks.spark.csv.package$CsvSchemaRDD$$anonfun$5$$anon$1.next(package.scala:147) at scala.collection.Iterator$$anon$11.next(Iterator.scala:328) at org.apache.spark.rdd.PairRDDFunctions$$anonfun$saveAsHadoopDataset$1$$anonfun$13$$anonfun$apply$6.apply$mcV$sp(PairRDDFunctions.scala:1109) at org.apache.spark.rdd.PairRDDFunctions$$anonfun$saveAsHadoopDataset$1$$anonfun$13$$anonfun$apply$6.apply(PairRDDFunctions.scala:1108) at org.apache.spark.rdd.PairRDDFunctions$$anonfun$saveAsHadoopDataset$1$$anonfun$13$$anonfun$apply$6.apply(PairRDDFunctions.scala:1108) at org.apache.spark.util.Utils$.tryWithSafeFinally(Utils.scala:1206) at org.apache.spark.rdd.PairRDDFunctions$$anonfun$saveAsHadoopDataset$1$$anonfun$13.apply(PairRDDFunctions.scala:1116) at org.apache.spark.rdd.PairRDDFunctions$$anonfun$saveAsHadoopDataset$1$$anonfun$13.apply(PairRDDFunctions.scala:1095) at org.apache.spark.scheduler.ResultTask.runTask(ResultTask.scala:66) at org.apache.spark.scheduler.Task.run(Task.scala:88) at org.apache.spark.executor.Executor$TaskRunner.run(Executor.scala:214) at java.util.concurrent.ThreadPoolExecutor.runWorker(ThreadPoolExecutor.java:1142) at java.util.concurrent.ThreadPoolExecutor$Worker.run(ThreadPoolExecutor.java:617) at java.lang.Thread.run(Thread.java:745) ---- So it looks like we have something wrong with our matching code and the only place we're matching anything is the +++<cite>+++Row+++</cite>+++ case class when we're mapping over +++<cite>+++rows+++</cite>+++. Although I thought price should be an integer I tweaked it to be a string just in case that was the issue: [source,scala] ---- case class BlogTransaction(price: Integer, date: String, postCode:String, paon:String, saon:String, street:String, locality:String, city:String, district:String, county:String) ... case Row(price: Integer, date: String, postCode:String, paon:String, saon:String, street:String, locality:String, city:String, district:String, county:String) => ---- changed to: [source,scala] ---- case class BlogTransaction(price: String, date: String, postCode:String, paon:String, saon:String, street:String, locality:String, city:String, district:String, county:String) ... case Row(price: String, date: String, postCode:String, paon:String, saon:String, street:String, locality:String, city:String, district:String, county:String) => ---- Attempt #2: [source,bash] ---- ./spark-1.5.0-bin-hadoop2.6/bin/spark-submit --class BlogApp --master local[8] --packages com.databricks:spark-csv_2.10:1.2.0 target/scala-2.10/simple-project_2.10-1.0.jar 15/10/27 23:01:35 WARN TaskSetManager: Lost task 6.0 in stage 1.0 (TID 7, localhost): TaskKilled (killed intentionally) Exception in thread "main" 15/10/27 23:01:35 WARN TaskSetManager: Lost task 1.0 in stage 1.0 (TID 2, localhost): TaskKilled (killed intentionally) org.apache.spark.SparkException: Job aborted due to stage failure: Task 2 in stage 1.0 failed 1 times, most recent failure: Lost task 2.0 in stage 1.0 (TID 3, localhost): scala.MatchError: [14850000,2013-11-13 00:00,SW3 4EU,9,,ORMONDE GATE,,LONDON,KENSINGTON AND CHELSEA,GREATER LONDON,9 / ORMONDE GATE / LONDON / KENSINGTON AND CHELSEA / GREATER LONDON / SW3 4EU] (of class org.apache.spark.sql.catalyst.expressions.GenericRow) at BlogApp$$anonfun$main$1.apply(BlogApp.scala:24) at BlogApp$$anonfun$main$1.apply(BlogApp.scala:24) at scala.collection.Iterator$$anon$11.next(Iterator.scala:328) at scala.collection.Iterator$$anon$11.next(Iterator.scala:328) at scala.collection.Iterator$$anon$11.next(Iterator.scala:328) at com.databricks.spark.csv.package$CsvSchemaRDD$$anonfun$5$$anon$1.next(package.scala:154) at com.databricks.spark.csv.package$CsvSchemaRDD$$anonfun$5$$anon$1.next(package.scala:147) at scala.collection.Iterator$$anon$11.next(Iterator.scala:328) at org.apache.spark.rdd.PairRDDFunctions$$anonfun$saveAsHadoopDataset$1$$anonfun$13$$anonfun$apply$6.apply$mcV$sp(PairRDDFunctions.scala:1109) at org.apache.spark.rdd.PairRDDFunctions$$anonfun$saveAsHadoopDataset$1$$anonfun$13$$anonfun$apply$6.apply(PairRDDFunctions.scala:1108) at org.apache.spark.rdd.PairRDDFunctions$$anonfun$saveAsHadoopDataset$1$$anonfun$13$$anonfun$apply$6.apply(PairRDDFunctions.scala:1108) at org.apache.spark.util.Utils$.tryWithSafeFinally(Utils.scala:1206) at org.apache.spark.rdd.PairRDDFunctions$$anonfun$saveAsHadoopDataset$1$$anonfun$13.apply(PairRDDFunctions.scala:1116) at org.apache.spark.rdd.PairRDDFunctions$$anonfun$saveAsHadoopDataset$1$$anonfun$13.apply(PairRDDFunctions.scala:1095) at org.apache.spark.scheduler.ResultTask.runTask(ResultTask.scala:66) at org.apache.spark.scheduler.Task.run(Task.scala:88) at org.apache.spark.executor.Executor$TaskRunner.run(Executor.scala:214) at java.util.concurrent.ThreadPoolExecutor.runWorker(ThreadPoolExecutor.java:1142) at java.util.concurrent.ThreadPoolExecutor$Worker.run(ThreadPoolExecutor.java:617) at java.lang.Thread.run(Thread.java:745) Driver stacktrace: at org.apache.spark.scheduler.DAGScheduler.org$apache$spark$scheduler$DAGScheduler$$failJobAndIndependentStages(DAGScheduler.scala:1280) at org.apache.spark.scheduler.DAGScheduler$$anonfun$abortStage$1.apply(DAGScheduler.scala:1268) at org.apache.spark.scheduler.DAGScheduler$$anonfun$abortStage$1.apply(DAGScheduler.scala:1267) at scala.collection.mutable.ResizableArray$class.foreach(ResizableArray.scala:59) at scala.collection.mutable.ArrayBuffer.foreach(ArrayBuffer.scala:47) at org.apache.spark.scheduler.DAGScheduler.abortStage(DAGScheduler.scala:1267) at org.apache.spark.scheduler.DAGScheduler$$anonfun$handleTaskSetFailed$1.apply(DAGScheduler.scala:697) at org.apache.spark.scheduler.DAGScheduler$$anonfun$handleTaskSetFailed$1.apply(DAGScheduler.scala:697) at scala.Option.foreach(Option.scala:236) at org.apache.spark.scheduler.DAGScheduler.handleTaskSetFailed(DAGScheduler.scala:697) at org.apache.spark.scheduler.DAGSchedulerEventProcessLoop.doOnReceive(DAGScheduler.scala:1493) at org.apache.spark.scheduler.DAGSchedulerEventProcessLoop.onReceive(DAGScheduler.scala:1455) at org.apache.spark.scheduler.DAGSchedulerEventProcessLoop.onReceive(DAGScheduler.scala:1444) at org.apache.spark.util.EventLoop$$anon$1.run(EventLoop.scala:48) at org.apache.spark.scheduler.DAGScheduler.runJob(DAGScheduler.scala:567) at org.apache.spark.SparkContext.runJob(SparkContext.scala:1813) at org.apache.spark.SparkContext.runJob(SparkContext.scala:1826) at org.apache.spark.SparkContext.runJob(SparkContext.scala:1903) at org.apache.spark.rdd.PairRDDFunctions$$anonfun$saveAsHadoopDataset$1.apply$mcV$sp(PairRDDFunctions.scala:1124) at org.apache.spark.rdd.PairRDDFunctions$$anonfun$saveAsHadoopDataset$1.apply(PairRDDFunctions.scala:1065) at org.apache.spark.rdd.PairRDDFunctions$$anonfun$saveAsHadoopDataset$1.apply(PairRDDFunctions.scala:1065) at org.apache.spark.rdd.RDDOperationScope$.withScope(RDDOperationScope.scala:147) at org.apache.spark.rdd.RDDOperationScope$.withScope(RDDOperationScope.scala:108) at org.apache.spark.rdd.RDD.withScope(RDD.scala:306) at org.apache.spark.rdd.PairRDDFunctions.saveAsHadoopDataset(PairRDDFunctions.scala:1065) at org.apache.spark.rdd.PairRDDFunctions$$anonfun$saveAsHadoopFile$4.apply$mcV$sp(PairRDDFunctions.scala:989) at org.apache.spark.rdd.PairRDDFunctions$$anonfun$saveAsHadoopFile$4.apply(PairRDDFunctions.scala:965) at org.apache.spark.rdd.PairRDDFunctions$$anonfun$saveAsHadoopFile$4.apply(PairRDDFunctions.scala:965) at org.apache.spark.rdd.RDDOperationScope$.withScope(RDDOperationScope.scala:147) at org.apache.spark.rdd.RDDOperationScope$.withScope(RDDOperationScope.scala:108) at org.apache.spark.rdd.RDD.withScope(RDD.scala:306) at org.apache.spark.rdd.PairRDDFunctions.saveAsHadoopFile(PairRDDFunctions.scala:965) at org.apache.spark.rdd.PairRDDFunctions$$anonfun$saveAsHadoopFile$1.apply$mcV$sp(PairRDDFunctions.scala:897) at org.apache.spark.rdd.PairRDDFunctions$$anonfun$saveAsHadoopFile$1.apply(PairRDDFunctions.scala:897) at org.apache.spark.rdd.PairRDDFunctions$$anonfun$saveAsHadoopFile$1.apply(PairRDDFunctions.scala:897) at org.apache.spark.rdd.RDDOperationScope$.withScope(RDDOperationScope.scala:147) at org.apache.spark.rdd.RDDOperationScope$.withScope(RDDOperationScope.scala:108) at org.apache.spark.rdd.RDD.withScope(RDD.scala:306) at org.apache.spark.rdd.PairRDDFunctions.saveAsHadoopFile(PairRDDFunctions.scala:896) at org.apache.spark.rdd.RDD$$anonfun$saveAsTextFile$1.apply$mcV$sp(RDD.scala:1426) at org.apache.spark.rdd.RDD$$anonfun$saveAsTextFile$1.apply(RDD.scala:1405) at org.apache.spark.rdd.RDD$$anonfun$saveAsTextFile$1.apply(RDD.scala:1405) at org.apache.spark.rdd.RDDOperationScope$.withScope(RDDOperationScope.scala:147) at org.apache.spark.rdd.RDDOperationScope$.withScope(RDDOperationScope.scala:108) at org.apache.spark.rdd.RDD.withScope(RDD.scala:306) at org.apache.spark.rdd.RDD.saveAsTextFile(RDD.scala:1405) at com.databricks.spark.csv.package$CsvSchemaRDD.saveAsCsvFile(package.scala:169) at com.databricks.spark.csv.DefaultSource.createRelation(DefaultSource.scala:165) at org.apache.spark.sql.execution.datasources.ResolvedDataSource$.apply(ResolvedDataSource.scala:170) at org.apache.spark.sql.DataFrameWriter.save(DataFrameWriter.scala:146) at org.apache.spark.sql.DataFrameWriter.save(DataFrameWriter.scala:137) at BlogApp$.main(BlogApp.scala:30) at BlogApp.main(BlogApp.scala) at sun.reflect.NativeMethodAccessorImpl.invoke0(Native Method) at sun.reflect.NativeMethodAccessorImpl.invoke(NativeMethodAccessorImpl.java:62) at sun.reflect.DelegatingMethodAccessorImpl.invoke(DelegatingMethodAccessorImpl.java:43) at java.lang.reflect.Method.invoke(Method.java:497) at org.apache.spark.deploy.SparkSubmit$.org$apache$spark$deploy$SparkSubmit$$runMain(SparkSubmit.scala:672) at org.apache.spark.deploy.SparkSubmit$.doRunMain$1(SparkSubmit.scala:180) at org.apache.spark.deploy.SparkSubmit$.submit(SparkSubmit.scala:205) at org.apache.spark.deploy.SparkSubmit$.main(SparkSubmit.scala:120) at org.apache.spark.deploy.SparkSubmit.main(SparkSubmit.scala) Caused by: scala.MatchError: [14850000,2013-11-13 00:00,SW3 4EU,9,,ORMONDE GATE,,LONDON,KENSINGTON AND CHELSEA,GREATER LONDON,9 / ORMONDE GATE / LONDON / KENSINGTON AND CHELSEA / GREATER LONDON / SW3 4EU] (of class org.apache.spark.sql.catalyst.expressions.GenericRow) at BlogApp$$anonfun$main$1.apply(BlogApp.scala:24) at BlogApp$$anonfun$main$1.apply(BlogApp.scala:24) at scala.collection.Iterator$$anon$11.next(Iterator.scala:328) at scala.collection.Iterator$$anon$11.next(Iterator.scala:328) at scala.collection.Iterator$$anon$11.next(Iterator.scala:328) at com.databricks.spark.csv.package$CsvSchemaRDD$$anonfun$5$$anon$1.next(package.scala:154) at com.databricks.spark.csv.package$CsvSchemaRDD$$anonfun$5$$anon$1.next(package.scala:147) at scala.collection.Iterator$$anon$11.next(Iterator.scala:328) at org.apache.spark.rdd.PairRDDFunctions$$anonfun$saveAsHadoopDataset$1$$anonfun$13$$anonfun$apply$6.apply$mcV$sp(PairRDDFunctions.scala:1109) at org.apache.spark.rdd.PairRDDFunctions$$anonfun$saveAsHadoopDataset$1$$anonfun$13$$anonfun$apply$6.apply(PairRDDFunctions.scala:1108) at org.apache.spark.rdd.PairRDDFunctions$$anonfun$saveAsHadoopDataset$1$$anonfun$13$$anonfun$apply$6.apply(PairRDDFunctions.scala:1108) at org.apache.spark.util.Utils$.tryWithSafeFinally(Utils.scala:1206) at org.apache.spark.rdd.PairRDDFunctions$$anonfun$saveAsHadoopDataset$1$$anonfun$13.apply(PairRDDFunctions.scala:1116) at org.apache.spark.rdd.PairRDDFunctions$$anonfun$saveAsHadoopDataset$1$$anonfun$13.apply(PairRDDFunctions.scala:1095) at org.apache.spark.scheduler.ResultTask.runTask(ResultTask.scala:66) at org.apache.spark.scheduler.Task.run(Task.scala:88) at org.apache.spark.executor.Executor$TaskRunner.run(Executor.scala:214) at java.util.concurrent.ThreadPoolExecutor.runWorker(ThreadPoolExecutor.java:1142) at java.util.concurrent.ThreadPoolExecutor$Worker.run(ThreadPoolExecutor.java:617) at java.lang.Thread.run(Thread.java:745) ---- Hmmm\....no improvement. At this point I realised I'd accidentally missed off the +++<cite>+++fullAddress+++</cite>+++ argument from the case statement so I added that in: [source,scala] ---- case class BlogTransaction(price: String, date: String, postCode:String, paon:String, saon:String, street:String, locality:String, city:String, district:String, county:String) ... case Row(price: String, date: String, postCode:String, paon:String, saon:String, street:String, locality:String, city:String, district:String, county:String) => ---- changed to: [source,scala] ---- case class BlogTransaction(price: String, date: String, postCode:String, paon:String, saon:String, street:String, locality:String, city:String, district:String, county:String, fullAddress:String) ... case Row(price: String, date: String, postCode:String, paon:String, saon:String, street:String, locality:String, city:String, district:String, county:String, fullAddress:String) => ---- Attempt #3: [source,bash] ---- ./spark-1.5.0-bin-hadoop2.6/bin/spark-submit --class BlogApp --master local[8] --packages com.databricks:spark-csv_2.10:1.2.0 target/scala-2.10/simple-project_2.10-1.0.jar ... 15/10/27 23:06:03 INFO DAGScheduler: Job 1 finished: saveAsTextFile at package.scala:169, took 39.665661 s ---- Hoorah, it took a bit of guess work but finally it's finally working! For completeness, here's the final version of the Spark job: [source,scala] ---- import org.apache.spark.sql.{SQLContext, _} import org.apache.spark.{SparkConf, SparkContext} case class BlogTransaction(price: Integer, date: String, postCode:String, paon:String, saon:String, street:String, locality:String, city:String, district:String, county:String, fullAddress:String) object BlogApp { def main(args: Array[String]) { val conf = new SparkConf().setAppName("Simple Application") val sc = new SparkContext(conf) val sqlContext = new SQLContext(sc) import sqlContext.implicits._ sqlContext.read.format("com.databricks.spark.csv").load("/Users/markneedham/projects/land-registry/pp-complete.csv").registerTempTable("transactions") val rows = sqlContext.sql("select C1,C2,C3,C7,C8,C9,C10,C11,C12,C13 from transactions where transactions.C3 = 'SW3 4EU'").map(x => Row.fromSeq(x.toSeq ++ Array(Array(x.get(4), x.get(3), x.get(5), x.get(6), x.get(7), x.get(8), x.get(9), x.get(2)) .map(x => x.toString) .filter(x => !x.isEmpty) .distinct .mkString(" / ")))) val path: String = "/tmp/tx-" + System.currentTimeMillis() + ".csv" rows.map { case Row(price: Integer, date: String, postCode:String, paon:String, saon:String, street:String, locality:String, city:String, district:String, county:String, fullAddress:String) => BlogTransaction(price, date, postCode, paon, saon, street, locality, city, district, county, fullAddress) } .toDF() .write .format("com.databricks.spark.csv") .save(path) } } ----

spark-matcherror-of-class-org-apache-spark-sql-catalyst-expressions-genericrow-spark

https://markhneedham.com/blog/2015/10/27/spark-matcherror-of-class-org-apache-spark-sql-catalyst-expressions-genericrow-spark

2015-10-09 06:34:45

jq: error - Cannot iterate over null (null)

I've been playing around with the https://stedolan.github.io/jq/[jq] library again over the past couple of days to convert the JSON from the http://api.stackexchange.com/[Stack Overflow API] into CSV and found myself needing to deal with an optional field. I've downloaded 100 or so questions and stored them as an array in a JSON array like so: [source,bash] ---- $ head -n 100 so.json [ { "has_more": true, "items": [ { "is_answered": false, "delete_vote_count": 0, "body_markdown": "...", "tags": [ "jdbc", "neo4j", "cypher", "spring-data-neo4j" ], "question_id": 33023306, "title": "How to delete multiple nodes by specific ID using Cypher", "down_vote_count": 0, "view_count": 8, "answers": [ { ... ] ---- I wrote the following command to try and extract the answer meta data and the corresponding question_id: [source,bash] ---- $ jq -r \ '.[] | .items[] | { question_id: .question_id, answer: .answers[] } | [.question_id, .answer.answer_id, .answer.title] | @csv' so.json 33023306,33024189,"How to delete multiple nodes by specific ID using Cypher" 33020796,33021958,"How do a general search across string properties in my nodes?" 33018818,33020068,"Neo4j match nodes related to all nodes in collection" 33018818,33024273,"Neo4j match nodes related to all nodes in collection" jq: error (at so.json:134903): Cannot iterate over null (null) ---- Unfortunately this results in an error since some questions haven't been answered yet and therefore don't have the 'answers' property. While reading https://stedolan.github.io/jq/manual/#ConditionalsandComparisons[the docs] I came across the alternative operation '//' which can be used to provide defaults - in this case I thought I could plugin an empty array of answers if a question hadn't been answered yet: [source,bash] ---- $ jq -r \ '.[] | .items[] | { question_id: .question_id, answer: (.answers[] // []) } | [.question_id, .answer.answer_id, .answer.title] | @csv' so.json 33023306,33024189,"How to delete multiple nodes by specific ID using Cypher" 33020796,33021958,"How do a general search across string properties in my nodes?" 33018818,33020068,"Neo4j match nodes related to all nodes in collection" 33018818,33024273,"Neo4j match nodes related to all nodes in collection" jq: error (at so.json:134903): Cannot iterate over null (null) ---- Still the same error! Reading down the page I noticed the ? operator which provides syntactic sugar for handling/catching errors. I gave it a try: [source,bash] ---- $ jq -r '.[] | .items[] | { question_id: .question_id, answer: .answers[]? } | [.question_id, .answer.answer_id, .answer.title] | @csv' so.json | head -n10 33023306,33024189,"How to delete multiple nodes by specific ID using Cypher" 33020796,33021958,"How do a general search across string properties in my nodes?" 33018818,33020068,"Neo4j match nodes related to all nodes in collection" 33018818,33024273,"Neo4j match nodes related to all nodes in collection" 33015714,33021482,"Upgrade of spring data neo4j 3.x to 4.x Relationship Operations" 33011477,33011721,"Why does Neo4j OGM delete method return void?" 33011102,33011565,"Neo4j and algorithms" 33011102,33013260,"Neo4j and algorithms" 33010859,33011505,"Importing data into an existing database in neo4j" 33009673,33010942,"How do I use Spring Data Neo4j to persist a Map (java.util.Map) object inside an NodeEntity?" ---- As far as I can tell we are just skipping any records that don't contain 'answers' which is exactly the behaviour I'm after so that's great - just what we need!

jq-error-cannot-iterate-over-null-null

https://markhneedham.com/blog/2015/10/09/jq-error-cannot-iterate-over-null-null

2015-10-31 23:58:22

Hadoop: HDFS - java.lang.NoSuchMethodError: org.apache.hadoop.fs.FSOutputSummer.<init>(Ljava/util/zip/Checksum;II)V

I wanted to write a little program to check that one machine could communicate a HDFS server running on the other and adapted https://wiki.apache.org/hadoop/HadoopDfsReadWriteExample[some code from the Hadoop wiki] as follows: [source,java] ---- package org.playground; import org.apache.hadoop.conf.Configuration; import org.apache.hadoop.fs.FSDataInputStream; import org.apache.hadoop.fs.FSDataOutputStream; import org.apache.hadoop.fs.FileSystem; import org.apache.hadoop.fs.Path; import java.io.IOException; public class HadoopDFSFileReadWrite { static void printAndExit(String str) { System.err.println( str ); System.exit(1); } public static void main (String[] argv) throws IOException { Configuration conf = new Configuration(); conf.addResource(new Path("/Users/markneedham/Downloads/core-site.xml")); FileSystem fs = FileSystem.get(conf); Path inFile = new Path("hdfs://192.168.0.11/user/markneedham/explore.R"); Path outFile = new Path("hdfs://192.168.0.11/user/markneedham/output-" + System.currentTimeMillis()); // Check if input/output are valid if (!fs.exists(inFile)) printAndExit("Input file not found"); if (!fs.isFile(inFile)) printAndExit("Input should be a file"); if (fs.exists(outFile)) printAndExit("Output already exists"); // Read from and write to new file byte buffer[] = new byte[256]; try ( FSDataInputStream in = fs.open( inFile ); FSDataOutputStream out = fs.create( outFile ) ) { int bytesRead = 0; while ( (bytesRead = in.read( buffer )) > 0 ) { out.write( buffer, 0, bytesRead ); } } catch ( IOException e ) { System.out.println( "Error while copying file" ); } } } ---- I initially thought I only had the following in my POM file: [source,xml] ---- <dependency> <groupId>org.apache.hadoop</groupId> <artifactId>hadoop-common</artifactId> <version>2.7.0</version> </dependency> ---- But when I ran the script I got the following exception: [source,text] ---- Exception in thread "main" java.lang.NoSuchMethodError: org.apache.hadoop.fs.FSOutputSummer.<init>(Ljava/util/zip/Checksum;II)V at org.apache.hadoop.hdfs.DFSOutputStream.<init>(DFSOutputStream.java:1553) at org.apache.hadoop.hdfs.DFSOutputStream.<init>(DFSOutputStream.java:1582) at org.apache.hadoop.hdfs.DFSOutputStream.newStreamForCreate(DFSOutputStream.java:1614) at org.apache.hadoop.hdfs.DFSClient.create(DFSClient.java:1465) at org.apache.hadoop.hdfs.DFSClient.create(DFSClient.java:1390) at org.apache.hadoop.hdfs.DistributedFileSystem$6.doCall(DistributedFileSystem.java:394) at org.apache.hadoop.hdfs.DistributedFileSystem$6.doCall(DistributedFileSystem.java:390) at org.apache.hadoop.fs.FileSystemLinkResolver.resolve(FileSystemLinkResolver.java:81) at org.apache.hadoop.hdfs.DistributedFileSystem.create(DistributedFileSystem.java:390) at org.apache.hadoop.hdfs.DistributedFileSystem.create(DistributedFileSystem.java:334) at org.apache.hadoop.fs.FileSystem.create(FileSystem.java:909) at org.apache.hadoop.fs.FileSystem.create(FileSystem.java:890) at org.apache.hadoop.fs.FileSystem.create(FileSystem.java:787) at org.apache.hadoop.fs.FileSystem.create(FileSystem.java:776) at org.playground.HadoopDFSFileReadWrite.main(HadoopDFSFileReadWrite.java:37) at sun.reflect.NativeMethodAccessorImpl.invoke0(Native Method) at sun.reflect.NativeMethodAccessorImpl.invoke(NativeMethodAccessorImpl.java:62) at sun.reflect.DelegatingMethodAccessorImpl.invoke(DelegatingMethodAccessorImpl.java:43) at java.lang.reflect.Method.invoke(Method.java:497) at com.intellij.rt.execution.application.AppMain.main(AppMain.java:140) ---- From following the stack trace I realised I'd made a mistake and had accidentally pulled in a dependency on hadoop-hdfs 2.4.1. If we don't have the hadoop-hdfs dependency we'd actually see this error instead: [source,text] ---- Exception in thread "main" java.io.IOException: No FileSystem for scheme: hdfs at org.apache.hadoop.fs.FileSystem.getFileSystemClass(FileSystem.java:2644) at org.apache.hadoop.fs.FileSystem.createFileSystem(FileSystem.java:2651) at org.apache.hadoop.fs.FileSystem.access$200(FileSystem.java:92) at org.apache.hadoop.fs.FileSystem$Cache.getInternal(FileSystem.java:2687) at org.apache.hadoop.fs.FileSystem$Cache.get(FileSystem.java:2669) at org.apache.hadoop.fs.FileSystem.get(FileSystem.java:371) at org.apache.hadoop.fs.FileSystem.get(FileSystem.java:170) at org.playground.HadoopDFSFileReadWrite.main(HadoopDFSFileReadWrite.java:22) at sun.reflect.NativeMethodAccessorImpl.invoke0(Native Method) at sun.reflect.NativeMethodAccessorImpl.invoke(NativeMethodAccessorImpl.java:62) at sun.reflect.DelegatingMethodAccessorImpl.invoke(DelegatingMethodAccessorImpl.java:43) at java.lang.reflect.Method.invoke(Method.java:497) at com.intellij.rt.execution.application.AppMain.main(AppMain.java:140) ---- Now let's add the correct version of the dependency and make sure it all works as expected: [source,xml] ---- <dependency> <groupId>org.apache.hadoop</groupId> <artifactId>hadoop-hdfs</artifactId> <version>2.7.0</version> <exclusions> <exclusion> <groupId>ch.qos.logback</groupId> <artifactId>logback-classic</artifactId> </exclusion> <exclusion> <groupId>javax.servlet</groupId> <artifactId>servlet-api</artifactId> </exclusion> </exclusions> </dependency> ---- When we run that a new file is created in HDFS on the other machine with the current timestamp: [source,bash] ---- $ date +%s000 1446336801000 $ hdfs dfs -ls ... -rw-r--r-- 3 markneedham supergroup 9249 2015-11-01 00:13 output-1446337098257 ... ----

hadoop-hdfs-ava-lang-nosuchmethoderror-org-apache-hadoop-fs-fsoutputsummer-ljavautilzipchecksumiiv

https://markhneedham.com/blog/2015/10/31/hadoop-hdfs-ava-lang-nosuchmethoderror-org-apache-hadoop-fs-fsoutputsummer-ljavautilzipchecksumiiv

2015-07-05 08:38:03

R: Wimbledon - How do the seeds get on?

Continuing on with the https://github.com/mneedham/neo4j-wimbledon/blob/master/wimbledon.csv[Wimbledon data set] I've been playing with I wanted to do some exploration on how the seeded players have fared over the years. Taking the last 10 years worth of data there have always had 32 seeds and with the following function we can feed in a seeding and get back the round they would be expected to reach: [source,r] ---- expected_round = function(seeding) { if(seeding == 1) { return("Winner") } else if(seeding == 2) { return("Finals") } else if(seeding <= 4) { return("Semi-Finals") } else if(seeding <= 8) { return("Quarter-Finals") } else if(seeding <= 16) { return("Round of 16") } else { return("Round of 32") } } > expected_round(1) [1] "Winner" > expected_round(4) [1] "Semi-Finals" ---- We can then have a look at each of the Wimbledon tournaments and work out how far they actually got. [source,r] ---- round_reached = function(player, main_matches) { furthest_match = main_matches %>% filter(winner == player | loser == player) %>% arrange(desc(round)) %>% head(1) return(ifelse(furthest_match$winner == player, "Winner", as.character(furthest_match$round))) } seeds = function(matches_to_consider) { winners = matches_to_consider %>% filter(!is.na(winner_seeding)) %>% select(name = winner, seeding = winner_seeding) %>% distinct() losers = matches_to_consider %>% filter( !is.na(loser_seeding)) %>% select(name = loser, seeding = loser_seeding) %>% distinct() return(rbind(winners, losers) %>% distinct() %>% mutate(name = as.character(name))) } ---- Let's have a look how the seeds got on last year: [source,r] ---- matches_to_consider = main_matches %>% filter(year == 2014) result = seeds(matches_to_consider) %>% group_by(name) %>% mutate(expected = expected_round(seeding), round = round_reached(name, matches_to_consider)) %>% ungroup() %>% arrange(seeding) rounds = c("Did not enter", "Round of 128", "Round of 64", "Round of 32", "Round of 16", "Quarter-Finals", "Semi-Finals", "Finals", "Winner") result$round = factor(result$round, levels = rounds, ordered = TRUE) result$expected = factor(result$expected, levels = rounds, ordered = TRUE) > result %>% head(10) Source: local data frame [10 x 4] name seeding expected round 1 Novak Djokovic 1 Winner Winner 2 Rafael Nadal 2 Finals Round of 16 3 Andy Murray 3 Semi-Finals Quarter-Finals 4 Roger Federer 4 Semi-Finals Finals 5 Stan Wawrinka 5 Quarter-Finals Quarter-Finals 6 Tomas Berdych 6 Quarter-Finals Round of 32 7 David Ferrer 7 Quarter-Finals Round of 64 8 Milos Raonic 8 Quarter-Finals Semi-Finals 9 John Isner 9 Round of 16 Round of 32 10 Kei Nishikori 10 Round of 16 Round of 16 ---- We'll wrap all of that code into the following function: [source,R] ---- expectations = function(y, matches) { matches_to_consider = matches %>% filter(year == y) result = seeds(matches_to_consider) %>% group_by(name) %>% mutate(expected = expected_round(seeding), round = round_reached(name, matches_to_consider)) %>% ungroup() %>% arrange(seeding) result$round = factor(result$round, levels = rounds, ordered = TRUE) result$expected = factor(result$expected, levels = rounds, ordered = TRUE) return(result) } ---- Next, instead of showing the round names it'd be cool to come up with numerical value indicating how well the player did: * -1 would mean they lost in the round before their seeding suggested e.g. seed 2 loses in Semi Final * 2 would mean they got 2 rounds further than they should have e.g. Seed 7 reaches the Final The http://www.markhneedham.com/blog/2015/07/02/r-calculating-the-difference-between-ordered-factor-variables/[+++<cite>+++unclass+++</cite>+++ function] comes to our rescue here: [source,r] ---- # expectations plot years = 2005:2014 exp = data.frame() for(y in years) { differences = (expectations(y, main_matches) %>% mutate(expected_n = unclass(expected), round_n = unclass(round), difference = round_n - expected_n))$difference %>% as.numeric() exp = rbind(exp, data.frame(year = rep(y, length(differences)), difference = differences)) } > exp %>% sample_n(10) Source: local data frame [10 x 6] name seeding expected_n round_n difference year 1 Tomas Berdych 6 6 5 -1 2011 2 Tomas Berdych 7 6 6 0 2013 3 Rafael Nadal 2 8 5 -3 2014 4 Fabio Fognini 16 5 4 -1 2014 5 Robin Soderling 13 5 5 0 2009 6 Jurgen Melzer 16 5 5 0 2010 7 Nicolas Almagro 19 4 2 -2 2010 8 Stan Wawrinka 14 5 3 -2 2011 9 David Ferrer 7 6 5 -1 2011 10 Mikhail Youzhny 14 5 5 0 2007 ---- We can then group by the 'difference' column to see how seeds are getting on as a whole: [source,r] ---- > exp %>% count(difference) Source: local data frame [9 x 2] difference n 1 -5 2 2 -4 7 3 -3 24 4 -2 70 5 -1 66 6 0 85 7 1 43 8 2 17 9 3 4 library(ggplot2) ggplot(aes(x = difference, y = n), data = exp %>% count(difference)) + geom_bar(stat = "identity") + scale_x_continuous(limits=c(min(potential), max(potential) + 1)) ---- image::{{<siteurl>}}/uploads/2015/07/2015-07-04_00-45-02.png[2015 07 04 00 45 02,400] So from this visualisation we can see that the most common outcome for a seed is that they reach the round they were expected to reach. There are still a decent number of seeds who do 1 or 2 rounds worse than expected as well though. https://twitter.com/tonkouts[Antonios] suggested doing some analysis of how the seeds fared on a year by year basis - we'll start by looking at what % of them exactly achieved their seeding: [source,R] ---- exp$correct_pred = 0 exp$correct_pred[dt$difference==0] = 1 exp %>% group_by(year) %>% summarise(MeanDiff = mean(difference), PrcCorrect = mean(correct_pred), N=n()) Source: local data frame [10 x 4] year MeanDiff PrcCorrect N 1 2005 -0.6562500 0.2187500 32 2 2006 -0.8125000 0.2812500 32 3 2007 -0.4838710 0.4193548 31 4 2008 -0.9677419 0.2580645 31 5 2009 -0.3750000 0.2500000 32 6 2010 -0.7187500 0.4375000 32 7 2011 -0.7187500 0.0937500 32 8 2012 -0.7500000 0.2812500 32 9 2013 -0.9375000 0.2500000 32 10 2014 -0.7187500 0.1875000 32 ---- Some years are better than others - we can use a https://stat.ethz.ch/R-manual/R-devel/library/stats/html/chisq.test.html[chisq test] to see whether there are any significant differences between the years: [source,r] ---- tbl = table(exp$year, exp$correct_pred) tbl > chisq.test(tbl) Pearson's Chi-squared test data: tbl X-squared = 14.9146, df = 9, p-value = 0.09331 ---- This looks for at least one statistically significant different between the years, although it doesn't look like there are any. We can also try doing a comparison of each year against all the others: [source,r] ---- > pairwise.prop.test(tbl) Pairwise comparisons using Pairwise comparison of proportions data: tbl 2005 2006 2007 2008 2009 2010 2011 2012 2013 2006 1.00 - - - - - - - - 2007 1.00 1.00 - - - - - - - 2008 1.00 1.00 1.00 - - - - - - 2009 1.00 1.00 1.00 1.00 - - - - - 2010 1.00 1.00 1.00 1.00 1.00 - - - - 2011 1.00 1.00 0.33 1.00 1.00 0.21 - - - 2012 1.00 1.00 1.00 1.00 1.00 1.00 1.00 - - 2013 1.00 1.00 1.00 1.00 1.00 1.00 1.00 1.00 - 2014 1.00 1.00 1.00 1.00 1.00 1.00 1.00 1.00 1.00 P value adjustment method: holm ---- 2007/2011 and 2010/2011 show the biggest differences but they're still not significant. Since we have so few data items in each bucket there has to be a really massive difference for it to be significant. The data I used in this post is available on https://gist.github.com/mneedham/190381d14f6f84b801a2[this gist] if you want to look into it and come up with your own analysis.

r-wimbledon-how-do-the-seeds-get-on

https://markhneedham.com/blog/2015/07/05/r-wimbledon-how-do-the-seeds-get-on

2015-07-02 22:55:01

R: Calculating the difference between ordered factor variables

In my continued exploration of https://github.com/mneedham/neo4j-wimbledon/blob/master/wimbledon.csv[Wimbledon data] I wanted to work out whether a player had done as well as their seeding suggested they should. I therefore wanted to work out the difference between the round they reached and the round they were expected to reach. A 'round' in the dataset is an ordered factor variable. These are all the possible values: [source,r] ---- rounds = c("Did not enter", "Round of 128", "Round of 64", "Round of 32", "Round of 16", "Quarter-Finals", "Semi-Finals", "Finals", "Winner") ---- And if we want to factorise a couple of strings into this factor we would do it like this: [source,r] ---- round = factor("Finals", levels = rounds, ordered = TRUE) expected = factor("Winner", levels = rounds, ordered = TRUE) > round [1] Finals 9 Levels: Did not enter < Round of 128 < Round of 64 < Round of 32 < Round of 16 < Quarter-Finals < ... < Winner > expected [1] Winner 9 Levels: Did not enter < Round of 128 < Round of 64 < Round of 32 < Round of 16 < Quarter-Finals < ... < Winner ---- In this case the difference between the actual round and expected round should be -1 - the player was expected to win the tournament but lost in the final. We can calculate that differnce by http://stackoverflow.com/questions/7611810/converting-a-factor-to-numeric-without-losing-information-r-as-numeric-doesn[calling the +++<cite>+++unclass+++</cite>+++ function on each variable]: [source,r] ---- > unclass(round) - unclass(expected) [1] -1 attr(,"levels") [1] "Did not enter" "Round of 128" "Round of 64" "Round of 32" "Round of 16" "Quarter-Finals" [7] "Semi-Finals" "Finals" "Winner" ---- That still seems to have some remnants of the factor variable so to get rid of that we can cast it to a numeric value: [source,r] ---- > as.numeric(unclass(round) - unclass(expected)) [1] -1 ---- And that's it! We can now go and apply this calculation to all seeds to see how they got on.

r-calculating-the-difference-between-ordered-factor-variables

https://markhneedham.com/blog/2015/07/02/r-calculating-the-difference-between-ordered-factor-variables

2015-07-16 06:40:26

Neo4j: The football transfers graph

Given we're still in http://www.bbc.co.uk/sport/0/football/33531715[pre season] http://www.bbc.co.uk/sport/0/football/33528451[transfer madness] as far as European football is concerned I thought it'd be interesting to put together a football transfers graph to see whether there are any interesting insights to be had. It took me a while to find an appropriate source but I eventually came across http://www.transfermarkt.co.uk/premier-league/sommertransfers/wettbewerb/GB1/saison_id/2015[transfermarkt.co.uk] which contains transfers going back at least as far as the start of the Premier League in 1992. I wrote a quick Python script to create a CSV file of all the transfers. This is what the file looks like: [source,bash] ---- $ head -n 10 data/transfers.csv player,from_team,from_team_id,to_team,to_team_id,fee,season Martin Keown,Everton,29,Arsenal FC,11,"2,10 Mill. £",1992-1993 John Jensen,Bröndby IF,206,Arsenal FC,11,"1,12 Mill. £",1992-1993 Alan Miller,Birmingham,337,Arsenal FC,11,,1992-1993 Jim Will,Sheffield Utd.,350,Arsenal FC,11,,1992-1993 David Rocastle,Arsenal FC,11,Leeds,399,"1,68 Mill. £",1992-1993 Perry Groves,Arsenal FC,11,Southampton FC,180,595 Th. £,1992-1993 Ty Gooden,Arsenal FC,11,Wycombe Wand.,2805,?,1992-1993 Geraint Williams,Derby,22,Ipswich Town,677,525 Th. £,1992-1993 Jason Winters,Chelsea U21,9250,Ipswich Town,677,?,1992-1993 ---- I'm going to create the following graph and then we'll write some queries which explore chains of transfers involving players and clubs. image::{{<siteurl>}}/uploads/2015/07/2015-07-15_07-28-11.png[2015 07 15 07 28 11,599] I wrote a few import scripts using Neo4j's http://neo4j.com/docs/stable/query-load-csv.html[LOAD CSV] command, having set up the appropriate indexes first: [source,cypher] ---- create index on :Team(id); create index on :Season(name); create index on :Transfer(description); create index on :Player(name); ---- [source,cypher] ---- // teams load csv with headers from "file:///Users/markneedham/projects/football-transfers/data/teams.csv" as row merge (team:Team {id: toint(row.team_id)}) on create set team.name = row.team; // seasons load csv with headers from "file:///Users/markneedham/projects/football-transfers/data/transfers.csv" as row merge (season:Season {name: row.season}) ON CREATE SET season.starts = toint(split(season.name, "-")[0]); // players load csv with headers from "file:///Users/markneedham/projects/football-transfers/data/transfers.csv" as row merge (player:Player {name: row.player}); // transfers load csv with headers from "file:///Users/markneedham/projects/football-transfers/data/transfers.csv" as row match (from:Team {id: toint(row.from_team_id)}) match (to:Team {id: toint(row.to_team_id)}) match (season:Season {name: row.season}) match (player:Player {name: row.player}) merge (transfer:Transfer {description: row.player + " from " + from.name + " to " + to.name}) merge (transfer)-[:FROM_TEAM]->(from) merge (transfer)-[:TO_TEAM]->(to) merge (transfer)-[:IN_SEASON]->(season) merge (transfer)-[:PLAYER]->(player); // connect transfers match (season)<-[:IN_SEASON]-(transfer:Transfer)-[:PLAYER]->(player) WITH player, season, transfer ORDER BY player.name, season.starts WITH player, COLLECT({s: season, t: transfer}) AS transfers UNWIND range(0, length(transfers)-2) AS idx WITH player, transfers[idx] AS t1, transfers[idx +1] AS t2 WITH player, t1.t AS t1, t2.t AS t2 MERGE (t1)-[:NEXT]->(t2); ---- All the files and scripts are on https://gist.github.com/mneedham/2c0b28ebe2314cb5eab5[this gist] if you want to play around with the data. The only thing you'll need to change is the file path on each of the 'LOAD CSV' lines. The 'connect transfers' query is a bit more complicated than the others - in that one we're first ordering the transfers in ascending order grouped by player and then http://www.markhneedham.com/blog/2015/06/04/neo4j-cypher-step-by-step-to-creating-a-linked-list-of-adjacent-nodes-using-unwind/[creating a linked list] of a player's transfers. Now that we've got the data loaded let's find out which player was transferred the most: [source,cypher] ---- match path = (:Transfer)-[:NEXT*0..]->(transfer:Transfer) where NOT (transfer)-[:NEXT]->() RETURN path ORDER BY LENGTH(path) DESC LIMIT 1 ---- image::{{<siteurl>}}/uploads/2015/07/graph-22.png[Graph 22,599] Which other players have moved teams frequently? [source,cypher] ---- match path = (first:Transfer)-[:NEXT*0..]->(transfer:Transfer), (player)<-[:PLAYER]-(transfer) where NOT ((transfer)-[:NEXT]->()) AND NOT ((first)<-[:NEXT]-()) RETURN player.name, LENGTH(path) AS numberOfTransfers ORDER BY numberOfTransfers DESC LIMIT 10 ==> +--------------------------------------+ ==> | player.name | numberOfTransfers | ==> +--------------------------------------+ ==> | "Craig Bellamy" | 7 | ==> | "David Unsworth" | 6 | ==> | "Andrew Cole" | 6 | ==> | "Peter Crouch" | 6 | ==> | "Les Ferdinand" | 5 | ==> | "Kevin Phillips" | 5 | ==> | "Mark Hughes" | 5 | ==> | "Tommy Wright" | 4 | ==> | "Carl Tiler" | 4 | ==> | "Don Hutchison" | 4 | ==> +--------------------------------------+ ==> 10 rows ---- What are the most frequent combinations of clubs involved in transfers? [source,cypher] ---- match (from)<-[:FROM_TEAM]-(t:Transfer)-[:TO_TEAM]->(to), (t)-[:PLAYER]->(p) RETURN from.name, to.name, COUNT(*) AS times, COLLECT(p.name) AS players ORDER BY times DESC LIMIT 10 ==> +------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------+ ==> | from.name | to.name | times | players | ==> +------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------+ ==> | "West Ham United" | "Queens Park Rangers" | 13 | ["Keith Rowland","Iain Dowie","Tim Breacker","Ludek Miklosko","Bertie Brayley","Terrell Forbes","Steve Lomas","Hogan Ephraim","Nigel Quashie","Danny Gabbidon","Kieron Dyer","Robert Green","Gary O'Neil"] | ==> | "Tottenham Hotspur" | "Portsmouth FC" | 12 | ["Paul Walsh","Andy Turner","Rory Allen","Justin Edinburgh","Tim Sherwood","Teddy Sheringham","Noé Pamarot","Pedro Mendes","Sean Davis","Jermain Defoe","Younès Kaboul","Kevin-Prince Boateng"] | ==> | "Liverpool FC" | "West Ham United" | 12 | ["Julian Dicks","David Burrows","Mike Marsh","Don Hutchison","Neil Ruddock","Titi Camara","Rob Jones","Rigobert Song","Craig Bellamy","Joe Cole","Andy Carroll","Stewart Downing"] | ==> | "Manchester United" | "Everton FC" | 9 | ["Andrey Kanchelskis","John O'Kane","Jesper Blomqvist","Phil Neville","Tim Howard","Louis Saha","Darron Gibson","Sam Byrne","Tom Cleverley"] | ==> | "Newcastle United" | "West Ham United" | 9 | ["Paul Kitson","Shaka Hislop","Stuart Pearce","Wayne Quinn","Lee Bowyer","Kieron Dyer","Scott Parker","Nolberto Solano","Kevin Nolan"] | ==> | "Blackburn Rovers" | "Leicester City" | 9 | ["Steve Agnew","Tim Flowers","Callum Davidson","John Curtis","Keith Gillespie","Craig Hignett","Nils-Eric Johansson","Bruno Berner","Paul Gallagher"] | ==> | "Chelsea FC" | "Southampton FC" | 8 | ["Ken Monkou","Kerry Dixon","Neil Shipperley","Mark Hughes","Paul Hughes","Graeme Le Saux","Jack Cork","Ryan Bertrand"] | ==> | "Birmingham City" | "Coventry City" | 8 | ["David Rennie","John Gayle","Liam Daish","Gary Breen","Stern John","Julian Gray","Lee Carsley","Gary McSheffrey"] | ==> | "Southampton FC" | "Fulham FC" | 8 | ["Micky Adams","Kevin Moore","Terry Hurlock","Maik Taylor","Alan Neilson","Luís Boa Morte","Antti Niemi","Chris Baird"] | ==> | "Portsmouth FC" | "Stoke City" | 8 | ["Kevin Harper","Lewis Buxton","Anthony Pulis","Vincent Péricard","Asmir Begovic","Marc Wilson","Elliot Wheeler","Alex Grant"] | ==> +------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------+ ==> 10 rows ---- Are there ever situations where players get transferred in both directions? [source,cypher] ---- match (from)<-[:FROM_TEAM]-(t:Transfer)-[:TO_TEAM]->(to), (t)-[:PLAYER]->(player) where id(from) < id(to) WITH from, to, COUNT(*) AS times, COLLECT(player.name) AS players match (to)<-[:FROM_TEAM]-(t:Transfer)-[:TO_TEAM]->(from), (t)-[:PLAYER]->(player) RETURN from.name, to.name, times, COUNT(*) as otherWayTimes, players, COLLECT(player.name) AS otherWayPlayers ORDER BY times + otherWayTimes DESC LIMIT 10 ==> +-----------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------+ ==> | from.name | to.name | times | otherWayTimes | players | otherWayPlayers | ==> +-----------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------+ ==> | "Tottenham Hotspur" | "Portsmouth FC" | 12 | 5 | ["Paul Walsh","Andy Turner","Rory Allen","Justin Edinburgh","Tim Sherwood","Teddy Sheringham","Noé Pamarot","Pedro Mendes","Sean Davis","Jermain Defoe","Younès Kaboul","Kevin-Prince Boateng"] | ["Jermain Defoe","Niko Kranjcar","Younès Kaboul","Peter Crouch","Darren Anderton"] | ==> | "West Ham United" | "Liverpool FC" | 4 | 12 | ["Julian Dicks","Daniel Sjölund","Yossi Benayoun","Javier Mascherano"] | ["Stewart Downing","Andy Carroll","Joe Cole","Craig Bellamy","Rigobert Song","Titi Camara","Rob Jones","Neil Ruddock","Don Hutchison","Julian Dicks","Mike Marsh","David Burrows"] | ==> | "West Ham United" | "Queens Park Rangers" | 13 | 2 | ["Keith Rowland","Iain Dowie","Tim Breacker","Ludek Miklosko","Bertie Brayley","Terrell Forbes","Steve Lomas","Hogan Ephraim","Nigel Quashie","Danny Gabbidon","Kieron Dyer","Robert Green","Gary O'Neil"] | ["Andy Impey","Trevor Sinclair"] | ==> | "West Ham United" | "Tottenham Hotspur" | 5 | 8 | ["Jermain Defoe","Frédéric Kanouté","Michael Carrick","Jimmy Walker","Scott Parker"] | ["Sergiy Rebrov","Mauricio Taricco","Calum Davenport","Les Ferdinand","Matthew Etherington","Bobby Zamora","Ilie Dumitrescu","Mark Robson"] | ==> | "West Ham United" | "Portsmouth FC" | 8 | 5 | ["Martin Allen","Adrian Whitbread","Marc Keller","Svetoslav Todorov","Hayden Foxe","Shaka Hislop","Sébastien Schemmel","Hayden Mullins"] | ["Stephen Henderson","Teddy Sheringham","Shaka Hislop","Marc Keller","Lee Chapman"] | ==> | "Newcastle United" | "West Ham United" | 9 | 3 | ["Paul Kitson","Shaka Hislop","Stuart Pearce","Wayne Quinn","Lee Bowyer","Kieron Dyer","Scott Parker","Nolberto Solano","Kevin Nolan"] | ["Demba Ba","Lee Bowyer","David Terrier"] | ==> | "Birmingham City" | "Coventry City" | 8 | 4 | ["David Rennie","John Gayle","Liam Daish","Gary Breen","Stern John","Julian Gray","Lee Carsley","Gary McSheffrey"] | ["Scott Dann","David Burrows","Peter Ndlovu","David Smith"] | ==> | "Manchester City" | "Portsmouth FC" | 8 | 4 | ["Paul Walsh","Carl Griffiths","Fitzroy Simpson","Eyal Berkovic","David James","Andrew Cole","Sylvain Distin","Tal Ben Haim"] | ["Benjani","Gerry Creaney","Kit Symons","Paul Walsh"] | ==> | "Blackburn Rovers" | "Southampton FC" | 5 | 6 | ["David Speedie","Stuart Ripley","James Beattie","Kevin Davies","Zak Jones"] | ["Zak Jones","Egil Östenstad","Kevin Davies","Alan Shearer","Jeff Kenna","Tim Flowers"] | ==> | "AFC Bournemouth" | "West Ham United" | 3 | 8 | ["Keith Rowland","Paul Mitchell","Scott Mean"] | ["Steve Jones","Matt Holland","Mohammed Berthé","Scott Mean","Paul Mitchell","Jamie Victory","Mark Watson","Stephen Purches"] | ==> +-----------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------+ ---- Any players who go back to the same club they were at previously? [source,cypher] ---- match (player:Player)<-[:PLAYER]-(t1:Transfer)-[:FROM_TEAM]->(from)<-[:TO_TEAM]-(t2:Transfer)-[:FROM_TEAM]->(to), (t2)-[:PLAYER]->(player), (t1)-[:TO_TEAM]->(to) WHERE ID(to) < ID(from) WITH player, COLLECT([ from.name, " ⥄ ", to.name]) AS teams RETURN player.name, REDUCE(acc = [], item in teams | acc + REDUCE(acc2 = "", i in item | acc2 + i)) AS thereAndBack ORDER BY LENGTH(thereAndBack) DESC LIMIT 10 ==> +-------------------------------------------------------------------------------------+ ==> | player.name | thereAndBack | ==> +-------------------------------------------------------------------------------------+ ==> | "Mark Stein" | ["Stoke City ⥄ Chelsea FC","Ipswich Town ⥄ Chelsea FC"] | ==> | "Peter Beagrie" | ["Bradford City ⥄ Everton FC","Bradford City ⥄ Wigan Athletic"] | ==> | "Richard Dryden" | ["Southampton FC ⥄ Stoke City","Southampton FC ⥄ Swindon Town"] | ==> | "Robbie Elliott" | ["Bolton Wanderers ⥄ Newcastle United"] | ==> | "Elliot Grandin" | ["Blackpool FC ⥄ Crystal Palace"] | ==> | "Robert Fleck" | ["Chelsea FC ⥄ Norwich City"] | ==> | "Paul Walsh" | ["Portsmouth FC ⥄ Manchester City"] | ==> | "Rick Holden" | ["Manchester City ⥄ Oldham Athletic"] | ==> | "Gary McAllister" | ["Liverpool FC ⥄ Coventry City"] | ==> | "Lee Bowyer" | ["West Ham United ⥄ Newcastle United"] | ==> +-------------------------------------------------------------------------------------+ ---- That's all I've got for now - if you can think of any other interesting avenues to explore let me know and I'll take a look.

neo4j-the-football-transfers-graph

https://markhneedham.com/blog/2015/07/16/neo4j-the-football-transfers-graph

2015-07-28 21:04:58

Neo4j: MERGE'ing on super nodes

In my continued playing with the https://data.cityofchicago.org/Public-Safety/Crimes-2001-to-present/ijzp-q8t2[Chicago crime data set] I wanted to connect the crimes committed to their position in the http://gis.chicagopolice.org/clearmap_crime_sums/crime_types.html[FBI crime type] hierarchy. These are the sub graphs that I want to connect: image::{{<siteurl>}}/uploads/2015/07/2015-07-26_22-19-04.png[2015 07 26 22 19 04,300] We have a 'fbiCode' on each 'Crime' node which indicates which 'Crime Sub Category' the crime belongs to. I started with the following query to connect the nodes together: [source,cypher] ---- MATCH (crime:Crime) WITH crime SKIP {skip} LIMIT 10000 MATCH (subCat:SubCategory {code: crime.fbiCode}) MERGE (crime)-[:CATEGORY]->(subCat) RETURN COUNT(*) AS crimesProcessed ---- I had this running inside a Python script which incremented 'skip' by 10,000 on each iteration as long as 'crimesProcessed' came back with a value > 0. To start with the 'CATEGORY' relationships were being created very quickly but it slowed down quite noticeably about 1 million nodes in. I profiled the queries but the query plans didn't show anything obviously wrong. My suspicion was that I had a super node problem where the cypher run time was iterating through all of the sub category's relationships to check whether one of them pointed to the crime on the other side of the 'MERGE' statement. I cancelled the import job and wrote a query to check how many relationships each sub category had. It varied from 1,000 to 93,000 somewhat confirming my suspicion. https://twitter.com/mesirii[Michael] suggested tweaking the query to use the shortestpath function to check for the existence of the relationship and then use the 'CREATE' clause to create it if it didn't exist. The neat thing about the shortestpath function is that it will start from the side with the lowest cardinality and as soon as it finds a relationship it will stop searching. Let's have a look at that version of the query: [source,cypher] ---- MATCH (crime:Crime) WITH crime SKIP {skip} LIMIT 10000 MATCH (subCat:SubCategory {code: crime.fbiCode}) WITH crime, subCat, shortestPath((crime)-[:CATEGORY]->(subCat)) AS path FOREACH(ignoreMe IN CASE WHEN path is NULL THEN [1] ELSE [] END | CREATE (crime)-[:CATEGORY]->(subCat)) RETURN COUNT(*) ---- This worked much better - 10,000 nodes processed in ~ 2.5 seconds - and the time remained constant as more relationships were added. This allowed me to create all the category nodes but we can actually do even better if we use _http://neo4j.com/docs/stable/query-create-unique.html[CREATE UNIQUE]_ instead of _MERGE_ [source,cypher] ---- MATCH (crime:Crime) WITH crime SKIP {skip} LIMIT 10000 MATCH (subCat:SubCategory {code: crime.fbiCode}) CREATE UNIQUE (crime)-[:CATEGORY]->(subCat) RETURN COUNT(*) AS crimesProcessed ---- Using this query 10,000 nodes took ~ 250ms -900ms second to process which means we can process all the nodes in 5-6 minutes - good times! I'm not super familiar with the 'CREATE UNIQUE' code so I'm not sure that it's always a good substitute for 'MERGE' but on this occasion it does the job. The lesson for me here is that if a query is taking longer than you think it should try and use other constructs / a combination of other constructs and see whether things improve - they just might!

neo4j-mergeing-on-super-nodes

https://markhneedham.com/blog/2015/07/28/neo4j-mergeing-on-super-nodes

2015-07-28 20:05:47

Python: Difference between two datetimes in milliseconds

I've been doing a bit of adhoc measurement of some cypher queries executed via http://py2neo.org/2.0/[py2neo] and wanted to work out how many milliseconds each query was taking end to end. I thought there'd be an obvious way of doing this but if there is it's evaded me so far and I ended up http://stackoverflow.com/questions/1345827/how-do-i-find-the-time-difference-between-two-datetime-objects-in-python[calculating the different between two _datetime_ objects] which gave me the following timedelta object: ~~~python >>> import datetime >>> start = datetime.datetime.now() >>> end = datetime.datetime.now() >>> end - start datetime.timedelta(0, 3, 519319) ~~~ The 3 parts of this object are 'days', 'seconds' and 'microseconds' which I found quite strange! These are the methods/attributes we have available to us: ~~~python >>> dir(end - start) ['__abs__', '__add__', '__class__', '__delattr__', '__div__', '__doc__', '__eq__', '__floordiv__', '__format__', '__ge__', '__getattribute__', '__gt__', '__hash__', '__init__', '__le__', '__lt__', '__mul__', '__ne__', '__neg__', '__new__', '__nonzero__', '__pos__', '__radd__', '__rdiv__', '__reduce__', '__reduce_ex__', '__repr__', '__rfloordiv__', '__rmul__', '__rsub__', '__setattr__', '__sizeof__', '__str__', '__sub__', '__subclasshook__', 'days', 'max', 'microseconds', 'min', 'resolution', 'seconds', 'total_seconds'] ~~~ There's no 'milliseconds' on there so we'll have to calculate it from what we do have: ~~~python >>> diff = end - start >>> elapsed_ms = (diff.days * 86400000) + (diff.seconds * 1000) + (diff.microseconds / 1000) >>> elapsed_ms 3519 ~~~ Or we could do the following slightly simpler calculation: ~~~python >>> diff.total_seconds() * 1000 3519.319 ~~~ And now back to the query profiling!

python-difference-between-two-datetimes-in-milliseconds

https://markhneedham.com/blog/2015/07/28/python-difference-between-two-datetimes-in-milliseconds

2015-07-17 23:34:52

R: Blog post frequency anomaly detection

I came across Twitter's https://github.com/twitter/AnomalyDetection[anomaly detection library] last year but haven't yet had a reason to take it for a test run so having got my blog post frequency data into shape I thought it'd be fun to run it through the algorithm. I wanted to see if it would detect any periods of time when the number of posts differed significantly - I don't really have an action I'm going to take based on the results, it's curiosity more than anything else! First we need to get the library installed. It's not on CRAN so we need to use devtools to install it from the github repository: ~~~ ~~~r install.packages("devtools") devtools::install_github("twitter/AnomalyDetection") library(AnomalyDetection) ~~~ The expected data format is two columns - one containing a time stamp and the other a count. e.g. using the 'raw_data' data frame that is in scope when you add the library: ~~~r > library(dplyr) > raw_data %>% head() timestamp count 1 1980-09-25 14:01:00 182.478 2 1980-09-25 14:02:00 176.231 3 1980-09-25 14:03:00 183.917 4 1980-09-25 14:04:00 177.798 5 1980-09-25 14:05:00 165.469 6 1980-09-25 14:06:00 181.878 ~~~ In our case the timestamps will be the start date of a week and the count the number of posts in that week. But first let's get some practice calling the anomaly function using the canned data: ~~~r res = AnomalyDetectionTs(raw_data, max_anoms=0.02, direction='both', plot=TRUE) res$plot ~~~ image::{{<siteurl>}}/uploads/2015/07/2015-07-18_00-09-22.png[2015 07 18 00 09 22,400] From this visualisation we learn that we should expect both high and low outliers to be identified. Let's give it a try with the blog post publication data. We need to get the data into shape so we'll start by getting a count of the number of blog posts by (week, year) pair: ~~~r > df %>% sample_n(5) title date 1425 Coding: Copy/Paste then refactor 2009-10-31 07:54:31 783 Neo4j 2.0.0-M06 \-> 2.0.0-RC1: Working with path expressions 2013-11-23 10:30:41 960 R: Removing for loops 2015-04-18 23:53:20 966 R: dplyr - Error in (list: invalid subscript type 'double' 2015-04-27 22:34:43 343 Parsing XML from the unix terminal/shell 2011-09-03 23:42:11 > byWeek = df %>% mutate(year = year(date), week = week(date)) %>% group_by(week, year) %>% summarise(n = n()) %>% ungroup() %>% arrange(desc(n)) > byWeek %>% sample_n(5) Source: local data frame [5 x 3] week year n 1 44 2009 6 2 37 2011 4 3 39 2012 3 4 7 2013 4 5 6 2010 6 ~~~ Great. The next step is to translate this data frame into one containing a date representing the start of that week and the number of posts: ~~~r > data = byWeek %>% mutate(start_of_week = calculate_start_of_week(week, year)) %>% filter(start_of_week > ymd("2008-07-01")) %>% select(start_of_week, n) > data %>% sample_n(5) Source: local data frame [5 x 2] start_of_week n 1 2010-09-10 4 2 2013-04-09 4 3 2010-04-30 6 4 2012-03-11 3 5 2014-12-03 3 ~~~ We're now ready to plug it into the anomaly detection function: ~~~R res = AnomalyDetectionTs(data, max_anoms=0.02, direction='both', plot=TRUE) res$plot ~~~ image::{{<siteurl>}}/uploads/2015/07/2015-07-18_00-24-20.png[2015 07 18 00 24 20,500] Interestingly I don't seem to have any low end anomalies - there were a couple of really high frequency weeks when I first started writing and I think one of the other weeks contains a New Year's Eve when I was particularly bored! If we group by month instead only the very first month stands out as an outlier: ~~~r data = byMonth %>% mutate(start_of_month = ymd(paste(year, month, 1, sep="-"))) %>% filter(start_of_month > ymd("2008-07-01")) %>% select(start_of_month, n) res = AnomalyDetectionTs(data, max_anoms=0.02, direction='both', #longterm = TRUE, plot=TRUE) res$plot ~~~ image::{{<siteurl>}}/uploads/2015/07/2015-07-18_00-34-02.png[2015 07 18 00 34 02,500] I'm not sure what else to do as far as anomaly detection goes but if you have any ideas please let me know!

r-blog-post-frequency-anomaly-detection

https://markhneedham.com/blog/2015/07/17/r-blog-post-frequency-anomaly-detection

2015-07-10 22:01:58

R: Date for given week/year

As I mentioned in my http://www.markhneedham.com/blog/2015/07/07/python-converting-wordpress-posts-in-csv-format/[last couple] http://www.markhneedham.com/blog/2015/07/09/r-dplyr-error-cannot-modify-grouping-variable/[of blog posts] I've been looking at the data behind this blog and I wanted to plot a chart showing the number of posts per week since the blog started. I started out with a data frame with posts and publication date: [source,r] ---- > library(dplyr) > df = read.csv("posts.csv") > df$date = ymd_hms(df$date) > df %>% sample_n(10) title date 538 Nygard Big Data Model: The Investigation Stage 2012-10-10 00:00:36 341 The read-only database 2011-08-29 23:32:26 1112 CSS in Internet Explorer - Some lessons learned 2008-10-31 15:24:51 143 Coding: Mutating parameters 2010-08-26 07:47:23 433 Scala: Counting number of inversions (via merge sort) for an unsorted collection 2012-03-20 06:53:18 618 neo4j/cypher: SQL style GROUP BY functionality 2013-02-17 21:05:27 1111 Testing Hibernate mappings: Setting up test data 2008-10-30 13:24:14 462 neo4j: What question do you want to answer? 2012-05-05 13:20:41 1399 Book Club: Design Sense (Michael Feathers) 2009-09-29 14:42:29 494 Bash Shell: Reusing parts of previous commands 2012-07-05 23:42:35 ---- The first step was to add a couple of columns representing the week and year for the publication date. The 'lubridate' library came in handy here: [source,r] ---- byWeek = df %>% mutate(year = year(date), week = week(date)) %>% group_by(week, year) %>% summarise(n = n()) %>% ungroup() %>% arrange(desc(n)) > byWeek Source: local data frame [352 x 3] week year n 1 33 2008 14 2 35 2008 11 3 53 2012 11 4 9 2013 10 5 12 2013 9 6 21 2009 9 7 22 2009 9 8 38 2013 9 9 40 2008 9 10 48 2012 9 .. ... ... .. ---- The next step is to calculate the start date of each of those weeks so that we can plot the counts on a continuous date scale. I spent a while searching how to do this before realising that the 'week' function I used before can set the week for a given data as well. Let's get to work: [source,r] ---- calculate_start_of_week = function(week, year) { date <- ymd(paste(year, 1, 1, sep="-")) week(date) = week return(date) } > calculate_start_of_week(c(1,2,3), c(2015,2014,2013)) [1] "2015-01-01 UTC" "2014-01-08 UTC" "2013-01-15 UTC" ---- And now let's transform our data frame and plot the counts: [source,r] ---- ggplot(aes(x=start_of_week, y=n, group=1), data = byWeek %>% mutate(start_of_week = calculate_start_of_week(week, year))) + geom_line() ---- image::{{<siteurl>}}/uploads/2015/07/2015-07-10_22-43-54.png[2015 07 10 22 43 54,599] It's a bit erratic as you can see. Some of this can be explained by the fact that I do in fact post in an erratic way while some of it is explained by the fact that some weeks only have a few days if they start on the 29th onwards.

r-date-for-given-weekyear

https://markhneedham.com/blog/2015/07/10/r-date-for-given-weekyear

2015-07-19 19:44:59

R: Bootstrap confidence intervals

I recently came across an interesting post on Julia Evans' blog showing how to http://jvns.ca/blog/2015/07/04/bootstrap-confidence-intervals/[generate a bigger set of data points by sampling the small set of data points] that we actually have using *bootstrapping*. Julia's examples are all in Python so I thought it'd be a fun exercise to translate them into R. We're doing the bootstrapping to simulate the number of no-shows for a flight so we can work out how many seats we can overbook the plane by. We start out with a small sample of no-shows and work off the assumption that it's ok to kick someone off a flight 5% of the time. Let's work out how many people that'd be for our initial sample: [source,r] ---- > data = c(0, 1, 3, 2, 8, 2, 3, 4) > quantile(data, 0.05) 5% 0.35 ---- 0.35 people! That's not a particularly useful result so we're going to resample the initial data set 10,000 times, taking the 5%ile each time and see if we come up with something better: We're going to use the +++<cite>+++sample+++</cite>+++ function with replacement to generate our resamples: [source,r] ---- > sample(data, replace = TRUE) [1] 0 3 2 8 8 0 8 0 > sample(data, replace = TRUE) [1] 2 2 4 3 4 4 2 2 ---- Now let's write a function to do that multiple times: [source,r] ---- library(ggplot) bootstrap_5th_percentile = function(data, n_bootstraps) { return(sapply(1:n_bootstraps, function(iteration) quantile(sample(data, replace = TRUE), 0.05))) } values = bootstrap_5th_percentile(data, 10000) ggplot(aes(x = value), data = data.frame(value = values)) + geom_histogram(binwidth=0.25) ---- image::{{<siteurl>}}/uploads/2015/07/2015-07-19_18-05-48.png[2015 07 19 18 05 48,300] So this visualisation is telling us that we can oversell by 0-2 people but we don't know an exact number. Let's try the same exercise but with a bigger initial data set of 1,000 values rather than just 8. First we'll generate a distribution (with a mean of 5 and standard deviation of 2) and visualise it: [source,r] ---- library(dplyr) df = data.frame(value = rnorm(1000,5, 2)) df = df %>% filter(value >= 0) %>% mutate(value = as.integer(round(value))) ggplot(aes(x = value), data = df) + geom_histogram(binwidth=1) ---- image::{{<siteurl>}}/uploads/2015/07/2015-07-19_18-09-15.png[2015 07 19 18 09 15,300] Our distribution seems to have a lot more values around 4 & 5 whereas the Python version has a flatter distribution - I'm not sure why that is so if you have any ideas let me know. In any case, let's check the 5%ile for this data set: [source,R] ---- > quantile(df$value, 0.05) 5% 2 ---- Cool! Now at least we have an integer value rather than the 0.35 we got earlier. Finally let's do some bootstrapping over our new distribution and see what 5%ile we come up with: [source,r] ---- resampled = bootstrap_5th_percentile(df$value, 10000) byValue = data.frame(value = resampled) %>% count(value) > byValue Source: local data frame [3 x 2] value n 1 1.0 3 2 1.7 2 3 2.0 9995 ggplot(aes(x = value, y = n), data = byValue) + geom_bar(stat = "identity") ---- image::{{<siteurl>}}/uploads/2015/07/2015-07-19_18-23-29.png[2015 07 19 18 23 29,300] '2' is by far the most popular 5%ile here although it seems weighted more towards that value than with Julia's Python version, which I imagine is because we seem to have sampled from a slightly different distribution.

r-bootstrap-confidence-intervals

https://markhneedham.com/blog/2015/07/19/r-bootstrap-confidence-intervals

2015-07-21 06:11:25

Neo4j 2.2.3: neo4j-import - Encoder StringEncoder[2] returned an illegal encoded value 0

I've been playing around with the Chicago crime data set again while preparing for a Neo4j webinar next week and while running the http://neo4j.com/docs/stable/import-tool.html[import tool] ran into the following exception: [source,bash] ---- Importing the contents of these files into tmp/crimes.db: Nodes: /Users/markneedham/projects/neo4j-spark-chicago/tmp/crimes.csv /Users/markneedham/projects/neo4j-spark-chicago/tmp/beats.csv /Users/markneedham/projects/neo4j-spark-chicago/tmp/primaryTypes.csv /Users/markneedham/projects/neo4j-spark-chicago/tmp/locations.csv Relationships: /Users/markneedham/projects/neo4j-spark-chicago/tmp/crimesBeats.csv /Users/markneedham/projects/neo4j-spark-chicago/tmp/crimesPrimaryTypes.csv /Users/markneedham/projects/neo4j-spark-chicago/tmp/crimesLocationsCleaned.csv Available memory: Free machine memory: 263.17 MB Max heap memory : 3.56 GB Nodes [*>:17.41 MB/s-------------------------|PROPERTIES(3)=|NODE:3|LABEL SCAN----|v:36.30 MB/s(2)===] 3MImport error: Panic called, so exiting java.lang.RuntimeException: Panic called, so exiting at org.neo4j.unsafe.impl.batchimport.staging.AbstractStep.assertHealthy(AbstractStep.java:200) at org.neo4j.unsafe.impl.batchimport.staging.AbstractStep.await(AbstractStep.java:191) at org.neo4j.unsafe.impl.batchimport.staging.ProcessorStep.receive(ProcessorStep.java:98) at org.neo4j.unsafe.impl.batchimport.staging.ProcessorStep.sendDownstream(ProcessorStep.java:224) at org.neo4j.unsafe.impl.batchimport.staging.ProcessorStep.access$400(ProcessorStep.java:42) at org.neo4j.unsafe.impl.batchimport.staging.ProcessorStep$Sender.send(ProcessorStep.java:250) at org.neo4j.unsafe.impl.batchimport.LabelScanStorePopulationStep.process(LabelScanStorePopulationStep.java:60) at org.neo4j.unsafe.impl.batchimport.LabelScanStorePopulationStep.process(LabelScanStorePopulationStep.java:37) at org.neo4j.unsafe.impl.batchimport.staging.ProcessorStep$4.run(ProcessorStep.java:120) at org.neo4j.unsafe.impl.batchimport.staging.ProcessorStep$4.run(ProcessorStep.java:102) at org.neo4j.unsafe.impl.batchimport.executor.DynamicTaskExecutor$Processor.run(DynamicTaskExecutor.java:237) Caused by: java.lang.IllegalStateException: Encoder StringEncoder[2] returned an illegal encoded value 0 at org.neo4j.unsafe.impl.batchimport.cache.idmapping.string.EncodingIdMapper.encode(EncodingIdMapper.java:229) at org.neo4j.unsafe.impl.batchimport.cache.idmapping.string.EncodingIdMapper.put(EncodingIdMapper.java:208) at org.neo4j.unsafe.impl.batchimport.NodeEncoderStep.process(NodeEncoderStep.java:77) at org.neo4j.unsafe.impl.batchimport.NodeEncoderStep.process(NodeEncoderStep.java:43) ... 3 more ---- I narrowed the problem down to a specific file and from tracing the code learned that this exception happens when we've ended up with a node that doesn't have an id. I guessed that this might be due to there being an empty column somewhere in my CSV file so I did a bit of grepping: [source,bash] ---- $ grep -rn "\"\"" tmp/locations.csv tmp/locations.csv:11:"",Location ---- We can now narrow down the import to just the one line and see if we still get the exception: [source,bash] ---- $ cat foo.csv id:ID(Location),:LABEL "",Location $ ./neo4j-community-2.2.3/bin/neo4j-import --into tmp/foo --nodes foo.csv Importing the contents of these files into tmp/foo: Nodes: /Users/markneedham/projects/neo4j-spark-chicago/foo.csv Available memory: Free machine memory: 2.22 GB Max heap memory : 3.56 GB Nodes Import error: Encoder StringEncoder[2] returned an illegal encoded value 0 ---- Yep, same error. Now we can clean up our CSV file and try again: [source,bash] ---- $ grep -v "\"\"" foo.csv > fooCleaned.csv # I put in a few real records so we can see them import $ cat fooCleaned.csv id:ID(Location),:LABEL "RAILROAD PROPERTY",Location "NEWSSTAND",Location "SCHOOL, PRIVATE, BUILDING",Location $ ./neo4j-community-2.2.3/bin/neo4j-import --into tmp/foo --nodes fooCleaned.csv Importing the contents of these files into tmp/foo: Nodes: /Users/markneedham/projects/neo4j-spark-chicago/fooCleaned.csv Available memory: Free machine memory: 1.23 GB Max heap memory : 3.56 GB Nodes [*>:??-------------------------------------------------|PROPE|NODE:7.63 MB-----------------|LA] 10k Done in 110ms Prepare node index [*DETECT:7.63 MB-------------------------------------------------------------------------------] 0 Done in 60ms Calculate dense nodes [*>:??-----------------------------------------------------------------------------------------] 0 Done in 10ms Relationships [*>:??-----------------------------------------------------------------------------------------] 0 Done in 11ms Node --> Relationship [*v:??-----------------------------------------------------------------------------------------] 10k Done in 1ms Relationship --> Relationship [*>:??-----------------------------------------------------------------------------------------] 0 Done in 11ms Node counts [>:|*COUNT:76.29 MB----------------------------------------------------------------------------] 10k Done in 46ms Relationship counts [*>:??-----------------------------------------------------------------------------------------] 0 Done in 12ms IMPORT DONE in 1s 576ms. Imported: 3 nodes 0 relationships 3 properties ---- Sweet! We're back in business.

neo4j-2-2-3-neo4j-import-encoder-stringencoder2-returned-an-illegal-encoded-value-0

https://markhneedham.com/blog/2015/07/21/neo4j-2-2-3-neo4j-import-encoder-stringencoder2-returned-an-illegal-encoded-value-0

2015-07-07 06:28:01

Python: Converting WordPress posts in CSV format

Over the weekend I wanted to look into the Wordpress data behind this blog (very meta!) and wanted to get the data in CSV format so I could do some analysis in R. image::{{<siteurl>}}/uploads/2015/07/2015-07-07_06-59-02.png[2015 07 07 06 59 02,313] I found a couple of WordPress CSV plugins but unfortunately I couldn't get any of them to work and ended up working with the raw XML data that WordPress produces when you 'export' a blog. I had the problem of the http://beerpla.net/2012/04/13/how-to-fix-incomplete-wordpress-wxr-exports/[export being incomplete] which I 'solved' by importing the posts in two parts of a few years each. I then spent quite a few hours struggling to get the data into shape using R's http://blog.rstudio.org/2014/11/24/rvest-easy-web-scraping-with-r/[rvest] library but eventually decided to do the scraping using Python's beautifulsoup and save it to a CSV file for analysis in R. The structure of the XML that we want to extract is as follows: [source,text] ---- <rss version="2.0" xmlns:excerpt="http://wordpress.org/export/1.2/excerpt/" xmlns:content="http://purl.org/rss/1.0/modules/content/" xmlns:wfw="http://wellformedweb.org/CommentAPI/" xmlns:dc="http://purl.org/dc/elements/1.1/" xmlns:wp="http://wordpress.org/export/1.2/" > ... <channel> <item> <title>First thoughts on Ruby...</title> <link>http://www.markhneedham.com/blog/2006/08/29/first-thoughts-on-ruby/</link> <pubDate>Tue, 29 Aug 2006 13:31:05 +0000</pubDate> ... ---- I wrote the following script to parse the files: [source,python] ---- from bs4 import BeautifulSoup from soupselect import select from dateutil import parser import csv def read_page(page): return BeautifulSoup(open(page, 'r').read()) with open("posts.csv", "w") as file: writer = csv.writer(file, delimiter=",") writer.writerow(["title", "date"]) for row in select(read_page("part2.xml"), "item"): title = select(row, "title")[0].text.encode("utf-8") date = parser.parse(select(row, "pubdate")[0].text) writer.writerow([title, date]) for row in select(read_page("part1.xml"), "item"): title = select(row, "title")[0].text.encode("utf-8") date = parser.parse(select(row, "pubdate")[0].text) writer.writerow([title, date]) ---- We end up with a CSV file that looks like this: [source,bash] ---- $ head -n 10 posts.csv title,date Functional C#: Writing a 'partition' function,2010-02-01 23:34:02+00:00 Coding: Wrapping/not wrapping 3rd party libraries and DSLs,2010-02-02 23:54:21+00:00 Functional C#: LINQ vs Method chaining,2010-02-05 18:06:28+00:00 F#: function keyword,2010-02-07 02:54:13+00:00 Willed vs Forced designs,2010-02-08 22:48:05+00:00 Functional C#: Extracting a higher order function with generics,2010-02-08 23:17:47+00:00 Javascript: File encoding when using string.replace,2010-02-10 00:02:02+00:00 F#: Inline functions and statically resolved type parameters,2010-02-10 23:06:14+00:00 Javascript: Passing functions around with call and apply,2010-02-12 20:18:02+00:00 ---- Let's quickly look over the data in R and check it's being correctly exported: [source,r] ---- require(dplyr) require(lubridate) df = read.csv("posts.csv") > df %>% count() Source: local data frame [1 x 1] n 1 1501 ---- So we've exported 1501 posts. Let's cross check with the WordPress dashboard: image::{{<siteurl>}}/uploads/2015/07/2015-07-07_07-06-02.png[2015 07 07 07 06 02,598] We've gained two extra posts! A bit more exploration of the WordPress dashboard reveals that there are actually 2 draft posts lying around. We probably want to remove those from the export and luckily there's a 'status' tag for each post that we can check. We want to make sure it doesn't have the value 'draft': [source,python] ---- from bs4 import BeautifulSoup from soupselect import select from dateutil import parser import csv def read_page(page): return BeautifulSoup(open(page, 'r').read()) with open("posts.csv", "w") as file: writer = csv.writer(file, delimiter=",") writer.writerow(["title", "date"]) for row in select(read_page("part2.xml"), "item"): if (not row.find("wp:status")) or row.find("wp:status").text != "draft": title = select(row, "title")[0].text.encode("utf-8") date = parser.parse(select(row, "pubdate")[0].text) writer.writerow([title, date]) for row in select(read_page("part1.xml"), "item"): if (not row.find("wp:status")) or row.find("wp:status").text != "draft": title = select(row, "title")[0].text.encode("utf-8") date = parser.parse(select(row, "pubdate")[0].text) writer.writerow([title, date]) ---- I also had to check if that tag actually existed since there were a couple of posts which didn't have it but had been published. If we check the resulting CSV file in R we can see that we've now got all the posts: [source,r] ---- > df = read.csv("posts.csv") > df %>% count() Source: local data frame [1 x 1] n 1 1499 ---- Now we're ready to test a couple of hypotheses that I have but that's for another post!

python-converting-wordpress-posts-in-csv-format

https://markhneedham.com/blog/2015/07/07/python-converting-wordpress-posts-in-csv-format

2015-07-09 05:55:33

R: dplyr - Error: cannot modify grouping variable

I've been doing some exploration of the posts made on this blog and I thought I'd start with answering a simple question - on which dates did I write the most posts? I started with a data frame containing each post and the date it was published: [source,r] ---- > library(dplyr) > df %>% sample_n(5) title date 1148 Taiichi Ohno's Workplace Management: Book Review 2008-12-08 14:14:48 158 Rails: Faking a delete method with 'form_for' 2010-09-20 18:52:15 331 Retrospectives: The 4 L's Retrospective 2011-07-25 21:00:30 1035 msbuild - Use OutputPath instead of OutDir 2008-08-14 18:54:03 1181 The danger of commenting out code 2009-01-17 06:02:33 ---- To find the most popular days for blog posts we can write the following aggregation function: [source,r] ---- > df %>% mutate(day = as.Date(date)) %>% count(day) %>% arrange(desc(n)) Source: local data frame [1,140 x 2] day n 1 2012-12-31 6 2 2014-05-31 6 3 2008-08-08 5 4 2013-01-27 5 5 2009-08-24 4 6 2012-06-24 4 7 2012-09-30 4 8 2012-10-27 4 9 2012-11-24 4 10 2013-02-28 4 ---- So we can see a couple of days with 6 posts, a couple with 5 posts, a few more with 4 posts and then presumably loads of days with 1 post. I thought it'd be cool if we could blog a histogram which had on the x axis the number of posts and on the y axis how many days that number of posts occurred e.g. for an x value of 6 (posts) we'd have a y value of 2 (occurrences). My initial attempt was this: [source,r] ---- > df %>% mutate(day = as.Date(date)) %>% count(day) %>% count(n) Error: cannot modify grouping variable ---- Unfortunately that isn't allowed. I tried ungrouping and then counting again: [source,r] ---- df %>% mutate(day = as.Date(date)) %>% count(day) %>% ungroup() %>% count(n) Error: cannot modify grouping variable ---- Still no luck. I did a bit of googlign around and came across a post which http://stackoverflow.com/questions/30243299/dplyr-error-cannot-modify-grouping-variable-even-when-first-applying-ungroup[suggested using a combination of group_by + mutate or group_by + summarize]. I tried the mutate approach first: [source,r] ---- > df %>% mutate(day = as.Date(date)) %>% + group_by(day) %>% mutate(n = n()) %>% ungroup() %>% sample_n(5) title Source: local data frame [5 x 4] title date day n 1 QCon London 2009: DDD & BDD - Dan North 2009-03-13 15:28:04 2009-03-13 2 2 Onboarding: Sketch the landscape 2013-02-15 07:36:06 2013-02-15 1 3 Ego Depletion 2013-06-04 23:16:29 2013-06-04 1 4 Clean Code: Book Review 2008-09-15 09:52:33 2008-09-15 1 5 Dreyfus Model: More thoughts 2009-08-10 10:36:51 2009-08-10 1 ---- That keeps around the 'title' which is a bit annoying. We can get rid of it using a distinct on 'day' if we want and if we also implement the second part of the function we end up with the following: [source,r] ---- > df %>% mutate(day = as.Date(date)) %>% group_by(day) %>% mutate(n = n()) %>% distinct(day) %>% ungroup() %>% group_by(n) %>% mutate(c = n()) %>% distinct(n) Source: local data frame [6 x 5] Groups: n title date day n c 1 Functional C#: Writing a 'partition' function 2010-02-01 23:34:02 2010-02-01 1 852 2 Willed vs Forced designs 2010-02-08 22:48:05 2010-02-08 2 235 3 TDD: Testing collections 2010-07-28 06:05:25 2010-07-28 3 41 4 Creating a Samba share between Ubuntu and Mac OS X 2012-06-24 00:40:35 2012-06-24 4 8 5 Gamification and Software: Some thoughts 2012-12-31 10:57:19 2012-12-31 6 2 6 Python/numpy: Selecting specific column in 2D array 2013-01-27 02:10:10 2013-01-27 5 2 ---- Annoyingly we've still got the 'title', 'date' and 'day' columns hanging around which we'd need to get rid of with a call to 'select'. The code also feels quite icky, especially the use of distinct in a couple of places. In fact we can simplify the code if we use summarize instead of mutate: [source,r] ---- > df %>% mutate(day = as.Date(date)) %>% group_by(day) %>% summarize(n = n()) %>% ungroup() %>% group_by(n) %>% summarize(c = n()) Source: local data frame [6 x 2] n c 1 1 852 2 2 235 3 3 41 4 4 8 5 5 2 6 6 2 ---- And we've got also rid of the extra columns in the bargain which is great! And now we can plot our histogram: [source,r] ---- > library(ggplot2) > post_frequencies = df %>% mutate(day = as.Date(date)) %>% group_by(day) %>% summarize(n = n()) %>% ungroup() %>% group_by(n) %>% summarize(c = n()) > ggplot(aes(x = n, y = c), data = post_frequencies) + geom_bar(stat = "identity") ---- image::{{<siteurl>}}/uploads/2015/07/2015-07-09_06-44-47.png[2015 07 09 06 44 47,520] In this case we don't actually need to do the second grouping to create the bar chart since ggplot will do it for us if we feed it the following data: [source,r] ---- . ggplot(aes(x = n), data = df %>% mutate(day = as.Date(date)) %>% group_by(day) %>% summarize(n = n()) %>% ungroup()) + geom_bar(binwidth = 1) + scale_x_continuous(limits=c(1, 6)) ---- image::{{<siteurl>}}/uploads/2015/07/2015-07-09_06-55-12.png[2015 07 09 06 55 12,517] Still, it's good to know how!

r-dplyr-error-cannot-modify-grouping-variable

https://markhneedham.com/blog/2015/07/09/r-dplyr-error-cannot-modify-grouping-variable

2015-07-30 06:23:03

Neo4j: Cypher - Removing consecutive duplicates

When writing Cypher queries I sometimes find myself wanting to remove consecutive duplicates in collections that I've joined together. e.g we might start with the following query where 1 and 7 appear consecutively: [source,cypher] ---- RETURN [1,1,2,3,4,5,6,7,7,8] AS values ==> +-----------------------+ ==> | values | ==> +-----------------------+ ==> | [1,1,2,3,4,5,6,7,7,8] | ==> +-----------------------+ ==> 1 row ---- We want to end up with [1,2,3,4,5,6,7,8]. We can start by exploding our array and putting consecutive elements next to each other: [source,cypher] ---- WITH [1,1,2,3,4,5,6,7,7,8] AS values UNWIND RANGE(0, LENGTH(values) - 2) AS idx RETURN idx, idx+1, values[idx], values[idx+1] ==> +-------------------------------------------+ ==> | idx | idx+1 | values[idx] | values[idx+1] | ==> +-------------------------------------------+ ==> | 0 | 1 | 1 | 1 | ==> | 1 | 2 | 1 | 2 | ==> | 2 | 3 | 2 | 3 | ==> | 3 | 4 | 3 | 4 | ==> | 4 | 5 | 4 | 5 | ==> | 5 | 6 | 5 | 6 | ==> | 6 | 7 | 6 | 7 | ==> | 7 | 8 | 7 | 7 | ==> | 8 | 9 | 7 | 8 | ==> +-------------------------------------------+ ==> 9 rows ---- Next we can filter out rows which have the same values since that means they have consecutive duplicates: [source,cypher] ---- WITH [1,1,2,3,4,5,6,7,7,8] AS values UNWIND RANGE(0, LENGTH(values) - 2) AS idx WITH values[idx] AS a, values[idx+1] AS b WHERE a <> b RETURN a,b ==> +-------+ ==> | a | b | ==> +-------+ ==> | 1 | 2 | ==> | 2 | 3 | ==> | 3 | 4 | ==> | 4 | 5 | ==> | 5 | 6 | ==> | 6 | 7 | ==> | 7 | 8 | ==> +-------+ ==> 7 rows ---- Now we need to join the collection back together again. Most of the values we want are in field 'b' but we also need to grab the first value from field 'a': [source,cypher] ---- WITH [1,1,2,3,4,5,6,7,7,8] AS values UNWIND RANGE(0, LENGTH(values) - 2) AS idx WITH values[idx] AS a, values[idx+1] AS b WHERE a <> b RETURN COLLECT(a)[0] + COLLECT(b) AS noDuplicates ==> +-------------------+ ==> | noDuplicates | ==> +-------------------+ ==> | [1,2,3,4,5,6,7,8] | ==> +-------------------+ ==> 1 row ---- What about if we have more than 2 duplicates in a row? [source,cypher] ---- WITH [1,1,1,2,3,4,5,5,6,7,7,8] AS values UNWIND RANGE(0, LENGTH(values) - 2) AS idx WITH values[idx] AS a, values[idx+1] AS b WHERE a <> b RETURN COLLECT(a)[0] + COLLECT(b) AS noDuplicates ==> +-------------------+ ==> | noDuplicates | ==> +-------------------+ ==> | [1,2,3,4,5,6,7,8] | ==> +-------------------+ ==> 1 row ---- Still happy, good times! Of course if we have a non consecutive duplicate that wouldn't be removed: [source,cypher] ---- WITH [1,1,1,2,3,4,5,5,6,7,7,8,1] AS values UNWIND RANGE(0, LENGTH(values) - 2) AS idx WITH values[idx] AS a, values[idx+1] AS b WHERE a <> b RETURN COLLECT(a)[0] + COLLECT(b) AS noDuplicates ==> +---------------------+ ==> | noDuplicates | ==> +---------------------+ ==> | [1,2,3,4,5,6,7,8,1] | ==> +---------------------+ ==> 1 row ----

neo4j-cypher-removing-consecutive-duplicates

https://markhneedham.com/blog/2015/07/30/neo4j-cypher-removing-consecutive-duplicates

2015-07-23 06:15:11

Neo4j: Loading JSON documents with Cypher

One of the most commonly asked questions I get asked is how to load JSON documents into Neo4j and although Cypher doesn't have a 'LOAD JSON' command we can still get JSON data into the graph. http://gist.asciidoctor.org/?dropbox-14493611%2Fblog%2Fadoc%2Fload_json.adoc[Michael shows how to do this from various languages in this blog post] and I recently wanted to load a JSON document that I generated from http://gis.chicagopolice.org/clearmap_crime_sums/crime_types.html[Chicago crime types]. This is a snippet of the JSON document: [source,json] ---- { "categories": [ { "name": "Index Crime", "sub_categories": [ { "code": "01A", "description": "Homicide 1st & 2nd Degree" } ] }, { "name": "Non-Index Crime", "sub_categories": [ { "code": "01B", "description": "Involuntary Manslaughter" } ] }, { "name": "Violent Crime", "sub_categories": [ { "code": "01A", "description": "Homicide 1st & 2nd Degree" } ] } ] } ---- We want to create the following graph structure from this document: image::{{<siteurl>}}/uploads/2015/07/2015-07-23_06-46-50.png[2015 07 23 06 46 50,300] We can then connect the crimes to the appropriate sub category and write aggregation queries that drill down from the category. To do this we're going to have to pass the JSON document to Neo4j via its HTTP API rather than through the browser. Luckily there are http://neo4j.com/developer/language-guides/[drivers available for {insert your favourite language here}] so we should still be good. Python is my current goto language so I'm going to use http://py2neo.org/2.0/[py2neo] to load the data in. Let's start by writing a simple query which passes our JSON document in and gets it straight back. Note that I've updated my Neo4j password to be 'foobar' - replace that with your equivalent if you're following along: [source,python] ---- import json from py2neo import Graph, authenticate # replace 'foobar' with your password authenticate("localhost:7474", "neo4j", "foobar") graph = Graph() with open('categories.json') as data_file: json = json.load(data_file) query = """ RETURN {json} """ # Send Cypher query. print graph.cypher.execute(query, json = json) ---- [source,bash] ---- $ python import_categories.py | document ---+---------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------- 1 | {u'categories': [{u'name': u'Index Crime', u'sub_categories': [{u'code': u'01A', u'description': u'Homicide 1st & 2nd Degree'}, {u'code': u'02', u'description': u'Criminal Sexual Assault'}, {u'code': u'03', u'description': u'Robbery'}, {u'code': u'04A', u'description': u'Aggravated Assault'}, {u'code': u'04B', u'description': u'Aggravated Battery'}, {u'code': u'05', u'description': u'Burglary'}, {u'code': u'06', u'description': u'Larceny'}, {u'code': u'07', u'description': u'Motor Vehicle Theft'}, {u'code': u'09', u'description': u'Arson'}]}, {u'name': u'Non-Index Crime', u'sub_categories': [{u'code': u'01B', u'description': u'Involuntary Manslaughter'}, {u'code': u'08A', u'description': u'Simple Assault'}, {u'code': u'08B', u'description': u'Simple Battery'}, {u'code': u'10', u'description': u'Forgery & Counterfeiting'}, {u'code': u'11', u'description': u'Fraud'}, {u'code': u'12', u'description': u'Embezzlement'}, {u'code': u'13', u'description': u'Stolen Property'}, {u'code': u'14', u'description': u'Vandalism'}, {u'code': u'15', u'description': u'Weapons Violation'}, {u'code': u'16', u'description': u'Prostitution'}, {u'code': u'17', u'description': u'Criminal Sexual Abuse'}, {u'code': u'18', u'description': u'Drug Abuse'}, {u'code': u'19', u'description': u'Gambling'}, {u'code': u'20', u'description': u'Offenses Against Family'}, {u'code': u'22', u'description': u'Liquor License'}, {u'code': u'24', u'description': u'Disorderly Conduct'}, {u'code': u'26', u'description': u'Misc Non-Index Offense'}]}, {u'name': u'Violent Crime', u'sub_categories': [{u'code': u'01A', u'description': u'Homicide 1st & 2nd Degree'}, {u'code': u'02', u'description': u'Criminal Sexual Assault'}, {u'code': u'03', u'description': u'Robbery'}, {u'code': u'04A', u'description': u'Aggravated Assault'}, {u'code': u'04B', u'description': u'Aggravated Battery'}]}]} ---- It's a bit ugly but we can see that everything's there! Our next step is to extract each category into its own row. We can do this by accessing the 'categories' key in our JSON document and then calling the +++<cite>+++http://neo4j.com/docs/stable/query-unwind.html[UNWIND]+++</cite>+++ function which allows us to expand a collection into a sequence of rows: [source,python] ---- query = """ WITH {json} AS document UNWIND document.categories AS category RETURN category.name """ ---- [source,bash] ---- $ python import_categories.py | category.name ---+----------------- 1 | Index Crime 2 | Non-Index Crime 3 | Violent Crime ---- Now we can create a node for each of those categories. We'll use the +++<cite>+++MERGE+++</cite>+++ command so that we can run this script multiple times without ending up with repeat categories: [source,python] ---- query = """ WITH {json} AS document UNWIND document.categories AS category MERGE (:CrimeCategory {name: category.name}) """ ---- Let's quickly check those categories were correctly imported: [source,cypher] ---- match (category:CrimeCategory) return category ---- image::{{<siteurl>}}/uploads/2015/07/graph-23.png[Graph 23,251] Looking good so far - now for the sub categories. We're going to use the +++<cite>+++UNWIND+++</cite>+++ function to help us out here as well: [source,python] ---- query = """ WITH {json} AS document UNWIND document.categories AS category UNWIND category.sub_categories AS subCategory RETURN category.name, subCategory.code, subCategory.description """ ---- [source,bash] ---- $ python import_categories.py | category.name | subCategory.code | subCategory.description ----+-----------------+------------------+--------------------------- 1 | Index Crime | 01A | Homicide 1st & 2nd Degree 2 | Index Crime | 02 | Criminal Sexual Assault 3 | Index Crime | 03 | Robbery 4 | Index Crime | 04A | Aggravated Assault 5 | Index Crime | 04B | Aggravated Battery 6 | Index Crime | 05 | Burglary 7 | Index Crime | 06 | Larceny 8 | Index Crime | 07 | Motor Vehicle Theft 9 | Index Crime | 09 | Arson 10 | Non-Index Crime | 01B | Involuntary Manslaughter 11 | Non-Index Crime | 08A | Simple Assault 12 | Non-Index Crime | 08B | Simple Battery 13 | Non-Index Crime | 10 | Forgery & Counterfeiting 14 | Non-Index Crime | 11 | Fraud 15 | Non-Index Crime | 12 | Embezzlement 16 | Non-Index Crime | 13 | Stolen Property 17 | Non-Index Crime | 14 | Vandalism 18 | Non-Index Crime | 15 | Weapons Violation 19 | Non-Index Crime | 16 | Prostitution 20 | Non-Index Crime | 17 | Criminal Sexual Abuse 21 | Non-Index Crime | 18 | Drug Abuse 22 | Non-Index Crime | 19 | Gambling 23 | Non-Index Crime | 20 | Offenses Against Family 24 | Non-Index Crime | 22 | Liquor License 25 | Non-Index Crime | 24 | Disorderly Conduct 26 | Non-Index Crime | 26 | Misc Non-Index Offense 27 | Violent Crime | 01A | Homicide 1st & 2nd Degree 28 | Violent Crime | 02 | Criminal Sexual Assault 29 | Violent Crime | 03 | Robbery 30 | Violent Crime | 04A | Aggravated Assault 31 | Violent Crime | 04B | Aggravated Battery ---- Let's give sub categories the +++<cite>+++MERGE+++</cite>+++ treatment too: [source,python] ---- query = """ WITH {json} AS document UNWIND document.categories AS category UNWIND category.sub_categories AS subCategory MERGE (c:CrimeCategory {name: category.name}) MERGE (sc:SubCategory {code: subCategory.code}) ON CREATE SET sc.description = subCategory.description MERGE (c)-[:CHILD]->(sc) """ ---- And finally let's write a query to check what we've imported: [source,cypher] ---- match (category:CrimeCategory)-[:CHILD]->(subCategory) return * ---- image::{{<siteurl>}}/uploads/2015/07/graph-24.png[Graph 24,599] I hadn't realised before running this query is that some sub categories sit under multiple categories so that's quite an interesting insight. The https://github.com/mneedham/neo4j-spark-chicago/blob/master/import_categories.py[final Python script] is available on github - any questions let me know.

neo4j-loading-json-documents-with-cypher

https://markhneedham.com/blog/2015/07/23/neo4j-loading-json-documents-with-cypher

2015-07-15 06:20:07

Python: UnicodeDecodeError: 'ascii' codec can't decode byte 0xe2 in position 0: ordinal not in range(128)

I was recently doing some text scrubbing and had difficulty working out how to remove the '†' character from strings. e.g. I had a string like this: [source,python] ---- >>> u'foo †' u'foo \u2020' ---- I wanted to get rid of the '†' character and then strip any trailing spaces so I'd end up with the string 'foo'. I tried to do this in one call to 'replace': [source,python] ---- >>> u'foo †'.replace(" †", "") Traceback (most recent call last): File "<stdin>", line 1, in <module> UnicodeDecodeError: 'ascii' codec can't decode byte 0xe2 in position 1: ordinal not in range(128) ---- It took me a while to work out that "† " was being treated as ASCII rather than UTF-8. Let's fix that: [source,python] ---- >>> u'foo †'.replace(u' †', "") u'foo' ---- I think the following call to +++<cite>+++unicode+++</cite>+++, which I've http://www.markhneedham.com/blog/2015/05/21/python-unicodeencodeerror-ascii-codec-cant-encode-character-uxfc-in-position-11-ordinal-not-in-range128/[written about before], is equivalent: [source,python] ---- >>> u'foo †'.replace(unicode(' †', "utf-8"), "") u'foo' ---- Now back to the scrubbing!

python-unicodedecodeerror-ascii-codec-cant-decode-byte-0xe2-in-position-0-ordinal-not-in-range128

https://markhneedham.com/blog/2015/07/15/python-unicodedecodeerror-ascii-codec-cant-decode-byte-0xe2-in-position-0-ordinal-not-in-range128

2015-07-12 08:30:40

R: Filling in missing dates with 0s

I wanted to plot a chart showing the number of blog posts published by month and started with the following code which makes use of http://cran.r-project.org/web/packages/zoo/index.html[zoo's] 'as.yearmon' function to add the appropriate column and grouping: [source,r] ---- > library(zoo) > library(dplyr) > df %>% sample_n(5) title date 888 R: Converting a named vector to a data frame 2014-10-31 23:47:26 144 Rails: Populating a dropdown list using 'form_for' 2010-08-31 01:22:14 615 Onboarding: Sketch the landscape 2013-02-15 07:36:06 28 Javascript: The 'new' keyword 2010-03-06 15:16:02 1290 Coding Dojo #16: Reading SUnit code 2009-05-28 23:23:19 > posts_by_date = df %>% mutate(year_mon = as.Date(as.yearmon(date))) %>% count(year_mon) > posts_by_date %>% head(5) year_mon n 1 2006-08-01 1 2 2006-09-01 4 3 2008-02-01 4 4 2008-07-01 2 5 2008-08-01 38 ---- I then plugged the new data frame into ggplot to get the chart: [source,R] ---- > ggplot(aes(x = year_mon, y = n), data = posts_by_date) + geom_line() ---- image::{{<siteurl>}}/uploads/2015/07/2015-07-12_09-07-47.png[2015 07 12 09 07 47,519] The problem with this chart is that it's showing there being 4 posts per month for all the dates between September 2006 and February 2008 even though I didn't write anything! It's doing the same thing between February 2008 and July 2008 too. We can fix that by http://stackoverflow.com/questions/14424565/how-to-fill-date-gaps-in-data-frame[filling in the gaps] with 0s. First we'll create a vector containing every month in the data range contained by our data frame: [source,r] ---- > all_dates = seq(as.Date(as.yearmon(min(df$date))), as.Date(as.yearmon(max(df$date))), by="month") > all_dates [1] "2006-08-01" "2006-09-01" "2006-10-01" "2006-11-01" "2006-12-01" "2007-01-01" "2007-02-01" "2007-03-01" [9] "2007-04-01" "2007-05-01" "2007-06-01" "2007-07-01" "2007-08-01" "2007-09-01" "2007-10-01" "2007-11-01" [17] "2007-12-01" "2008-01-01" "2008-02-01" "2008-03-01" "2008-04-01" "2008-05-01" "2008-06-01" "2008-07-01" [25] "2008-08-01" "2008-09-01" "2008-10-01" "2008-11-01" "2008-12-01" "2009-01-01" "2009-02-01" "2009-03-01" [33] "2009-04-01" "2009-05-01" "2009-06-01" "2009-07-01" "2009-08-01" "2009-09-01" "2009-10-01" "2009-11-01" [41] "2009-12-01" "2010-01-01" "2010-02-01" "2010-03-01" "2010-04-01" "2010-05-01" "2010-06-01" "2010-07-01" [49] "2010-08-01" "2010-09-01" "2010-10-01" "2010-11-01" "2010-12-01" "2011-01-01" "2011-02-01" "2011-03-01" [57] "2011-04-01" "2011-05-01" "2011-06-01" "2011-07-01" "2011-08-01" "2011-09-01" "2011-10-01" "2011-11-01" [65] "2011-12-01" "2012-01-01" "2012-02-01" "2012-03-01" "2012-04-01" "2012-05-01" "2012-06-01" "2012-07-01" [73] "2012-08-01" "2012-09-01" "2012-10-01" "2012-11-01" "2012-12-01" "2013-01-01" "2013-02-01" "2013-03-01" [81] "2013-04-01" "2013-05-01" "2013-06-01" "2013-07-01" "2013-08-01" "2013-09-01" "2013-10-01" "2013-11-01" [89] "2013-12-01" "2014-01-01" "2014-02-01" "2014-03-01" "2014-04-01" "2014-05-01" "2014-06-01" "2014-07-01" [97] "2014-08-01" "2014-09-01" "2014-10-01" "2014-11-01" "2014-12-01" "2015-01-01" "2015-02-01" "2015-03-01" [105] "2015-04-01" "2015-05-01" "2015-06-01" "2015-07-01" ---- Now we need to create a data frame containing those dates and merge it with the original: [source,r] ---- posts_by_date_clean = merge(data.frame(date = all_dates), posts_by_date, by.x='date', by.y='year_mon', all.x=T, all.y=T) > posts_by_date_clean %>% head() date n 1 2006-08-01 1 2 2006-09-01 4 3 2006-10-01 NA 4 2006-11-01 NA 5 2006-12-01 NA 6 2007-01-01 NA ---- We've still got some 'NA' values in there which won't plot so well. Let's set those to 0 and then try and plot our chart again: [source,r] ---- > posts_by_date_clean$n[is.na(posts_by_date_clean$n)] = 0 > ggplot(aes(x = date, y = n), data = posts_by_date_clean) + geom_line() ---- image::{{<siteurl>}}/uploads/2015/07/2015-07-12_09-17-10.png[2015 07 12 09 17 10,524] Much better!

r-filling-in-missing-dates-with-0s

https://markhneedham.com/blog/2015/07/12/r-filling-in-missing-dates-with-0s

2015-07-12 09:53:04

R: I write more in the last week of the month, or do I?

I've been writing on this blog for almost 7 years and have always believed that I write more frequently towards the end of a month. Now that I've got all the data I thought it'd be interesting to test that belief. I started with a data frame containing each post and its publication date and added an extra column which works out how many weeks from the end of the month that post was written: [source,r] ---- > df %>% sample_n(5) title date 946 Python: Equivalent to flatMap for flattening an array of arrays 2015-03-23 00:45:00 175 Ruby: Hash default value 2010-10-16 14:02:37 375 Java/Scala: Runtime.exec hanging/in 'pipe_w' state 2011-11-20 20:20:08 1319 Coding Dojo #18: Groovy Bowling Game 2009-06-26 08:15:23 381 Continuous Delivery: Removing manual scenarios 2011-12-05 23:13:34 calculate_start_of_week = function(week, year) { date <- ymd(paste(year, 1, 1, sep="-")) week(date) = week return(date) } tidy_df = df %>% mutate(year = year(date), week = week(date), week_in_month = ceiling(day(date) / 7), max_week = max(week_in_month), weeks_from_end = max_week - week_in_month, start_of_week = calculate_start_of_week(week, year)) > tidy_df %>% select(date, weeks_from_end, start_of_week) %>% sample_n(5) date weeks_from_end start_of_week 1023 2008-08-08 21:16:02 3 2008-08-05 800 2014-01-31 06:51:06 0 2014-01-29 859 2014-08-14 10:24:52 3 2014-08-13 107 2010-07-10 22:49:52 3 2010-07-09 386 2011-12-20 23:57:51 2 2011-12-17 ---- Next I want to get a count of how many posts were published in a given week. The following code does that transformation for us: [source,r] ---- weeks_from_end_counts = tidy_df %>% group_by(start_of_week, weeks_from_end) %>% summarise(count = n()) > weeks_from_end_counts Source: local data frame [540 x 4] Groups: start_of_week, weeks_from_end start_of_week weeks_from_end year count 1 2006-08-27 0 2006 1 2 2006-08-27 4 2006 3 3 2006-09-03 4 2006 1 4 2008-02-05 3 2008 2 5 2008-02-12 3 2008 2 6 2008-07-15 2 2008 1 7 2008-07-22 1 2008 1 8 2008-08-05 3 2008 8 9 2008-08-12 2 2008 5 10 2008-08-12 3 2008 9 .. ... ... ... ... ---- We group by both 'start_of_week' and 'weeks_from_end' because we could have posts published in the same week but different month and we want to capture that difference. Now we can run a correlation on the data frame to see if there's any relationship between 'count' and 'weeks_from_end': [source,r] ---- > cor(weeks_from_end_counts %>% ungroup() %>% select(weeks_from_end, count)) weeks_from_end count weeks_from_end 1.00000000 -0.08253569 count -0.08253569 1.00000000 ---- This suggests there's a slight negative correlation between the two variables i.e. 'count' decreases as 'weeks_from_end' increases. Let's plug the data frame into a linear model to see how good 'weeks_from_end' is as a predictor of 'count': [source,r] ---- > fit = lm(count ~ weeks_from_end, weeks_from_end_counts) > summary(fit) Call: lm(formula = count ~ weeks_from_end, data = weeks_from_end_counts) Residuals: Min 1Q Median 3Q Max -2.0000 -1.5758 -0.5758 1.1060 8.0000 Coefficients: Estimate Std. Error t value Pr(>|t|) (Intercept) 3.00000 0.13764 21.795 <2e-16 *** weeks_from_end -0.10605 0.05521 -1.921 0.0553 . --- Signif. codes: 0 ‘***’ 0.001 ‘**’ 0.01 ‘*’ 0.05 ‘.’ 0.1 ‘ ’ 1 Residual standard error: 1.698 on 538 degrees of freedom Multiple R-squared: 0.006812, Adjusted R-squared: 0.004966 F-statistic: 3.69 on 1 and 538 DF, p-value: 0.05527 ---- We see a similar result here. The effect of 'weeks_from_end' is worth 0.1 posts per week with a p value of 0.0553 so it's on the border line of being significant. We also have a very low 'R squared' value which suggests the 'weeks_from_end' isn't explaining much of the variation in the data which makes sense given that we didn't see much of a correlation. If we charged on and wanted to predict the number of posts likely to be published in a given week we could run the predict function like this: [source,r] ---- > predict(fit, data.frame(weeks_from_end=c(1,2,3,4,5))) 1 2 3 4 5 2.893952 2.787905 2.681859 2.575812 2.469766 ---- Obviously it's a bit flawed since we could plug in any numeric value we want, even ones that don't make any sense, and it'd still come back with a prediction: [source,r] ---- > predict(fit, data.frame(weeks_from_end=c(30 ,-10))) 1 2 -0.181394 4.060462 ---- I think we'd probably protect against that with a function wrapping our call to predict that doesn't allow 'weeks_from_end' to be greater than 5 or less than 0. So far it looks like my belief is incorrect! I'm a bit dubious about my calculation of 'weeks_from_end' though - it's not completely capturing what I want since in some months the last week only contains a couple of days. Next I'm going to explore whether it makes any difference if I calculate that value by counting the number of days back from the last day of the month rather than using week number.

r-i-write-more-in-the-last-week-of-the-month-or-do-i

https://markhneedham.com/blog/2015/07/12/r-i-write-more-in-the-last-week-of-the-month-or-do-i

2015-07-25 23:05:33

Neo4j: From JSON to CSV to LOAD CSV via jq

In my last blog post I showed http://www.markhneedham.com/blog/2015/07/23/neo4j-loading-json-documents-with-cypher/[how to import a Chicago crime categories & sub categories JSON document] using Neo4j's cypher query language via the py2neo driver. While this is a good approach for people with a developer background, many of the users I encounter aren't developers and favour using Cypher via the Neo4j browser. If we're going to do this we'll need to transform our JSON document into a CSV file so that we can use the LOAD CSV command on it. Michael pointed me to the http://stedolan.github.io/jq/[jq] tool which comes in very handy. To recap, this is a part of the JSON file: [source,text] ---- { "categories": [ { "name": "Index Crime", "sub_categories": [ { "code": "01A", "description": "Homicide 1st & 2nd Degree" }, ] }, { "name": "Non-Index Crime", "sub_categories": [ { "code": "01B", "description": "Involuntary Manslaughter" }, ] }, { "name": "Violent Crime", "sub_categories": [ { "code": "01A", "description": "Homicide 1st & 2nd Degree" }, ] } ] } ---- We want to get one row for each sub category which contains three columns - category name, sub category code, sub category description. First we need to pull out the categories: [source,bash] ---- $ jq ".categories[]" categories.json { "name": "Index Crime", "sub_categories": [ { "code": "01A", "description": "Homicide 1st & 2nd Degree" }, ] } { "name": "Non-Index Crime", "sub_categories": [ { "code": "01B", "description": "Involuntary Manslaughter" }, ] } { "name": "Violent Crime", "sub_categories": [ { "code": "01A", "description": "Homicide 1st & 2nd Degree" }, ] } ---- Next we want to create a row for each sub category with the category alongside it. We can use the pipe function to combine the two selectors: [source,bash] ---- $ jq ".categories[] | {name: .name, sub_category: .sub_categories[]}" categories.json { "name": "Index Crime", "sub_category": { "code": "01A", "description": "Homicide 1st & 2nd Degree" } } ... { "name": "Non-Index Crime", "sub_category": { "code": "01B", "description": "Involuntary Manslaughter" } } ... { "name": "Violent Crime", "sub_category": { "code": "01A", "description": "Homicide 1st & 2nd Degree" } } ---- Now we want to un-nest the sub category: [source,bash] ---- $ jq ".categories[] | {name: .name, sub_category: .sub_categories[]} | [.name, .sub_category.code, .sub_category.description]" categories.json [ "Index Crime", "01A", "Homicide 1st & 2nd Degree" ] [ "Non-Index Crime", "01B", "Involuntary Manslaughter" ] [ "Violent Crime", "01A", "Homicide 1st & 2nd Degree" ] ---- And finally let's use the +++<cite>+++@csv+++</cite>+++ filter to generate CSV lines: [source,bash] ---- $ jq ".categories[] | {name: .name, sub_category: .sub_categories[]} | [.name, .sub_category.code, .sub_category.description] | @csv" categories.json "\"Index Crime\",\"01A\",\"Homicide 1st & 2nd Degree\"" "\"Index Crime\",\"02\",\"Criminal Sexual Assault\"" "\"Index Crime\",\"03\",\"Robbery\"" "\"Index Crime\",\"04A\",\"Aggravated Assault\"" "\"Index Crime\",\"04B\",\"Aggravated Battery\"" "\"Index Crime\",\"05\",\"Burglary\"" "\"Index Crime\",\"06\",\"Larceny\"" "\"Index Crime\",\"07\",\"Motor Vehicle Theft\"" "\"Index Crime\",\"09\",\"Arson\"" "\"Non-Index Crime\",\"01B\",\"Involuntary Manslaughter\"" "\"Non-Index Crime\",\"08A\",\"Simple Assault\"" "\"Non-Index Crime\",\"08B\",\"Simple Battery\"" "\"Non-Index Crime\",\"10\",\"Forgery & Counterfeiting\"" "\"Non-Index Crime\",\"11\",\"Fraud\"" "\"Non-Index Crime\",\"12\",\"Embezzlement\"" "\"Non-Index Crime\",\"13\",\"Stolen Property\"" "\"Non-Index Crime\",\"14\",\"Vandalism\"" "\"Non-Index Crime\",\"15\",\"Weapons Violation\"" "\"Non-Index Crime\",\"16\",\"Prostitution\"" "\"Non-Index Crime\",\"17\",\"Criminal Sexual Abuse\"" "\"Non-Index Crime\",\"18\",\"Drug Abuse\"" "\"Non-Index Crime\",\"19\",\"Gambling\"" "\"Non-Index Crime\",\"20\",\"Offenses Against Family\"" "\"Non-Index Crime\",\"22\",\"Liquor License\"" "\"Non-Index Crime\",\"24\",\"Disorderly Conduct\"" "\"Non-Index Crime\",\"26\",\"Misc Non-Index Offense\"" "\"Violent Crime\",\"01A\",\"Homicide 1st & 2nd Degree\"" "\"Violent Crime\",\"02\",\"Criminal Sexual Assault\"" "\"Violent Crime\",\"03\",\"Robbery\"" "\"Violent Crime\",\"04A\",\"Aggravated Assault\"" "\"Violent Crime\",\"04B\",\"Aggravated Battery\"" ---- The only annoying thing about this output is that all the double quotes are escaped. We can sort that out by passing the '-r' flag when we call jq: [source,bash] ---- $ jq -r ".categories[] | {name: .name, sub_category: .sub_categories[]} | [.name, .sub_category.code, .sub_category.description] | @csv" categories.json "Index Crime","01A","Homicide 1st & 2nd Degree" "Index Crime","02","Criminal Sexual Assault" "Index Crime","03","Robbery" "Index Crime","04A","Aggravated Assault" "Index Crime","04B","Aggravated Battery" "Index Crime","05","Burglary" "Index Crime","06","Larceny" "Index Crime","07","Motor Vehicle Theft" "Index Crime","09","Arson" "Non-Index Crime","01B","Involuntary Manslaughter" "Non-Index Crime","08A","Simple Assault" "Non-Index Crime","08B","Simple Battery" "Non-Index Crime","10","Forgery & Counterfeiting" "Non-Index Crime","11","Fraud" "Non-Index Crime","12","Embezzlement" "Non-Index Crime","13","Stolen Property" "Non-Index Crime","14","Vandalism" "Non-Index Crime","15","Weapons Violation" "Non-Index Crime","16","Prostitution" "Non-Index Crime","17","Criminal Sexual Abuse" "Non-Index Crime","18","Drug Abuse" "Non-Index Crime","19","Gambling" "Non-Index Crime","20","Offenses Against Family" "Non-Index Crime","22","Liquor License" "Non-Index Crime","24","Disorderly Conduct" "Non-Index Crime","26","Misc Non-Index Offense" "Violent Crime","01A","Homicide 1st & 2nd Degree" "Violent Crime","02","Criminal Sexual Assault" "Violent Crime","03","Robbery" "Violent Crime","04A","Aggravated Assault" "Violent Crime","04B","Aggravated Battery" ---- Excellent. The only thing left is to write a header and then direct the output into a CSV file and get it into Neo4j: [source,bash] ---- $ echo "category,sub_category_code,sub_category_description" > categories.csv $ jq -r ".categories[] | {name: .name, sub_category: .sub_categories[]} | [.name, .sub_category.code, .sub_category.description] | @csv " categories.json >> categories.csv ---- [source,bash] ---- $ head -n10 categories.csv category,sub_category_code,sub_category_description "Index Crime","01A","Homicide 1st & 2nd Degree" "Index Crime","02","Criminal Sexual Assault" "Index Crime","03","Robbery" "Index Crime","04A","Aggravated Assault" "Index Crime","04B","Aggravated Battery" "Index Crime","05","Burglary" "Index Crime","06","Larceny" "Index Crime","07","Motor Vehicle Theft" "Index Crime","09","Arson" ---- [source,cypher] ---- LOAD CSV WITH HEADERS FROM "file:///Users/markneedham/projects/neo4j-spark-chicago/categories.csv" AS row MERGE (c:CrimeCategory {name: row.category}) MERGE (sc:SubCategory {code: row.sub_category_code}) ON CREATE SET sc.description = row.sub_category_description MERGE (c)-[:CHILD]->(sc) ---- And that's it! image::{{<siteurl>}}/uploads/2015/07/graph-25.png[Graph 25,598]

neo4j-from-json-to-csv-to-load-csv-via-jq

https://markhneedham.com/blog/2015/07/25/neo4j-from-json-to-csv-to-load-csv-via-jq

2015-09-27 22:02:07

R: data.table - Comparing adjacent rows

As part of my exploration of the https://data.gov.uk/dataset/land-registry-monthly-price-paid-data[Land Registry price paid data set] I wanted to compare the difference between consecutive sales of properties. This means we need to group the sales by a property identifier and then get the previous sale price into a column on each row unless it's the first sale in which case we'll have 'NA'. We can do this by creating a +++<cite>+++http://stackoverflow.com/questions/26291988/r-how-to-create-a-lag-variable-for-each-by-group[lag]+++</cite>+++ variable. I'll use a simpler data set which is very similar in structure to the Land Registry's to demonstrate: [source,r] ---- > blogDT = data.table(name = c("Property 1","Property 1","Property 1","Property 2","Property 2","Property 2"), price = c(10000, 12500, 18000, 245000, 512000, 1000000)) > blogDT name price 1: Property 1 10000 2: Property 1 12500 3: Property 1 18000 4: Property 2 245000 5: Property 2 512000 6: Property 2 1000000 ---- We want to group by the 'name' column and then have the price on row 1 show on row 2, the price on row 2 on row 3, the price on row 4 on row 5 and the price on row 5 on row 6. To do that we'll introduce a 'lag.price' column: [source,r] ---- > blogDT[, lag.price := c(NA, price[-.N]), by = name] > blogDT name price lag.price 1: Property 1 10000 NA 2: Property 1 12500 10000 3: Property 1 18000 12500 4: Property 2 245000 NA 5: Property 2 512000 245000 6: Property 2 1000000 512000 ---- Next let's calculate the difference between the two prices: [source,r] ---- > blogDT[, diff := price - lag.price] > blogDT name price lag.price diff 1: Property 1 10000 NA NA 2: Property 1 12500 10000 2500 3: Property 1 18000 12500 5500 4: Property 2 245000 NA NA 5: Property 2 512000 245000 267000 6: Property 2 1000000 512000 488000 ---- Finally let's order the data table by the biggest price gains: [source,r] ---- > blogDT[order(-diff)] name price lag.price diff 1: Property 2 1000000 512000 488000 2: Property 2 512000 245000 267000 3: Property 1 18000 12500 5500 4: Property 1 12500 10000 2500 5: Property 1 10000 NA NA 6: Property 2 245000 NA NA ----

r-data-table-comparing-adjacent-rows

https://markhneedham.com/blog/2015/09/27/r-data-table-comparing-adjacent-rows

2015-09-21 22:30:51

SparkR: Add new column to data frame by concatenating other columns

Continuing with my http://www.markhneedham.com/blog/2015/09/21/sparkr-error-in-invokejavaisstatic-true-classname-methodname-java-lang-classnotfoundexception-failed-to-load-class-for-data-source-csv/[exploration of the Land Registry open data set using SparkR] I wanted to see which road in the UK has had the most property sales over the last 20 years. To recap, this is what the data frame looks like: [source,bash] ---- ./spark-1.5.0-bin-hadoop2.6/bin/sparkR --packages com.databricks:spark-csv_2.11:1.2.0 > sales <- read.df(sqlContext, "pp-complete.csv", "com.databricks.spark.csv", header="false") > head(sales) C0 C1 C2 C3 C4 C5 1 {0C7ADEF5-878D-4066-B785-0000003ED74A} 163000 2003-02-21 00:00 UB5 4PJ T N 2 {35F67271-ABD4-40DA-AB09-00000085B9D3} 247500 2005-07-15 00:00 TA19 9DD D N 3 {B20B1C74-E8E1-4137-AB3E-0000011DF342} 320000 2010-09-10 00:00 W4 1DZ F N 4 {7D6B0915-C56B-4275-AF9B-00000156BCE7} 104000 1997-08-27 00:00 NE61 2BH D N 5 {47B60101-B64C-413D-8F60-000002F1692D} 147995 2003-05-02 00:00 PE33 0RU D N 6 {51F797CA-7BEB-4958-821F-000003E464AE} 110000 2013-03-22 00:00 NR35 2SF T N C6 C7 C8 C9 C10 C11 1 F 106 READING ROAD NORTHOLT NORTHOLT 2 F 58 ADAMS MEADOW ILMINSTER ILMINSTER 3 L 58 WHELLOCK ROAD LONDON 4 F 17 WESTGATE MORPETH MORPETH 5 F 4 MASON GARDENS WEST WINCH KING'S LYNN 6 F 5 WILD FLOWER WAY DITCHINGHAM BUNGAY C12 C13 C14 1 EALING GREATER LONDON A 2 SOUTH SOMERSET SOMERSET A 3 EALING GREATER LONDON A 4 CASTLE MORPETH NORTHUMBERLAND A 5 KING'S LYNN AND WEST NORFOLK NORFOLK A 6 SOUTH NORFOLK NORFOLK A ---- https://www.gov.uk/guidance/about-the-price-paid-data#explanations-of-column-headers-in-the-ppd[This document] explains the data stored in each field and for this particular query we're interested in fields C9-C12. The plan is to group the data frame by those fields and then sort by frequency in descending order. When grouping by multiple fields it tends to be easiest to create a new field which concatenates them all and then group by that. I started with the following: [source,bash] ---- > sales$address = paste(sales$C9, sales$C10, sales$C11, sales$C12, sep=", ") Error in as.character.default(<S4 object of class "Column">) : no method for coercing this S4 class to a vector ---- Not so successful! Next I went even more primitive: [source,bash] ---- > sales$address = sales$C9 + ", " + sales$C10 + ", " + sales$C11 + ", " + sales$C12 > head(sales) C0 C1 C2 C3 C4 C5 1 {0C7ADEF5-878D-4066-B785-0000003ED74A} 163000 2003-02-21 00:00 UB5 4PJ T N 2 {35F67271-ABD4-40DA-AB09-00000085B9D3} 247500 2005-07-15 00:00 TA19 9DD D N 3 {B20B1C74-E8E1-4137-AB3E-0000011DF342} 320000 2010-09-10 00:00 W4 1DZ F N 4 {7D6B0915-C56B-4275-AF9B-00000156BCE7} 104000 1997-08-27 00:00 NE61 2BH D N 5 {47B60101-B64C-413D-8F60-000002F1692D} 147995 2003-05-02 00:00 PE33 0RU D N 6 {51F797CA-7BEB-4958-821F-000003E464AE} 110000 2013-03-22 00:00 NR35 2SF T N C6 C7 C8 C9 C10 C11 1 F 106 READING ROAD NORTHOLT NORTHOLT 2 F 58 ADAMS MEADOW ILMINSTER ILMINSTER 3 L 58 WHELLOCK ROAD LONDON 4 F 17 WESTGATE MORPETH MORPETH 5 F 4 MASON GARDENS WEST WINCH KING'S LYNN 6 F 5 WILD FLOWER WAY DITCHINGHAM BUNGAY C12 C13 C14 address 1 EALING GREATER LONDON A NA 2 SOUTH SOMERSET SOMERSET A NA 3 EALING GREATER LONDON A NA 4 CASTLE MORPETH NORTHUMBERLAND A NA 5 KING'S LYNN AND WEST NORFOLK NORFOLK A NA 6 SOUTH NORFOLK NORFOLK A NA ---- That at least compiled but all addresses were 'NA' which isn't what we want. After a bit of searching I realised that there was a https://spark.apache.org/docs/latest/api/R/concat_ws.html[concat function] that I could use for exactly this task: [source,bash] ---- > sales$address = concat_ws(sep=", ", sales$C9, sales$C10, sales$C11, sales$C12) > head(sales) C0 C1 C2 C3 C4 C5 1 {0C7ADEF5-878D-4066-B785-0000003ED74A} 163000 2003-02-21 00:00 UB5 4PJ T N 2 {35F67271-ABD4-40DA-AB09-00000085B9D3} 247500 2005-07-15 00:00 TA19 9DD D N 3 {B20B1C74-E8E1-4137-AB3E-0000011DF342} 320000 2010-09-10 00:00 W4 1DZ F N 4 {7D6B0915-C56B-4275-AF9B-00000156BCE7} 104000 1997-08-27 00:00 NE61 2BH D N 5 {47B60101-B64C-413D-8F60-000002F1692D} 147995 2003-05-02 00:00 PE33 0RU D N 6 {51F797CA-7BEB-4958-821F-000003E464AE} 110000 2013-03-22 00:00 NR35 2SF T N C6 C7 C8 C9 C10 C11 1 F 106 READING ROAD NORTHOLT NORTHOLT 2 F 58 ADAMS MEADOW ILMINSTER ILMINSTER 3 L 58 WHELLOCK ROAD LONDON 4 F 17 WESTGATE MORPETH MORPETH 5 F 4 MASON GARDENS WEST WINCH KING'S LYNN 6 F 5 WILD FLOWER WAY DITCHINGHAM BUNGAY C12 C13 C14 1 EALING GREATER LONDON A 2 SOUTH SOMERSET SOMERSET A 3 EALING GREATER LONDON A 4 CASTLE MORPETH NORTHUMBERLAND A 5 KING'S LYNN AND WEST NORFOLK NORFOLK A 6 SOUTH NORFOLK NORFOLK A address 1 READING ROAD, NORTHOLT, NORTHOLT, EALING 2 ADAMS MEADOW, ILMINSTER, ILMINSTER, SOUTH SOMERSET 3 WHELLOCK ROAD, , LONDON, EALING 4 WESTGATE, MORPETH, MORPETH, CASTLE MORPETH 5 MASON GARDENS, WEST WINCH, KING'S LYNN, KING'S LYNN AND WEST NORFOLK 6 WILD FLOWER WAY, DITCHINGHAM, BUNGAY, SOUTH NORFOLK ---- That's more like it! Now let's see which streets have sold the most properties: [source,bash] ---- > byAddress = summarize(groupBy(sales, sales$address), count = n(sales$address)) > head(arrange(byAddress, desc(byAddress$count)), 10) address count 1 BARBICAN, LONDON, LONDON, CITY OF LONDON 1398 2 CHRISTCHURCH ROAD, BOURNEMOUTH, BOURNEMOUTH, BOURNEMOUTH 1313 3 MAIDA VALE, LONDON, LONDON, CITY OF WESTMINSTER 1305 4 ROTHERHITHE STREET, LONDON, LONDON, SOUTHWARK 1253 5 SLOANE AVENUE, LONDON, LONDON, KENSINGTON AND CHELSEA 1219 6 THE STRAND, BRIGHTON MARINA VILLAGE, BRIGHTON, BRIGHTON AND HOVE 1218 7 FAIRFIELD ROAD, LONDON, LONDON, TOWER HAMLETS 1217 8 QUEENSTOWN ROAD, , LONDON, WANDSWORTH 1153 9 UPPER RICHMOND ROAD, LONDON, LONDON, WANDSWORTH 1123 10 QUEENSTOWN ROAD, LONDON, LONDON, WANDSWORTH 1079 ---- Next we'll drill into the data further but that's for another post.

sparkr-add-new-column-to-data-frame-by-concatenating-other-columns

https://markhneedham.com/blog/2015/09/21/sparkr-add-new-column-to-data-frame-by-concatenating-other-columns

2015-09-21 22:06:44

SparkR: Error in invokeJava(isStatic = TRUE, className, methodName, ...) : java.lang.ClassNotFoundException: Failed to load class for data source: csv.

I've been wanting to play around with https://spark.apache.org/docs/latest/sparkr.html[SparkR] for a while and over the weekend deciding to explore a large link:[Land Registry CSV file] containing all the sales of properties in the UK over the last 20 years. First I started up the SparkR shell with the CSV package loaded in: ~~~bash ./spark-1.5.0-bin-hadoop2.6/bin/sparkR --packages com.databricks:spark-csv_2.11:1.2.0 ~~~ Next I tried to read the CSV file into a Spark data frame by modifying one of the examples from the tutorial: ~~~bash > sales \<- read.df(sqlContext, "pp-complete.csv", "csv") 15/09/20 19:13:02 ERROR RBackendHandler: loadDF on org.apache.spark.sql.api.r.SQLUtils failed Error in invokeJava(isStatic = TRUE, className, methodName, \...) : java.lang.ClassNotFoundException: Failed to load class for data source: csv. at org.apache.spark.sql.execution.datasources.ResolvedDataSource$.lookupDataSource(ResolvedDataSource.scala:67) at org.apache.spark.sql.execution.datasources.ResolvedDataSource$.apply(ResolvedDataSource.scala:87) at org.apache.spark.sql.DataFrameReader.load(DataFrameReader.scala:114) at org.apache.spark.sql.api.r.SQLUtils$.loadDF(SQLUtils.scala:156) at org.apache.spark.sql.api.r.SQLUtils.loadDF(SQLUtils.scala) at sun.reflect.NativeMethodAccessorImpl.invoke0(Native Method) at sun.reflect.NativeMethodAccessorImpl.invoke(NativeMethodAccessorImpl.java:62) at sun.reflect.DelegatingMethodAccessorImpl.invoke(DelegatingMethodAccessorImpl.java:43) at java.lang.reflect.Method.invoke(Method.java:497) at org.apache.spark.api.r.RBackendHandler.handleMethodCall(RBackendHandler.scala:132) at org.apache.spark.api.r.RBackendHandler.channelRead0(RBackendHandler.scala:79) at org.apache.spark.api.r.RBackendH ~~~ As far as I can tell I have loaded in the CSV data source so I'm not sure why that doesn't work. However, I came across this https://github.com/databricks/spark-csv/issues/79[github issue] which suggested passing in the full package name as the 3rd argument of 'read.df' rather than just 'csv': ~~~bash > sales \<- read.df(sqlContext, "pp-complete.csv", "com.databricks.spark.csv", header="false") > sales DataFrame[C0:string, C1:string, C2:string, C3:string, C4:string, C5:string, C6:string, C7:string, C8:string, C9:string, C10:string, C11:string, C12:string, C13:string, C14:string] ~~~ And that worked much better! We can now carry on and do some slicing and dicing of the data to see if there are any interesting insights.

sparkr-error-in-invokejavaisstatic-true-classname-methodname-java-lang-classnotfoundexception-failed-to-load-class-for-data-source-csv

https://markhneedham.com/blog/2015/09/21/sparkr-error-in-invokejavaisstatic-true-classname-methodname-java-lang-classnotfoundexception-failed-to-load-class-for-data-source-csv

2015-09-30 05:54:54

IntelliJ 14.1.5: Unable to import maven project

After a recent IntelliJ upgrade I've been running into the following error when trying to attach the sources of any library being pulled in via Maven: ____ Unable to import maven project ____ It seems like this is a recent issue in the 14.x series and luckily is http://stackoverflow.com/questions/30569909/unable-to-import-maven-project-in-intellij14[reasonably easy to fix] by https://youtrack.jetbrains.com/issue/IDEA-140208[adding the following flag] to the VM options passed to the Maven importer: [source,text] ---- -Didea.maven3.use.compat.resolver ---- And this is where you need to add it: image::{{<siteurl>}}/uploads/2015/09/2015-09-30_00-18-17.png[2015 09 30 00 18 17,500] Cmd + , \-> Build, Execution, Deployment \-> Build Tools \-> Maven \-> Importing

intellij-14-1-5-unable-to-import-maven-project

https://markhneedham.com/blog/2015/09/30/intellij-14-1-5-unable-to-import-maven-project

2015-09-25 06:28:29

R: Querying a 20 million line CSV file - data.table vs data frame

As I mentioned in a couple of blog posts already, I've been exploring the https://data.gov.uk/dataset/land-registry-monthly-price-paid-data[Land Registry price paid data set] and although I've initially been using SparkR I was curious how easy it would be to explore the data set using plain R. I thought I'd start out by loading the data into a data frame and run the same queries using deployer. I've come across Hadley Wickham's https://github.com/hadley/readr[readr] library before but hadn't used it and since I needed to load a 20 million line CSV file this seemed the perfect time to give it a try. ____ The goal of readr is to provide a fast and friendly way to read tabular data into R. ____ Let's' get started: [source,r] ---- > library(readr) > system.time(read_csv("pp-complete.csv", col_names = FALSE)) user system elapsed 127.367 21.957 159.963 > df = read_csv("pp-complete.csv", col_names = FALSE) ---- So it took a little over 2 minutes to process the CSV file into a data frame. Let's take a quick look at its contents: [source,r] ---- > head(df) Source: local data frame [6 x 16] X1 X2 X3 X4 X5 X6 X7 X8 X9 (chr) (int) (date) (chr) (chr) (chr) (chr) (chr) (chr) 1 {0C7ADEF5-878D-4066-B785-0000003ED74A} 163000 <NA> UB5 4PJ T N F 106 2 {35F67271-ABD4-40DA-AB09-00000085B9D3} 247500 <NA> TA19 9DD D N F 58 3 {B20B1C74-E8E1-4137-AB3E-0000011DF342} 320000 <NA> W4 1DZ F N L 58 4 {7D6B0915-C56B-4275-AF9B-00000156BCE7} 104000 <NA> NE61 2BH D N F 17 5 {47B60101-B64C-413D-8F60-000002F1692D} 147995 <NA> PE33 0RU D N F 4 6 {51F797CA-7BEB-4958-821F-000003E464AE} 110000 <NA> NR35 2SF T N F 5 Variables not shown: X10 (chr), X11 (chr), X12 (chr), X13 (chr), X14 (chr), X15 (chr), address (chr) ---- Now let's query the data frame to see which postcode has the highest average sale price. We'll need to group by the 'X4' column before applying some aggregate functions: [source,R] ---- > library(dplyr) > system.time(df %>% group_by(X4) %>% summarise(total = sum(as.numeric(X2)), count = n(), ave = total / count) %>% arrange(desc(ave))) user system elapsed 122.557 1.135 124.211 Source: local data frame [1,164,396 x 4] X4 total count ave (chr) (dbl) (int) (dbl) 1 SW7 1DW 39000000 1 39000000 2 SW1W 0NH 32477000 1 32477000 3 W1K 7PX 27000000 1 27000000 4 SW1Y 6HD 24750000 1 24750000 5 SW6 1BA 18000000 1 18000000 6 SW1X 7EE 101505645 6 16917608 7 N6 4LA 16850000 1 16850000 8 EC4N 5AE 16500000 1 16500000 9 W8 7EA 82075000 6 13679167 10 W1K 1DP 13500000 1 13500000 ---- What about if instead of the average price by post code we want to find the most expensive property ever sold instead? [source,R] ---- > system.time(df %>% group_by(X4) %>% summarise(max = max(X2)) %>% arrange(desc(max))) user system elapsed 35.438 0.478 36.026 Source: local data frame [1,164,396 x 2] X4 max (chr) (int) 1 SW10 9SU 54959000 2 SW7 1QJ 50000000 3 SW1X 8HG 46013365 4 SW7 1DW 39000000 5 SW1W 0NH 32477000 6 SW1X 7LJ 29350000 7 W8 7EA 27900000 8 SW3 3SR 27750000 9 W1K 7PX 27000000 10 SW1X 7EE 25533000 .. ... ... ---- Interestingly that one was much quicker than the first one even though it seems like we only did slightly less work. At this point I mentioned my experiment to https://twitter.com/a5hok[Ashok] who suggested I give https://github.com/Rdatatable/data.table[data.table] a try to see if that fared any better. I'd not used it before but was able to get it http://stackoverflow.com/questions/1727772/quickly-reading-very-large-tables-as-dataframes-in-r[up and running reasonably quickly]: [source,r] ---- > library(data.table) > system.time(fread("pp-complete.csv", header = FALSE)) Read 20075122 rows and 15 (of 15) columns from 3.221 GB file in 00:01:05 user system elapsed 59.324 5.798 68.956 > dt = fread("pp-complete.csv", header = FALSE) > head(dt) V1 V2 V3 V4 V5 V6 V7 V8 V9 1: {0C7ADEF5-878D-4066-B785-0000003ED74A} 163000 2003-02-21 00:00 UB5 4PJ T N F 106 2: {35F67271-ABD4-40DA-AB09-00000085B9D3} 247500 2005-07-15 00:00 TA19 9DD D N F 58 3: {B20B1C74-E8E1-4137-AB3E-0000011DF342} 320000 2010-09-10 00:00 W4 1DZ F N L 58 4: {7D6B0915-C56B-4275-AF9B-00000156BCE7} 104000 1997-08-27 00:00 NE61 2BH D N F 17 5: {47B60101-B64C-413D-8F60-000002F1692D} 147995 2003-05-02 00:00 PE33 0RU D N F 4 6: {51F797CA-7BEB-4958-821F-000003E464AE} 110000 2013-03-22 00:00 NR35 2SF T N F 5 V10 V11 V12 V13 V14 V15 1: READING ROAD NORTHOLT NORTHOLT EALING GREATER LONDON A 2: ADAMS MEADOW ILMINSTER ILMINSTER SOUTH SOMERSET SOMERSET A 3: WHELLOCK ROAD LONDON EALING GREATER LONDON A 4: WESTGATE MORPETH MORPETH CASTLE MORPETH NORTHUMBERLAND A 5: MASON GARDENS WEST WINCH KING'S LYNN KING'S LYNN AND WEST NORFOLK NORFOLK A 6: WILD FLOWER WAY DITCHINGHAM BUNGAY SOUTH NORFOLK NORFOLK A ---- So we've already gained one minute in the parsing time which is pretty nice. Let's try and find the postcode with the highest average price: [source,r] ---- > dt[,list(length(V2), sum(V2)), by=V4][, V2 / V1, by=V4][order(-V1)][1:10] Error in sum(V2) : invalid 'type' (character) of argument ---- Hmmm, seems like we need to make column 'V2' numeric. Let's do that: [source,r] ---- > dt = dt[, V2:= as.numeric(V2)] > dt[,list(length(V2), sum(V2)), by=V4][, V2 / V1, by=V4][order(-V1)][1:10] user system elapsed 5.108 0.670 6.183 V4 V1 1: SW7 1DW 39000000 2: SW1W 0NH 32477000 3: W1K 7PX 27000000 4: SW1Y 6HD 24750000 5: SW6 1BA 18000000 6: SW1X 7EE 16917608 7: N6 4LA 16850000 8: EC4N 5AE 16500000 9: W8 7EA 13679167 10: W1K 1DP 13500000 ---- That's quite a bit faster than our data frame version - ~5 seconds compared to ~2 minutes. We have lost the total sales and number of sales columns but I expect that's just because my data.table foo is weak and we could keep them if we wanted. But a good start in terms of execution time. Now let's try the maximum sale price by post code query: [source,r] ---- > system.time(dt[,list(max(V2)), by=V4][order(-V1)][1:10]) user system elapsed 3.684 0.358 4.132 V4 V1 1: SW10 9SU 54959000 2: SW7 1QJ 50000000 3: SW1X 8HG 46013365 4: SW7 1DW 39000000 5: SW1W 0NH 32477000 6: SW1X 7LJ 29350000 7: W8 7EA 27900000 8: SW3 3SR 27750000 9: W1K 7PX 27000000 10: SW1X 7EE 25533000 ---- We've got the same results as before and this time it took ~4 seconds compared to ~35 seconds. We can actually do even better if we set the postcode column as a key: [source,r] ---- > setkey(dt, V4) > system.time(dt[,list(length(V2), sum(V2)), by=V4][, V2 / V1, by=V4][order(-V1)][1:10]) user system elapsed 1.500 0.047 1.548 > system.time(dt[,list(max(V2)), by=V4][order(-V1)][1:10]) user system elapsed 0.578 0.026 0.604 ---- And that's as far as I've got with my experiment. If there's anything else I can do to make either of the versions quicker do let me know in the comments. Oh and for a bit of commentary on what we can learn from the queries\...Knightsbridge is a seriously expensive area to live!

r-querying-a-20-million-line-csv-file-data-table-vs-data-frame

https://markhneedham.com/blog/2015/09/25/r-querying-a-20-million-line-csv-file-data-table-vs-data-frame

2015-08-04 06:35:40

Spark: pyspark/Hadoop - py4j.protocol.Py4JJavaError: An error occurred while calling o23.load.: org.apache.hadoop.ipc.RemoteException: Server IPC version 9 cannot communicate with client version 4

I've been playing around with https://spark.apache.org/docs/0.9.0/python-programming-guide.html[pyspark] - Spark's Python library - and I wanted to execute the following job which takes a file from my local HDFS and then counts how many times each FBI code appears using https://spark.apache.org/docs/1.3.0/sql-programming-guide.html[Spark SQL]: [source,python] ---- from pyspark import SparkContext from pyspark.sql import SQLContext sc = SparkContext("local", "Simple App") sqlContext = SQLContext(sc) file = "hdfs://localhost:9000/user/markneedham/Crimes_-_2001_to_present.csv" sqlContext.load(source="com.databricks.spark.csv", header="true", path = file).registerTempTable("crimes") rows = sqlContext.sql("select `FBI Code` AS fbiCode, COUNT(*) AS times FROM crimes GROUP BY `FBI Code` ORDER BY times DESC").collect() for row in rows: print("{0} -> {1}".format(row.fbiCode, row.times)) ---- I submitted the job and waited: [source,bash] ---- $ ./spark-1.3.0-bin-hadoop1/bin/spark-submit --driver-memory 5g --packages com.databricks:spark-csv_2.10:1.1.0 fbi_spark.py ... Traceback (most recent call last): File "/Users/markneedham/projects/neo4j-spark-chicago/fbi_spark.py", line 11, in <module> sqlContext.load(source="com.databricks.spark.csv", header="true", path = file).registerTempTable("crimes") File "/Users/markneedham/projects/neo4j-spark-chicago/spark-1.3.0-bin-hadoop1/python/pyspark/sql/context.py", line 482, in load df = self._ssql_ctx.load(source, joptions) File "/Users/markneedham/projects/neo4j-spark-chicago/spark-1.3.0-bin-hadoop1/python/lib/py4j-0.8.2.1-src.zip/py4j/java_gateway.py", line 538, in __call__ File "/Users/markneedham/projects/neo4j-spark-chicago/spark-1.3.0-bin-hadoop1/python/lib/py4j-0.8.2.1-src.zip/py4j/protocol.py", line 300, in get_return_value py4j.protocol.Py4JJavaError: An error occurred while calling o23.load. : org.apache.hadoop.ipc.RemoteException: Server IPC version 9 cannot communicate with client version 4 at org.apache.hadoop.ipc.Client.call(Client.java:1070) at org.apache.hadoop.ipc.RPC$Invoker.invoke(RPC.java:225) at com.sun.proxy.$Proxy7.getProtocolVersion(Unknown Source) at org.apache.hadoop.ipc.RPC.getProxy(RPC.java:396) at org.apache.hadoop.ipc.RPC.getProxy(RPC.java:379) at org.apache.hadoop.hdfs.DFSClient.createRPCNamenode(DFSClient.java:119) at org.apache.hadoop.hdfs.DFSClient.<init>(DFSClient.java:238) at org.apache.hadoop.hdfs.DFSClient.<init>(DFSClient.java:203) at org.apache.hadoop.hdfs.DistributedFileSystem.initialize(DistributedFileSystem.java:89) at org.apache.hadoop.fs.FileSystem.createFileSystem(FileSystem.java:1386) at org.apache.hadoop.fs.FileSystem.access$200(FileSystem.java:66) at org.apache.hadoop.fs.FileSystem$Cache.get(FileSystem.java:1404) at org.apache.hadoop.fs.FileSystem.get(FileSystem.java:254) at org.apache.hadoop.fs.Path.getFileSystem(Path.java:187) at org.apache.hadoop.mapred.FileInputFormat.listStatus(FileInputFormat.java:176) at org.apache.hadoop.mapred.FileInputFormat.getSplits(FileInputFormat.java:208) at org.apache.spark.rdd.HadoopRDD.getPartitions(HadoopRDD.scala:203) at org.apache.spark.rdd.RDD$$anonfun$partitions$2.apply(RDD.scala:219) at org.apache.spark.rdd.RDD$$anonfun$partitions$2.apply(RDD.scala:217) at scala.Option.getOrElse(Option.scala:120) at org.apache.spark.rdd.RDD.partitions(RDD.scala:217) at org.apache.spark.rdd.MapPartitionsRDD.getPartitions(MapPartitionsRDD.scala:32) at org.apache.spark.rdd.RDD$$anonfun$partitions$2.apply(RDD.scala:219) at org.apache.spark.rdd.RDD$$anonfun$partitions$2.apply(RDD.scala:217) at scala.Option.getOrElse(Option.scala:120) at org.apache.spark.rdd.RDD.partitions(RDD.scala:217) at org.apache.spark.rdd.RDD.take(RDD.scala:1156) at org.apache.spark.rdd.RDD.first(RDD.scala:1189) at com.databricks.spark.csv.CsvRelation.firstLine$lzycompute(CsvRelation.scala:129) at com.databricks.spark.csv.CsvRelation.firstLine(CsvRelation.scala:127) at com.databricks.spark.csv.CsvRelation.inferSchema(CsvRelation.scala:109) at com.databricks.spark.csv.CsvRelation.<init>(CsvRelation.scala:62) at com.databricks.spark.csv.DefaultSource.createRelation(DefaultSource.scala:115) at com.databricks.spark.csv.DefaultSource.createRelation(DefaultSource.scala:40) at com.databricks.spark.csv.DefaultSource.createRelation(DefaultSource.scala:28) at org.apache.spark.sql.sources.ResolvedDataSource$.apply(ddl.scala:290) at org.apache.spark.sql.SQLContext.load(SQLContext.scala:679) at org.apache.spark.sql.SQLContext.load(SQLContext.scala:667) at sun.reflect.NativeMethodAccessorImpl.invoke0(Native Method) at sun.reflect.NativeMethodAccessorImpl.invoke(NativeMethodAccessorImpl.java:62) at sun.reflect.DelegatingMethodAccessorImpl.invoke(DelegatingMethodAccessorImpl.java:43) at java.lang.reflect.Method.invoke(Method.java:497) at py4j.reflection.MethodInvoker.invoke(MethodInvoker.java:231) at py4j.reflection.ReflectionEngine.invoke(ReflectionEngine.java:379) at py4j.Gateway.invoke(Gateway.java:259) at py4j.commands.AbstractCommand.invokeMethod(AbstractCommand.java:133) at py4j.commands.CallCommand.execute(CallCommand.java:79) at py4j.GatewayConnection.run(GatewayConnection.java:207) at java.lang.Thread.run(Thread.java:745) ---- It looks like my http://glennklockwood.blogspot.co.uk/2014/06/spark-on-supercomputers-few-notes.html[Hadoop Client and Server are using different versions] which in fact they are! We can see from the name of the spark folder that I'm using Hadoop 1.x there and if we check the local Hadoop version we'll notice it's using the 2.x seris: [source,bash] ---- $ hadoop version Hadoop 2.6.0 ---- In this case the easiest fix is use a version of Spark that's compiled against Hadoop 2.6 which as of now means Spark 1.4.1. Let's try and run our job again: [source,bash] ---- $ ./spark-1.4.1-bin-hadoop2.6/bin/spark-submit --driver-memory 5g --packages com.databricks:spark-csv_2.10:1.1.0 fbi_spark.py 06 -> 859197 08B -> 653575 14 -> 488212 18 -> 457782 26 -> 431316 05 -> 257310 07 -> 197404 08A -> 188964 03 -> 157706 11 -> 112675 04B -> 103961 04A -> 60344 16 -> 47279 15 -> 40361 24 -> 31809 10 -> 22467 17 -> 17555 02 -> 17008 20 -> 15190 19 -> 10878 22 -> 8847 09 -> 6358 01A -> 4830 13 -> 1561 12 -> 835 01B -> 16 ---- And it's working!

spark-pysparkhadoop-py4j-protocol-py4jjavaerror-an-error-occurred-while-calling-o23-load-org-apache-hadoop-ipc-remoteexception-server-ipc-version-9-cannot-communicate-with-client-version-4

https://markhneedham.com/blog/2015/08/04/spark-pysparkhadoop-py4j-protocol-py4jjavaerror-an-error-occurred-while-calling-o23-load-org-apache-hadoop-ipc-remoteexception-server-ipc-version-9-cannot-communicate-with-client-version-4

2015-08-02 18:08:47

Spark: Processing CSV files using Databricks Spark CSV Library

Last year I wrote about http://www.markhneedham.com/blog/2014/11/16/spark-parse-csv-file-and-group-by-column-value/[exploring the Chicago crime data set using Spark and the OpenCSV parser] and while this worked well, a few months ago I noticed that there's now a https://github.com/databricks/spark-csv[spark-csv] library which I should probably use instead. I thought it'd be a fun exercise to translate my code to use it. So to recap our goal: we want to count how many times each type of crime has been committed. I have a more up to date version of the crimes file now so the numbers won't be exactly the same. First let's launch the spark-shell and register our CSV file as a temporary table so we can query it as if it was a SQL table: [source,bash] ---- $ ./spark-1.3.0-bin-hadoop1/bin/spark-shell scala> import org.apache.spark.sql.SQLContext import org.apache.spark.sql.SQLContext scala> val crimeFile = "/Users/markneedham/Downloads/Crimes_-_2001_to_present.csv" crimeFile: String = /Users/markneedham/Downloads/Crimes_-_2001_to_present.csv scala> val sqlContext = new SQLContext(sc) sqlContext: org.apache.spark.sql.SQLContext = org.apache.spark.sql.SQLContext@9746157 scala> sqlContext.load("com.databricks.spark.csv", Map("path" -> crimeFile, "header" -> "true")).registerTempTable("crimes") java.lang.RuntimeException: Failed to load class for data source: com.databricks.spark.csv at scala.sys.package$.error(package.scala:27) at org.apache.spark.sql.sources.ResolvedDataSource$.lookupDataSource(ddl.scala:268) at org.apache.spark.sql.sources.ResolvedDataSource$.apply(ddl.scala:279) at org.apache.spark.sql.SQLContext.load(SQLContext.scala:679) at java.lang.reflect.Method.invoke(Method.java:497) at org.apache.spark.repl.SparkIMain$ReadEvalPrint.call(SparkIMain.scala:1065) at org.apache.spark.repl.SparkIMain$Request.loadAndRun(SparkIMain.scala:1338) at org.apache.spark.repl.SparkIMain.loadAndRunReq$1(SparkIMain.scala:840) at org.apache.spark.repl.SparkIMain.interpret(SparkIMain.scala:871) at org.apache.spark.repl.SparkIMain.interpret(SparkIMain.scala:819) at org.apache.spark.repl.SparkILoop.reallyInterpret$1(SparkILoop.scala:856) at org.apache.spark.repl.SparkILoop.interpretStartingWith(SparkILoop.scala:901) at org.apache.spark.repl.SparkILoop.command(SparkILoop.scala:813) at org.apache.spark.repl.SparkILoop.processLine$1(SparkILoop.scala:656) at org.apache.spark.repl.SparkILoop.innerLoop$1(SparkILoop.scala:664) at org.apache.spark.repl.SparkILoop.org$apache$spark$repl$SparkILoop$$loop(SparkILoop.scala:669) at org.apache.spark.repl.SparkILoop$$anonfun$org$apache$spark$repl$SparkILoop$$process$1.apply$mcZ$sp(SparkILoop.scala:996) at org.apache.spark.repl.SparkILoop$$anonfun$org$apache$spark$repl$SparkILoop$$process$1.apply(SparkILoop.scala:944) at org.apache.spark.repl.SparkILoop$$anonfun$org$apache$spark$repl$SparkILoop$$process$1.apply(SparkILoop.scala:944) at scala.tools.nsc.util.ScalaClassLoader$.savingContextLoader(ScalaClassLoader.scala:135) at org.apache.spark.repl.SparkILoop.org$apache$spark$repl$SparkILoop$$process(SparkILoop.scala:944) at org.apache.spark.repl.SparkILoop.process(SparkILoop.scala:1058) at org.apache.spark.repl.Main$.main(Main.scala:31) at org.apache.spark.repl.Main.main(Main.scala) at sun.reflect.NativeMethodAccessorImpl.invoke0(Native Method) at sun.reflect.NativeMethodAccessorImpl.invoke(NativeMethodAccessorImpl.java:62) at sun.reflect.DelegatingMethodAccessorImpl.invoke(DelegatingMethodAccessorImpl.java:43) at java.lang.reflect.Method.invoke(Method.java:497) at org.apache.spark.deploy.SparkSubmit$.org$apache$spark$deploy$SparkSubmit$$runMain(SparkSubmit.scala:569) at org.apache.spark.deploy.SparkSubmit$.doRunMain$1(SparkSubmit.scala:166) at org.apache.spark.deploy.SparkSubmit$.submit(SparkSubmit.scala:189) at org.apache.spark.deploy.SparkSubmit$.main(SparkSubmit.scala:110) at org.apache.spark.deploy.SparkSubmit.main(SparkSubmit.scala) ---- I've actually forgotten to tell spark-shell about the CSV package so let's restart the shell and pass it as an argument: [source,bash] ---- $ ./spark-1.3.0-bin-hadoop1/bin/spark-shell --packages com.databricks:spark-csv_2.10:1.1.0 scala> import org.apache.spark.sql.SQLContext import org.apache.spark.sql.SQLContext scala> val crimeFile = "/Users/markneedham/Downloads/Crimes_-_2001_to_present.csv" crimeFile: String = /Users/markneedham/Downloads/Crimes_-_2001_to_present.csv scala> val sqlContext = new SQLContext(sc) sqlContext: org.apache.spark.sql.SQLContext = org.apache.spark.sql.SQLContext@44587c44 scala> sqlContext.load("com.databricks.spark.csv", Map("path" -> crimeFile, "header" -> "true")).registerTempTable("crimes") ... 15/08/02 18:57:46 INFO TaskSchedulerImpl: Removed TaskSet 0.0, whose tasks have all completed, from pool 15/08/02 18:57:46 INFO DAGScheduler: Stage 0 (first at CsvRelation.scala:129) finished in 0.207 s 15/08/02 18:57:46 INFO DAGScheduler: Job 0 finished: first at CsvRelation.scala:129, took 0.267327 s ---- Now we can write a simple SQL query on our 'crimes' table to find the most popular crime types: [source,bash] ---- scala> sqlContext.sql( """ select `Primary Type` as primaryType, COUNT(*) AS times from crimes group by `Primary Type` order by times DESC """).save("/tmp/agg.csv", "com.databricks.spark.csv") ---- That spits out a load of CSV 'part files' into +++<cite>+++/tmp/agg.csv+++</cite>+++ so let's bring in the merge function that we've used previously to combine these into one CSV file: [source,scala] ---- scala> import org.apache.hadoop.conf.Configuration scala> import org.apache.hadoop.fs._ scala> def merge(srcPath: String, dstPath: String): Unit = { val hadoopConfig = new Configuration() val hdfs = FileSystem.get(hadoopConfig) FileUtil.copyMerge(hdfs, new Path(srcPath), hdfs, new Path(dstPath), false, hadoopConfig, null) } scala> merge("/tmp/agg.csv", "agg.csv") ---- And finally let's browse the contents of our new CSV file: [source,bash] ---- $ cat agg.csv THEFT,1206745 BATTERY,1066110 CRIMINAL DAMAGE,672782 NARCOTICS,662257 OTHER OFFENSE,360824 ASSAULT,354583 BURGLARY,343443 MOTOR VEHICLE THEFT,278014 ROBBERY,218190 DECEPTIVE PRACTICE,197477 CRIMINAL TRESPASS,171363 PROSTITUTION,65660 WEAPONS VIOLATION,56218 PUBLIC PEACE VIOLATION,42446 OFFENSE INVOLVING CHILDREN,37010 CRIM SEXUAL ASSAULT,21346 SEX OFFENSE,21305 GAMBLING,13704 LIQUOR LAW VIOLATION,13264 INTERFERENCE WITH PUBLIC OFFICER,11366 ARSON,9642 HOMICIDE,7192 KIDNAPPING,6029 INTIMIDATION,3443 STALKING,2760 OBSCENITY,331 PUBLIC INDECENCY,123 OTHER NARCOTIC VIOLATION,106 CONCEALED CARRY LICENSE VIOLATION,34 NON-CRIMINAL,31 NON - CRIMINAL,25 RITUALISM,23 HUMAN TRAFFICKING,9 NON-CRIMINAL (SUBJECT SPECIFIED),3 DOMESTIC VIOLENCE,1 ---- Great! We've got the same output with much less code which is always a #win.

spark-processing-csv-files-using-databricks-spark-csv-library

https://markhneedham.com/blog/2015/08/02/spark-processing-csv-files-using-databricks-spark-csv-library

2015-08-11 06:59:50

Java: Jersey - java.lang.NoSuchMethodError: com.sun.jersey.core.reflection.ReflectionHelper. getContextClassLoaderPA()Ljava/security/PrivilegedAction;

I've been trying to put some tests around an https://github.com/mneedham/dummy-unmanaged-extension[Neo4j unmanaged extension] I've been working on and ran into the following stack trace when launching the server using the http://neo4j.com/docs/stable/server-unmanaged-extensions-testing.html[Neo4j test harness]: [source,java] ---- public class ExampleResourceTest { @Rule public Neo4jRule neo4j = new Neo4jRule() .withFixture("CREATE (:Person {name: 'Mark'})") .withFixture("CREATE (:Person {name: 'Nicole'})") .withExtension( "/unmanaged", ExampleResource.class ); @Test public void shouldReturnAllTheNodes() { // Given URI serverURI = neo4j.httpURI(); // When HTTP.Response response = HTTP.GET(serverURI.resolve("/unmanaged/example/people").toString()); // Then assertEquals(200, response.status()); List content = response.content(); assertEquals(2, content.size()); } } ---- [source,text] ---- 07:51:32.985 [main] WARN o.e.j.u.component.AbstractLifeCycle - FAILED o.e.j.s.ServletContextHandler@29eda4f8{/unmanaged,null,STARTING}: java.lang.NoSuchMethodError: com.sun.jersey.core.reflection.ReflectionHelper.getContextClassLoaderPA()Ljava/security/PrivilegedAction; java.lang.NoSuchMethodError: com.sun.jersey.core.reflection.ReflectionHelper.getContextClassLoaderPA()Ljava/security/PrivilegedAction; at com.sun.jersey.spi.scanning.AnnotationScannerListener.<init>(AnnotationScannerListener.java:94) ~[jersey-server-1.19.jar:1.19] at com.sun.jersey.spi.scanning.PathProviderScannerListener.<init>(PathProviderScannerListener.java:59) ~[jersey-server-1.19.jar:1.19] at com.sun.jersey.api.core.ScanningResourceConfig.init(ScanningResourceConfig.java:79) ~[jersey-server-1.19.jar:1.19] at com.sun.jersey.api.core.PackagesResourceConfig.init(PackagesResourceConfig.java:104) ~[jersey-server-1.19.jar:1.19] at com.sun.jersey.api.core.PackagesResourceConfig.<init>(PackagesResourceConfig.java:78) ~[jersey-server-1.19.jar:1.19] at com.sun.jersey.api.core.PackagesResourceConfig.<init>(PackagesResourceConfig.java:89) ~[jersey-server-1.19.jar:1.19] at com.sun.jersey.spi.container.servlet.WebComponent.createResourceConfig(WebComponent.java:696) ~[jersey-servlet-1.17.1.jar:1.17.1] at com.sun.jersey.spi.container.servlet.WebComponent.createResourceConfig(WebComponent.java:674) ~[jersey-servlet-1.17.1.jar:1.17.1] at com.sun.jersey.spi.container.servlet.WebComponent.init(WebComponent.java:203) ~[jersey-servlet-1.17.1.jar:1.17.1] at com.sun.jersey.spi.container.servlet.ServletContainer.init(ServletContainer.java:374) ~[jersey-servlet-1.17.1.jar:1.17.1] at com.sun.jersey.spi.container.servlet.ServletContainer.init(ServletContainer.java:557) ~[jersey-servlet-1.17.1.jar:1.17.1] at javax.servlet.GenericServlet.init(GenericServlet.java:244) ~[javax.servlet-api-3.1.0.jar:3.1.0] at org.eclipse.jetty.servlet.ServletHolder.initServlet(ServletHolder.java:612) ~[jetty-servlet-9.2.4.v20141103.jar:9.2.4.v20141103] at org.eclipse.jetty.servlet.ServletHolder.initialize(ServletHolder.java:395) ~[jetty-servlet-9.2.4.v20141103.jar:9.2.4.v20141103] at org.eclipse.jetty.servlet.ServletHandler.initialize(ServletHandler.java:871) ~[jetty-servlet-9.2.4.v20141103.jar:9.2.4.v20141103] at org.eclipse.jetty.servlet.ServletContextHandler.startContext(ServletContextHandler.java:298) ~[jetty-servlet-9.2.4.v20141103.jar:9.2.4.v20141103] at org.eclipse.jetty.server.handler.ContextHandler.doStart(ContextHandler.java:741) ~[jetty-server-9.2.4.v20141103.jar:9.2.4.v20141103] at org.eclipse.jetty.util.component.AbstractLifeCycle.start(AbstractLifeCycle.java:68) [jetty-util-9.2.4.v20141103.jar:9.2.4.v20141103] at org.eclipse.jetty.util.component.ContainerLifeCycle.start(ContainerLifeCycle.java:132) [jetty-util-9.2.4.v20141103.jar:9.2.4.v20141103] at org.eclipse.jetty.util.component.ContainerLifeCycle.doStart(ContainerLifeCycle.java:114) [jetty-util-9.2.4.v20141103.jar:9.2.4.v20141103] at org.eclipse.jetty.server.handler.AbstractHandler.doStart(AbstractHandler.java:61) [jetty-server-9.2.4.v20141103.jar:9.2.4.v20141103] at org.eclipse.jetty.util.component.AbstractLifeCycle.start(AbstractLifeCycle.java:68) [jetty-util-9.2.4.v20141103.jar:9.2.4.v20141103] at org.eclipse.jetty.util.component.ContainerLifeCycle.start(ContainerLifeCycle.java:132) [jetty-util-9.2.4.v20141103.jar:9.2.4.v20141103] at org.eclipse.jetty.server.Server.start(Server.java:387) [jetty-server-9.2.4.v20141103.jar:9.2.4.v20141103] at org.eclipse.jetty.util.component.ContainerLifeCycle.doStart(ContainerLifeCycle.java:114) [jetty-util-9.2.4.v20141103.jar:9.2.4.v20141103] at org.eclipse.jetty.server.handler.AbstractHandler.doStart(AbstractHandler.java:61) [jetty-server-9.2.4.v20141103.jar:9.2.4.v20141103] at org.eclipse.jetty.server.Server.doStart(Server.java:354) [jetty-server-9.2.4.v20141103.jar:9.2.4.v20141103] at org.eclipse.jetty.util.component.AbstractLifeCycle.start(AbstractLifeCycle.java:68) [jetty-util-9.2.4.v20141103.jar:9.2.4.v20141103] at org.neo4j.server.web.Jetty9WebServer.startJetty(Jetty9WebServer.java:381) [neo4j-server-2.2.3.jar:2.2.3] at org.neo4j.server.web.Jetty9WebServer.start(Jetty9WebServer.java:184) [neo4j-server-2.2.3.jar:2.2.3] at org.neo4j.server.AbstractNeoServer.startWebServer(AbstractNeoServer.java:474) [neo4j-server-2.2.3.jar:2.2.3] at org.neo4j.server.AbstractNeoServer.start(AbstractNeoServer.java:230) [neo4j-server-2.2.3.jar:2.2.3] at org.neo4j.harness.internal.InProcessServerControls.start(InProcessServerControls.java:59) [neo4j-harness-2.2.3.jar:2.2.3] at org.neo4j.harness.internal.InProcessServerBuilder.newServer(InProcessServerBuilder.java:72) [neo4j-harness-2.2.3.jar:2.2.3] at org.neo4j.harness.junit.Neo4jRule$1.evaluate(Neo4jRule.java:64) [neo4j-harness-2.2.3.jar:2.2.3] at org.junit.rules.RunRules.evaluate(RunRules.java:20) [junit-4.11.jar:na] at org.junit.runners.ParentRunner.runLeaf(ParentRunner.java:271) [junit-4.11.jar:na] at org.junit.runners.BlockJUnit4ClassRunner.runChild(BlockJUnit4ClassRunner.java:70) [junit-4.11.jar:na] at org.junit.runners.BlockJUnit4ClassRunner.runChild(BlockJUnit4ClassRunner.java:50) [junit-4.11.jar:na] at org.junit.runners.ParentRunner$3.run(ParentRunner.java:238) [junit-4.11.jar:na] at org.junit.runners.ParentRunner$1.schedule(ParentRunner.java:63) [junit-4.11.jar:na] at org.junit.runners.ParentRunner.runChildren(ParentRunner.java:236) [junit-4.11.jar:na] at org.junit.runners.ParentRunner.access$000(ParentRunner.java:53) [junit-4.11.jar:na] at org.junit.runners.ParentRunner$2.evaluate(ParentRunner.java:229) [junit-4.11.jar:na] at org.junit.runners.ParentRunner.run(ParentRunner.java:309) [junit-4.11.jar:na] at org.junit.runner.JUnitCore.run(JUnitCore.java:160) [junit-4.11.jar:na] at com.intellij.junit4.JUnit4IdeaTestRunner.startRunnerWithArgs(JUnit4IdeaTestRunner.java:78) [junit-rt.jar:na] at com.intellij.rt.execution.junit.JUnitStarter.prepareStreamsAndStart(JUnitStarter.java:212) [junit-rt.jar:na] at com.intellij.rt.execution.junit.JUnitStarter.main(JUnitStarter.java:68) [junit-rt.jar:na] at sun.reflect.NativeMethodAccessorImpl.invoke0(Native Method) ~[na:1.8.0_51] at sun.reflect.NativeMethodAccessorImpl.invoke(NativeMethodAccessorImpl.java:62) ~[na:1.8.0_51] at sun.reflect.DelegatingMethodAccessorImpl.invoke(DelegatingMethodAccessorImpl.java:43) ~[na:1.8.0_51] at java.lang.reflect.Method.invoke(Method.java:497) ~[na:1.8.0_51] at com.intellij.rt.execution.application.AppMain.main(AppMain.java:140) [idea_rt.jar:na] 07:51:32.991 [main] WARN o.e.j.u.component.AbstractLifeCycle - FAILED org.eclipse.jetty.server.handler.HandlerList@2b22a1cc[o.e.j.s.h.MovedContextHandler@5e671e20{/,null,AVAILABLE}, o.e.j.s.ServletContextHandler@29eda4f8{/unmanaged,null,STARTING}, o.e.j.s.ServletContextHandler@62573c86{/db/manage,null,null}, o.e.j.s.ServletContextHandler@2418ba04{/db/data,null,null}, o.e.j.w.WebAppContext@14229fa7{/browser,jar:file:/Users/markneedham/.m2/repository/org/neo4j/app/neo4j-browser/2.2.3/neo4j-browser-2.2.3.jar!/browser,null}, o.e.j.s.ServletContextHandler@2ab0702e{/,null,null}]: java.lang.NoSuchMethodError: com.sun.jersey.core.reflection.ReflectionHelper.getContextClassLoaderPA()Ljava/security/PrivilegedAction; java.lang.NoSuchMethodError: com.sun.jersey.core.reflection.ReflectionHelper.getContextClassLoaderPA()Ljava/security/PrivilegedAction; at com.sun.jersey.spi.scanning.AnnotationScannerListener.<init>(AnnotationScannerListener.java:94) ~[jersey-server-1.19.jar:1.19] at com.sun.jersey.spi.scanning.PathProviderScannerListener.<init>(PathProviderScannerListener.java:59) ~[jersey-server-1.19.jar:1.19] at com.sun.jersey.api.core.ScanningResourceConfig.init(ScanningResourceConfig.java:79) ~[jersey-server-1.19.jar:1.19] at com.sun.jersey.api.core.PackagesResourceConfig.init(PackagesResourceConfig.java:104) ~[jersey-server-1.19.jar:1.19] at com.sun.jersey.api.core.PackagesResourceConfig.<init>(PackagesResourceConfig.java:78) ~[jersey-server-1.19.jar:1.19] at com.sun.jersey.api.core.PackagesResourceConfig.<init>(PackagesResourceConfig.java:89) ~[jersey-server-1.19.jar:1.19] at com.sun.jersey.spi.container.servlet.WebComponent.createResourceConfig(WebComponent.java:696) ~[jersey-servlet-1.17.1.jar:1.17.1] at com.sun.jersey.spi.container.servlet.WebComponent.createResourceConfig(WebComponent.java:674) ~[jersey-servlet-1.17.1.jar:1.17.1] at com.sun.jersey.spi.container.servlet.WebComponent.init(WebComponent.java:203) ~[jersey-servlet-1.17.1.jar:1.17.1] at com.sun.jersey.spi.container.servlet.ServletContainer.init(ServletContainer.java:374) ~[jersey-servlet-1.17.1.jar:1.17.1] at com.sun.jersey.spi.container.servlet.ServletContainer.init(ServletContainer.java:557) ~[jersey-servlet-1.17.1.jar:1.17.1] at javax.servlet.GenericServlet.init(GenericServlet.java:244) ~[javax.servlet-api-3.1.0.jar:3.1.0] at org.eclipse.jetty.servlet.ServletHolder.initServlet(ServletHolder.java:612) ~[jetty-servlet-9.2.4.v20141103.jar:9.2.4.v20141103] at org.eclipse.jetty.servlet.ServletHolder.initialize(ServletHolder.java:395) ~[jetty-servlet-9.2.4.v20141103.jar:9.2.4.v20141103] at org.eclipse.jetty.servlet.ServletHandler.initialize(ServletHandler.java:871) ~[jetty-servlet-9.2.4.v20141103.jar:9.2.4.v20141103] at org.eclipse.jetty.servlet.ServletContextHandler.startContext(ServletContextHandler.java:298) ~[jetty-servlet-9.2.4.v20141103.jar:9.2.4.v20141103] at org.eclipse.jetty.server.handler.ContextHandler.doStart(ContextHandler.java:741) ~[jetty-server-9.2.4.v20141103.jar:9.2.4.v20141103] at org.eclipse.jetty.util.component.AbstractLifeCycle.start(AbstractLifeCycle.java:68) [jetty-util-9.2.4.v20141103.jar:9.2.4.v20141103] at org.eclipse.jetty.util.component.ContainerLifeCycle.start(ContainerLifeCycle.java:132) [jetty-util-9.2.4.v20141103.jar:9.2.4.v20141103] at org.eclipse.jetty.util.component.ContainerLifeCycle.doStart(ContainerLifeCycle.java:114) [jetty-util-9.2.4.v20141103.jar:9.2.4.v20141103] at org.eclipse.jetty.server.handler.AbstractHandler.doStart(AbstractHandler.java:61) [jetty-server-9.2.4.v20141103.jar:9.2.4.v20141103] at org.eclipse.jetty.util.component.AbstractLifeCycle.start(AbstractLifeCycle.java:68) [jetty-util-9.2.4.v20141103.jar:9.2.4.v20141103] at org.eclipse.jetty.util.component.ContainerLifeCycle.start(ContainerLifeCycle.java:132) [jetty-util-9.2.4.v20141103.jar:9.2.4.v20141103] at org.eclipse.jetty.server.Server.start(Server.java:387) [jetty-server-9.2.4.v20141103.jar:9.2.4.v20141103] at org.eclipse.jetty.util.component.ContainerLifeCycle.doStart(ContainerLifeCycle.java:114) [jetty-util-9.2.4.v20141103.jar:9.2.4.v20141103] at org.eclipse.jetty.server.handler.AbstractHandler.doStart(AbstractHandler.java:61) [jetty-server-9.2.4.v20141103.jar:9.2.4.v20141103] at org.eclipse.jetty.server.Server.doStart(Server.java:354) [jetty-server-9.2.4.v20141103.jar:9.2.4.v20141103] at org.eclipse.jetty.util.component.AbstractLifeCycle.start(AbstractLifeCycle.java:68) [jetty-util-9.2.4.v20141103.jar:9.2.4.v20141103] at org.neo4j.server.web.Jetty9WebServer.startJetty(Jetty9WebServer.java:381) [neo4j-server-2.2.3.jar:2.2.3] at org.neo4j.server.web.Jetty9WebServer.start(Jetty9WebServer.java:184) [neo4j-server-2.2.3.jar:2.2.3] at org.neo4j.server.AbstractNeoServer.startWebServer(AbstractNeoServer.java:474) [neo4j-server-2.2.3.jar:2.2.3] at org.neo4j.server.AbstractNeoServer.start(AbstractNeoServer.java:230) [neo4j-server-2.2.3.jar:2.2.3] at org.neo4j.harness.internal.InProcessServerControls.start(InProcessServerControls.java:59) [neo4j-harness-2.2.3.jar:2.2.3] at org.neo4j.harness.internal.InProcessServerBuilder.newServer(InProcessServerBuilder.java:72) [neo4j-harness-2.2.3.jar:2.2.3] at org.neo4j.harness.junit.Neo4jRule$1.evaluate(Neo4jRule.java:64) [neo4j-harness-2.2.3.jar:2.2.3] at org.junit.rules.RunRules.evaluate(RunRules.java:20) [junit-4.11.jar:na] at org.junit.runners.ParentRunner.runLeaf(ParentRunner.java:271) [junit-4.11.jar:na] at org.junit.runners.BlockJUnit4ClassRunner.runChild(BlockJUnit4ClassRunner.java:70) [junit-4.11.jar:na] at org.junit.runners.BlockJUnit4ClassRunner.runChild(BlockJUnit4ClassRunner.java:50) [junit-4.11.jar:na] at org.junit.runners.ParentRunner$3.run(ParentRunner.java:238) [junit-4.11.jar:na] at org.junit.runners.ParentRunner$1.schedule(ParentRunner.java:63) [junit-4.11.jar:na] at org.junit.runners.ParentRunner.runChildren(ParentRunner.java:236) [junit-4.11.jar:na] at org.junit.runners.ParentRunner.access$000(ParentRunner.java:53) [junit-4.11.jar:na] at org.junit.runners.ParentRunner$2.evaluate(ParentRunner.java:229) [junit-4.11.jar:na] at org.junit.runners.ParentRunner.run(ParentRunner.java:309) [junit-4.11.jar:na] at org.junit.runner.JUnitCore.run(JUnitCore.java:160) [junit-4.11.jar:na] at com.intellij.junit4.JUnit4IdeaTestRunner.startRunnerWithArgs(JUnit4IdeaTestRunner.java:78) [junit-rt.jar:na] at com.intellij.rt.execution.junit.JUnitStarter.prepareStreamsAndStart(JUnitStarter.java:212) [junit-rt.jar:na] at com.intellij.rt.execution.junit.JUnitStarter.main(JUnitStarter.java:68) [junit-rt.jar:na] at sun.reflect.NativeMethodAccessorImpl.invoke0(Native Method) ~[na:1.8.0_51] at sun.reflect.NativeMethodAccessorImpl.invoke(NativeMethodAccessorImpl.java:62) ~[na:1.8.0_51] at sun.reflect.DelegatingMethodAccessorImpl.invoke(DelegatingMethodAccessorImpl.java:43) ~[na:1.8.0_51] at java.lang.reflect.Method.invoke(Method.java:497) ~[na:1.8.0_51] at com.intellij.rt.execution.application.AppMain.main(AppMain.java:140) [idea_rt.jar:na] 07:51:33.013 [main] INFO o.e.jetty.server.ServerConnector - Started ServerConnector@19962194{HTTP/1.1}{localhost:7475} 07:51:33.014 [main] WARN o.e.j.u.component.AbstractLifeCycle - FAILED org.eclipse.jetty.server.Server@481e91b6: java.lang.NoSuchMethodError: com.sun.jersey.core.reflection.ReflectionHelper.getContextClassLoaderPA()Ljava/security/PrivilegedAction; java.lang.NoSuchMethodError: com.sun.jersey.core.reflection.ReflectionHelper.getContextClassLoaderPA()Ljava/security/PrivilegedAction; at com.sun.jersey.spi.scanning.AnnotationScannerListener.<init>(AnnotationScannerListener.java:94) ~[jersey-server-1.19.jar:1.19] at com.sun.jersey.spi.scanning.PathProviderScannerListener.<init>(PathProviderScannerListener.java:59) ~[jersey-server-1.19.jar:1.19] at com.sun.jersey.api.core.ScanningResourceConfig.init(ScanningResourceConfig.java:79) ~[jersey-server-1.19.jar:1.19] at com.sun.jersey.api.core.PackagesResourceConfig.init(PackagesResourceConfig.java:104) ~[jersey-server-1.19.jar:1.19] at com.sun.jersey.api.core.PackagesResourceConfig.<init>(PackagesResourceConfig.java:78) ~[jersey-server-1.19.jar:1.19] at com.sun.jersey.api.core.PackagesResourceConfig.<init>(PackagesResourceConfig.java:89) ~[jersey-server-1.19.jar:1.19] at com.sun.jersey.spi.container.servlet.WebComponent.createResourceConfig(WebComponent.java:696) ~[jersey-servlet-1.17.1.jar:1.17.1] at com.sun.jersey.spi.container.servlet.WebComponent.createResourceConfig(WebComponent.java:674) ~[jersey-servlet-1.17.1.jar:1.17.1] at com.sun.jersey.spi.container.servlet.WebComponent.init(WebComponent.java:203) ~[jersey-servlet-1.17.1.jar:1.17.1] at com.sun.jersey.spi.container.servlet.ServletContainer.init(ServletContainer.java:374) ~[jersey-servlet-1.17.1.jar:1.17.1] at com.sun.jersey.spi.container.servlet.ServletContainer.init(ServletContainer.java:557) ~[jersey-servlet-1.17.1.jar:1.17.1] at javax.servlet.GenericServlet.init(GenericServlet.java:244) ~[javax.servlet-api-3.1.0.jar:3.1.0] at org.eclipse.jetty.servlet.ServletHolder.initServlet(ServletHolder.java:612) ~[jetty-servlet-9.2.4.v20141103.jar:9.2.4.v20141103] at org.eclipse.jetty.servlet.ServletHolder.initialize(ServletHolder.java:395) ~[jetty-servlet-9.2.4.v20141103.jar:9.2.4.v20141103] at org.eclipse.jetty.servlet.ServletHandler.initialize(ServletHandler.java:871) ~[jetty-servlet-9.2.4.v20141103.jar:9.2.4.v20141103] at org.eclipse.jetty.servlet.ServletContextHandler.startContext(ServletContextHandler.java:298) ~[jetty-servlet-9.2.4.v20141103.jar:9.2.4.v20141103] at org.eclipse.jetty.server.handler.ContextHandler.doStart(ContextHandler.java:741) ~[jetty-server-9.2.4.v20141103.jar:9.2.4.v20141103] at org.eclipse.jetty.util.component.AbstractLifeCycle.start(AbstractLifeCycle.java:68) ~[jetty-util-9.2.4.v20141103.jar:9.2.4.v20141103] at org.eclipse.jetty.util.component.ContainerLifeCycle.start(ContainerLifeCycle.java:132) ~[jetty-util-9.2.4.v20141103.jar:9.2.4.v20141103] at org.eclipse.jetty.util.component.ContainerLifeCycle.doStart(ContainerLifeCycle.java:114) ~[jetty-util-9.2.4.v20141103.jar:9.2.4.v20141103] at org.eclipse.jetty.server.handler.AbstractHandler.doStart(AbstractHandler.java:61) ~[jetty-server-9.2.4.v20141103.jar:9.2.4.v20141103] at org.eclipse.jetty.util.component.AbstractLifeCycle.start(AbstractLifeCycle.java:68) ~[jetty-util-9.2.4.v20141103.jar:9.2.4.v20141103] at org.eclipse.jetty.util.component.ContainerLifeCycle.start(ContainerLifeCycle.java:132) ~[jetty-util-9.2.4.v20141103.jar:9.2.4.v20141103] at org.eclipse.jetty.server.Server.start(Server.java:387) ~[jetty-server-9.2.4.v20141103.jar:9.2.4.v20141103] at org.eclipse.jetty.util.component.ContainerLifeCycle.doStart(ContainerLifeCycle.java:114) ~[jetty-util-9.2.4.v20141103.jar:9.2.4.v20141103] at org.eclipse.jetty.server.handler.AbstractHandler.doStart(AbstractHandler.java:61) ~[jetty-server-9.2.4.v20141103.jar:9.2.4.v20141103] at org.eclipse.jetty.server.Server.doStart(Server.java:354) ~[jetty-server-9.2.4.v20141103.jar:9.2.4.v20141103] at org.eclipse.jetty.util.component.AbstractLifeCycle.start(AbstractLifeCycle.java:68) ~[jetty-util-9.2.4.v20141103.jar:9.2.4.v20141103] at org.neo4j.server.web.Jetty9WebServer.startJetty(Jetty9WebServer.java:381) [neo4j-server-2.2.3.jar:2.2.3] at org.neo4j.server.web.Jetty9WebServer.start(Jetty9WebServer.java:184) [neo4j-server-2.2.3.jar:2.2.3] at org.neo4j.server.AbstractNeoServer.startWebServer(AbstractNeoServer.java:474) [neo4j-server-2.2.3.jar:2.2.3] at org.neo4j.server.AbstractNeoServer.start(AbstractNeoServer.java:230) [neo4j-server-2.2.3.jar:2.2.3] at org.neo4j.harness.internal.InProcessServerControls.start(InProcessServerControls.java:59) [neo4j-harness-2.2.3.jar:2.2.3] at org.neo4j.harness.internal.InProcessServerBuilder.newServer(InProcessServerBuilder.java:72) [neo4j-harness-2.2.3.jar:2.2.3] at org.neo4j.harness.junit.Neo4jRule$1.evaluate(Neo4jRule.java:64) [neo4j-harness-2.2.3.jar:2.2.3] at org.junit.rules.RunRules.evaluate(RunRules.java:20) [junit-4.11.jar:na] at org.junit.runners.ParentRunner.runLeaf(ParentRunner.java:271) [junit-4.11.jar:na] at org.junit.runners.BlockJUnit4ClassRunner.runChild(BlockJUnit4ClassRunner.java:70) [junit-4.11.jar:na] at org.junit.runners.BlockJUnit4ClassRunner.runChild(BlockJUnit4ClassRunner.java:50) [junit-4.11.jar:na] at org.junit.runners.ParentRunner$3.run(ParentRunner.java:238) [junit-4.11.jar:na] at org.junit.runners.ParentRunner$1.schedule(ParentRunner.java:63) [junit-4.11.jar:na] at org.junit.runners.ParentRunner.runChildren(ParentRunner.java:236) [junit-4.11.jar:na] at org.junit.runners.ParentRunner.access$000(ParentRunner.java:53) [junit-4.11.jar:na] at org.junit.runners.ParentRunner$2.evaluate(ParentRunner.java:229) [junit-4.11.jar:na] at org.junit.runners.ParentRunner.run(ParentRunner.java:309) [junit-4.11.jar:na] at org.junit.runner.JUnitCore.run(JUnitCore.java:160) [junit-4.11.jar:na] at com.intellij.junit4.JUnit4IdeaTestRunner.startRunnerWithArgs(JUnit4IdeaTestRunner.java:78) [junit-rt.jar:na] at com.intellij.rt.execution.junit.JUnitStarter.prepareStreamsAndStart(JUnitStarter.java:212) [junit-rt.jar:na] at com.intellij.rt.execution.junit.JUnitStarter.main(JUnitStarter.java:68) [junit-rt.jar:na] at sun.reflect.NativeMethodAccessorImpl.invoke0(Native Method) ~[na:1.8.0_51] at sun.reflect.NativeMethodAccessorImpl.invoke(NativeMethodAccessorImpl.java:62) ~[na:1.8.0_51] at sun.reflect.DelegatingMethodAccessorImpl.invoke(DelegatingMethodAccessorImpl.java:43) ~[na:1.8.0_51] at java.lang.reflect.Method.invoke(Method.java:497) ~[na:1.8.0_51] at com.intellij.rt.execution.application.AppMain.main(AppMain.java:140) [idea_rt.jar:na] org.neo4j.server.ServerStartupException: Starting Neo4j Server failed: com.sun.jersey.core.reflection.ReflectionHelper.getContextClassLoaderPA()Ljava/security/PrivilegedAction; at org.neo4j.server.AbstractNeoServer.start(AbstractNeoServer.java:258) at org.neo4j.harness.internal.InProcessServerControls.start(InProcessServerControls.java:59) at org.neo4j.harness.internal.InProcessServerBuilder.newServer(InProcessServerBuilder.java:72) at org.neo4j.harness.junit.Neo4jRule$1.evaluate(Neo4jRule.java:64) at org.junit.rules.RunRules.evaluate(RunRules.java:20) at org.junit.runners.ParentRunner.runLeaf(ParentRunner.java:271) at org.junit.runners.BlockJUnit4ClassRunner.runChild(BlockJUnit4ClassRunner.java:70) at org.junit.runners.BlockJUnit4ClassRunner.runChild(BlockJUnit4ClassRunner.java:50) at org.junit.runners.ParentRunner$3.run(ParentRunner.java:238) at org.junit.runners.ParentRunner$1.schedule(ParentRunner.java:63) at org.junit.runners.ParentRunner.runChildren(ParentRunner.java:236) at org.junit.runners.ParentRunner.access$000(ParentRunner.java:53) at org.junit.runners.ParentRunner$2.evaluate(ParentRunner.java:229) at org.junit.runners.ParentRunner.run(ParentRunner.java:309) at org.junit.runner.JUnitCore.run(JUnitCore.java:160) at com.intellij.junit4.JUnit4IdeaTestRunner.startRunnerWithArgs(JUnit4IdeaTestRunner.java:78) at com.intellij.rt.execution.junit.JUnitStarter.prepareStreamsAndStart(JUnitStarter.java:212) at com.intellij.rt.execution.junit.JUnitStarter.main(JUnitStarter.java:68) at sun.reflect.NativeMethodAccessorImpl.invoke0(Native Method) at sun.reflect.NativeMethodAccessorImpl.invoke(NativeMethodAccessorImpl.java:62) at com.intellij.rt.execution.application.AppMain.main(AppMain.java:140) Caused by: java.lang.NoSuchMethodError: com.sun.jersey.core.reflection.ReflectionHelper.getContextClassLoaderPA()Ljava/security/PrivilegedAction; at com.sun.jersey.spi.scanning.AnnotationScannerListener.<init>(AnnotationScannerListener.java:94) at com.sun.jersey.spi.scanning.PathProviderScannerListener.<init>(PathProviderScannerListener.java:59) at com.sun.jersey.api.core.ScanningResourceConfig.init(ScanningResourceConfig.java:79) at com.sun.jersey.api.core.PackagesResourceConfig.init(PackagesResourceConfig.java:104) at com.sun.jersey.api.core.PackagesResourceConfig.<init>(PackagesResourceConfig.java:78) at com.sun.jersey.api.core.PackagesResourceConfig.<init>(PackagesResourceConfig.java:89) at com.sun.jersey.spi.container.servlet.WebComponent.createResourceConfig(WebComponent.java:696) at com.sun.jersey.spi.container.servlet.WebComponent.createResourceConfig(WebComponent.java:674) at com.sun.jersey.spi.container.servlet.WebComponent.init(WebComponent.java:203) at com.sun.jersey.spi.container.servlet.ServletContainer.init(ServletContainer.java:374) at com.sun.jersey.spi.container.servlet.ServletContainer.init(ServletContainer.java:557) at javax.servlet.GenericServlet.init(GenericServlet.java:244) at org.eclipse.jetty.servlet.ServletHolder.initServlet(ServletHolder.java:612) at org.eclipse.jetty.servlet.ServletHolder.initialize(ServletHolder.java:395) at org.eclipse.jetty.servlet.ServletHandler.initialize(ServletHandler.java:871) at org.eclipse.jetty.servlet.ServletContextHandler.startContext(ServletContextHandler.java:298) at org.eclipse.jetty.server.handler.ContextHandler.doStart(ContextHandler.java:741) at org.eclipse.jetty.util.component.AbstractLifeCycle.start(AbstractLifeCycle.java:68) at org.eclipse.jetty.util.component.ContainerLifeCycle.start(ContainerLifeCycle.java:132) at org.eclipse.jetty.util.component.ContainerLifeCycle.doStart(ContainerLifeCycle.java:114) at org.eclipse.jetty.server.handler.AbstractHandler.doStart(AbstractHandler.java:61) at org.eclipse.jetty.util.component.AbstractLifeCycle.start(AbstractLifeCycle.java:68) at org.eclipse.jetty.util.component.ContainerLifeCycle.start(ContainerLifeCycle.java:132) at org.eclipse.jetty.server.Server.start(Server.java:387) at org.eclipse.jetty.util.component.ContainerLifeCycle.doStart(ContainerLifeCycle.java:114) at org.eclipse.jetty.server.handler.AbstractHandler.doStart(AbstractHandler.java:61) at org.eclipse.jetty.server.Server.doStart(Server.java:354) at org.eclipse.jetty.util.component.AbstractLifeCycle.start(AbstractLifeCycle.java:68) at org.neo4j.server.web.Jetty9WebServer.startJetty(Jetty9WebServer.java:381) at org.neo4j.server.web.Jetty9WebServer.start(Jetty9WebServer.java:184) at org.neo4j.server.AbstractNeoServer.startWebServer(AbstractNeoServer.java:474) at org.neo4j.server.AbstractNeoServer.start(AbstractNeoServer.java:230) ... 22 more ---- I was a bit baffled at first but if we look closely about 5 - 10 lines down the stack trace we can see the mistake I've made: [source,text] ---- at com.sun.jersey.api.core.PackagesResourceConfig.(PackagesResourceConfig.java:89) ~[jersey-server-1.19.jar:1.19] at com.sun.jersey.spi.container.servlet.WebComponent.createResourceConfig(WebComponent.java:696) ~[jersey-servlet-1.17.1.jar:1.17.1] at com.sun.jersey.spi.container.servlet.WebComponent.createResourceConfig(WebComponent.java:674) ~[jersey-servlet-1.17.1.jar:1.17.1] ---- We have different versions of the Jersey libraries. Since we're writing the extension for Neo4j 2.2.3 we can quickly check which version it depends on: [source,bash] ---- $ ls -alh neo4j-community-2.2.3/system/lib/ | grep jersey -rwxr-xr-x@ 1 markneedham staff 426K 22 Jun 04:57 jersey-core-1.19.jar -rwxr-xr-x@ 1 markneedham staff 52K 22 Jun 05:02 jersey-multipart-1.19.jar -rwxr-xr-x@ 1 markneedham staff 686K 22 Jun 05:02 jersey-server-1.19.jar -rwxr-xr-x@ 1 markneedham staff 126K 22 Jun 05:02 jersey-servlet-1.19.jar ---- So we should't have 1.17.1 in our project and it was easy enough to find my mistake by looking in the pom file: [source,xml] ---- <properties> ... <jersey.version>1.17.1</jersey.version> ... </properties> <dependencies> ... <dependency> <groupId>com.sun.jersey</groupId> <artifactId>jersey-servlet</artifactId> <version>${jersey.version}</version> </dependency> ... </dependencies> ---- Also easy enough to fix! [source,xml] ---- <properties> ... <jersey.version>1.19</jersey.version> ... </properties> ---- You can see https://github.com/mneedham/dummy-unmanaged-extension/blob/master/src/test/java/org/neo4j/unmanaged/ExampleResourceTest.java[an example of this working on github].

java-jersey-java-lang-nosuchmethoderror-com-sun-jersey-core-reflection-reflectionhelper-getcontextclassloaderpaljavasecurityprivilegedaction

https://markhneedham.com/blog/2015/08/11/java-jersey-java-lang-nosuchmethoderror-com-sun-jersey-core-reflection-reflectionhelper-getcontextclassloaderpaljavasecurityprivilegedaction

2015-08-10 23:57:01

Neo4j 2.2.3: Unmanaged extensions - Creating gzipped streamed responses with Jetty

Back in 2013 I wrote a http://www.markhneedham.com/blog/2013/07/08/jax-rs-streaming-a-response-using-streamingoutput/[couple of] http://www.markhneedham.com/blog/2013/07/08/neo4j-unmanaged-extension-creating-gzipped-streamed-responses-with-jetty/[blog posts] showing examples of an unmanaged extension which had a streamed and gzipped response but two years on I realised they were a bit out of date and deserved a refresh. When writing unmanaged extensions in Neo4j a good rule of thumb is to try and reduce the amount of objects you keep hanging around. In this context this means that we should stream our response to the client as quickly as possible rather than building it up in memory and sending it in one go. The http://neo4j.com/docs/2.2.3/server-unmanaged-extensions.html#server-unmanaged-extensions-streaming[documentation has a good example showing how to stream a list of colleagues] but in this blog post we'll look at how to do something simpler - we'll create a couple of nodes representing people and then write an unmanaged extension to return them. We'll first create an unmanaged extension which runs a cypher query, iterates through the rows returned and sends them to the client: [source,java] ---- @Path("/example") public class ExampleResource { private final GraphDatabaseService db; private static final ObjectMapper OBJECT_MAPPER = new ObjectMapper(); public ExampleResource(@Context GraphDatabaseService db) { this.db = db; } @GET @Produces(MediaType.APPLICATION_JSON) @Path("/people") public Response allNodes() throws IOException { StreamingOutput stream = streamQueryResponse("MATCH (n:Person) RETURN n.name AS name"); return Response.ok().entity(stream).type(MediaType.APPLICATION_JSON).build(); } private StreamingOutput streamQueryResponse(final String query) { return new StreamingOutput() { @Override public void write(OutputStream os) throws IOException, WebApplicationException { JsonGenerator jg = OBJECT_MAPPER.getJsonFactory().createJsonGenerator(os, JsonEncoding.UTF8); jg.writeStartArray(); writeQueryResultTo(query, jg); jg.writeEndArray(); jg.flush(); jg.close(); } }; } private void writeQueryResultTo(String query, JsonGenerator jg) throws IOException { try (Result result = db.execute(query)) { while (result.hasNext()) { Map<String, Object> row = result.next(); jg.writeStartObject(); for (Map.Entry<String, Object> entry : row.entrySet()) { jg.writeFieldName(entry.getKey()); jg.writeString(entry.getValue().toString()); } jg.writeEndObject(); } } } } ---- There's nothing too complicated going on here although notice that we make much more fine grained calls to the JSON Library rather than created a JSON object in memory and calling +++<cite>+++ObjectMapper#writeValueAsString+++</cite>+++ on it. To get this to work we'd build a JAR containing this class, put that into the plugins folder and then add the following property to +++<cite>+++conf/neo4j-server.properties+++</cite>+++ (or the Neo4j desktop equivalent) before restarting the server: [source,text] ---- org.neo4j.server.thirdparty_jaxrs_classes=org.neo4j.unmanaged=/unmanaged ---- We can then test it out like this: [source,bash] ---- $ curl http://localhost:7474/unmanaged/example/people [{"name":"Mark"},{"name":"Nicole"}] ---- I've put in a couple of test people nodes - full instructions are available on the https://github.com/mneedham/dummy-unmanaged-extension[github README page]. Next we want to make it possible to send that response in the gzip format. To do that we need to add a GzipFilter to the Neo4j lifecycle. This class has moved to a different namespace in Jetty 9 which Neo4j 2.2.3 depends on, but the following class does the job: [source,java] ---- import org.eclipse.jetty.servlets.GzipFilter; public class GZipInitialiser implements SPIPluginLifecycle { private WebServer webServer; @Override public Collection<Injectable<?>> start(NeoServer neoServer) { webServer = getWebServer(neoServer); GzipFilter filter = new GzipFilter(); webServer.addFilter(filter, "/*"); return Collections.emptyList(); } private WebServer getWebServer(final NeoServer neoServer) { if (neoServer instanceof AbstractNeoServer) { return ((AbstractNeoServer) neoServer).getWebServer(); } throw new IllegalArgumentException("expected AbstractNeoServer"); } @Override public Collection<Injectable<?>> start(GraphDatabaseService graphDatabaseService, Configuration configuration) { throw new IllegalAccessError(); } @Override public void stop() { } } ---- I needed to include the +++<cite>+++jersey-servlets+++</cite>+++ JAR in my unmanaged extension JAR in order for this to work correctly. Once we redeploy the JAR and restart Neo4j we can try making the same request as above but with a gzip header: [source,bash] ---- $ curl -v -H "Accept-Encoding:gzip,deflate" http://localhost:7474/unmanaged/example/people ��V�K�MU�R�M,�V�Ձ��2��sR�jcf(�# ---- We can unpack that on the fly by piping it through gunzip to check we get a sensible result: [source,bash] ---- $ curl -v -H "Accept-Encoding:gzip,deflate" http://localhost:7474/unmanaged/example/people | gunzip [{"name":"Mark"},{"name":"Nicole"}] ---- And there we have it - a gzipped streamed response. https://github.com/mneedham/dummy-unmanaged-extension[All the code is on github] so give it a try and give me a shout if it doesn't work. The fastest way to get me is probably on our new shiny http://neo4j-users-slack-invite.herokuapp.com/[neo4j-users Slack group].

neo4j-2-2-3-unmanaged-extensions-creating-gzipped-streamed-responses-with-jetty

https://markhneedham.com/blog/2015/08/10/neo4j-2-2-3-unmanaged-extensions-creating-gzipped-streamed-responses-with-jetty

2015-08-19 23:27:42

Python: Extracting Excel spreadsheet into CSV files

I've been playing around with the http://data.gov.uk/dataset/road-accidents-safety-data[Road Safety open data set] and the download comes with several CSV files and an excel spreadsheet containing the legend. There are 45 sheets in total and each of them looks like this: image::{{<siteurl>}}/uploads/2015/08/2015-08-17_23-33-19.png[2015 08 17 23 33 19] I wanted to create a CSV file for each sheet so that I can import the data set into Neo4j using the http://neo4j.com/docs/stable/query-load-csv.html[LOAD CSV command]. I came across the http://www.python-excel.org/[Python Excel website] which pointed me at the +++<cite>+++xlrd+++</cite>+++ library since I'm working with a pre 2010 Excel file. The https://secure.simplistix.co.uk/svn/xlrd/trunk/xlrd/doc/xlrd.html?p=4966#sheet.Cell-class[main documentation is very extensive] but I found the https://github.com/python-excel/xlrd[github example] much easier to follow. I ended up with the following script which iterates through all but the first two sheets in the spreadsheet - the first two sheets contain instructions rather than data: [source,python] ---- from xlrd import open_workbook import csv wb = open_workbook('Road-Accident-Safety-Data-Guide-1979-2004.xls') for i in range(2, wb.nsheets): sheet = wb.sheet_by_index(i) print sheet.name with open("data/%s.csv" %(sheet.name.replace(" ","")), "w") as file: writer = csv.writer(file, delimiter = ",") print sheet, sheet.name, sheet.ncols, sheet.nrows header = [cell.value for cell in sheet.row(0)] writer.writerow(header) for row_idx in range(1, sheet.nrows): row = [int(cell.value) if isinstance(cell.value, float) else cell.value for cell in sheet.row(row_idx)] writer.writerow(row) ---- I've replaced spaces in the sheet name so that the file name on a disk is a bit easier to work with. For some reason the numeric values were all floats whereas I wanted them as ints so I had to explicitly apply that transformation. Here are a few examples of what the CSV files look like: [source,bash] ---- $ cat data/1stPointofImpact.csv code,label 0,Did not impact 1,Front 2,Back 3,Offside 4,Nearside -1,Data missing or out of range $ cat data/RoadType.csv code,label 1,Roundabout 2,One way street 3,Dual carriageway 6,Single carriageway 7,Slip road 9,Unknown 12,One way street/Slip road -1,Data missing or out of range $ cat data/Weather.csv code,label 1,Fine no high winds 2,Raining no high winds 3,Snowing no high winds 4,Fine + high winds 5,Raining + high winds 6,Snowing + high winds 7,Fog or mist 8,Other 9,Unknown -1,Data missing or out of range ---- And that's it. Not too difficult!

python-extracting-excel-spreadsheet-into-csv-files

https://markhneedham.com/blog/2015/08/19/python-extracting-excel-spreadsheet-into-csv-files

2015-08-19 23:27:28

Unix: Stripping first n bytes in a file / Byte Order Mark (BOM)

I've previously written http://www.markhneedham.com/blog/2012/10/07/mac-os-x-removing-byte-order-mark-with-an-editor/[a couple] http://www.markhneedham.com/blog/2012/09/03/a-rogue-357273277-utf-8-byte-order-mark/[of blog posts] showing how to strip out the byte order mark (BOM) from CSV files to make loading them into Neo4j easier and today I came across another way to clean up the file using tail. The BOM is 3 bytes long at the beginning of the file so if we know that a file contains it then we can strip out those first 3 bytes tail like this: [source,bash] ---- $ time tail -c +4 Casualty7904.csv > Casualty7904_stripped.csv real 0m31.945s user 0m31.370s sys 0m0.518s ---- The -c command is described thus; [source,bash] ---- -c number The location is number bytes. ---- So in this case we start reading at byte 4 (i.e. skipping the first 3 bytes) and then direct the output into a new file. Although using tail is quite simple, it took 30 seconds to process a 300MB CSV file which might actually be slower than opening the file with a Hex editor and manually deleting the bytes!

unix-stripping-first-n-bytes-in-a-file-byte-order-mark-bom

https://markhneedham.com/blog/2015/08/19/unix-stripping-first-n-bytes-in-a-file-byte-order-mark-bom

2015-08-21 20:59:37

Neo4j: Summarising neo4j-shell output

</p> I frequently find myself trying to optimise a set of cypher queries and I tend to group them together in a script that I fed to the http://neo4j.com/docs/stable/shell.html[Neo4j shell]. </p> When tweaking the queries it's easy to make a mistake and end up not creating the same data so I decided to write a script which will show me the aggregates of all the commands executed. I want to see the number of constraints created, indexes added, nodes, relationships and properties created. The first 2 don't need to match across the scripts but the latter 3 should be the same. I put together the following script: [source,python] ---- import re import sys from tabulate import tabulate lines = sys.stdin.readlines() def search(term, line): m = re.match(term + ": (.*)", line) return (int(m.group(1)) if m else 0) nodes_created, relationships_created, constraints_added, indexes_added, labels_added, properties_set = 0, 0, 0, 0, 0, 0 for line in lines: nodes_created = nodes_created + search("Nodes created", line) relationships_created = relationships_created + search("Relationships created", line) constraints_added = constraints_added + search("Constraints added", line) indexes_added = indexes_added + search("Indexes added", line) labels_added = labels_added + search("Labels added", line) properties_set = properties_set + search("Properties set", line) time_match = re.match("real.*([0-9]+m[0-9]+\.[0-9]+s)$", line) if time_match: time = time_match.group(1) table = [ ["Constraints added", constraints_added], ["Indexes added", indexes_added], ["Nodes created", nodes_created], ["Relationships created", relationships_created], ["Labels added", labels_added], ["Properties set", properties_set], ["Time", time] ] print tabulate(table) ---- Its input is the piped output of the neo4j-shell command which will contain a description of all the queries it executed. [source,bash] ---- $ cat import.sh #!/bin/sh { ./neo4j-community-2.2.3/bin/neo4j stop; } 2>&1 rm -rf neo4j-community-2.2.3/data/graph.db/ { ./neo4j-community-2.2.3/bin/neo4j start; } 2>&1 { time ./neo4j-community-2.2.3/bin/neo4j-shell --file $1; } 2>&1 ---- We can use the script in two ways. Either we can pipe the output of our shell straight into it and just get the summary e.g. [source,bash] ---- $ ./import.sh local.import.optimised.cql | python summarise.py --------------------- --------- Constraints added 5 Indexes added 1 Nodes created 13249 Relationships created 32227 Labels added 21715 Properties set 36480 Time 0m17.595s --------------------- --------- ---- \...or we can make use of the 'tee' function in Unix and pipe the output into stdout and into the file and then either tail the file on another window or inspect it afterwards to see the detailed timings. e.g. [source,bash] ---- $ ./import.sh local.import.optimised.cql | tee /tmp/output.txt | python summarise.py --------------------- --------- Constraints added 5 Indexes added 1 Nodes created 13249 Relationships created 32227 Labels added 21715 Properties set 36480 Time 0m11.428s --------------------- --------- ---- [source,bash] ---- $ tail -f /tmp/output.txt +-------------+ | appearances | +-------------+ | 3771 | +-------------+ 1 row Nodes created: 3439 Properties set: 3439 Labels added: 3439 289 ms +------------------------------------+ | appearances -> player, match, team | +------------------------------------+ | 3771 | +------------------------------------+ 1 row Relationships created: 10317 1006 ms ... ---- My only dependency is the tabulate package to get the pretty table: [source,bash] ---- $ cat requirements.txt tabulate==0.7.5 ---- The https://github.com/mneedham/neo4j-bbc/blob/master/local.import.optimised.cql[cypher script] I'm running creates a BBC football graph which is available as a https://github.com/mneedham/neo4j-bbc[github project]. Feel free to grab it and play around - any problems let me know!

neo4j-summarising-neo4j-shell-output

https://markhneedham.com/blog/2015/08/21/neo4j-summarising-neo4j-shell-output

2015-08-08 22:50:41

Record Linkage: Playing around with Duke

I've become quite interesting in https://en.wikipedia.org/wiki/Record_linkage[record linkage] recently and came across the https://github.com/larsga/Duke[Duke] project which provides some tools to help solve this problem. I thought I'd give it a try. The typical problem when doing record linkage is that we have two records from different data sets which represent the same entity but don't have a common key that we can use to merge them together. We therefore need to come up with a heuristic that will allow us to do so. Duke has a few examples showing it in action and I decided to go with the https://github.com/larsga/Duke/wiki/LinkingCountries[linking countries] one. Here we have countries from Dbpedia and the Mondial database and we want to link them together. The first thing we need to do is build the project: [source,bash] ---- export JAVA_HOME=`/usr/libexec/java_home` mvn clean package -DskipTests ---- At the time of writing this will put a zip fail containing everything we need at +++<cite>+++duke-dist/target/+++</cite>+++. Let's unpack that: [source,bash] ---- unzip duke-dist/target/duke-dist-1.3-SNAPSHOT-bin.zip ---- Next we need to download the data files and Duke configuration file: [source,bash] ---- wget https://raw.githubusercontent.com/larsga/Duke/master/doc/example-data/countries-dbpedia.csv wget https://raw.githubusercontent.com/larsga/Duke/master/doc/example-data/countries.xml wget https://raw.githubusercontent.com/larsga/Duke/master/doc/example-data/countries-mondial.csv wget https://raw.githubusercontent.com/larsga/Duke/master/doc/example-data/countries-test.txt ---- Now we're ready to give it a go: [source,bash] ---- java -cp "duke-dist-1.3-SNAPSHOT/lib/*" no.priv.garshol.duke.Duke --testfile=countries-test.txt --testdebug --showmatches countries.xml ... NO MATCH FOR: ID: '7706', NAME: 'guatemala', AREA: '108890', CAPITAL: 'guatemala city', MATCH 0.9825124555160142 ID: '10052', NAME: 'pitcairn islands', AREA: '47', CAPITAL: 'adamstown', ID: 'http://dbpedia.org/resource/Pitcairn_Islands', NAME: 'pitcairn islands', AREA: '47', CAPITAL: 'adamstown', Correct links found: 200 / 218 (91.7%) Wrong links found: 0 / 24 (0.0%) Unknown links found: 0 Percent of links correct 100.0%, wrong 0.0%, unknown 0.0% Records with no link: 18 Precision 100.0%, recall 91.74311926605505%, f-number 0.9569377990430622 ---- We can look in +++<cite>+++countries.xml+++</cite>+++ to see how the similarity between records is being calculated: [source,xml] ---- <schema> <threshold>0.7</threshold> ... <property> <name>NAME</name> <comparator>no.priv.garshol.duke.comparators.Levenshtein</comparator> <low>0.09</low> <high>0.93</high> </property> <property> <name>AREA</name> <comparator>no.priv.garshol.duke.comparators.NumericComparator</comparator> <low>0.04</low> <high>0.73</high> </property> <property> <name>CAPITAL</name> <comparator>no.priv.garshol.duke.comparators.Levenshtein</comparator> <low>0.12</low> <high>0.61</high> </property> </schema> ---- So we're working out similarity of the capital city and country by calculating their Levenshtein distance i.e. the minimum number of single-character edits required to change one word into the other This works very well if there is a typo or difference in spelling in one of the data sets. However, I was curious what would happen if the country had two completely different names e.g Cote d'Ivoire is sometimes know as Ivory Coast. Let's try changing the country name in one of the files: [source,text] ---- "19147","Cote dIvoire","Yamoussoukro","322460" ---- [source,bash] ---- java -cp "duke-dist-1.3-SNAPSHOT/lib/*" no.priv.garshol.duke.Duke --testfile=countries-test.txt --testdebug --showmatches countries.xml NO MATCH FOR: ID: '19147', NAME: 'ivory coast', AREA: '322460', CAPITAL: 'yamoussoukro', ---- I also tried it out with the http://www.bbc.co.uk/sport/0/football/33744640[BBC] and http://www.espnfc.co.uk/gamecast/statistics/id/422662/statistics.html[ESPN] match reports of the Man Utd vs Tottenham match - the BBC references players by surname, while ESPN has their full names. When I compared the full name against surname using the Levenshtein comparator there were no matches as you'd expect. I had to split the ESPN names up into first name and surname to get the linking to work. Equally when I varied the team name's to be 'Man Utd' rather than 'Manchester United' and 'Tottenham' rather than 'Tottenham Hotspur' that didn't work either. I think I probably need to write a domain specific comparator but I'm also curious whether I could come up with a bunch of training examples and then train a model to detect what makes two records similar. It'd be less deterministic but perhaps more robust.

record-linkage-playing-around-with-duke

https://markhneedham.com/blog/2015/08/08/record-linkage-playing-around-with-duke

2015-08-06 21:11:44

Spark: Convert RDD to DataFrame

As I mentioned in a previous blog post I've been http://www.markhneedham.com/blog/2015/08/02/spark-processing-csv-files-using-databricks-spark-csv-library/[playing around with the Databricks Spark CSV library] and wanted to take a CSV file, clean it up and then write out a new CSV file containing some of the columns. I started by processing the CSV file and writing it into a temporary table: [source,scala] ---- import org.apache.spark.sql.{SQLContext, Row, DataFrame} val sqlContext = new SQLContext(sc) val crimeFile = "Crimes_-_2001_to_present.csv" sqlContext.load("com.databricks.spark.csv", Map("path" -> crimeFile, "header" -> "true")).registerTempTable("crimes") ---- I wanted to get to the point where I could call the following function which writes a DataFrame to disk: [source,scala] ---- private def createFile(df: DataFrame, file: String, header: String): Unit = { FileUtil.fullyDelete(new File(file)) val tmpFile = "tmp/" + System.currentTimeMillis() + "-" + file df.distinct.save(tmpFile, "com.databricks.spark.csv") } ---- The first file only needs to contain the primary type of crime, which we can extract with the following query: [source,scala] ---- val rows = sqlContext.sql("select `Primary Type` as primaryType FROM crimes LIMIT 10") rows.collect() res4: Array[org.apache.spark.sql.Row] = Array([ASSAULT], [ROBBERY], [CRIMINAL DAMAGE], [THEFT], [THEFT], [BURGLARY], [THEFT], [BURGLARY], [THEFT], [CRIMINAL DAMAGE]) ---- Some of the primary types have trailing spaces which I want to get rid of. As far as I can tell Spark's variant of SQL doesn't have the LTRIM or RTRIM functions but we can map over 'rows' and use the String 'trim' function instead: [source,scala] ---- rows.map { case Row(primaryType: String) => Row(primaryType.trim) } res8: org.apache.spark.rdd.RDD[org.apache.spark.sql.Row] = MapPartitionsRDD[29] at map at DataFrame.scala:776 ---- Now we've got an RDD of Rows which we need to convert back to a DataFrame again. 'sqlContext' has a function which we might be able to use: [source,scala] ---- sqlContext.createDataFrame(rows.map { case Row(primaryType: String) => Row(primaryType.trim) }) <console>:27: error: overloaded method value createDataFrame with alternatives: [A <: Product](data: Seq[A])(implicit evidence$4: reflect.runtime.universe.TypeTag[A])org.apache.spark.sql.DataFrame <and> [A <: Product](rdd: org.apache.spark.rdd.RDD[A])(implicit evidence$3: reflect.runtime.universe.TypeTag[A])org.apache.spark.sql.DataFrame cannot be applied to (org.apache.spark.rdd.RDD[org.apache.spark.sql.Row]) sqlContext.createDataFrame(rows.map { case Row(primaryType: String) => Row(primaryType.trim) }) ^ ---- These are the signatures we can choose from: image::{{<siteurl>}}/uploads/2015/08/2015-08-06_21-58-12.png[2015 08 06 21 58 12,250] If we want to pass in an RDD of type Row we're going to have to define a StructType or we can convert each row into something more strongly typed: [source,scala] ---- case class CrimeType(primaryType: String) sqlContext.createDataFrame(rows.map { case Row(primaryType: String) => CrimeType(primaryType.trim) }) res14: org.apache.spark.sql.DataFrame = [primaryType: string] ---- Great, we've got our DataFrame which we can now plug into the 'createFile' function like so: [source,scala] ---- createFile( sqlContext.createDataFrame(rows.map { case Row(primaryType: String) => CrimeType(primaryType.trim) }), "/tmp/crimeTypes.csv", "crimeType:ID(CrimeType)") ---- We can actually http://alvincjin.blogspot.co.uk/2015/03/dataframe-in-spark13.html[do better though]! Since we've got an RDD of a specific class we can make use of the 'rddToDataFrameHolder' implicit function and then the 'toDF' function on 'DataFrameHolder'. This is what the code looks like: [source,scala] ---- import sqlContext.implicits._ createFile( rows.map { case Row(primaryType: String) => CrimeType(primaryType.trim) }.toDF(), "/tmp/crimeTypes.csv", "crimeType:ID(CrimeType)") ---- And we're done!

spark-convert-rdd-to-dataframe

https://markhneedham.com/blog/2015/08/06/spark-convert-rdd-to-dataframe

2015-08-15 15:55:32

Unix: Redirecting stderr to stdout

I've been trying to optimise some Neo4j import queries over the last couple of days and as part of the script I've been executed I wanted to redirect the output of a couple of commands into a file to parse afterwards. I started with the following script which doesn't do any explicit redirection of the output: [source,bash] ---- #!/bin/sh ./neo4j-community-2.2.3/bin/neo4j start ---- Now let's run that script and redirect the output to a file: [source,bash] ---- $ ./foo.sh > /tmp/output.txt Unable to find any JVMs matching version "1.7". $ cat /tmp/output.txt Starting Neo4j Server...WARNING: not changing user process [48230]... waiting for server to be ready.... OK. http://localhost:7474/ is ready. ---- So the line about not finding a matching JVM is being printed to stderr. That's reasonably easy to fix: [source,bash] ---- #!/bin/sh ./neo4j-community-2.2.3/bin/neo4j start 2>&1 ---- Let's run the script again: [source,bash] ---- $ ./foo.sh > /tmp/output.txt $ cat /tmp/output.txt Unable to find any JVMs matching version "1.7". Starting Neo4j Server...WARNING: not changing user process [47989]... waiting for server to be ready.... OK. http://localhost:7474/ is ready. ---- Great, that worked as expected. Next I extended the script to stop Neo4j, delete all it's data, start it again and execute a cypher script: [source,bash] ---- #!/bin/sh ./neo4j-community-2.2.3/bin/neo4j start 2>&1 rm -rf neo4j-community-2.2.3/data/graph.db/ ./neo4j-community-2.2.3/bin/neo4j start 2>&1 time ./neo4j-community-2.2.3/bin/neo4j-shell --file foo.cql 2>&1 ---- Let's run that script and redirect the output: [source,bash] ---- $ ./foo.sh > /tmp/output.txt Unable to find any JVMs matching version "1.7". real 0m0.604s user 0m0.334s sys 0m0.054s $ cat /tmp/output.txt Unable to find any JVMs matching version "1.7". Another server-process is running with [50614], cannot start a new one. Exiting. Unable to find any JVMs matching version "1.7". Another server-process is running with [50614], cannot start a new one. Exiting. +---------+ | "hello" | +---------+ | "hello" | +---------+ 1 row 4 ms ---- It looks like our stderr \-> stdout redirection on the last line didn't work. My understanding is that the 'time' command http://stackoverflow.com/questions/2408981/how-cant-i-redirect-the-output-of-time-command[swallows all the arguments that follow] whereas we want the redirection to be run afterwards. We can work our way around this problem by putting the actual command in a code block and redirected the output of that: [source,bash] ---- #!/bin/sh ./neo4j-community-2.2.3/bin/neo4j start 2>&1 rm -rf neo4j-community-2.2.3/data/graph.db/ ./neo4j-community-2.2.3/bin/neo4j start 2>&1 { time ./neo4j-community-2.2.3/bin/neo4j-shell --file foo.cql; } 2>&1 ---- [source,bash] ---- $ ./foo.sh > /tmp/output.txt $ cat /tmp/output.txt Unable to find any JVMs matching version "1.7". Another server-process is running with [50614], cannot start a new one. Exiting. Unable to find any JVMs matching version "1.7". Another server-process is running with [50614], cannot start a new one. Exiting. Unable to find any JVMs matching version "1.7". +---------+ | "hello" | +---------+ | "hello" | +---------+ 1 row 4 ms real 0m0.615s user 0m0.316s sys 0m0.050s ---- Much better!

unix-redirecting-stderr-to-stdout

https://markhneedham.com/blog/2015/08/15/unix-redirecting-stderr-to-stdout

2015-08-13 19:30:51

Sed: Using environment variables

I've been playing around with the http://www.markhneedham.com/blog/2015/05/16/neo4j-bbc-football-live-text-fouls-graph/[BBC football data set] that I wrote about a couple of months ago and I wanted to write some code that would take the import script and replace all instances of remote URIs with a file system path. For example the import file contains several lines similar to this: [source,text] ---- LOAD CSV WITH HEADERS FROM "https://raw.githubusercontent.com/mneedham/neo4j-bbc/master/data/matches.csv" AS row ---- And I want that to read: [source,text] ---- LOAD CSV WITH HEADERS FROM "file:///Users/markneedham/repos/neo4j-bbc/data/matches.csv" AS row ---- The start of that path also happens to be my working directory: [source,bash] ---- $ echo $PWD /Users/markneedham/repos/neo4j-bbc ---- So I wanted to write a script that would look for occurrences of 'https://raw.githubusercontent.com/mneedham/neo4j-bbc/master' and replace it with $PWD. I'm a fan of Sed so I thought I'd try and use it to solve my problem. The first thing we can do to make life easy is to change the default delimiter. Sed usually uses '/' to separate parts of the command but since we're using URIs that's going to be horrible so we'll use an underscore instead. For a first cut I tried just removing that first part of the URI but not replacing it with anything in particular: [source,bash] ---- $ sed 's_https://raw.githubusercontent.com/mneedham/neo4j-bbc/master__' import.cql $ sed 's_https://raw.githubusercontent.com/mneedham/neo4j-bbc/master__' import.cql | grep LOAD LOAD CSV WITH HEADERS FROM "/data/matches.csv" AS row LOAD CSV WITH HEADERS FROM "/data/players.csv" AS row LOAD CSV WITH HEADERS FROM "/data/players.csv" AS row LOAD CSV WITH HEADERS FROM "/data/fouls.csv" AS row LOAD CSV WITH HEADERS FROM "/data/attempts.csv" AS row LOAD CSV WITH HEADERS FROM "/data/attempts.csv" AS row LOAD CSV WITH HEADERS FROM "/data/corners.csv" AS row LOAD CSV WITH HEADERS FROM "/data/corners.csv" AS row LOAD CSV WITH HEADERS FROM "/data/cards.csv" AS row LOAD CSV WITH HEADERS FROM "/data/cards.csv" AS row LOAD CSV WITH HEADERS FROM "/data/subs.csv" AS row ---- Cool! That worked as expected. Now let's try and replace it with $PWD: [source,bash] ---- $ sed 's_https://raw.githubusercontent.com/mneedham/neo4j-bbc/master_file://$PWD_' import.cql | grep LOAD LOAD CSV WITH HEADERS FROM "file://$PWD/data/matches.csv" AS row LOAD CSV WITH HEADERS FROM "file://$PWD/data/players.csv" AS row LOAD CSV WITH HEADERS FROM "file://$PWD/data/players.csv" AS row LOAD CSV WITH HEADERS FROM "file://$PWD/data/fouls.csv" AS row LOAD CSV WITH HEADERS FROM "file://$PWD/data/attempts.csv" AS row LOAD CSV WITH HEADERS FROM "file://$PWD/data/attempts.csv" AS row LOAD CSV WITH HEADERS FROM "file://$PWD/data/corners.csv" AS row LOAD CSV WITH HEADERS FROM "file://$PWD/data/corners.csv" AS row LOAD CSV WITH HEADERS FROM "file://$PWD/data/cards.csv" AS row LOAD CSV WITH HEADERS FROM "file://$PWD/data/cards.csv" AS row LOAD CSV WITH HEADERS FROM "file://$PWD/data/subs.csv" AS row ---- Hmmm that didn't work as expected. The $PWD is being treated as a literal instead of being evaluated like we want it to be. It turns out this is http://askubuntu.com/questions/76808/how-to-use-variables-in-sed-command[a popular question on Stack Overflow] and there are lots of suggestions - I tried a few of them and found that single quotes did the trick: [source,bash] ---- $ sed 's_https://raw.githubusercontent.com/mneedham/neo4j-bbc/master_file://'$PWD'_' import.cql | grep LOAD LOAD CSV WITH HEADERS FROM "file:///Users/markneedham/repos/neo4j-bbc/data/matches.csv" AS row LOAD CSV WITH HEADERS FROM "file:///Users/markneedham/repos/neo4j-bbc/data/players.csv" AS row LOAD CSV WITH HEADERS FROM "file:///Users/markneedham/repos/neo4j-bbc/data/players.csv" AS row LOAD CSV WITH HEADERS FROM "file:///Users/markneedham/repos/neo4j-bbc/data/fouls.csv" AS row LOAD CSV WITH HEADERS FROM "file:///Users/markneedham/repos/neo4j-bbc/data/attempts.csv" AS row LOAD CSV WITH HEADERS FROM "file:///Users/markneedham/repos/neo4j-bbc/data/attempts.csv" AS row LOAD CSV WITH HEADERS FROM "file:///Users/markneedham/repos/neo4j-bbc/data/corners.csv" AS row LOAD CSV WITH HEADERS FROM "file:///Users/markneedham/repos/neo4j-bbc/data/corners.csv" AS row LOAD CSV WITH HEADERS FROM "file:///Users/markneedham/repos/neo4j-bbc/data/cards.csv" AS row LOAD CSV WITH HEADERS FROM "file:///Users/markneedham/repos/neo4j-bbc/data/cards.csv" AS row LOAD CSV WITH HEADERS FROM "file:///Users/markneedham/repos/neo4j-bbc/data/subs.csv" AS row ---- We could also use double quotes everywhere if we prefer: [source,bash] ---- $ sed "s_https://raw.githubusercontent.com/mneedham/neo4j-bbc/master_file://"$PWD"_" import.cql | grep LOAD LOAD CSV WITH HEADERS FROM "file:///Users/markneedham/repos/neo4j-bbc/data/matches.csv" AS row LOAD CSV WITH HEADERS FROM "file:///Users/markneedham/repos/neo4j-bbc/data/players.csv" AS row LOAD CSV WITH HEADERS FROM "file:///Users/markneedham/repos/neo4j-bbc/data/players.csv" AS row LOAD CSV WITH HEADERS FROM "file:///Users/markneedham/repos/neo4j-bbc/data/fouls.csv" AS row LOAD CSV WITH HEADERS FROM "file:///Users/markneedham/repos/neo4j-bbc/data/attempts.csv" AS row LOAD CSV WITH HEADERS FROM "file:///Users/markneedham/repos/neo4j-bbc/data/attempts.csv" AS row LOAD CSV WITH HEADERS FROM "file:///Users/markneedham/repos/neo4j-bbc/data/corners.csv" AS row LOAD CSV WITH HEADERS FROM "file:///Users/markneedham/repos/neo4j-bbc/data/corners.csv" AS row LOAD CSV WITH HEADERS FROM "file:///Users/markneedham/repos/neo4j-bbc/data/cards.csv" AS row LOAD CSV WITH HEADERS FROM "file:///Users/markneedham/repos/neo4j-bbc/data/cards.csv" AS row LOAD CSV WITH HEADERS FROM "file:///Users/markneedham/repos/neo4j-bbc/data/subs.csv" AS row ----

sed-using-environment-variables

https://markhneedham.com/blog/2015/08/13/sed-using-environment-variables

2015-01-10 08:39:15

Python: gensim - clang: error: unknown argument: '-mno-fused-madd' [-Wunused-command-line-argument-hard-error-in-future]

While working through https://www.kaggle.com/c/word2vec-nlp-tutorial/details/part-2-word-vectors[part 2 of Kaggle's bag of words tutorial] I needed to install the https://radimrehurek.com/gensim/[gensim] library and initially ran into the following error: [source,text] ---- $ pip install gensim ... cc -fno-strict-aliasing -fno-common -dynamic -arch x86_64 -arch i386 -g -Os -pipe -fno-common -fno-strict-aliasing -fwrapv -mno-fused-madd -DENABLE_DTRACE -DMACOSX -DNDEBUG -Wall -Wstrict-prototypes -Wshorten-64-to-32 -DNDEBUG -g -fwrapv -Os -Wall -Wstrict-prototypes -DENABLE_DTRACE -arch x86_64 -arch i386 -pipe -I/Users/markneedham/projects/neo4j-himym/himym/build/gensim/gensim/models -I/System/Library/Frameworks/Python.framework/Versions/2.7/include/python2.7 -I/Users/markneedham/projects/neo4j-himym/himym/lib/python2.7/site-packages/numpy/core/include -c ./gensim/models/word2vec_inner.c -o build/temp.macosx-10.9-intel-2.7/./gensim/models/word2vec_inner.o clang: error: unknown argument: '-mno-fused-madd' [-Wunused-command-line-argument-hard-error-in-future] clang: note: this will be a hard error (cannot be downgraded to a warning) in the future command 'cc' failed with exit status 1 an integer is required Traceback (most recent call last): File "<string>", line 1, in <module> File "/Users/markneedham/projects/neo4j-himym/himym/build/gensim/setup.py", line 166, in <module> include_package_data=True, File "/System/Library/Frameworks/Python.framework/Versions/2.7/lib/python2.7/distutils/core.py", line 152, in setup dist.run_commands() File "/System/Library/Frameworks/Python.framework/Versions/2.7/lib/python2.7/distutils/dist.py", line 953, in run_commands self.run_command(cmd) File "/System/Library/Frameworks/Python.framework/Versions/2.7/lib/python2.7/distutils/dist.py", line 972, in run_command cmd_obj.run() File "/Users/markneedham/projects/neo4j-himym/himym/lib/python2.7/site-packages/setuptools/command/install.py", line 59, in run return orig.install.run(self) File "/System/Library/Frameworks/Python.framework/Versions/2.7/lib/python2.7/distutils/command/install.py", line 573, in run self.run_command('build') File "/System/Library/Frameworks/Python.framework/Versions/2.7/lib/python2.7/distutils/cmd.py", line 326, in run_command self.distribution.run_command(command) File "/System/Library/Frameworks/Python.framework/Versions/2.7/lib/python2.7/distutils/dist.py", line 972, in run_command cmd_obj.run() File "/System/Library/Frameworks/Python.framework/Versions/2.7/lib/python2.7/distutils/command/build.py", line 127, in run self.run_command(cmd_name) File "/System/Library/Frameworks/Python.framework/Versions/2.7/lib/python2.7/distutils/cmd.py", line 326, in run_command self.distribution.run_command(command) File "/System/Library/Frameworks/Python.framework/Versions/2.7/lib/python2.7/distutils/dist.py", line 972, in run_command cmd_obj.run() File "/Users/markneedham/projects/neo4j-himym/himym/build/gensim/setup.py", line 71, in run "There was an issue with your platform configuration - see above.") TypeError: an integer is required ---------------------------------------- Cleaning up... Command /Users/markneedham/projects/neo4j-himym/himym/bin/python -c "import setuptools, tokenize;__file__='/Users/markneedham/projects/neo4j-himym/himym/build/gensim/setup.py';exec(compile(getattr(tokenize, 'open', open)(__file__).read().replace('\r\n', '\n'), __file__, 'exec'))" install --record /var/folders/sb/6zb6j_7n6bz1jhhplc7c41n00000gn/T/pip-i8aeKR-record/install-record.txt --single-version-externally-managed --compile --install-headers /Users/markneedham/projects/neo4j-himym/himym/include/site/python2.7 failed with error code 1 in /Users/markneedham/projects/neo4j-himym/himym/build/gensim Storing debug log for failure in /Users/markneedham/.pip/pip.log ---- The exception didn't make much sense to me but I came across https://kaspermunck.github.io/2014/03/fixing-clang-error/[a blog post which explained it:</p>] ____ The Apple LLVM compiler in Xcode 5.1 treats unrecognized command-line options as errors. This issue has been seen when building both Python native extensions and Ruby Gems, where some invalid compiler options are currently specified. ____ The author suggests this only became a problem with XCode 5.1 so I'm surprised I hadn't come across it sooner since I haven't upgraded XCode in a long time. We can work around the problem by telling the compiler to treat extra command line arguments as a warning rather than an error ~~~bash export ARCHFLAGS=-Wno-error=unused-command-line-argument-hard-error-in-future ~~~ Now it installs with no problems.

python-gensim-clang-error-unknown-argument-mno-fused-madd-wunused-command-line-argument-hard-error-in-future

https://markhneedham.com/blog/2015/01/10/python-gensim-clang-error-unknown-argument-mno-fused-madd-wunused-command-line-argument-hard-error-in-future

2015-01-10 08:48:04

Python: scikit-learn: ImportError: cannot import name __check_build

In https://www.kaggle.com/c/word2vec-nlp-tutorial/details/part-3-more-fun-with-word-vectors[part 3 of Kaggle's series on text analytics] I needed to install http://scikit-learn.org/stable/[scikit-learn] and having done so ran into the following error when trying to use one of its classes: [source,python] ---- >>> from sklearn.feature_extraction.text import CountVectorizer Traceback (most recent call last): File "<stdin>", line 1, in <module> File "/Users/markneedham/projects/neo4j-himym/himym/lib/python2.7/site-packages/sklearn/__init__.py", line 37, in <module> from . import __check_build ImportError: cannot import name __check_build ---- This error doesn't reveal very much but I found that when I exited the REPL and tried the same command again I got a different error which was a bit more useful: [source,python] ---- >>> from sklearn.feature_extraction.text import CountVectorizer Traceback (most recent call last): File "<stdin>", line 1, in <module> File "/Users/markneedham/projects/neo4j-himym/himym/lib/python2.7/site-packages/sklearn/__init__.py", line 38, in <module> from .base import clone File "/Users/markneedham/projects/neo4j-himym/himym/lib/python2.7/site-packages/sklearn/base.py", line 10, in <module> from scipy import sparse ImportError: No module named scipy ---- The fix for this is now obvious: [source,python] ---- $ pip install scipy ---- And I can now load +++<cite>+++CountVectorizer+++</cite>+++ without any problem: [source,python] ---- $ python Python 2.7.5 (default, Aug 25 2013, 00:04:04) [GCC 4.2.1 Compatible Apple LLVM 5.0 (clang-500.0.68)] on darwin Type "help", "copyright", "credits" or "license" for more information. >>> from sklearn.feature_extraction.text import CountVectorizer ----

python-scikit-learn-importerror-cannot-import-name-__check_build

https://markhneedham.com/blog/2015/01/10/python-scikit-learn-importerror-cannot-import-name-__check_build

2015-01-10 01:22:56

Python NLTK/Neo4j: Analysing the transcripts of How I Met Your Mother

After reading https://twitter.com/emileifrem[Emil's] http://dataconomy.com/discovering-the-power-of-dark-data/[blog post about dark data] a few weeks ago I became intrigued about trying to find some structure in free text data and I thought How I met your mother's transcripts would be a good place to start. I found a website which has the http://transcripts.foreverdreaming.org/viewforum.php?f=177[transcripts for all the episodes] and then having manually downloaded the two pages which listed all the episodes, wrote a script to grab each of the transcripts so I could use them on my machine. I wanted to learn a bit of Python and my colleague https://twitter.com/technige[Nigel] pointed me towards the http://docs.python-requests.org/en/latest/[requests] and http://www.crummy.com/software/BeautifulSoup/bs4/doc/[BeautifulSoup] libraries to help me with my task. The script to grab the transcripts looks like this: [source,python] ---- import requests from bs4 import BeautifulSoup from soupselect import select episodes = {} for i in range(1,3): page = open("data/transcripts/page-" + str(i) + ".html", 'r') soup = BeautifulSoup(page.read()) for row in select(soup, "td.topic-titles a"): parts = row.text.split(" - ") episodes[parts[0]] = {"title": parts[1], "link": row.get("href")} for key, value in episodes.iteritems(): parts = key.split("x") season = int(parts[0]) episode = int(parts[1]) filename = "data/transcripts/S%d-Ep%d" %(season, episode) print filename with open(filename, 'wb') as handle: headers = {'User-Agent': 'Mozilla/5.0 (Macintosh; Intel Mac OS X 10_9_2) AppleWebKit/537.36 (KHTML, like Gecko) Chrome/39.0.2171.95 Safari/537.36'} response = requests.get("http://transcripts.foreverdreaming.org" + value["link"], headers = headers) if response.ok: for block in response.iter_content(1024): if not block: break handle.write(block) ---- _the files containing the lists of episodes are named 'page-1' and 'page-2'_ The code is reasonably simple - we find all the links inside the table, put them in a dictionary and then iterate through the dictionary and download the files to disk. The code to save the file is a bit of a monstrosity but there didn't seem to be a 'save' method that I could use. Having downloaded the files, I thought through all sorts of clever things I could do, including generating a bag of words model for each episode or performing sentiment analysis on each sentence which I'd https://www.kaggle.com/c/word2vec-nlp-tutorial[learnt about from a Kaggle tutorial]. In the end I decided to start simple and extract all the words from the transcripts and count many times a word occurred in a given episode. I ended up with the following script which created a dictionary of (episode \-> words + occurrences): [source,python] ---- import csv import nltk import re from bs4 import BeautifulSoup from soupselect import select from nltk.corpus import stopwords from collections import Counter from nltk.tokenize import word_tokenize def count_words(words): tally=Counter() for elem in words: tally[elem] += 1 return tally episodes_dict = {} with open('data/import/episodes.csv', 'r') as episodes: reader = csv.reader(episodes, delimiter=',') reader.next() for row in reader: print row transcript = open("data/transcripts/S%s-Ep%s" %(row[3], row[1])).read() soup = BeautifulSoup(transcript) rows = select(soup, "table.tablebg tr td.post-body div.postbody") raw_text = rows[0] [ad.extract() for ad in select(raw_text, "div.ads-topic")] [ad.extract() for ad in select(raw_text, "div.t-foot-links")] text = re.sub("[^a-zA-Z]", " ", raw_text.text.strip()) words = [w for w in nltk.word_tokenize(text) if not w.lower() in stopwords.words("english")] episodes_dict[row[0]] = count_words(words) ---- Next I wanted to explore the data a bit to see which words occurred across episodes or which word occurred most frequently and realised that this would be a much easier task if I stored the data somewhere. s/somewhere/in Neo4j Neo4j's query language, Cypher, has a really nice ETL-esque tool called 'LOAD CSV' for loading in CSV files (as the name suggests!) so I added some code to save my words to disk: [source,python] ---- with open("data/import/words.csv", "w") as words: writer = csv.writer(words, delimiter=",") writer.writerow(["EpisodeId", "Word", "Occurrences"]) for episode_id, words in episodes_dict.iteritems(): for word in words: writer.writerow([episode_id, word, words[word]]) ---- This is what the CSV file contents look like: [source,bash] ---- $ head -n 10 data/import/words.csv EpisodeId,Word,Occurrences 165,secondly,1 165,focus,1 165,baby,1 165,spiders,1 165,go,4 165,apartment,1 165,buddy,1 165,Exactly,1 165,young,1 ---- Now we need to write some Cypher to get the data into Neo4j: [source,cypher] ---- // words LOAD CSV WITH HEADERS FROM "file:/Users/markneedham/projects/neo4j-himym/data/import/words.csv" AS row MERGE (word:Word {value: row.Word}) ---- [source,cypher] ---- // episodes LOAD CSV WITH HEADERS FROM "file:/Users/markneedham/projects/neo4j-himym/data/import/words.csv" AS row MERGE (episode:Episode {id: TOINT(row.EpisodeId)}) ---- [source,cypher] ---- // words to episodes LOAD CSV WITH HEADERS FROM "file:/Users/markneedham/projects/neo4j-himym/data/import/words.csv" AS row MATCH (word:Word {value: row.Word}) MATCH (episode:Episode {id: TOINT(row.EpisodeId)}) MERGE (word)-[:USED_IN_EPISODE {times: TOINT(row.Occurrences) }]->(episode); ---- Having done that we can write some simple queries to explore the words used in How I met your mother: [source,cypher] ---- MATCH (word:Word)-[r:USED_IN_EPISODE]->(episode) RETURN word.value, COUNT(episode) AS episodes, SUM(r.times) AS occurrences ORDER BY occurrences DESC LIMIT 10 ==> +-------------------------------------+ ==> | word.value | episodes | occurrences | ==> +-------------------------------------+ ==> | "Ted" | 207 | 11437 | ==> | "Barney" | 208 | 8052 | ==> | "Marshall" | 208 | 7236 | ==> | "Robin" | 205 | 6626 | ==> | "Lily" | 207 | 6330 | ==> | "m" | 208 | 4777 | ==> | "re" | 208 | 4097 | ==> | "know" | 208 | 3489 | ==> | "Oh" | 197 | 3448 | ==> | "like" | 208 | 2498 | ==> +-------------------------------------+ ==> 10 rows ---- The main 5 characters occupy the top 5 positions which is probably what you'd expect. I'm not sure why 'm' and 're' are in the next two position s - I expect that might be scraping gone wrong! Our next query might focus around checking which character is referred to the post in each episode: [source,cypher] ---- WITH ["Ted", "Barney", "Robin", "Lily", "Marshall"] as mainCharacters MATCH (word:Word) WHERE word.value IN mainCharacters MATCH (episode:Episode)<-[r:USED_IN_EPISODE]-(word) WITH episode, word, r ORDER BY episode.id, r.times DESC WITH episode, COLLECT({word: word.value, times: r.times})[0] AS topWord RETURN episode.id, topWord.word AS word, topWord.times AS occurrences LIMIT 10 ==> +---------------------------------------+ ==> | episode.id | word | occurrences | ==> +---------------------------------------+ ==> | 72 | "Barney" | 75 | ==> | 143 | "Ted" | 16 | ==> | 43 | "Lily" | 74 | ==> | 156 | "Ted" | 12 | ==> | 206 | "Barney" | 23 | ==> | 50 | "Marshall" | 51 | ==> | 113 | "Ted" | 76 | ==> | 178 | "Barney" | 21 | ==> | 182 | "Barney" | 22 | ==> | 67 | "Ted" | 84 | ==> +---------------------------------------+ ==> 10 rows ---- If we dig into it further there's actually quite a bit of variety in the number of times the top character in each episode is mentioned which again probably says something about the data: [source,cypher] ---- WITH ["Ted", "Barney", "Robin", "Lily", "Marshall"] as mainCharacters MATCH (word:Word) WHERE word.value IN mainCharacters MATCH (episode:Episode)<-[r:USED_IN_EPISODE]-(word) WITH episode, word, r ORDER BY episode.id, r.times DESC WITH episode, COLLECT({word: word.value, times: r.times})[0] AS topWord RETURN MIN(topWord.times), MAX(topWord.times), AVG(topWord.times), STDEV(topWord.times) ==> +-------------------------------------------------------------------------------------+ ==> | MIN(topWord.times) | MAX(topWord.times) | AVG(topWord.times) | STDEV(topWord.times) | ==> +-------------------------------------------------------------------------------------+ ==> | 3 | 259 | 63.90865384615385 | 42.36255207691068 | ==> +-------------------------------------------------------------------------------------+ ==> 1 row ---- Obviously this is a very simple way of deriving structure from text, here are some of the things I want to try out next: * Detecting common phrases/memes/phrases used in the show (e.g. the yellow umbrella) - this should be possible by creating different length n-grams and then searching for those phrases across the corpus. * Pull out scenes - some of the transcripts use the keyword 'scene' to denote this although some of them don't. Depending how many transcripts contain scene demarkations perhaps we could train a classifier to detect where scenes should be in the transcripts which don't have scenes. * Analyse who talks to each other or who talks about each other most frequently * Create a graph of conversations as my colleagues http://maxdemarzi.com/2012/08/10/summarize-opinions-with-a-graph-part-1/[Max] and http://jexp.github.io/blog/html/simple_nlp_with_graphs.html[Michael] have previously blogged about.

python-nltkneo4j-analysing-the-transcripts-of-how-i-met-your-mother

https://markhneedham.com/blog/2015/01/10/python-nltkneo4j-analysing-the-transcripts-of-how-i-met-your-mother

2015-01-19 00:24:23

Python/NLTK: Finding the most common phrases in How I Met Your Mother

Following on from http://www.markhneedham.com/blog/2015/01/10/python-nltkneo4j-analysing-the-transcripts-of-how-i-met-your-mother/[last week's blog post] where I found the most popular words in How I met your mother transcripts, in this post we'll have a look at how we can pull out sentences and then phrases from our corpus. The first thing I did was tweak the scraping script to pull out the sentences spoken by characters in the transcripts.</p> Each dialogue is separated by two line breaks so we use that as our separator. I also manually skimmed through the transcripts and found out which tags we need to strip out. I ended up with the following: [source,python] ---- import csv import nltk import re import bs4 from bs4 import BeautifulSoup, NavigableString from soupselect import select from nltk.corpus import stopwords from collections import Counter from nltk.tokenize import word_tokenize episodes_dict = {} def strip_tags(soup, invalid_tags): for tag in invalid_tags: for match in soup.findAll(tag): match.replaceWithChildren() return soup def extract_sentences(html): clean = [] brs_in_a_row = 0 temp = "" for item in raw_text.contents: if item.name == "br": brs_in_a_row = brs_in_a_row + 1 else: temp = temp + item if brs_in_a_row == 2: clean.append(temp) temp = "" brs_in_a_row = 0 return clean speakers = [] with open('data/import/episodes.csv', 'r') as episodes_file, \ open("data/import/sentences.csv", 'w') as sentences_file: reader = csv.reader(episodes_file, delimiter=',') reader.next() writer = csv.writer(sentences_file, delimiter=',') writer.writerow(["SentenceId", "EpisodeId", "Season", "Episode", "Sentence"]) sentence_id = 1 for row in reader: transcript = open("data/transcripts/S%s-Ep%s" %(row[3], row[1])).read() soup = BeautifulSoup(transcript) rows = select(soup, "table.tablebg tr td.post-body div.postbody") raw_text = rows[0] [ad.extract() for ad in select(raw_text, "div.ads-topic")] [ad.extract() for ad in select(raw_text, "div.t-foot-links")] [ad.extract() for ad in select(raw_text, "hr")] for tag in ['strong', 'em', "a"]: for match in raw_text.findAll(tag): match.replace_with_children() print row for sentence in [ item.encode("utf-8").strip() for item in extract_sentences(raw_text.contents) ]: writer.writerow([sentence_id, row[0], row[3], row[1], sentence]) sentence_id = sentence_id + 1 ---- Here's a preview of the sentences CSV file: [source,python] ---- $ head -n 10 data/import/sentences.csv SentenceId,EpisodeId,Season,Episode,Sentence 1,1,1,1,Pilot 2,1,1,1,Scene One 3,1,1,1,[Title: The Year 2030] 4,1,1,1,"Narrator: Kids, I'm going to tell you an incredible story. The story of how I met your mother" 5,1,1,1,Son: Are we being punished for something? 6,1,1,1,Narrator: No 7,1,1,1,"Daughter: Yeah, is this going to take a while?" 8,1,1,1,"Narrator: Yes. (Kids are annoyed) Twenty-five years ago, before I was dad, I had this whole other life." 9,1,1,1,"(Music Plays, Title ""How I Met Your Mother"" appears)" ---- The next step is to iterate through each of those sentences and create some http://www.nltk.org/_modules/nltk/util.html[n-grams] to capture the common phrases in the transcripts. ____ In the fields of computational linguistics and probability, an n-gram is a contiguous sequence of n items from a given sequence of text or speech. ____ Python's http://www.nltk.org/[nltk] library has a function that makes this easy e.g. [source,python] ---- >>> import nltk >>> tokens = nltk.word_tokenize("I want to be in an n gram") >>> tokens ['I', 'want', 'to', 'be', 'in', 'an', 'n', 'gram'] >>> nltk.util.ngrams(tokens, 2) [('I', 'want'), ('want', 'to'), ('to', 'be'), ('be', 'in'), ('in', 'an'), ('an', 'n'), ('n', 'gram')] >>> nltk.util.ngrams(tokens, 3) [('I', 'want', 'to'), ('want', 'to', 'be'), ('to', 'be', 'in'), ('be', 'in', 'an'), ('in', 'an', 'n'), ('an', 'n', 'gram')] ---- If we do a similar thing of HIMYM transcripts while stripping out the speaker's name - lines are mostly in the form "Speaker:Sentence" - we end up with the following top phrases: [source,python] ---- import nltk import csv import string import re from collections import Counter non_speaker = re.compile('[A-Za-z]+: (.*)') def extract_phrases(text, phrase_counter, length): for sent in nltk.sent_tokenize(text): strip_speaker = non_speaker.match(sent) if strip_speaker is not None: sent = strip_speaker.group(1) words = nltk.word_tokenize(sent) for phrase in nltk.util.ngrams(words, length): phrase_counter[bphrase] += 1 phrase_counter = Counter() with open("data/import/sentences.csv", "r") as sentencesfile: reader = csv.reader(sentencesfile, delimiter=",") reader.next() for sentence in reader: extract_phrases(sentence[4], phrase_counter, 3) most_common_phrases = phrase_counter.most_common(50) for k,v in most_common_phrases: print '{0: <5}'.format(v), k ---- And if we run that: [source,bash] ---- $ python extract_phrases.py 1123 (',', 'I', "'m") 1099 ('I', 'do', "n't") 1005 (',', 'it', "'s") 535 ('I', 'ca', "n't") 523 ('I', "'m", 'not') 507 ('I', 'mean', ',') 507 (',', 'you', "'re") 459 (',', 'that', "'s") 458 ('2030', ')', ':') 454 ('(', '2030', ')') 453 ('Ted', '(', '2030') 449 ('I', "'m", 'sorry') ... 247 ('I', 'have', 'to') 247 ('No', ',', 'I') 246 ("'s", 'gon', 'na') 241 (',', 'I', "'ll") 229 ('I', "'m", 'going') 226 ('do', "n't", 'want') 226 ('It', "'s", 'not') ---- I noticed that quite a few of the phrases had punctuation in so my next step was to get rid of any of the phrases that had any punctuation in. I updated +++<cite>+++extract_phrases+++</cite>+++ like so: [source,python] ---- def extract_phrases(text, phrase_counter, length): for sent in nltk.sent_tokenize(text): strip_speaker = non_speaker.match(sent) if strip_speaker is not None: sent = strip_speaker.group(1) words = nltk.word_tokenize(sent) for phrase in nltk.util.ngrams(words, length): if all(word not in string.punctuation for word in phrase): phrase_counter[phrase] += 1 ---- Let's run it again: [source,bash] ---- $ python extract_phrases.py 1099 ('I', 'do', "n't") 535 ('I', 'ca', "n't") 523 ('I', "'m", 'not') 449 ('I', "'m", 'sorry') 414 ('do', "n't", 'know') 383 ('Ted', 'from', '2030') 338 ("'m", 'gon', 'na') 334 ('I', "'m", 'gon') 300 ('gon', 'na', 'be') 279 ('END', 'OF', 'FLASHBACK') 267 ("'re", 'gon', 'na') ... 155 ('It', "'s", 'just') 151 ('at', 'the', 'bar') 150 ('a', 'lot', 'of') 147 ("'re", 'going', 'to') 144 ('I', 'have', 'a') 142 ('I', "'m", 'so') 138 ('do', "n't", 'have') 137 ('I', 'think', 'I') 136 ('not', 'gon', 'na') 136 ('I', 'can', 'not') 135 ('and', 'I', "'m") ---- Next I wanted to display each phrase as a string rather than a tuple which was http://stackoverflow.com/questions/21948019/python-untokenize-a-sentence[more difficult than I expected]. I ended up with the following function which almost does the job: [source,python] ---- def untokenize(ngram): tokens = list(ngram) return "".join([" "+i if not i.startswith("'") and \ i not in string.punctuation and \ i != "n't" else i for i in tokens]).strip() ---- I updated +++<cite>+++extract_phrases+++</cite>+++ to use that function: [source,python] ---- def extract_phrases(text, phrase_counter, length): for sent in nltk.sent_tokenize(text): strip_speaker = non_speaker.match(sent) if strip_speaker is not None: sent = strip_speaker.group(1) words = nltk.word_tokenize(sent) for phrase in nltk.util.ngrams(words, length): if all(word not in string.punctuation for word in phrase): phrase_counter[untokenize(phrase)] += 1 ---- Let's go again: [source,bash] ---- $ python extract_phrases.py 1099 I don't 535 I can't 523 I'm not 449 I'm sorry 414 don't know 383 Ted from 2030 338 'm gon na 334 I'm gon 300 gon na be 279 END OF FLASHBACK ... 151 at the bar 150 a lot of 147 're going to 144 I have a 142 I'm so 138 don't have 137 I think I 136 not gon na 136 I can not 135 and I'm ---- These were some of the interesting things that stood out for me and deserve further digging into: * A lot of the most popular phrases begin with 'I' - it would be interesting to filter those sentences to find the general sentiment. * The 'untokenize' function struggles to reconstruct the slang phrase 'gonna' into a single word. * 'Ted from 2030' is actually a speaker which doesn't follow the expected regex pattern and so wasn't filtered out. * 'END OF FLASHBACK' shows quite high up and pulling out those flashbacks would probably be an interesting feature to extract to see which episodes reference each other. * 'Marshall and Lily' and 'Lily and Marshall' show up on the list - it would be interesting to explore the frequency of pairs of other characters. The https://github.com/mneedham/neo4j-himym[code is all on github] if you want to play with it.

pythonnltk-finding-the-most-common-phrases-in-how-i-met-your-mother

https://markhneedham.com/blog/2015/01/19/pythonnltk-finding-the-most-common-phrases-in-how-i-met-your-mother

2015-01-30 00:27:53

R: ggplot2 - Each group consist of only one observation. Do you need to adjust the group aesthetic?

I've been playing around with some weather data over the last couple of days which I aggregated down to the average temperature per month over the last 4 years and stored in a CSV file. This is what the file looks like: [source,bash] ---- $ cat /tmp/averageTemperatureByMonth.csv "month","aveTemperature" "January",6.02684563758389 "February",5.89380530973451 "March",7.54838709677419 "April",10.875 "May",13.3064516129032 "June",15.9666666666667 "July",18.8387096774194 "August",18.3709677419355 "September",16.2583333333333 "October",13.4596774193548 "November",9.19166666666667 "December",7.01612903225806 ---- I wanted to create a simple line chart which would show the months of the year in ascending order with the appropriate temperature. My first attempt was the following: [source,rstats] ---- df = read.csv("/tmp/averageTemperatureByMonth.csv") df$month = factor(df$month, month.name) ggplot(aes(x = month, y = aveTemperature), data = df) + geom_line( ) + ggtitle("Temperature by month") ---- which resulted in the following error: [source,text] ---- geom_path: Each group consist of only one observation. Do you need to adjust the group aesthetic? ---- My understanding is that the points don't get joined up by default because the variable on the x axis is not a continuous one but rather a factor variable. One way to work around this problem is to make it numeric, like so: [source,rstats] ---- ggplot(aes(x = as.numeric(month), y = aveTemperature), data = df) + geom_line( ) + ggtitle("Temperature by month") ---- which results in the following chart: image::{{<siteurl>}}/uploads/2015/01/2015-01-30_00-25-18.png[2015 01 30 00 25 18,600] This isn't bad but it'd be much nicer if we could have the month names along the bottom instead. It turns out we can but we need to https://kohske.wordpress.com/2010/12/27/faq-geom_line-doesnt-draw-lines/[specify a group] that each point belongs to. ggplot will then connects points which belong to the same group. In this case we don't really have one so we'll define a dummy one instead: [source,rstats] ---- ggplot(aes(x = month, y = aveTemperature, group=1), data = df) + geom_line( ) + ggtitle("Temperature by month") ---- And now we get the visualisation we want: image::{{<siteurl>}}/uploads/2015/01/2015-01-29_23-28-23.png[2015 01 29 23 28 23,600]

r-ggplot2-each-group-consist-of-only-one-observation-do-you-need-to-adjust-the-group-aesthetic

https://markhneedham.com/blog/2015/01/30/r-ggplot2-each-group-consist-of-only-one-observation-do-you-need-to-adjust-the-group-aesthetic

2015-01-30 21:29:00

Python/matpotlib: Plotting occurrences of the main characters in How I Met Your Mother

Normally when I'm playing around with data sets in R I get out ggplot2 to plot some charts to get a feel for the data but having spent quite a bit of time with Python and How I met your mother transcripts I haven't created a single plot. I thought I'd better change change that. After a bit of searching around it seems that http://matplotlib.org/[matplotlib] is the go to library for this job and I thought an interesting thing to plot would be how often each of the main characters appear in each episode across the show. I've already got all the https://github.com/mneedham/neo4j-himym/blob/master/data/import/sentences.csv[sentences] from each episode as well as the https://github.com/mneedham/neo4j-himym/blob/master/data/import/episodes.csv[list of episodes] pulled out into CSV files so we can start from there. This is a sample of the sentences file: [source,bash] ---- $ head -n 10 data/import/sentences.csv SentenceId,EpisodeId,Season,Episode,Sentence 1,1,1,1,Pilot 2,1,1,1,Scene One 3,1,1,1,[Title: The Year 2030] 4,1,1,1,"Narrator: Kids, I'm going to tell you an incredible story. The story of how I met your mother" 5,1,1,1,Son: Are we being punished for something? 6,1,1,1,Narrator: No 7,1,1,1,"Daughter: Yeah, is this going to take a while?" 8,1,1,1,"Narrator: Yes. (Kids are annoyed) Twenty-five years ago, before I was dad, I had this whole other life." 9,1,1,1,"(Music Plays, Title ""How I Met Your Mother"" appears)" ---- My first step was to transform the CSV file into an array of words grouped by episode. I created a dictionary, iterated over the CSV file and then used nltk's word tokeniser to pull out words from sentences: [source,python] ---- import csv from collections import defaultdict episodes = defaultdict(list) with open("data/import/sentences.csv", "r") as sentencesfile: reader = csv.reader(sentencesfile, delimiter = ",") reader.next() for row in reader: episodes[row[1]].append([ word for word in nltk.word_tokenize(row[4].lower())] ) ---- Let's have a quick look what's in our dictionary: [source,python] ---- >>> episodes.keys()[:10] ['165', '133', '132', '131', '130', '137', '136', '135', '134', '139'] ---- We've got some episode numbers as we'd expect. Now let's have a look at some of the words for one of the episodes: [source,python] ---- >>> episodes["165"][5] ['\xe2\x99\xaa', 'how', 'i', 'met', 'your', 'mother', '8x05', '\xe2\x99\xaa'] ---- So we've got an list of lists of words for each episode but https://radimrehurek.com/gensim/[gensim] (which I wanted to play around with) requires a single array of words per document. I transformed the data into the appropriate format and fed it into a gensim Dictionary: [source,python] ---- from gensim import corpora texts = [] for id, episode in episodes.iteritems(): texts.append([item for sublist in episode for item in sublist]) dictionary = corpora.Dictionary(texts) ---- If we peek into 'texts' we can see that the list has been flattened: [source,python] ---- >>> texts[0][10:20] ['a', 'bit', 'of', 'a', 'dog', ',', 'and', 'even', 'though', 'he'] ---- We'll now convert our dictionary of words into a sparse vector which contains pairs of word ids and the number of time they occur: [source,python] ---- corpus = [dictionary.doc2bow(text) for text in texts] ---- Let's try and find out how many times the word 'ted' occurs in our corpus. First we need to find out the word id for 'ted': [source,python] ---- >>> dictionary.token2id["ted"] 551 ---- I don't know how to look up the word id directly from the corpus but you can get back an individual document (episode) and its words quite easily: [source,python] ---- >>> corpus[0][:5] [(0, 8), (1, 1), (2, 2), (3, 13), (4, 20)] ---- We can then convert that into a dictionary and look up our word: [source,python] ---- >>> dict(corpus[0]).get(551) 16 ---- So 'ted' occurs 16 times in the first episode. If we generify that code we end up with the following: [source,python] ---- words = ["ted", "robin", "barney", "lily", "marshall"] words_dict = dict() for word in words: word_id = dictionary.token2id[word] counts = [] for episode in corpus: count = dict(episode).get(word_id) or 0 counts.append(count) words_dict[word] = counts ---- There's quite a lot of counts in there so let's just preview the first 5 episodes: [source,python] ---- >>> for word, counts in words_dict.iteritems(): print word, counts[:5] lily [3, 20, 47, 26, 41] marshall [8, 25, 63, 27, 34] barney [9, 94, 58, 92, 102] ted [16, 46, 66, 32, 44] robin [18, 43, 25, 24, 34] ---- Now it's time to bring out matplotlib and make this visual! I initially put all the characters on one chart but it looks very messy and there's a lot of overlap so I decided on separate charts. The only thing I had to do to achieve this was call +++<cite>+++plt.figure()+++</cite>+++ at the beginning of the loop to create a new plot: [source,python] ---- import matplotlib matplotlib.use('TkAgg') import matplotlib.pyplot as plt pylab.show() for word, counts in words_dict.iteritems(): plt.figure() plt.plot(counts) plt.legend([word], loc='upper left') plt.ylabel('occurrences') plt.xlabel('episode') plt.xlim(0, 208) plt.savefig('images/%s.png' % (word), dpi=200) ---- This generates plots like this: image::{{<siteurl>}}/uploads/2015/01/2015-01-30_21-15-03.png[2015 01 30 21 15 03,600] This is good but I thought it'd be interesting to put in the season demarcations to see if that could give any more insight. We can call the function +++<cite>+++plt.axvline+++</cite>+++ and pass in the appropriate episode number to achieve this effect but I needed to know the episode ID for the last episode in each season which required a bit of code: [source,python] ---- import pandas as pd df = pd.read_csv('data/import/episodes.csv', index_col=False, header=0) last_episode_in_season = list(df.groupby("Season").max()["NumberOverall"]) >>> last_episode_in_season [22, 44, 64, 88, 112, 136, 160, 184, 208] ---- Now let's plug that into matplotlib: [source,python] ---- for word, counts in words_dict.iteritems(): plt.figure() plt.plot(counts) for episode in last_episode_in_season: plt.axvline(x=episode, color = "red") plt.legend([word], loc='upper left') plt.ylabel('occurrences') plt.xlabel('episode') plt.xlim(0, 208) plt.savefig('images/%s.png' % (word), dpi=200) ---- image::{{<siteurl>}}/uploads/2015/01/2015-01-30_21-10-52.png[2015 01 30 21 10 52,569] The last thing I wanted to do is get all the plots on the same scale for which I needed to get the maximum number of occurrences of any character in any episode. It was easier than I expected: [source,python] ---- >>> y_max = max([max(count) for count in words_dict.values()]) >>> y_max 260 ---- And now let's plot again: [source,python] ---- for word, counts in words_dict.iteritems(): plt.figure() plt.plot(counts) for episode in last_episode_in_season: plt.axvline(x=episode, color = "red") plt.legend([word], loc='upper left') plt.ylabel('occurrences') plt.xlabel('episode') plt.xlim(0, 208) plt.ylim(0, y_max) plt.savefig('images/%s.png' % (word), dpi=200) ---- Our charts are now easy to compare: image::{{<siteurl>}}/uploads/2015/01/2015-01-30_21-23-48.png[2015 01 30 21 23 48,575] image::{{<siteurl>}}/uploads/2015/01/2015-01-30_21-24-03.png[2015 01 30 21 24 03,551] For some reason there's a big spike of the word 'ted' in the middle of the 7th season - I'm clearly not a big enough fan to know why that is but it's a spike of 30% over the next highest value. The data isn't perfect - some of the episodes list the speaker of the sentence and some don't so it may be that the spikes indicate that rather than anything else. I find it's always nice to do a bit of visual exploration of the data anyway and now I know it's possible to do so pretty easily in Python land.

pythonmatpotlib-plotting-occurrences-of-the-main-characters-in-how-i-met-your-mother

https://markhneedham.com/blog/2015/01/30/pythonmatpotlib-plotting-occurrences-of-the-main-characters-in-how-i-met-your-mother

2015-01-12 23:16:58

Python: Counter - ValueError: too many values to unpack

I recently came across https://docs.python.org/2/library/collections.html#collections.Counter[Python's Counter tool] which makes it really easy to count the number of occurrences of items in a list. In my case I was trying to work out how many times words occurred in a corpus so I had something like the following: [source,python] ---- >> from collections import Counter >> counter = Counter(["word1", "word2", "word3", "word1"]) >> print counter Counter({'word1': 2, 'word3': 1, 'word2': 1}) ---- I wanted to write a for loop to iterate over the counter and print the (key, value) pairs and started with the following: [source,python] ---- >>> for key, value in counter: ... print key, value ... Traceback (most recent call last): File "<stdin>", line 1, in <module> ValueError: too many values to unpack ---- I'm not sure why I expected this to work but in fact since Counter is a sub class of dict we need to call https://docs.python.org/2/library/stdtypes.html#dict.iteritems[iteritems] to get an iterator of pairs rather than just keys. The following does the job: [source,python] ---- >>> for key, value in counter.iteritems(): ... print key, value ... word1 2 word3 1 word2 1 ---- Hopefully future Mark will remember this!

python-counter-valueerror-too-many-values-to-unpack

https://markhneedham.com/blog/2015/01/12/python-counter-valueerror-too-many-values-to-unpack

2015-01-22 00:25:24

Python/pdfquery: Scraping the FIFA World Player of the Year votes PDF into shape

Last week the http://www.fifa.com/ballon-dor/[FIFA Ballon d'Or 2014] was announced and along with the announcement of the winner the http://es.fifa.com/mm/document/ballon-dor/playeroftheyear-men/02/50/58/45/fboaward_menplayer2014_neutral.pdf[individual votes were also made available]. Unfortunately they weren't made open in a way that Ben Wellington (of http://iquantny.tumblr.com/[IQuantNY] fame) http://iquantny.tumblr.com/post/108236949969/why-open-data-is-still-too-closed-my-tedxnewyork?utm_content=buffera7566&utm_medium=social&utm_source=twitter.com&utm_campaign=buffer[would approve of] - the choice of format for the data is a PDF file! I wanted to extract this data to play around with it but I wanted to automate the extraction as I'd done when http://www.markhneedham.com/blog/2014/12/09/r-cleaning-up-plotting-google-trends-data/[working with Google Trends data]. I had a quick look for PDF scraping libraries in Python and R and eventually settled on Python's https://pypi.python.org/pypi/pdfquery[pdfquery], mainly because there was lots of documentation which made it easy to get started. One way you scrape data from a PDF is by locating an element on the page and then grabbing everything within a bounded box relative to that element. In my case I had 17 pages all of which had a heading for each of six columns. image::{{<siteurl>}}/uploads/2015/01/2015-01-22_00-08-18.png[2015 01 22 00 08 18,600] I wanted to grab the data in each of those columns but initially struggled working out what elements I should be looking for until I came across the following function which allows you to dump an XML version of the PDF to disk: [source,python] ---- import pdfquery pdf = pdfquery.PDFQuery("fboaward_menplayer2014_neutral.pdf") pdf.load() pdf.tree.write("/tmp/yadda", pretty_print=True) ---- The output looks like this: [source,bash] ---- $ head -n 10 /tmp/yadda <pdfxml ModDate="D:20150110224554+01'00'" CreationDate="D:20150110224539+01'00'" Producer="Microsoft&#174; Excel&#174; 2010" Creator="Microsoft&#174; Excel&#174; 2010"> <LTPage bbox="[0, 0, 841.8, 595.2]" height="595.2" pageid="1" rotate="0" width="841.8" x0="0" x1="841.8" y0="0" y1="595.2" page_index="0" page_label=""> <LTAnon> </LTAnon> <LTTextLineHorizontal bbox="[31.08, 546.15, 122.524, 556.59]" height="10.44" width="91.444" word_margin="0.1" x0="31.08" x1="122.524" y0="546.15" y1="556.59"><LTTextBoxHorizontal bbox="[31.08, 546.15, 122.524, 556.59]" height="10.44" index="0" width="91.444" x0="31.08" x1="122.524" y0="546.15" y1="556.59">FIFA Ballon d'Or 2014 </LTTextBoxHorizontal></LTTextLineHorizontal> <LTAnon> </LTAnon> <LTAnon> </LTAnon> <LTAnon> </LTAnon> <LTAnon> </LTAnon> <LTAnon> </LTAnon> <LTAnon> </LTAnon> ---- Having scanned through the file I realised that what I needed to do was locate the 'LTTextLineHorizontal' element for each heading and then grab all the 'LTTextLineHorizontal' elements that appeared in that column. I started out by trying to grab the 'Name' column on the first page: [source,python] ---- >>> name_element = pdf.pq('LTPage[pageid=\'1\'] LTTextLineHorizontal:contains("Name")')[0] >>> name_element.text 'Name ' ---- Next I needed to get the other elements in that column. With a bit of trial and error I ended up with the following code: [source,python] ---- x = float(name_element.get('x0')) y = float(name_element.get('y0')) cells = pdf.extract( [ ('with_parent','LTPage[pageid=\'1\']'), ('cells', 'LTTextLineHorizontal:in_bbox("%s,%s,%s,%s")' % (x, y-500, x+150, y)) ]) >>> [cell.text.encode('utf-8').strip() for cell in cells['cells']] ['Amiri Islam', 'Cana Lorik', 'Bougherra Madjid', 'Luvu Rafe Talalelei', 'Sonejee Masand Oscar', 'Amaral Felisberto', 'Liddie Ryan', 'Griffith Quinton', 'Messi Lionel', 'Berezovskiy Roman', 'Breinburg Reinhard', 'Jedinak Mile', 'Fuchs Christian', 'Sadigov Rashad', 'Gavin Christie', 'Hasan Mohamed', 'Mamun Md Mamnul Islam', 'Burgess Romelle', 'Kalachou Tsimafei', 'Komany Vincent', 'Eiley Dalton', 'Nusum John', 'Tshering Passang', 'Raldes Ronald', 'D\xc5\xbeeko Edin', 'Da Silva Santos Junior Neymar', 'Ceasar Troy', 'Popov Ivelin', 'Kabore Charles', 'Ntibazonkiza Saidi', 'Kouch Sokumpheak'] ---- I cleaned that up and generified it to work for any page and for columns of different widths. This is what the function looks like: [source,python] ---- def extract_cells(page, header, cell_width): name_element = pdf.pq('LTPage[pageid=\'%s\'] LTTextLineHorizontal:contains("%s")' % (page, header))[0] x = float(name_element.get('x0')) y = float(name_element.get('y0')) cells = pdf.extract( [ ('with_parent','LTPage[pageid=\'%s\']' %(page)), ('cells', 'LTTextLineHorizontal:in_bbox("%s,%s,%s,%s")' % (x, y-500, x+cell_width, y)) ]) return [cell.text.encode('utf-8').strip() for cell in cells['cells']] ---- We can then call that for each column on the page and zip together the resulting arrays to get a tuple for each row: [source,python] ---- roles = extract_cells(1, "Vote", 50) countries = extract_cells(1, "Country", 150) voters = extract_cells(1, "Name", 170) first = extract_cells(1, "First (5 points)", 150) second = extract_cells(1, "Second (3 points)", 150) third = extract_cells(1, "Third (1 point)", 130) >>> for vote in zip(roles, countries, voters, first, second, third)[:5]: print vote ('Captain', 'Afghanistan', 'Amiri Islam', 'Messi Lionel', 'Cristiano Ronaldo', 'Ibrahimovic Zlatan') ('Captain', 'Albania', 'Cana Lorik', 'Cristiano Ronaldo', 'Robben Arjen', 'Mueller Thomas') ('Captain', 'Algeria', 'Bougherra Madjid', 'Cristiano Ronaldo', 'Robben Arjen', 'Benzema Karim') ('Captain', 'American Samoa', 'Luvu Rafe Talalelei', 'Neymar', 'Robben Arjen', 'Cristiano Ronaldo') ('Captain', 'Andorra', 'Sonejee Masand Oscar', 'Cristiano Ronaldo', 'Mueller Thomas', 'Kroos Toni') ---- The next step was to write out each of those rows to a CSV file so we can use it from another program. The full script looks like this: [source,python] ---- import pdfquery import csv def extract_cells(page, header, cell_width): name_element = pdf.pq('LTPage[pageid=\'%s\'] LTTextLineHorizontal:contains("%s")' % (page, header))[0] x = float(name_element.get('x0')) y = float(name_element.get('y0')) cells = pdf.extract( [ ('with_parent','LTPage[pageid=\'%s\']' %(page)), ('cells', 'LTTextLineHorizontal:in_bbox("%s,%s,%s,%s")' % (x, y-500, x+cell_width, y)) ]) return [cell.text.encode('utf-8').strip() for cell in cells['cells']] if __name__ == "__main__": pdf = pdfquery.PDFQuery("fboaward_menplayer2014_neutral.pdf") pdf.load() pdf.tree.write("/tmp/yadda", pretty_print=True) pages_in_pdf = len(pdf.pq('LTPage')) with open('votes.csv', 'w') as votesfile: writer = csv.writer(votesfile, delimiter=",") writer.writerow(["Role", "Country", "Voter", "FirstPlace", "SecondPlace", "ThirdPlace"]) for page in range(1, pages_in_pdf + 1): print page roles = extract_cells(page, "Vote", 50) countries = extract_cells(page, "Country", 150) voters = extract_cells(page, "Name", 170) first = extract_cells(page, "First (5 points)", 150) second = extract_cells(page, "Second (3 points)", 150) third = extract_cells(page, "Third (1 point)", 130) votes = zip(roles, countries, voters, first, second, third) print votes for vote in votes: writer.writerow(list(vote)) ---- The https://github.com/mneedham/fifa[code is on github] if you want to play around with it or if you just want to grab the https://github.com/mneedham/fifa/blob/master/votes.csv[votes data] that's there too.

pythonpdfquery-scraping-the-fifa-world-player-of-the-year-votes-pdf-into-shape

https://markhneedham.com/blog/2015/01/22/pythonpdfquery-scraping-the-fifa-world-player-of-the-year-votes-pdf-into-shape

2015-01-25 12:47:01

Python: Find the highest value in a group

In my continued playing around with a How I met your mother data set I needed to find out the last episode that happened in a season so that I could use it in a chart I wanted to plot. I had this CSV file containing each of the episodes: [source,bash] ---- $ head -n 10 data/import/episodes.csv NumberOverall,NumberInSeason,Episode,Season,DateAired,Timestamp 1,1,/wiki/Pilot,1,"September 19, 2005",1127084400 2,2,/wiki/Purple_Giraffe,1,"September 26, 2005",1127689200 3,3,/wiki/Sweet_Taste_of_Liberty,1,"October 3, 2005",1128294000 4,4,/wiki/Return_of_the_Shirt,1,"October 10, 2005",1128898800 5,5,/wiki/Okay_Awesome,1,"October 17, 2005",1129503600 6,6,/wiki/Slutty_Pumpkin,1,"October 24, 2005",1130108400 7,7,/wiki/Matchmaker,1,"November 7, 2005",1131321600 8,8,/wiki/The_Duel,1,"November 14, 2005",1131926400 9,9,/wiki/Belly_Full_of_Turkey,1,"November 21, 2005",1132531200 ---- I started out by parsing the CSV file into a dictionary of (seasons \-> episode ids): [source,python] ---- import csv from collections import defaultdict seasons = defaultdict(list) with open("data/import/episodes.csv", "r") as episodesfile: reader = csv.reader(episodesfile, delimiter = ",") reader.next() for row in reader: seasons[int(row[3])].append(int(row[0])) print seasons ---- which outputs the following: [source,bash] ---- $ python blog.py defaultdict(<type 'list'>, { 1: [1, 2, 3, 4, 5, 6, 7, 8, 9, 10, 11, 12, 13, 14, 15, 16, 17, 18, 19, 20, 21, 22], 2: [23, 24, 25, 26, 27, 28, 29, 30, 31, 32, 33, 34, 35, 36, 37, 38, 39, 40, 41, 42, 43, 44], 3: [45, 46, 47, 48, 49, 50, 51, 52, 53, 54, 55, 56, 57, 58, 59, 60, 61, 62, 63, 64], 4: [65, 66, 67, 68, 69, 70, 71, 72, 73, 74, 75, 76, 77, 78, 79, 80, 81, 82, 83, 84, 85, 86, 87, 88], 5: [89, 90, 91, 92, 93, 94, 95, 96, 97, 98, 99, 100, 101, 102, 103, 104, 105, 106, 107, 108, 109, 110, 111, 112], 6: [113, 114, 115, 116, 117, 118, 119, 120, 121, 122, 123, 124, 125, 126, 127, 128, 129, 130, 131, 132, 133, 134, 135, 136], 7: [137, 138, 139, 140, 141, 142, 143, 144, 145, 146, 147, 148, 149, 150, 151, 152, 153, 154, 155, 156, 157, 158, 159, 160], 8: [161, 162, 163, 164, 165, 166, 167, 168, 169, 170, 171, 172, 173, 174, 175, 176, 177, 178, 179, 180, 181, 182, 183, 184], 9: [185, 186, 187, 188, 189, 190, 191, 192, 193, 194, 195, 196, 197, 198, 199, 200, 201, 202, 203, 204, 205, 206, 207, 208]}) ---- It's reasonably easy to transform that into a dictionary of (season \-> max episode id) with the following couple of lines: [source,python] ---- for season, episode_ids in seasons.iteritems(): seasons[season] = max(episode_ids) >>> print seasons defaultdict(<type 'list'>, {1: 22, 2: 44, 3: 64, 4: 88, 5: 112, 6: 136, 7: 160, 8: 184, 9: 208}) ---- This works fine but it felt very much like a https://github.com/hadley/dplyr[dplyr] problem to me so I wanted to see whether I could write something cleaner using http://pandas.pydata.org/[pandas]. I started out by capturing the seasons and episode ids in separate lists and then building up a DataFrame: [source,python] ---- import pandas as pd from pandas import DataFrame seasons, episode_ids = [], [] with open("data/import/episodes.csv", "r") as episodesfile: reader = csv.reader(episodesfile, delimiter = ",") reader.next() for row in reader: seasons.append(int(row[3])) episode_ids.append(int(row[0])) df = DataFrame.from_items([('Season', seasons), ('EpisodeId', episode_ids)]) >>> print df.groupby("Season").max()["EpisodeId"] Season 1 22 2 44 3 64 4 88 5 112 6 136 7 160 8 184 9 208 ---- Or we can simplify that and read the CSV file directly into a DataFrame: [source,python] ---- df = pd.read_csv('data/import/episodes.csv', index_col=False, header=0) >>> print df.groupby("Season").max()["NumberOverall"] Season 1 22 2 44 3 64 4 88 5 112 6 136 7 160 8 184 9 208 ---- Pretty neat. I need to get more into pandas.

python-find-the-highest-value-in-a-group

https://markhneedham.com/blog/2015/01/25/python-find-the-highest-value-in-a-group

2015-06-03 05:52:08

R: ggplot geom_density - Error in exists(name, envir = env, mode = mode) : argument "env" is missing, with no default

Continuing on from yesterday's blog post where I http://www.markhneedham.com/blog/2015/06/02/r-dplyr-removing-empty-rows/[worked out how to clean up the Think Bayes Price is Right data set], the next task was to plot a distribution of the prices of show case items. To recap, this is what the data frame we're working with looks like: [source,R] ---- library(dplyr) df2011 = read.csv("~/projects/rLearning/showcases.2011.csv", na.strings = c("", "NA")) df2011 = df2011 %>% na.omit() > df2011 %>% head() X Sep..19 Sep..20 Sep..21 Sep..22 Sep..23 Sep..26 Sep..27 Sep..28 Sep..29 Sep..30 Oct..3 3 Showcase 1 50969 21901 32815 44432 24273 30554 20963 28941 25851 28800 37703 4 Showcase 2 45429 34061 53186 31428 22320 24337 41373 45437 41125 36319 38752 6 Bid 1 42000 14000 32000 27000 18750 27222 25000 35000 22500 21300 21567 7 Bid 2 34000 59900 45000 38000 23000 18525 32000 45000 32000 27500 23800 9 Difference 1 8969 7901 815 17432 5523 3332 -4037 -6059 3351 7500 16136 10 Difference 2 11429 -25839 8186 -6572 -680 5812 9373 437 9125 8819 14952 ... ---- So our goal is to plot the density of the 'Showcase 1' items. Unfortunately those aren't currently stored in a way that makes this easy for us. We need to flip the data frame so that we have a row for each date/price type/price: [source,text] ---- PriceType Date Price Showcase 1 Sep..19 50969 Showcase 2 Sep..19 21901 ... Showcase 1 Sep..20 45429 Showcase 2 Sep..20 34061 ---- The http://had.co.nz/reshape/introduction.pdf[reshape] library's +++<cite>+++melt+++</cite>+++ function is our friend here: [source,R] ---- library(reshape) meltedDf = melt(df2011, id=c("X")) > meltedDf %>% sample_n(10) X variable value 643 Showcase 1 Feb..24 27883 224 Showcase 2 Nov..10 34089 1062 Difference 2 Jun..4 9962 770 Showcase 2 Mar..28 39620 150 Difference 2 Oct..24 9137 431 Difference 1 Jan..4 7516 345 Bid 1 Dec..12 21569 918 Difference 2 May.1 -2093 536 Showcase 2 Jan..31 30918 502 Bid 2 Jan..23 27000 ---- Now we need to plug this into ggplot. We'll start by just plotting all the prices for showcase 1: [source,r] ---- > ggplot(aes(x = value), data = meltedDf %>% filter(X == "Showcase 1")) + geom_density() Error in exists(name, envir = env, mode = mode) : argument "env" is missing, with no default ---- This error usually means that you've http://stackoverflow.com/questions/24708188/r-error-in-qplot-from-ggplot2-argument-env-is-missing-with-no-default[passed an empty data set] to ggplot which isn't the case here, but if we extract the values column we can see the problem: [source,r] ---- > meltedDf$value[1:10] [1] "50969" "45429" "42000" "34000" "8969" "11429" "21901" "34061" "14000" "59900" ---- They are all strings! Making it very difficult to plot a density curve which relies on the data being continuous. Let's fix that and try again: [source,r] ---- meltedDf$value = as.numeric(meltedDf$value) ggplot(aes(x = value), data = meltedDf %>% filter(X == "Showcase 1")) + geom_density() ---- image::{{<siteurl>}}/uploads/2015/06/2015-06-03_06-46-48.png[2015 06 03 06 46 48,400] If we want to show the curves for both showcases we can tweak our code slightly: [source,R] ---- ggplot(meltedDf %>% filter(grepl("Showcase", X)), aes(x = value, colour = X)) + geom_density() + theme(legend.position="top") ---- image::{{<siteurl>}}/uploads/2015/06/2015-06-03_06-50-35.png[2015 06 03 06 50 35,400] Et voila!

r-ggplot-geom_density-error-in-existsname-envir-env-mode-mode-argument-env-is-missing-with-no-default

https://markhneedham.com/blog/2015/06/03/r-ggplot-geom_density-error-in-existsname-envir-env-mode-mode-argument-env-is-missing-with-no-default

2015-06-04 22:17:34

Neo4j: Cypher - Step by step to creating a linked list of adjacent nodes using UNWIND

In late 2013 I wrote a post showing http://www.markhneedham.com/blog/2013/11/22/neo4j-cypher-creating-relationships-between-nodes-from-adjacent-rows-in-a-query/[how to create a linked list connecting different football seasons together] using Neo4j's Cypher query language, a post I've frequently copy & pasted from! Now 18 months later, and using Neo4j 2.2 rather than 2.0, we can actually solve this problem in what I believe is a more intuitive way using the +++<cite>+++http://neo4j.com/docs/stable/query-unwind.html[UNWIND]+++</cite>+++ function. Credit for the idea goes to https://twitter.com/mesirii[Michael], I'm just the messenger. To recap, we had a collection of football seasons and we wanted to connect adjacent seasons to each other to allow easy querying between seasons. The following is the code we used: [source,cypher] ---- CREATE (:Season {name: "2013/2014", timestamp: 1375315200}) CREATE (:Season {name: "2012/2013", timestamp: 1343779200}) CREATE (:Season {name: "2011/2012", timestamp: 1312156800}) CREATE (:Season {name: "2010/2011", timestamp: 1280620800}) CREATE (:Season {name: "2009/2010", timestamp: 1249084800}) ---- [source,cypher] ---- MATCH (s:Season) WITH s ORDER BY s.timestamp WITH COLLECT(s) AS seasons FOREACH(i in RANGE(0, length(seasons)-2) | FOREACH(si in [seasons[i]] | FOREACH(si2 in [seasons[i+1]] | MERGE (si)-[:NEXT]->(si2)))) ---- Our goal is to replace those 3 +++<cite>+++FOREACH+++</cite>+++ loops with something a bit easier to understand. To start with, let's run the first part of the query to get some intuition of what we're trying to do: [source,cypher] ---- MATCH (s:Season) WITH s ORDER BY s.timestamp RETURN COLLECT(s) AS seasons ==> +-------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------+ ==> | seasons | ==> +-------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------+ ==> | [Node[1973]{timestamp:1249084800,name:"2009/2010"},Node[1972]{timestamp:1280620800,name:"2010/2011"},Node[1971]{timestamp:1312156800,name:"2011/2012"},Node[1970]{timestamp:1343779200,name:"2012/2013"},Node[1969]{timestamp:1375315200,name:"2013/2014"}] | ==> +-------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------+ ---- So at this point we've got all the seasons in an array going from 2009/2010 up to 2013/2014. We want to create a 'NEXT' relationship between 2009/2010 \-> 2010/2011, 2010/2011 \-> 2011/2012 and so on. To achieve this we need to get the adjacent seasons split into two columns, like so: [source,text] ---- 2009/2010 2010/2011 2010/2011 2011/2012 2011/2012 2012/2013 2012/2013 2013/2014 ---- If we can get the data into that format then we can apply a +++<cite>+++MERGE+++</cite>+++ between the two fields to create the 'NEXT' relationship. So how do we do that? If we were in Python we'd be calling for the zip function which we could apply like this: [source,python] ---- >>> seasons = ["2009/2010", "2010/2011", "2011/2012", "2012/2013", "2013/2014"] >>> zip(seasons, seasons[1:]) [('2009/2010', '2010/2011'), ('2010/2011', '2011/2012'), ('2011/2012', '2012/2013'), ('2012/2013', '2013/2014')] ---- Unfortunately we don't have an equivalent function in Cypher but we can achieve the same outcome by creating 2 columns with adjacent integer values. The +++<cite>+++RANGE+++</cite>+++ and +++<cite>+++UNWIND+++</cite>+++ functions are our friends here: [source,cypher] ---- return RANGE(0,4) ==> +-------------+ ==> | RANGE(0,4) | ==> +-------------+ ==> | [0,1,2,3,4] | ==> +-------------+ ---- [source,cypher] ---- UNWIND RANGE(0,4) as idx RETURN idx, idx +1; ==> +--------------+ ==> | idx | idx +1 | ==> +--------------+ ==> | 0 | 1 | ==> | 1 | 2 | ==> | 2 | 3 | ==> | 3 | 4 | ==> | 4 | 5 | ==> +--------------+ ==> 5 rows ---- Now all we need to do is plug this code into our original query where 'idx' and 'idx + 1' represent indexes into the array of seasons. We use a range which stops 1 element early since there isn't anywhere to connect our last season to: [source,cypher] ---- MATCH (s:Season) WITH s ORDER BY s.timestamp WITH COLLECT(s) AS seasons UNWIND RANGE(0,LENGTH(seasons) - 2) as idx RETURN seasons[idx], seasons[idx+1] ==> +-------------------------------------------------------------------------------------------------------+ ==> | seasons[idx] | seasons[idx+1] | ==> +-------------------------------------------------------------------------------------------------------+ ==> | Node[1973]{timestamp:1249084800,name:"2009/2010"} | Node[1972]{timestamp:1280620800,name:"2010/2011"} | ==> | Node[1972]{timestamp:1280620800,name:"2010/2011"} | Node[1971]{timestamp:1312156800,name:"2011/2012"} | ==> | Node[1971]{timestamp:1312156800,name:"2011/2012"} | Node[1970]{timestamp:1343779200,name:"2012/2013"} | ==> | Node[1970]{timestamp:1343779200,name:"2012/2013"} | Node[1969]{timestamp:1375315200,name:"2013/2014"} | ==> +-------------------------------------------------------------------------------------------------------+ ==> 4 rows ---- Now we've got all the adjacent seasons lined up we complete the query with a call to +++<cite>+++MERGE+++</cite>+++: [source,cypher] ---- MATCH (s:Season) WITH s ORDER BY s.timestamp WITH COLLECT(s) AS seasons UNWIND RANGE(0,LENGTH(seasons) - 2) as idx WITH seasons[idx] AS s1, seasons[idx+1] AS s2 MERGE (s1)-[:NEXT]->(s2) ==> +-------------------+ ==> | No data returned. | ==> +-------------------+ ==> Relationships created: 4 ---- And we're done. Hopefully I can remember this approach more than I did the initial one!

neo4j-cypher-step-by-step-to-creating-a-linked-list-of-adjacent-nodes-using-unwind

https://markhneedham.com/blog/2015/06/04/neo4j-cypher-step-by-step-to-creating-a-linked-list-of-adjacent-nodes-using-unwind

2015-06-05 21:25:25

Netty: Testing encoders/decoders

I've been working with Netty a bit recently and having built a pipeline of encoders/decoders as described in this http://seeallhearall.blogspot.co.uk/2012/05/netty-tutorial-part-1-introduction-to.html[excellent tutorial] wanted to test that the encoders and decoders were working without having to send real messages around. Luckily there is a https://netty.io/4.0/api/io/netty/channel/embedded/EmbeddedChannel.html[EmbeddedChannel] which makes our life very easy indeed. Let's say we've got a message 'Foo' that we want to send across the wire. It only contains a single integer value so we'll just send that and reconstruct 'Foo' on the other side. We might write the following encoder to do this: [source,java] ---- // Examples uses Netty 4.0.28.Final public static class MessageEncoder extends MessageToMessageEncoder<Foo> { @Override protected void encode( ChannelHandlerContext ctx, Foo msg, List<Object> out ) throws Exception { ByteBuf buf = ctx.alloc().buffer(); buf.writeInt( msg.value() ); out.add( buf ); } } public static class Foo { private Integer value; public Foo(Integer value) { this.value = value; } public int value() { return value; } } ---- So all we're doing is taking the 'value' field out of 'Foo' and putting it into the list which gets passed downstream. Let's write a test which simulates sending a 'Foo' message and has an empty decoder attempt to process the message: [source,java] ---- @Test public void shouldEncodeAndDecodeVoteRequest() { // given EmbeddedChannel channel = new EmbeddedChannel( new MessageEncoder(), new MessageDecoder() ); // when Foo foo = new Foo( 42 ); channel.writeOutbound( foo ); channel.writeInbound( channel.readOutbound() ); // then Foo returnedFoo = (Foo) channel.readInbound(); assertNotNull(returnedFoo); assertEquals( foo.value(), returnedFoo.value() ); } public static class MessageDecoder extends MessageToMessageDecoder<ByteBuf> { @Override protected void decode( ChannelHandlerContext ctx, ByteBuf msg, List<Object> out ) throws Exception { } } ---- So in the test we write 'Foo' to the outbound channel and then read it back into the inbound channel and then check what we've got. If we run that test now this is what we'll see: [source,java] ---- junit.framework.AssertionFailedError at NettyTest.shouldEncodeAndDecodeVoteRequest(NettyTest.java:28) ---- The message we get back is null which makes sense given that we didn't bother writing the decoder. Let's implement the decoder then: [source,java] ---- public static class MessageDecoder extends MessageToMessageDecoder<ByteBuf> { @Override protected void decode( ChannelHandlerContext ctx, ByteBuf msg, List<Object> out ) throws Exception { int value = msg.readInt(); out.add( new Foo(value) ); } } ---- Now if we run our test again it's all green and happy. We can now go and encode/decode some more complex structures and update our test accordingly.

netty-testing-encodersdecoders

https://markhneedham.com/blog/2015/06/05/netty-testing-encodersdecoders

2015-06-02 06:49:10

R: dplyr - removing empty rows

I'm still working my way through the exercises in Think Bayes and in Chapter 6 needed to do some cleaning of the data in a CSV file containing information about the Price is Right. I downloaded the file using wget: [source,bash] ---- wget http://www.greenteapress.com/thinkbayes/showcases.2011.csv ---- And then loaded it into R and explored the first few rows using dplyr [source,R] ---- library(dplyr) df2011 = read.csv("~/projects/rLearning/showcases.2011.csv") > df2011 %>% head(10) X Sep..19 Sep..20 Sep..21 Sep..22 Sep..23 Sep..26 Sep..27 Sep..28 Sep..29 Sep..30 Oct..3 1 5631K 5632K 5633K 5634K 5635K 5641K 5642K 5643K 5644K 5645K 5681K 2 3 Showcase 1 50969 21901 32815 44432 24273 30554 20963 28941 25851 28800 37703 4 Showcase 2 45429 34061 53186 31428 22320 24337 41373 45437 41125 36319 38752 5 ... ---- As you can see, we have some empty rows which we want to get rid of to ease future processing. I couldn't find an easy way to filter those out but what we can do instead is have empty columns converted to 'NA' and then filter those. First we need to tell +++<cite>+++read.csv+++</cite>+++ to treat empty columns as NA: [source,r] ---- df2011 = read.csv("~/projects/rLearning/showcases.2011.csv", na.strings = c("", "NA")) ---- And now we can filter them out using +++<cite>+++na.omit+++</cite>+++: [source,r] ---- df2011 = df2011 %>% na.omit() > df2011 %>% head(5) X Sep..19 Sep..20 Sep..21 Sep..22 Sep..23 Sep..26 Sep..27 Sep..28 Sep..29 Sep..30 Oct..3 3 Showcase 1 50969 21901 32815 44432 24273 30554 20963 28941 25851 28800 37703 4 Showcase 2 45429 34061 53186 31428 22320 24337 41373 45437 41125 36319 38752 6 Bid 1 42000 14000 32000 27000 18750 27222 25000 35000 22500 21300 21567 7 Bid 2 34000 59900 45000 38000 23000 18525 32000 45000 32000 27500 23800 9 Difference 1 8969 7901 815 17432 5523 3332 -4037 -6059 3351 7500 16136 ... ---- Much better!

r-dplyr-removing-empty-rows

https://markhneedham.com/blog/2015/06/02/r-dplyr-removing-empty-rows

2015-06-20 22:18:55

R: dplyr - segfault cause 'memory not mapped'

In my http://www.markhneedham.com/blog/2015/06/19/r-regex-capturing-multiple-matches-of-the-same-group/[continued playing around with web logs in R] I wanted to process the logs for a day and see what the most popular URIs were. I first read in all the lines using the +++<cite>+++read_lines+++</cite>+++ function in readr and put the vector it produced into a data frame so I could process it using dplyr. [source,r] ---- library(readr) dlines = data.frame(column = read_lines("~/projects/logs/2015-06-18-22-docs")) ---- In the previous post I showed some code to extract the URI from a log line. I extracted this code out into a function and adapted it so that I could pass in a list of values instead of a single value: [source,r] ---- extract_uri = function(log) { parts = str_extract_all(log, "\"[^\"]*\"") return(lapply(parts, function(p) str_match(p[1], "GET (.*) HTTP")[2] %>% as.character)) } ---- Next I ran the following function to count the number of times each URI appeared in the logs: [source,r] ---- library(dplyr) pages_viewed = dlines %>% mutate(uri = extract_uri(column)) %>% count(uri) %>% arrange(desc(n)) ---- This crashed my R process with the following error message: [source,r] ---- segfault cause 'memory not mapped' ---- I narrowed it down to a problem when doing a group by operation on the 'uri' field and came across https://github.com/hadley/dplyr/issues/322[this post] which suggested that it was handled more cleanly in more recently version of dplyr. I upgraded to 0.4.2 and tried again: [source,R] ---- ## Error in eval(expr, envir, enclos): cannot group column uri, of class 'list' ---- That makes more sense. We're probably returning a list from +++<cite>+++extract_uri+++</cite>+++ rather than a vector which would fit nicely back into the data frame. That's fixed easily enough by unlisting the result: [source,r] ---- extract_uri = function(log) { parts = str_extract_all(log, "\"[^\"]*\"") return(unlist(lapply(parts, function(p) str_match(p[1], "GET (.*) HTTP")[2] %>% as.character))) } ---- And now when we run the count function it's happy again, good times!

r-dplyr-segfault-cause-memory-not-mapped

https://markhneedham.com/blog/2015/06/20/r-dplyr-segfault-cause-memory-not-mapped

2015-06-27 22:36:50

R: dplyr - squashing multiple rows per group into one

I spent a bit of the day working on my http://www.markhneedham.com/blog/2015/06/25/r-scraping-wimbledon-draw-data/[Wimbledon] http://www.markhneedham.com/blog/2015/06/26/r-ggplot-show-discrete-scale-even-with-no-value/[data set] and the next thing I explored is all the people that have beaten Andy Murray in the tournament. The following dplyr query gives us the names of those people and the year the match took place: [source,r] ---- library(dplyr) > main_matches %>% filter(loser == "Andy Murray") %>% select(winner, year) winner year 1 Grigor Dimitrov 2014 2 Roger Federer 2012 3 Rafael Nadal 2011 4 Rafael Nadal 2010 5 Andy Roddick 2009 6 Rafael Nadal 2008 7 Marcos Baghdatis 2006 8 David Nalbandian 2005 ---- As you can see, Rafael Nadal shows up multiple times. I wanted to get one row per player and list all the years in a single column. This was my initial attempt: [source,r] ---- > main_matches %>% filter(loser == "Andy Murray") %>% group_by(winner) %>% summarise(years = paste(year)) Source: local data frame [6 x 2] winner years 1 Andy Roddick 2009 2 David Nalbandian 2005 3 Grigor Dimitrov 2014 4 Marcos Baghdatis 2006 5 Rafael Nadal 2011 6 Roger Federer 2012 ---- Unfortunately it just gives you the last matching row per group which isn't quite what we want.. I realised my mistake while trying to pass a vector into paste and noticing that a vector came back when I'd expected a string: [source,r] ---- > paste(c(2008,2009,2010)) [1] "2008" "2009" "2010" ---- The missing argument was 'collapse' - http://www.markhneedham.com/blog/2014/08/11/r-grouping-by-two-variables/[something I'd come across when using plyr] last year: [source,r] ---- > paste(c(2008,2009,2010), collapse=", ") [1] "2008, 2009, 2010" ---- Now, if we apply that to our original function: [source,r] ---- > main_matches %>% filter(loser == "Andy Murray") %>% group_by(winner) %>% summarise(years = paste(year, collapse=", ")) Source: local data frame [6 x 2] winner years 1 Andy Roddick 2009 2 David Nalbandian 2005 3 Grigor Dimitrov 2014 4 Marcos Baghdatis 2006 5 Rafael Nadal 2011, 2010, 2008 6 Roger Federer 2012 ---- That's exactly what we want. Let's tidy that up a bit: [source,r] ---- > main_matches %>% filter(loser == "Andy Murray") %>% group_by(winner) %>% arrange(year) %>% summarise(years = paste(year, collapse =","), times = length(year)) %>% arrange(desc(times), years) Source: local data frame [6 x 3] winner years times 1 Rafael Nadal 2008,2010,2011 3 2 David Nalbandian 2005 1 3 Marcos Baghdatis 2006 1 4 Andy Roddick 2009 1 5 Roger Federer 2012 1 6 Grigor Dimitrov 2014 1 ----

r-dplyr-squashing-multiple-rows-per-group-into-one

https://markhneedham.com/blog/2015/06/27/r-dplyr-squashing-multiple-rows-per-group-into-one

2015-06-27 22:47:22

R: Command line - Error in GenericTranslator$new : could not find function "loadMethod"

I've been reading https://pragprog.com/book/rmtpruby/text-processing-with-ruby[Text Processing with Ruby] over the last week or so and one of the ideas the author describes is setting up your scripts so you can run them directly from the command line. I wanted to do this with my Wimbledon R script and wrote the following script which uses the 'Rscript' executable so that R doesn't launch in interactive mode: _wimbledon_ [source,bash] ---- #!/usr/bin/env Rscript library(rvest) library(dplyr) library(stringr) library(readr) # stuff ---- Then I tried to run it: [source,bash] ---- $ time ./wimbledon ... Error in GenericTranslator$new : could not find function "loadMethod" Calls: write.csv ... html_extract_n -> <Anonymous> -> Map -> mapply -> <Anonymous> -> $ Execution halted real 0m1.431s user 0m1.127s sys 0m0.078s ---- As the error suggests, the script fails when trying to write to a CSV file - it looks like Rscript doesn't load in something from the core library that we need. It turns out https://groups.google.com/forum/#!topic/shiny-discuss/Gx2P_dhzM38[adding the following line] to our script is all we need: [source,bash] ---- library(methods) ---- So we end up with this: [source,bash] ---- #!/usr/bin/env Rscript library(methods) library(rvest) library(dplyr) library(stringr) library(readr) ---- And when we run that all is well!

r-command-line-error-in-generictranslatornew-could-not-find-function-loadmethod

https://markhneedham.com/blog/2015/06/27/r-command-line-error-in-generictranslatornew-could-not-find-function-loadmethod

2015-06-11 23:15:06

Neo4j: Using LOAD CSV to help explore CSV files

During the http://www.meetup.com/graphdb-london/events/222398970/[Neo4j How I met your mother hackathon] that we ran last week one of the attendees noticed that one of the CSV files we were importing wasn't creating as many records as they expected it to. This is typically the case when there's some odd quoting in the CSV file but we decided to look into it. The file in question was one containing references made in HIMYM. The first 5 lines look like this: ~~~bash $ head -n 5 data/import/references.csv ReferencedEpisodeId,ReferencingEpisodeId,ReferenceText 168,184,"Marshall will eventually hear back from the New York State Judicatory Committee in Something New, which will become a main plot point of Season 9." 168,169,Barney proclaiming to be done with Robin will be the focal point of Lobster Crawl. 58,57,"Barney finally confronts his saboteur (Abby, whom he slept with in Ten Sessions) in Everything Must Go." 58,63,"Barney finally confronts his saboteur (Abby, whom he slept with in Ten Sessions) in Everything Must Go." ~~~ And this is how many lines the Unix 'wc' command sees: ~~~bash $ wc -l data/import/references.csv 782 data/import/references.csv ~~~ So we might expect that there are going to be 782 records created if we import that file into Neo4j. Let's run a quick query in Neo4j to see what it thinks: ~~~cypher LOAD CSV WITH HEADERS FROM "file:///Users/markneedham/projects/neo4j-himym/data/import/references.csv" AS row return count(*) =\=> +----------+ =\=> | count(*) | =\=> +----------+ =\=> | 636 | =\=> +----------+ =\=> 1 row ~~~ So we have 146 less records than we expected which means Neo4j is treating multiple lines as one CSV line in some cases. Let's go back to the Unix command line to try and work out which lines those are. There must be some lines which start with part of the 'ReferenceText' rather than a 'ReferenceEpisodeId' so let's extract the first column and see what's going on there: ~~~bash $ cat data/import/references.csv | cut -d"," -f1 | grep -v '[0-9]\+$'| head -n 10 ReferencedEpisodeId This is the Mother's first on-screen appearance with the yellow umbrella. Previously she appeared in Lucky Penny This is the Mother's first on-screen appearance with the yellow umbrella. Previously she appeared in Lucky Penny This is the Mother's first on-screen appearance with the yellow umbrella. Previously she appeared in Lucky Penny This is the Mother's first on-screen appearance with the yellow umbrella. Previously she appeared in Lucky Penny This is the Mother's first on-screen appearance with the yellow umbrella. Previously she appeared in Lucky Penny This is the Mother's first on-screen appearance with the yellow umbrella. Previously she appeared in Lucky Penny This is the Mother's first on-screen appearance with the yellow umbrella. Previously she appeared in Lucky Penny This is the Mother's first on-screen appearance with the yellow umbrella. Previously she appeared in Lucky Penny Also ~~~ We've extracted the first column and then filter the output to only keep rows which don't contain all numbers which will be our rogue rows. Let's switch back to Neo4j land to see which rows it thinks contains these fragments of text: ~~~cypher LOAD CSV WITH HEADERS FROM "file:///Users/markneedham/projects/neo4j-himym/data/import/references.csv" AS row WITH row WHERE row.ReferenceText =~ ".*This is the Mother's first.*" RETURN row.ReferencedEpisodeId, row.ReferencingEpisodeId, row.ReferenceText =\=> +--------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------+ =\=> | row.ReferencedEpisodeId | row.ReferencingEpisodeId | row.ReferenceText | =\=> +--------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------+ =\=> | "45" | "37" | "This is the Mother's first on-screen appearance with the yellow umbrella. Previously she appeared in Lucky Penny, with her head obscured by a bridal veil. She is seen again in No Tomorrow, again hidden by the umbrella, her ankle is seen briefly in Girls Versus Suits, and she gets her first proper appearance in Something New, after which she appears throughout Season 9." | =\=> | "45" | "184" | "This is the Mother's first on-screen appearance with the yellow umbrella. Previously she appeared in Lucky Penny, with her head obscured by a bridal veil. She is seen again in No Tomorrow, again hidden by the umbrella, her ankle is seen briefly in Girls Versus Suits, and she gets her first proper appearance in Something New, after which she appears throughout Season 9." | =\=> +--------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------+ ~~~ Interestingly it only returns two rows containing that phrase whereas we see it at least 8 times. Initially I thought this was an issue with the LOAD CSV command but if we filter the rows to only return ones that have a 'ReferencedEpisodeId' of '45' then we do see them returned: ~~~cypher =\=> +---------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------+ =\=> | row | =\=> +---------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------+ =\=> | {ReferencedEpisodeId \-> "45", ReferencingEpisodeId \-> "53", ReferenceText \-> "The website counting down to the next slap (slapcountdown.com) that Marshall sends Barney reaches zero in Slapsgiving, when the third slap is delivered."} | =\=> | {ReferencedEpisodeId \-> "45", ReferencingEpisodeId \-> "55", ReferenceText \-> "Ted gets rid of his butterfly tramp stamp through ten weekly sessions of laser tattoo removal between The Platinum Rule and Ten Sessions, over the course of which he meets, asks out, and eventually starts dating his dermatologist, Stella Zinman."} | =\=> | {ReferencedEpisodeId \-> "45", ReferencingEpisodeId \-> "57", ReferenceText \-> "Ted gets rid of his butterfly tramp stamp through ten weekly sessions of laser tattoo removal between The Platinum Rule and Ten Sessions, over the course of which he meets, asks out, and eventually starts dating his dermatologist, Stella Zinman."} | =\=> | {ReferencedEpisodeId \-> "45", ReferencingEpisodeId \-> "56", ReferenceText \-> "Ted eventually acquires the yellow umbrella in No Tomorrow (after the Mother leaves it behind at the St. Patrick's Day party, as seen in How Your Mother Met Me), and leaves it in Cindy's and the Mother's apartment in Girls Versus Suits. The umbrella is also seen/referenced in many other episodes, including Right Place, Right Time, Big Days, and Farhampton. =\=> This is the Mother's first on-screen appearance with the yellow umbrella. Previously she appeared in Lucky Penny, with her head obscured by a bridal veil. She is seen again in No Tomorrow, again hidden by the umbrella, her ankle is seen briefly in Girls Versus Suits, and she gets her first proper appearance in Something New, after which she appears throughout Season 9."} | =\=> | {ReferencedEpisodeId \-> "45", ReferencingEpisodeId \-> "200", ReferenceText \-> "Ted eventually acquires the yellow umbrella in No Tomorrow (after the Mother leaves it behind at the St. Patrick's Day party, as seen in How Your Mother Met Me), and leaves it in Cindy's and the Mother's apartment in Girls Versus Suits. The umbrella is also seen/referenced in many other episodes, including Right Place, Right Time, Big Days, and Farhampton. =\=> This is the Mother's first on-screen appearance with the yellow umbrella. Previously she appeared in Lucky Penny, with her head obscured by a bridal veil. She is seen again in No Tomorrow, again hidden by the umbrella, her ankle is seen briefly in Girls Versus Suits, and she gets her first proper appearance in Something New, after which she appears throughout Season 9."} | =\=> | {ReferencedEpisodeId \-> "45", ReferencingEpisodeId \-> "100", ReferenceText \-> "Ted eventually acquires the yellow umbrella in No Tomorrow (after the Mother leaves it behind at the St. Patrick's Day party, as seen in How Your Mother Met Me), and leaves it in Cindy's and the Mother's apartment in Girls Versus Suits. The umbrella is also seen/referenced in many other episodes, including Right Place, Right Time, Big Days, and Farhampton. =\=> This is the Mother's first on-screen appearance with the yellow umbrella. Previously she appeared in Lucky Penny, with her head obscured by a bridal veil. She is seen again in No Tomorrow, again hidden by the umbrella, her ankle is seen briefly in Girls Versus Suits, and she gets her first proper appearance in Something New, after which she appears throughout Season 9."} | =\=> | {ReferencedEpisodeId \-> "45", ReferencingEpisodeId \-> "86", ReferenceText \-> "Ted eventually acquires the yellow umbrella in No Tomorrow (after the Mother leaves it behind at the St. Patrick's Day party, as seen in How Your Mother Met Me), and leaves it in Cindy's and the Mother's apartment in Girls Versus Suits. The umbrella is also seen/referenced in many other episodes, including Right Place, Right Time, Big Days, and Farhampton. =\=> This is the Mother's first on-screen appearance with the yellow umbrella. Previously she appeared in Lucky Penny, with her head obscured by a bridal veil. She is seen again in No Tomorrow, again hidden by the umbrella, her ankle is seen briefly in Girls Versus Suits, and she gets her first proper appearance in Something New, after which she appears throughout Season 9."} | =\=> | {ReferencedEpisodeId \-> "45", ReferencingEpisodeId \-> "113", ReferenceText \-> "Ted eventually acquires the yellow umbrella in No Tomorrow (after the Mother leaves it behind at the St. Patrick's Day party, as seen in How Your Mother Met Me), and leaves it in Cindy's and the Mother's apartment in Girls Versus Suits. The umbrella is also seen/referenced in many other episodes, including Right Place, Right Time, Big Days, and Farhampton. =\=> This is the Mother's first on-screen appearance with the yellow umbrella. Previously she appeared in Lucky Penny, with her head obscured by a bridal veil. She is seen again in No Tomorrow, again hidden by the umbrella, her ankle is seen briefly in Girls Versus Suits, and she gets her first proper appearance in Something New, after which she appears throughout Season 9."} | =\=> | {ReferencedEpisodeId \-> "45", ReferencingEpisodeId \-> "161", ReferenceText \-> "Ted eventually acquires the yellow umbrella in No Tomorrow (after the Mother leaves it behind at the St. Patrick's Day party, as seen in How Your Mother Met Me), and leaves it in Cindy's and the Mother's apartment in Girls Versus Suits. The umbrella is also seen/referenced in many other episodes, including Right Place, Right Time, Big Days, and Farhampton. =\=> This is the Mother's first on-screen appearance with the yellow umbrella. Previously she appeared in Lucky Penny, with her head obscured by a bridal veil. She is seen again in No Tomorrow, again hidden by the umbrella, her ankle is seen briefly in Girls Versus Suits, and she gets her first proper appearance in Something New, after which she appears throughout Season 9."} | =\=> | {ReferencedEpisodeId \-> "45", ReferencingEpisodeId \-> "37", ReferenceText \-> "Ted eventually acquires the yellow umbrella in No Tomorrow (after the Mother leaves it behind at the St. Patrick's Day party, as seen in How Your Mother Met Me), and leaves it in Cindy's and the Mother's apartment in Girls Versus Suits. The umbrella is also seen/referenced in many other episodes, including Right Place, Right Time, Big Days, and Farhampton. =\=> This is the Mother's first on-screen appearance with the yellow umbrella. Previously she appeared in Lucky Penny, with her head obscured by a bridal veil. She is seen again in No Tomorrow, again hidden by the umbrella, her ankle is seen briefly in Girls Versus Suits, and she gets her first proper appearance in Something New, after which she appears throughout Season 9."} | =\=> | {ReferencedEpisodeId \-> "45", ReferencingEpisodeId \-> "184", ReferenceText \-> "Ted eventually acquires the yellow umbrella in No Tomorrow (after the Mother leaves it behind at the St. Patrick's Day party, as seen in How Your Mother Met Me), and leaves it in Cindy's and the Mother's apartment in Girls Versus Suits. The umbrella is also seen/referenced in many other episodes, including Right Place, Right Time, Big Days, and Farhampton. =\=> This is the Mother's first on-screen appearance with the yellow umbrella. Previously she appeared in Lucky Penny, with her head obscured by a bridal veil. She is seen again in No Tomorrow, again hidden by the umbrella, her ankle is seen briefly in Girls Versus Suits, and she gets her first proper appearance in Something New, after which she appears throughout Season 9."} | =\=> | {ReferencedEpisodeId \-> "45", ReferencingEpisodeId \-> "37", ReferenceText \-> "This is the Mother's first on-screen appearance with the yellow umbrella. Previously she appeared in Lucky Penny, with her head obscured by a bridal veil. She is seen again in No Tomorrow, again hidden by the umbrella, her ankle is seen briefly in Girls Versus Suits, and she gets her first proper appearance in Something New, after which she appears throughout Season 9."} | =\=> | {ReferencedEpisodeId \-> "45", ReferencingEpisodeId \-> "184", ReferenceText \-> "This is the Mother's first on-screen appearance with the yellow umbrella. Previously she appeared in Lucky Penny, with her head obscured by a bridal veil. She is seen again in No Tomorrow, again hidden by the umbrella, her ankle is seen briefly in Girls Versus Suits, and she gets her first proper appearance in Something New, after which she appears throughout Season 9."} | =\=> | {ReferencedEpisodeId \-> "45", ReferencingEpisodeId \-> "63", ReferenceText \-> "Marshall makes other home-made websites in Everything Must Go (lilyandmarshallselltheirstuff.com) and The Sexless Innkeeper (itwasthebestnightever.com), where Lily and Future Ted mention it being a problem."} | =\=> | {ReferencedEpisodeId \-> "45", ReferencingEpisodeId \-> "92", ReferenceText \-> "Marshall makes other home-made websites in Everything Must Go (lilyandmarshallselltheirstuff.com) and The Sexless Innkeeper (itwasthebestnightever.com), where Lily and Future Ted mention it being a problem."} | =\=> +---------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------+ ~~~ So the actual problem is that the regex matcher doesn't deal with the new line in the string. Our next step is therefore to get rid of new lines within strings. I spent ages trying to find the appropriate command before coming across the http://www.unix.com/shell-programming-and-scripting/195671-replace-newline-character-between-double-quotes-space.html[following use of awk which does the job]: ~~~bash $ cat data/import/references.csv | awk '(NR-1)%2{$1=$1} {print $0}' RS=\" ORS=\" | wc -l 637 $ cat data/import/references.csv | awk '(NR-1)%2{$1=$1} {print $0}' RS=\" ORS=\" > data/import/refs.csv ~~~ Let's try the LOAD CSV command again: ~~~cypher LOAD CSV WITH HEADERS FROM "file:///Users/markneedham/projects/neo4j-himym/data/import/refs.csv" AS row WITH row WHERE row.ReferenceText =~ ".*This is the Mother's first.*" RETURN row.ReferencedEpisodeId, row.ReferencingEpisodeId, row.ReferenceText =\=> +-------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------+ =\=> | row.ReferencedEpisodeId | row.ReferencingEpisodeId | row.ReferenceText | =\=> +-------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------+ =\=> | "45" | "56" | "Ted eventually acquires the yellow umbrella in No Tomorrow (after the Mother leaves it behind at the St. Patrick's Day party, as seen in How Your Mother Met Me), and leaves it in Cindy's and the Mother's apartment in Girls Versus Suits. The umbrella is also seen/referenced in many other episodes, including Right Place, Right Time, Big Days, and Farhampton. This is the Mother's first on-screen appearance with the yellow umbrella. Previously she appeared in Lucky Penny, with her head obscured by a bridal veil. She is seen again in No Tomorrow, again hidden by the umbrella, her ankle is seen briefly in Girls Versus Suits, and she gets her first proper appearance in Something New, after which she appears throughout Season 9." | =\=> | "45" | "200" | "Ted eventually acquires the yellow umbrella in No Tomorrow (after the Mother leaves it behind at the St. Patrick's Day party, as seen in How Your Mother Met Me), and leaves it in Cindy's and the Mother's apartment in Girls Versus Suits. The umbrella is also seen/referenced in many other episodes, including Right Place, Right Time, Big Days, and Farhampton. This is the Mother's first on-screen appearance with the yellow umbrella. Previously she appeared in Lucky Penny, with her head obscured by a bridal veil. She is seen again in No Tomorrow, again hidden by the umbrella, her ankle is seen briefly in Girls Versus Suits, and she gets her first proper appearance in Something New, after which she appears throughout Season 9." | =\=> | "45" | "100" | "Ted eventually acquires the yellow umbrella in No Tomorrow (after the Mother leaves it behind at the St. Patrick's Day party, as seen in How Your Mother Met Me), and leaves it in Cindy's and the Mother's apartment in Girls Versus Suits. The umbrella is also seen/referenced in many other episodes, including Right Place, Right Time, Big Days, and Farhampton. This is the Mother's first on-screen appearance with the yellow umbrella. Previously she appeared in Lucky Penny, with her head obscured by a bridal veil. She is seen again in No Tomorrow, again hidden by the umbrella, her ankle is seen briefly in Girls Versus Suits, and she gets her first proper appearance in Something New, after which she appears throughout Season 9." | =\=> | "45" | "86" | "Ted eventually acquires the yellow umbrella in No Tomorrow (after the Mother leaves it behind at the St. Patrick's Day party, as seen in How Your Mother Met Me), and leaves it in Cindy's and the Mother's apartment in Girls Versus Suits. The umbrella is also seen/referenced in many other episodes, including Right Place, Right Time, Big Days, and Farhampton. This is the Mother's first on-screen appearance with the yellow umbrella. Previously she appeared in Lucky Penny, with her head obscured by a bridal veil. She is seen again in No Tomorrow, again hidden by the umbrella, her ankle is seen briefly in Girls Versus Suits, and she gets her first proper appearance in Something New, after which she appears throughout Season 9." | =\=> | "45" | "113" | "Ted eventually acquires the yellow umbrella in No Tomorrow (after the Mother leaves it behind at the St. Patrick's Day party, as seen in How Your Mother Met Me), and leaves it in Cindy's and the Mother's apartment in Girls Versus Suits. The umbrella is also seen/referenced in many other episodes, including Right Place, Right Time, Big Days, and Farhampton. This is the Mother's first on-screen appearance with the yellow umbrella. Previously she appeared in Lucky Penny, with her head obscured by a bridal veil. She is seen again in No Tomorrow, again hidden by the umbrella, her ankle is seen briefly in Girls Versus Suits, and she gets her first proper appearance in Something New, after which she appears throughout Season 9." | =\=> | "45" | "161" | "Ted eventually acquires the yellow umbrella in No Tomorrow (after the Mother leaves it behind at the St. Patrick's Day party, as seen in How Your Mother Met Me), and leaves it in Cindy's and the Mother's apartment in Girls Versus Suits. The umbrella is also seen/referenced in many other episodes, including Right Place, Right Time, Big Days, and Farhampton. This is the Mother's first on-screen appearance with the yellow umbrella. Previously she appeared in Lucky Penny, with her head obscured by a bridal veil. She is seen again in No Tomorrow, again hidden by the umbrella, her ankle is seen briefly in Girls Versus Suits, and she gets her first proper appearance in Something New, after which she appears throughout Season 9." | =\=> | "45" | "37" | "Ted eventually acquires the yellow umbrella in No Tomorrow (after the Mother leaves it behind at the St. Patrick's Day party, as seen in How Your Mother Met Me), and leaves it in Cindy's and the Mother's apartment in Girls Versus Suits. The umbrella is also seen/referenced in many other episodes, including Right Place, Right Time, Big Days, and Farhampton. This is the Mother's first on-screen appearance with the yellow umbrella. Previously she appeared in Lucky Penny, with her head obscured by a bridal veil. She is seen again in No Tomorrow, again hidden by the umbrella, her ankle is seen briefly in Girls Versus Suits, and she gets her first proper appearance in Something New, after which she appears throughout Season 9." | =\=> | "45" | "184" | "Ted eventually acquires the yellow umbrella in No Tomorrow (after the Mother leaves it behind at the St. Patrick's Day party, as seen in How Your Mother Met Me), and leaves it in Cindy's and the Mother's apartment in Girls Versus Suits. The umbrella is also seen/referenced in many other episodes, including Right Place, Right Time, Big Days, and Farhampton. This is the Mother's first on-screen appearance with the yellow umbrella. Previously she appeared in Lucky Penny, with her head obscured by a bridal veil. She is seen again in No Tomorrow, again hidden by the umbrella, her ankle is seen briefly in Girls Versus Suits, and she gets her first proper appearance in Something New, after which she appears throughout Season 9." | =\=> | "45" | "37" | "This is the Mother's first on-screen appearance with the yellow umbrella. Previously she appeared in Lucky Penny, with her head obscured by a bridal veil. She is seen again in No Tomorrow, again hidden by the umbrella, her ankle is seen briefly in Girls Versus Suits, and she gets her first proper appearance in Something New, after which she appears throughout Season 9." | =\=> | "45" | "184" | "This is the Mother's first on-screen appearance with the yellow umbrella. Previously she appeared in Lucky Penny, with her head obscured by a bridal veil. She is seen again in No Tomorrow, again hidden by the umbrella, her ankle is seen briefly in Girls Versus Suits, and she gets her first proper appearance in Something New, after which she appears throughout Season 9." | =\=> +-------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------+ ~~~ And there we go! == Update https://twitter.com/mesirii[Michael] pointed out that I could have used the dotall regex flag at the beginning of the regular expression in order to search across new lines without having to remove them! In that case the query would read like this: ~~~cypher LOAD CSV WITH HEADERS FROM "file:///Users/markneedham/projects/neo4j-himym/data/import/references.csv" AS row WITH row WHERE row.ReferenceText =~ "(?s).*This is the Mother.*" RETURN row =\=> +---------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------+ =\=> | row | =\=> +---------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------+ =\=> | {ReferencedEpisodeId \-> "45", ReferencingEpisodeId \-> "56", ReferenceText \-> "Ted eventually acquires the yellow umbrella in No Tomorrow (after the Mother leaves it behind at the St. Patrick's Day party, as seen in How Your Mother Met Me), and leaves it in Cindy's and the Mother's apartment in Girls Versus Suits. The umbrella is also seen/referenced in many other episodes, including Right Place, Right Time, Big Days, and Farhampton. =\=> This is the Mother's first on-screen appearance with the yellow umbrella. Previously she appeared in Lucky Penny, with her head obscured by a bridal veil. She is seen again in No Tomorrow, again hidden by the umbrella, her ankle is seen briefly in Girls Versus Suits, and she gets her first proper appearance in Something New, after which she appears throughout Season 9."} | =\=> | {ReferencedEpisodeId \-> "45", ReferencingEpisodeId \-> "200", ReferenceText \-> "Ted eventually acquires the yellow umbrella in No Tomorrow (after the Mother leaves it behind at the St. Patrick's Day party, as seen in How Your Mother Met Me), and leaves it in Cindy's and the Mother's apartment in Girls Versus Suits. The umbrella is also seen/referenced in many other episodes, including Right Place, Right Time, Big Days, and Farhampton. =\=> This is the Mother's first on-screen appearance with the yellow umbrella. Previously she appeared in Lucky Penny, with her head obscured by a bridal veil. She is seen again in No Tomorrow, again hidden by the umbrella, her ankle is seen briefly in Girls Versus Suits, and she gets her first proper appearance in Something New, after which she appears throughout Season 9."} | =\=> | {ReferencedEpisodeId \-> "45", ReferencingEpisodeId \-> "100", ReferenceText \-> "Ted eventually acquires the yellow umbrella in No Tomorrow (after the Mother leaves it behind at the St. Patrick's Day party, as seen in How Your Mother Met Me), and leaves it in Cindy's and the Mother's apartment in Girls Versus Suits. The umbrella is also seen/referenced in many other episodes, including Right Place, Right Time, Big Days, and Farhampton. =\=> This is the Mother's first on-screen appearance with the yellow umbrella. Previously she appeared in Lucky Penny, with her head obscured by a bridal veil. She is seen again in No Tomorrow, again hidden by the umbrella, her ankle is seen briefly in Girls Versus Suits, and she gets her first proper appearance in Something New, after which she appears throughout Season 9."} | =\=> | {ReferencedEpisodeId \-> "45", ReferencingEpisodeId \-> "86", ReferenceText \-> "Ted eventually acquires the yellow umbrella in No Tomorrow (after the Mother leaves it behind at the St. Patrick's Day party, as seen in How Your Mother Met Me), and leaves it in Cindy's and the Mother's apartment in Girls Versus Suits. The umbrella is also seen/referenced in many other episodes, including Right Place, Right Time, Big Days, and Farhampton. =\=> This is the Mother's first on-screen appearance with the yellow umbrella. Previously she appeared in Lucky Penny, with her head obscured by a bridal veil. She is seen again in No Tomorrow, again hidden by the umbrella, her ankle is seen briefly in Girls Versus Suits, and she gets her first proper appearance in Something New, after which she appears throughout Season 9."} | =\=> | {ReferencedEpisodeId \-> "45", ReferencingEpisodeId \-> "113", ReferenceText \-> "Ted eventually acquires the yellow umbrella in No Tomorrow (after the Mother leaves it behind at the St. Patrick's Day party, as seen in How Your Mother Met Me), and leaves it in Cindy's and the Mother's apartment in Girls Versus Suits. The umbrella is also seen/referenced in many other episodes, including Right Place, Right Time, Big Days, and Farhampton. =\=> This is the Mother's first on-screen appearance with the yellow umbrella. Previously she appeared in Lucky Penny, with her head obscured by a bridal veil. She is seen again in No Tomorrow, again hidden by the umbrella, her ankle is seen briefly in Girls Versus Suits, and she gets her first proper appearance in Something New, after which she appears throughout Season 9."} | =\=> | {ReferencedEpisodeId \-> "45", ReferencingEpisodeId \-> "161", ReferenceText \-> "Ted eventually acquires the yellow umbrella in No Tomorrow (after the Mother leaves it behind at the St. Patrick's Day party, as seen in How Your Mother Met Me), and leaves it in Cindy's and the Mother's apartment in Girls Versus Suits. The umbrella is also seen/referenced in many other episodes, including Right Place, Right Time, Big Days, and Farhampton. =\=> This is the Mother's first on-screen appearance with the yellow umbrella. Previously she appeared in Lucky Penny, with her head obscured by a bridal veil. She is seen again in No Tomorrow, again hidden by the umbrella, her ankle is seen briefly in Girls Versus Suits, and she gets her first proper appearance in Something New, after which she appears throughout Season 9."} | =\=> | {ReferencedEpisodeId \-> "45", ReferencingEpisodeId \-> "37", ReferenceText \-> "Ted eventually acquires the yellow umbrella in No Tomorrow (after the Mother leaves it behind at the St. Patrick's Day party, as seen in How Your Mother Met Me), and leaves it in Cindy's and the Mother's apartment in Girls Versus Suits. The umbrella is also seen/referenced in many other episodes, including Right Place, Right Time, Big Days, and Farhampton. =\=> This is the Mother's first on-screen appearance with the yellow umbrella. Previously she appeared in Lucky Penny, with her head obscured by a bridal veil. She is seen again in No Tomorrow, again hidden by the umbrella, her ankle is seen briefly in Girls Versus Suits, and she gets her first proper appearance in Something New, after which she appears throughout Season 9."} | =\=> | {ReferencedEpisodeId \-> "45", ReferencingEpisodeId \-> "184", ReferenceText \-> "Ted eventually acquires the yellow umbrella in No Tomorrow (after the Mother leaves it behind at the St. Patrick's Day party, as seen in How Your Mother Met Me), and leaves it in Cindy's and the Mother's apartment in Girls Versus Suits. The umbrella is also seen/referenced in many other episodes, including Right Place, Right Time, Big Days, and Farhampton. =\=> This is the Mother's first on-screen appearance with the yellow umbrella. Previously she appeared in Lucky Penny, with her head obscured by a bridal veil. She is seen again in No Tomorrow, again hidden by the umbrella, her ankle is seen briefly in Girls Versus Suits, and she gets her first proper appearance in Something New, after which she appears throughout Season 9."} | =\=> | {ReferencedEpisodeId \-> "45", ReferencingEpisodeId \-> "37", ReferenceText \-> "This is the Mother's first on-screen appearance with the yellow umbrella. Previously she appeared in Lucky Penny, with her head obscured by a bridal veil. She is seen again in No Tomorrow, again hidden by the umbrella, her ankle is seen briefly in Girls Versus Suits, and she gets her first proper appearance in Something New, after which she appears throughout Season 9."} | =\=> | {ReferencedEpisodeId \-> "45", ReferencingEpisodeId \-> "184", ReferenceText \-> "This is the Mother's first on-screen appearance with the yellow umbrella. Previously she appeared in Lucky Penny, with her head obscured by a bridal veil. She is seen again in No Tomorrow, again hidden by the umbrella, her ankle is seen briefly in Girls Versus Suits, and she gets her first proper appearance in Something New, after which she appears throughout Season 9."} | =\=> +---------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------+ ~~~

neo4j-using-load-csv-to-help-explore-csv-files

https://markhneedham.com/blog/2015/06/11/neo4j-using-load-csv-to-help-explore-csv-files

2015-06-11 21:38:32

Mac OS X: GNU sed - Hex string replacement / replacing new line characters

Recently I was working with a CSV file which contained both Windows and Unix line endings which was making it difficult to work with. The actual line endings were HEX '0A0D' i.e. Windows line breaks but there were also HEX 'OA' i.e. Unix line breaks within one of the columns. I wanted to get rid of the Unix line breaks and discovered that you can do http://stackoverflow.com/questions/7760717/hex-string-replacement-using-sed[HEX sequence replacement using the GNU version of sed] - unfortunately the Mac ships with the BSD version which doesn't have this functionaltiy. The first step was therefore to install the GNU version of sed. [source,bash] ---- brew install coreutils brew install gnu-sed --with-default-names ---- I wanted to replace my system sed so that's why I went with the '--with-default-names' flag - without that flag I believe the sed installation would be accessible as 'gs-sed'. The following is an example of what the lines in the file look like: [source,bash] ---- $ echo -e "Hello\x0AMark\x0A\x0D" Hello Mark ---- We want to get rid of the new line in between 'Hello' and 'Mark' but leave the other one be. I adapted one of the commands from http://backreference.org/2009/12/23/how-to-match-newlines-in-sed/[this tutorial] to look for lines which end in '0A' where that isn't followed by a '0D': [source,bash] ---- $ echo -e "Hello\x0AMark\x0A\x0D" | \ sed 'N;/\x0A[^\x0D]/s/\n/ /' Hello Mark ---- Let's go through the parts of the sed command: * +++<cite>+++N+++</cite>+++ - this creates a multiline pattern space by reading a new line of input and appending it to the contents of the pattern space. The two lines are separated by a new line. * +++<cite>+++/\x0A[{caret}\x0D]/+++</cite>+++ - this matches any lines which contain 'OA' not followed by 'OD' * +++<cite>+++/s/\n/ /+++</cite>+++ - this substitutes the new line character with a space for those matching lines from the previous command. Now let's check it works if we have multiple lines that we want to squash: [source,bash] ---- $ echo -e "Hello\x0AMark\x0A\x0DHello\x0AMichael\x0A\x0D" Hello Mark Hello Michael $ echo -e "Hello\x0AMark\x0A\x0DHello\x0AMichael\x0A\x0D" | \ sed 'N;/\x0A[^\x0D]/s/\n/ /' Hello Mark Hello Michael ---- Looks good! The actual file is a bit more nuanced so I've still got a bit more work to do but this is a good start.

mac-os-x-gnu-sed-hex-string-replacement-replacing-new-line-characters

https://markhneedham.com/blog/2015/06/11/mac-os-x-gnu-sed-hex-string-replacement-replacing-new-line-characters

2015-06-29 05:36:22

R: Speeding up the Wimbledon scraping job

Over the past few days I've written http://www.markhneedham.com/blog/2015/06/26/r-ggplot-show-discrete-scale-even-with-no-value/[a few] http://www.markhneedham.com/blog/2015/06/25/r-scraping-wimbledon-draw-data/[blog] http://www.markhneedham.com/blog/2015/06/28/r-dplyr-update-rows-with-earlierprevious-rows-values/[posts] about a https://github.com/mneedham/neo4j-wimbledon/blob/master/wimbledon.csv[Wimbledon data set] I've been building and after running the scripts a few times I noticed that it was taking much longer to run that I expected. To recap, I started out with the following function which takes in a URI and returns a data frame containing a row for each match: [source,r] ---- library(rvest) library(dplyr) scrape_matches1 = function(uri) { matches = data.frame() s = html(uri) rows = s %>% html_nodes("div#scoresResultsContent tr") i = 0 for(row in rows) { players = row %>% html_nodes("td.day-table-name a") seedings = row %>% html_nodes("td.day-table-seed") score = row %>% html_node("td.day-table-score a") flags = row %>% html_nodes("td.day-table-flag img") if(!is.null(score)) { player1 = players[1] %>% html_text() %>% str_trim() seeding1 = ifelse(!is.na(seedings[1]), seedings[1] %>% html_node("span") %>% html_text() %>% str_trim(), NA) flag1 = flags[1] %>% html_attr("alt") player2 = players[2] %>% html_text() %>% str_trim() seeding2 = ifelse(!is.na(seedings[2]), seedings[2] %>% html_node("span") %>% html_text() %>% str_trim(), NA) flag2 = flags[2] %>% html_attr("alt") matches = rbind(data.frame(winner = player1, winner_seeding = seeding1, winner_flag = flag1, loser = player2, loser_seeding = seeding2, loser_flag = flag2, score = score %>% html_text() %>% str_trim(), round = round), matches) } else { round = row %>% html_node("th") %>% html_text() } } return(matches) } ---- Let's run it to get an idea of the data that it returns: [source,r] ---- matches1 = scrape_matches1("http://www.atpworldtour.com/en/scores/archive/wimbledon/540/2014/results") > matches1 %>% filter(round %in% c("Finals", "Semi-Finals", "Quarter-Finals")) winner winner_seeding winner_flag loser loser_seeding loser_flag score round 1 Milos Raonic (8) CAN Nick Kyrgios (WC) AUS 674 62 64 764 Quarter-Finals 2 Roger Federer (4) SUI Stan Wawrinka (5) SUI 36 765 64 64 Quarter-Finals 3 Grigor Dimitrov (11) BUL Andy Murray (3) GBR 61 764 62 Quarter-Finals 4 Novak Djokovic (1) SRB Marin Cilic (26) CRO 61 36 674 62 62 Quarter-Finals 5 Roger Federer (4) SUI Milos Raonic (8) CAN 64 64 64 Semi-Finals 6 Novak Djokovic (1) SRB Grigor Dimitrov (11) BUL 64 36 762 767 Semi-Finals 7 Novak Djokovic (1) SRB Roger Federer (4) SUI 677 64 764 57 64 Finals ---- As I mentioned, it's quite slow but I thought I'd wrap it in +++<cite>+++system.time+++</cite>+++ so I could see exactly how long it was taking: [source,r] ---- > system.time(scrape_matches1("http://www.atpworldtour.com/en/scores/archive/wimbledon/540/2014/results")) user system elapsed 25.570 0.111 31.416 ---- About 30 seconds! The first thing I tried was downloading the file separately and running the function against the local file: [source,r] ---- > system.time(scrape_matches1("data/raw/2014.html")) user system elapsed 25.662 0.123 25.863 ---- Hmmm, that's only saved us 5 seconds so the bottleneck must be somewhere else. Still there's no point making a HTTP request every time we run the script so we'll stick with the local file version. While browsing http://cran.r-project.org/web/packages/rvest/rvest.pdf[rvest's vignette] I noticed a function called +++<cite>+++html_table+++</cite>+++ which I was curious about. I decided to try and replace some of my code with a call to that: [source,r] ---- matches2= html("data/raw/2014.html") %>% html_node("div#scoresResultsContent table.day-table") %>% html_table(header = FALSE) %>% mutate(X1 = ifelse(X1 == "", NA, X1)) %>% mutate(round = ifelse(grepl("\\([0-9]\\)|\\(", X1), NA, X1)) %>% mutate(round = na.locf(round)) %>% filter(!is.na(X8)) %>% select(winner = X3, winner_seeding = X1, loser = X7, loser_seeding = X5, score = X8, round) > matches2 %>% filter(round %in% c("Finals", "Semi-Finals", "Quarter-Finals")) winner winner_seeding loser loser_seeding score round 1 Novak Djokovic (1) Roger Federer (4) 677 64 764 57 64 Finals 2 Novak Djokovic (1) Grigor Dimitrov (11) 64 36 762 767 Semi-Finals 3 Roger Federer (4) Milos Raonic (8) 64 64 64 Semi-Finals 4 Novak Djokovic (1) Marin Cilic (26) 61 36 674 62 62 Quarter-Finals 5 Grigor Dimitrov (11) Andy Murray (3) 61 764 62 Quarter-Finals 6 Roger Federer (4) Stan Wawrinka (5) 36 765 64 64 Quarter-Finals 7 Milos Raonic (8) Nick Kyrgios (WC) 674 62 64 764 Quarter-Finals ---- I had to do some slightly clever stuff to get the 'round' column into shape using zoo's +++<cite>+++na.locf+++</cite>+++ function which I http://www.markhneedham.com/blog/2015/06/28/r-dplyr-update-rows-with-earlierprevious-rows-values/[wrote about previously.] Unfortunately I couldn't work out how to extract the flag with this version - that value is hidden in the 'alt' tag of an img and presumably +++<cite>+++html_table+++</cite>+++ is just grabbing the text value of each cell. This version is much quicker though! [source,r] ---- system.time(html("data/raw/2014.html") %>% html_node("div#scoresResultsContent table.day-table") %>% html_table(header = FALSE) %>% mutate(X1 = ifelse(X1 == "", NA, X1)) %>% mutate(round = ifelse(grepl("\\([0-9]\\)|\\(", X1), NA, X1)) %>% mutate(round = na.locf(round)) %>% filter(!is.na(X8)) %>% select(winner = X3, winner_seeding = X1, loser = X7, loser_seeding = X5, score = X8, round)) user system elapsed 0.545 0.002 0.548 ---- What I realised from writing this version is that I need to match all the columns with one call to +++<cite>+++html_nodes+++</cite>+++ rather than getting the row and then each column in a loop. I rewrote the function to do that: [source,r] ---- scrape_matches3 = function(uri) { s = html(uri) players = s %>% html_nodes("div#scoresResultsContent tr td.day-table-name a") seedings = s %>% html_nodes("div#scoresResultsContent tr td.day-table-seed") scores = s %>% html_nodes("div#scoresResultsContent tr td.day-table-score a") flags = s %>% html_nodes("div#scoresResultsContent tr td.day-table-flag img") %>% html_attr("alt") %>% str_trim() matches3 = data.frame( winner = sapply(seq(1,length(players),2), function(idx) players[[idx]] %>% html_text()), winner_seeding = sapply(seq(1,length(seedings),2), function(idx) seedings[[idx]] %>% html_text() %>% str_trim()), winner_flag = sapply(seq(1,length(flags),2), function(idx) flags[[idx]]), loser = sapply(seq(2,length(players),2), function(idx) players[[idx]] %>% html_text()), loser_seeding = sapply(seq(2,length(seedings),2), function(idx) seedings[[idx]] %>% html_text() %>% str_trim()), loser_flag = sapply(seq(2,length(flags),2), function(idx) flags[[idx]]), score = sapply(scores, function(score) score %>% html_text() %>% str_trim()) ) return(matches3) } ---- Let's run and time that to check we're getting back the right results in a timely manner: [source,r] ---- > matches3 %>% sample_n(10) winner winner_seeding winner_flag loser loser_seeding loser_flag score 70 David Ferrer (7) ESP Pablo Carreno Busta ESP 60 673 61 61 128 Alex Kuznetsov (26) USA Tim Smyczek (3) USA 46 63 63 63 220 Rogerio Dutra Silva BRA Kristijan Mesaros CRO 62 63 83 Kevin Anderson (20) RSA Aljaz Bedene (LL) GBR 63 75 62 73 Kei Nishikori (10) JPN Kenny De Schepper FRA 64 765 75 56 Roberto Bautista Agut (27) ESP Jan Hernych (Q) CZE 75 46 62 62 138 Ante Pavic CRO Marc Gicquel (29) FRA 46 63 765 64 174 Tim Puetz GER Ruben Bemelmans BEL 64 62 103 Lleyton Hewitt AUS Michal Przysiezny POL 62 6714 61 64 35 Roger Federer (4) SUI Gilles Muller (Q) LUX 63 75 63 > system.time(scrape_matches3("data/raw/2014.html")) user system elapsed 0.815 0.006 0.827 ---- It's still quick - a bit slower than +++<cite>+++html_table+++</cite>+++ but we can deal with that. As you can see, I also had to add some logic to separate the values for the winners and losers - the players, seeds, flags come back as as one big list. The odd rows represent the winner; the even rows the loser. Annoyingly we've now lost the 'round' column because that appears as a table heading so we can't extract it the same way. I ended up cheating a bit to get it to work by working out how many matches each round should contain and generated a vector with that number of entries: [source,r] ---- raw_rounds = s %>% html_nodes("th") %>% html_text() > raw_rounds [1] "Finals" "Semi-Finals" "Quarter-Finals" "Round of 16" "Round of 32" [6] "Round of 64" "Round of 128" "3rd Round Qualifying" "2nd Round Qualifying" "1st Round Qualifying" rounds = c( sapply(0:6, function(idx) rep(raw_rounds[[idx + 1]], 2 ** idx)) %>% unlist(), sapply(7:9, function(idx) rep(raw_rounds[[idx + 1]], 2 ** (idx - 3))) %>% unlist()) > rounds[1:10] [1] "Finals" "Semi-Finals" "Semi-Finals" "Quarter-Finals" "Quarter-Finals" "Quarter-Finals" "Quarter-Finals" [8] "Round of 16" "Round of 16" "Round of 16" ---- Let's put that code into the function and see if we end up with the same resulting data frame: [source,r] ---- scrape_matches4 = function(uri) { s = html(uri) players = s %>% html_nodes("div#scoresResultsContent tr td.day-table-name a") seedings = s %>% html_nodes("div#scoresResultsContent tr td.day-table-seed") scores = s %>% html_nodes("div#scoresResultsContent tr td.day-table-score a") flags = s %>% html_nodes("div#scoresResultsContent tr td.day-table-flag img") %>% html_attr("alt") %>% str_trim() raw_rounds = s %>% html_nodes("th") %>% html_text() rounds = c( sapply(0:6, function(idx) rep(raw_rounds[[idx + 1]], 2 ** idx)) %>% unlist(), sapply(7:9, function(idx) rep(raw_rounds[[idx + 1]], 2 ** (idx - 3))) %>% unlist()) matches4 = data.frame( winner = sapply(seq(1,length(players),2), function(idx) players[[idx]] %>% html_text()), winner_seeding = sapply(seq(1,length(seedings),2), function(idx) seedings[[idx]] %>% html_text() %>% str_trim()), winner_flag = sapply(seq(1,length(flags),2), function(idx) flags[[idx]]), loser = sapply(seq(2,length(players),2), function(idx) players[[idx]] %>% html_text()), loser_seeding = sapply(seq(2,length(seedings),2), function(idx) seedings[[idx]] %>% html_text() %>% str_trim()), loser_flag = sapply(seq(2,length(flags),2), function(idx) flags[[idx]]), score = sapply(scores, function(score) score %>% html_text() %>% str_trim()), round = rounds ) return(matches4) } matches4 = scrape_matches4("data/raw/2014.html") > matches4 %>% filter(round %in% c("Finals", "Semi-Finals", "Quarter-Finals")) winner winner_seeding winner_flag loser loser_seeding loser_flag score round 1 Novak Djokovic (1) SRB Roger Federer (4) SUI 677 64 764 57 64 Finals 2 Novak Djokovic (1) SRB Grigor Dimitrov (11) BUL 64 36 762 767 Semi-Finals 3 Roger Federer (4) SUI Milos Raonic (8) CAN 64 64 64 Semi-Finals 4 Novak Djokovic (1) SRB Marin Cilic (26) CRO 61 36 674 62 62 Quarter-Finals 5 Grigor Dimitrov (11) BUL Andy Murray (3) GBR 61 764 62 Quarter-Finals 6 Roger Federer (4) SUI Stan Wawrinka (5) SUI 36 765 64 64 Quarter-Finals 7 Milos Raonic (8) CAN Nick Kyrgios (WC) AUS 674 62 64 764 Quarter-Finals ---- We shouldn't have added much to the time but let's check: [source,r] ---- > system.time(scrape_matches4("data/raw/2014.html")) user system elapsed 0.816 0.004 0.824 ---- Sweet. We've saved ourselves 29 seconds per page as long as the number of rounds stayed constant over the years. For the 10 years that I've looked at it has but I expect if you go back further the draw sizes will have been different and our script would break. For now though this will do!

r-speeding-up-the-wimbledon-scraping-job

https://markhneedham.com/blog/2015/06/29/r-speeding-up-the-wimbledon-scraping-job

2015-06-28 22:30:08

R: dplyr - Update rows with earlier/previous rows values

Recently I had a data frame which contained a column which had mostly empty values: [source,r] ---- > data.frame(col1 = c(1,2,3,4,5), col2 = c("a", NA, NA , "b", NA)) col1 col2 1 1 a 2 2 <NA> 3 3 <NA> 4 4 b 5 5 <NA> ---- I wanted to fill in the NA values with the last non NA value from that column. So I want the data frame to look like this: [source,text] ---- 1 1 a 2 2 a 3 3 a 4 4 b 5 5 b ---- I spent ages searching around before I http://stackoverflow.com/questions/27207162/fill-in-na-based-on-the-last-non-na-value-for-each-group-in-r[came across the +++<cite>+++na.locf+++</cite>+++ function in the zoo library] which does the job: [source,r] ---- library(zoo) library(dplyr) > data.frame(col1 = c(1,2,3,4,5), col2 = c("a", NA, NA , "b", NA)) %>% do(na.locf(.)) col1 col2 1 1 a 2 2 a 3 3 a 4 4 b 5 5 b ---- This will fill in the missing values for every column, so if we had a third column with missing values it would populate those too: [source,r] ---- > data.frame(col1 = c(1,2,3,4,5), col2 = c("a", NA, NA , "b", NA), col3 = c("A", NA, "B", NA, NA)) %>% do(na.locf(.)) col1 col2 col3 1 1 a A 2 2 a A 3 3 a B 4 4 b B 5 5 b B ---- If we only want to populate 'col2' and leave 'col3' as it is we can apply the function specifically to that column: [source,r] ---- > data.frame(col1 = c(1,2,3,4,5), col2 = c("a", NA, NA , "b", NA), col3 = c("A", NA, "B", NA, NA)) %>% mutate(col2 = na.locf(col2)) col1 col2 col3 1 1 a A 2 2 a <NA> 3 3 a B 4 4 b <NA> 5 5 b <NA> ---- It's quite a neat function and certainly comes in helpful when cleaning up data sets which don't tend to be as uniform as you'd hope!

r-dplyr-update-rows-with-earlierprevious-rows-values

https://markhneedham.com/blog/2015/06/28/r-dplyr-update-rows-with-earlierprevious-rows-values

2015-06-17 17:23:10

Coding: Explore and retreat

When refactoring code or looking for the best way to integrate a new piece of functionality I generally favour a small steps/incremental approach but recent experiences have led me to believe that this isn't always the quickest approach. Sometimes it seems to make more sense to go on little discovery missions in the code, make some bigger steps and then if necessary retreat and revert our changes and apply the lessons learnt on our next discovery mission. This technique which isn't anything novel but I think is quite effective. https://twitter.com/mesirii[Michael] and I were recently looking at the http://www.ludowaltman.nl/slm/[Smart Local Moving algorithm] which is used for community detection in large networks and decided to refactor the code to make sure we understood how it worked. When we started the outline of the main class was https://github.com/mneedham/slm/blob/3e8468f3598ce9f61b2be32e2953890d1d497a4b/src/main/java/Network.java[like this]: [source,java] ---- public class Network implements Cloneable, Serializable { private static final long serialVersionUID = 1; private int numberOfNodes; private int[] firstNeighborIndex; private int[] neighbor; private double[] edgeWeight; private double totalEdgeWeightSelfLinks; private double[] nodeWeight; private int nClusters; private int[] cluster; private double[] clusterWeight; private int[] numberNodesPerCluster; private int[][] nodePerCluster; private boolean clusteringStatsAvailable; ... } ---- My initial approach was to put methods around things to make it a bit easier to understand and then step by step replace each of those fields with nodes and relationships. I spent the first couple of hours doing this and while it was making the code more readable it wasn't progressing very quickly and I wasn't much wiser about how the code worked. Michael and I paired on it for a few hours and he adopted a slightly different but more successful approach where we looked at slightly bigger chunks of code e.g. all the loops that used the +++<cite>+++firstNeighborIndex+++</cite>+++ field and then created a hypothesis of what that code was doing. In this case +++<cite>+++firstNeighborIndex+++</cite>+++ acts as an offset into +++<cite>+++neighbor+++</cite>+++ and is used to iterate through a node's relationships. We thought we could probably replace that with something more similar to the http://neo4j.com/[Neo4j] model where you have classes for nodes and relationships and a node has a method which returns a collection of relationships. We tried tearing out everywhere that used those two fields and replacing them with our new nodes/relationships code but that didn't work because we hadn't realised that +++<cite>+++edgeWeight+++</cite>+++ and +++<cite>+++nodeWeight+++</cite>+++ are also tied to the contents of the original fields. We therefore needed to retreat and try again. This time I put the new approach alongside the existing approach and then slowly replaced existing bits of code. Along the way I came up with other ideas about how to restructure the code, tried some more bigger leaps to validate my ideas and then moved back into incremental mode again. In summary I've found the combination of incrementally changing code and going on bigger exploratory missions works quite well. Now I'm trying to work out when each approach is appropriate and I'll write that up when I learn more! You can see my progress https://github.com/mneedham/slm/commits/master[via the github commits].

coding-explore-and-retreat

https://markhneedham.com/blog/2015/06/17/coding-explore-and-retreat

2015-06-19 21:38:47

R: Regex - capturing multiple matches of the same group

I've been playing around with some web logs using R and I wanted to extract everything that existed in double quotes within a logged entry. This is an example of a log entry that I want to parse: [source,r] ---- log = '2015-06-18-22:277:548311224723746831\t2015-06-18T22:00:11\t2015-06-18T22:00:05Z\t93317114\tip-127-0-0-1\t127.0.0.5\tUser\tNotice\tneo4j.com.access.log\t127.0.0.3 - - [18/Jun/2015:22:00:11 +0000] "GET /docs/stable/query-updating.html HTTP/1.1" 304 0 "http://neo4j.com/docs/stable/cypher-introduction.html" "Mozilla/5.0 (Macintosh; Intel Mac OS X 10_10_2) AppleWebKit/537.36 (KHTML, like Gecko) Chrome/43.0.2357.124 Safari/537.36"' ---- And I want to extract these 3 things: * /docs/stable/query-updating.html * http://neo4j.com/docs/stable/cypher-introduction.html * Mozilla/5.0 (Macintosh; Intel Mac OS X 10_10_2) AppleWebKit/537.36 (KHTML, like Gecko) Chrome/43.0.2357.124 Safari/537.36 i.e. the URI, the referrer and browser details. I'll be using the +++<cite>+++stringr+++</cite>+++ library which seems to work quite well for this type of work. To extract these values we need to find all the occurrences of double quotes and get the text inside those quotes. We might start by using the +++<cite>+++str_match+++</cite>+++ function: [source,r] ---- > library(stringr) > str_match(log, "\"[^\"]*\"") [,1] [1,] "\"GET /docs/stable/query-updating.html HTTP/1.1\"" ---- Unfortunately that only picked up the first occurrence of the pattern so we've got the URI but not the referrer or browser details. I tried +++<cite>+++str_extract+++</cite>+++ with similar results before I found +++<cite>+++str_extract_all+++</cite>+++ which does the job: [source,r] ---- > str_extract_all(log, "\"[^\"]*\"") [[1]] [1] "\"GET /docs/stable/query-updating.html HTTP/1.1\"" [2] "\"http://neo4j.com/docs/stable/cypher-introduction.html\"" [3] "\"Mozilla/5.0 (Macintosh; Intel Mac OS X 10_10_2) AppleWebKit/537.36 (KHTML, like Gecko) Chrome/43.0.2357.124 Safari/537.36\"" ---- We still need to do a bit of cleanup to get rid of the 'GET' and 'HTTP/1.1' in the URI and the quotes in all of them: [source,r] ---- parts = str_extract_all(log, "\"[^\"]*\"")[[1]] uri = str_match(parts[1], "GET (.*) HTTP")[2] referer = str_match(parts[2], "\"(.*)\"")[2] browser = str_match(parts[3], "\"(.*)\"")[2] > uri [1] "/docs/stable/query-updating.html" > referer [1] "https://www.google.com/" > browser [1] "Mozilla/5.0 (Macintosh; Intel Mac OS X 10_10_2) AppleWebKit/537.36 (KHTML, like Gecko) Chrome/43.0.2357.124 Safari/537.36" ---- We could then go on to split out the browser string into its sub components but that'll do for now!

r-regex-capturing-multiple-matches-of-the-same-group

https://markhneedham.com/blog/2015/06/19/r-regex-capturing-multiple-matches-of-the-same-group

2015-06-26 22:48:17

R: ggplot - Show discrete scale even with no value

As I mentioned in a previous blog post, I've been http://www.markhneedham.com/blog/2015/06/25/r-scraping-wimbledon-draw-data/[scraping data for the Wimbledon tennis tournament], and having got the data for the last ten years I wrote a query using dplyr to find out how players did each year over that period. I ended up with the following functions to filter my data frame of all the matches: [source,r] ---- round_reached = function(player, main_matches) { furthest_match = main_matches %>% filter(winner == player | loser == player) %>% arrange(desc(round)) %>% head(1) return(ifelse(furthest_match$winner == player, "Winner", as.character(furthest_match$round))) } player_performance = function(name, matches) { player = data.frame() for(y in 2005:2014) { round = round_reached(name, filter(matches, year == y)) if(length(round) == 1) { player = rbind(player, data.frame(year = y, round = round)) } else { player = rbind(player, data.frame(year = y, round = "Did not enter")) } } return(player) } ---- When we call that function we see the following output: [source,r] ---- > player_performance("Andy Murray", main_matches) year round 1 2005 Round of 32 2 2006 Round of 16 3 2007 Did not enter 4 2008 Quarter-Finals 5 2009 Semi-Finals 6 2010 Semi-Finals 7 2011 Semi-Finals 8 2012 Finals 9 2013 Winner 10 2014 Quarter-Finals ---- I wanted to create a chart showing Murray's progress over the years with the round reached on the y axis and the year on the x axis. In order to do this I had to make sure the 'round' column was being treated as a factor variable: [source,r] ---- df = player_performance("Andy Murray", main_matches) rounds = c("Did not enter", "Round of 128", "Round of 64", "Round of 32", "Round of 16", "Quarter-Finals", "Semi-Finals", "Finals", "Winner") df$round = factor(df$round, levels = rounds) > df$round [1] Round of 32 Round of 16 Did not enter Quarter-Finals Semi-Finals Semi-Finals Semi-Finals [8] Finals Winner Quarter-Finals Levels: Did not enter Round of 128 Round of 64 Round of 32 Round of 16 Quarter-Finals Semi-Finals Finals Winner ---- Now that we've got that we can plot his progress: [source,r] ---- ggplot(aes(x = year, y = round, group=1), data = df) + geom_point() + geom_line() + scale_x_continuous(breaks=df$year) + scale_y_discrete(breaks = rounds) ---- image::{{<siteurl>}}/uploads/2015/06/2015-06-26_23-37-32.png[2015 06 26 23 37 32,522] This is a good start but we've lost the rounds which don't have a corresponding entry on the x axis. I'd like to keep them so it's easier to compare the performance of different players. It turns out that http://stackoverflow.com/questions/9818835/geom-boxplot-from-ggplot2-forcing-an-empty-level-to-appear[all we need to do] is pass 'drop = FALSE' to +++<cite>+++scale_y_discrete+++</cite>+++ and it will work exactly as we want: [source,r] ---- ggplot(aes(x = year, y = round, group=1), data = df) + geom_point() + geom_line() + scale_x_continuous(breaks=df$year) + scale_y_discrete(breaks = rounds, drop = FALSE) ---- image::{{<siteurl>}}/uploads/2015/06/2015-06-26_23-41-01.png[2015 06 26 23 41 01,523] Neat. Now let's have a look at the performances of some of the other top players: [source,r] ---- draw_chart = function(player, main_matches){ df = player_performance(player, main_matches) df$round = factor(df$round, levels = rounds) ggplot(aes(x = year, y = round, group=1), data = df) + geom_point() + geom_line() + scale_x_continuous(breaks=df$year) + scale_y_discrete(breaks = rounds, drop=FALSE) + ggtitle(player) + theme(axis.text.x=element_text(angle=90, hjust=1)) } a = draw_chart("Andy Murray", main_matches) b = draw_chart("Novak Djokovic", main_matches) c = draw_chart("Rafael Nadal", main_matches) d = draw_chart("Roger Federer", main_matches) library(gridExtra) grid.arrange(a,b,c,d, ncol=2) ---- image::{{<siteurl>}}/uploads/2015/06/2015-06-26_23-46-15.png[2015 06 26 23 46 15,526] And that's all for now!

r-ggplot-show-discrete-scale-even-with-no-value

https://markhneedham.com/blog/2015/06/26/r-ggplot-show-discrete-scale-even-with-no-value

2015-06-21 22:07:49

R: Scraping Neo4j release dates with rvest

As part of my log analysis I wanted to get the Neo4j release dates which are accessible from the http://neo4j.com/release-notes[release notes] and decided to try out http://blog.rstudio.org/2014/11/24/rvest-easy-web-scraping-with-r/[Hadley Wickham's rvest scraping library] which he released at the end of 2014. rvest is based on Python's http://www.crummy.com/software/BeautifulSoup/bs4/doc/[beautifulsoup] which has become my scraping library of choice so I didn't find it too difficult to pick up. To start with we need to download the release notes locally so we don't have to go over the network when we're doing our scraping: [source,r] ---- download.file("http://neo4j.com/release-notes/page/1", "release-notes.html") download.file("http://neo4j.com/release-notes/page/2", "release-notes2.html") ---- We want to parse those pages back and return the rows which contain version numbers and release dates. The HTML looks like this: image::{{<siteurl>}}/uploads/2015/06/2015-06-21_22-57-20.png[2015 06 21 22 57 20,400] We can get the rows with the following code: [source,r] ---- library(rvest) library(dplyr) page1 <- html("release-notes.html") page2 <- html("release-notes2.html") rows = c(page1 %>% html_nodes("div.small-12 div.row"), page2 %>% html_nodes("div.small-12 div.row") ) > rows %>% head(1) [[1]] <div class="row"> <h3 class="entry-title"><a href="http://neo4j.com/release-notes/neo4j-2-2-2/">Latest Release: Neo4j 2.2.2</a></h3> <h6>05/21/2015</h6> <p>Neo4j 2.2.2 is a maintenance release, with critical improvements.</p> <p>Notably, this release:</p> <ul><li>Provides support for running Neo4j on Oracle and OpenJDK Java 8 runtimes</li> <li>Resolves an issue that prevented the Neo4j Browser from loading in the latest Chrome release (43.0.2357.65).</li> <li>Corrects the behavior of the <code>:sysinfo</code> (aka <code>:play sysinfo</code>) browser directive.</li> <li>Improves the <a href="http://neo4j.com/docs/2.2.2/import-tool.html">import tool</a> handling of values containing newlines, and adds support f...</li></ul><a href="http://neo4j.com/release-notes/neo4j-2-2-2/">Read full notes →</a> </div> ---- Now we need to loop through the rows and pull out just the version and release date. I wrote the following function to do this and strip out any extra text that we're not interested in: [source,r] ---- generate_releases = function(rows) { releases = data.frame() for(row in rows) { version = row %>% html_node("h3.entry-title") date = row %>% html_node("h6") if(!is.null(version) && !is.null(date)) { version = version %>% html_text() version = gsub("Latest Release: ", "", version) version = gsub("Neo4j ", "", version) releases = rbind(releases, data.frame(version = version, date = date %>% html_text())) } } return(releases) } > generate_releases(rows) version date 1 2.2.2 05/21/2015 2 2.2.1 04/14/2015 3 2.1.8 04/01/2015 4 2.2.0 03/25/2015 5 2.1.7 02/03/2015 6 2.1.6 11/25/2014 7 1.9.9 10/13/2014 8 2.1.5 09/30/2014 9 2.1.4 09/04/2014 10 2.1.3 07/28/2014 11 2.0.4 07/08/2014 12 1.9.8 06/19/2014 13 2.1.2 06/11/2014 14 2.0.3 04/30/2014 15 2.0.1 02/04/2014 16 2.0.2 04/15/2014 17 1.9.7 04/11/2014 18 1.9.6 02/03/2014 19 2.0 12/11/2013 20 1.9.5 11/11/2013 21 1.9.4 09/19/2013 22 1.9.3 08/30/2013 23 1.9.2 07/16/2013 24 1.9.1 06/24/2013 25 1.9 05/13/2013 26 1.8.3 // ---- Finally I wanted to convert the 'date' column to be in R date format and get rid of the 1.8.3 row since it doesn't contain a date. lubridate is my goto library for date manipulation in R so we'll use that here: [source,r] ---- library(lubridate) > generate_releases(rows) %>% mutate(date = mdy(date)) %>% filter(!is.na(date)) version date 1 2.2.2 2015-05-21 2 2.2.1 2015-04-14 3 2.1.8 2015-04-01 4 2.2.0 2015-03-25 5 2.1.7 2015-02-03 6 2.1.6 2014-11-25 7 1.9.9 2014-10-13 8 2.1.5 2014-09-30 9 2.1.4 2014-09-04 10 2.1.3 2014-07-28 11 2.0.4 2014-07-08 12 1.9.8 2014-06-19 13 2.1.2 2014-06-11 14 2.0.3 2014-04-30 15 2.0.1 2014-02-04 16 2.0.2 2014-04-15 17 1.9.7 2014-04-11 18 1.9.6 2014-02-03 19 2.0 2013-12-11 20 1.9.5 2013-11-11 21 1.9.4 2013-09-19 22 1.9.3 2013-08-30 23 1.9.2 2013-07-16 24 1.9.1 2013-06-24 25 1.9 2013-05-13 ---- We could then easily see how many releases there were by year: [source,r] ---- releasesByDate = generate_releases(rows) %>% mutate(date = mdy(date)) %>% filter(!is.na(date)) > releasesByDate %>% mutate(year = year(date)) %>% count(year) Source: local data frame [3 x 2] year n 1 2013 7 2 2014 13 3 2015 5 ---- Or by month: [source,r] ---- > releasesByDate %>% mutate(month = month(date)) %>% count(month) Source: local data frame [11 x 2] month n 1 2 3 2 3 1 3 4 5 4 5 2 5 6 3 6 7 3 7 8 1 8 9 3 9 10 1 10 11 2 11 12 1 ---- Previous to this quick bit of hacking I'd always turned to Ruby or Python whenever I wanted to scrape a dataset but it looks like rvest makes R a decent option for this type of work now. Good times!

r-scraping-neo4j-release-dates-with-rvest

https://markhneedham.com/blog/2015/06/21/r-scraping-neo4j-release-dates-with-rvest

2015-06-30 22:26:39

R: write.csv - unimplemented type 'list' in 'EncodeElement'

Everyone now and then I want to serialise an R data frame to a CSV file so I can easily load it up again if my R environment crashes without having to recalculate everything but recently ran into the following error: [source,r] ---- > write.csv(foo, "/tmp/foo.csv", row.names = FALSE) Error in .External2(C_writetable, x, file, nrow(x), p, rnames, sep, eol, : unimplemented type 'list' in 'EncodeElement' ---- If we take a closer look at the data frame in question it looks ok: [source,r] ---- > foo col1 col2 1 1 a 2 2 b 3 3 c ---- However, one of the columns contains a list in each cell and we need to find out which one it is. I've found the quickest way is to run the +++<cite>+++typeof+++</cite>+++ function over each column: [source,r] ---- > typeof(foo$col1) [1] "double" > typeof(foo$col2) [1] "list" ---- So 'col2' is the problem one which isn't surprising if you consider the way I created 'foo': [source,r] ---- library(dplyr) foo = data.frame(col1 = c(1,2,3)) %>% mutate(col2 = list("a", "b", "c")) ---- If we do have a list that we want to add to the data frame we need to convert it to a vector first so we don't run into this type of problem: [source,R] ---- foo = data.frame(col1 = c(1,2,3)) %>% mutate(col2 = list("a", "b", "c") %>% unlist()) ---- And now we can write to the CSV file: [source,r] ---- write.csv(foo, "/tmp/foo.csv", row.names = FALSE) ---- [source,bash] ---- $ cat /tmp/foo.csv "col1","col2" 1,"a" 2,"b" 3,"c" ---- And that's it!

r-write-csv-unimplemented-type-list-in-encodeelement

https://markhneedham.com/blog/2015/06/30/r-write-csv-unimplemented-type-list-in-encodeelement

2015-06-08 22:23:32

Unix: Converting a file of values into a comma separated list

I recently had a bunch of values in a file that I wanted to paste into a Java program which required a comma separated list of strings. This is what the file looked like: [source,bash] ---- $ cat foo2.txt | head -n 5 1.0 1.0 1.0 1.0 1.0 ---- And the idea is that we would end up with something like this: [source,text] ---- "1.0","1.0","1.0","1.0","1.0" ---- The first thing we need to do is quote each of the values. I found a http://www.unix.com/unix-for-dummies-questions-and-answers/138445-unix-command-insert-double-quotes-delimited-file.html[nice way to do this using sed]: [source,bash] ---- $ sed 's/.*/"&"/g' foo2.txt | head -n 5 "1.0" "1.0" "1.0" "1.0" "1.0" ---- Now that we've got all the values quoted we need to get rid of the new lines and replace them with commas. The way I'd normally do this is using 'tr' and then just not copy the final comma\... [source,bash] ---- $ sed 's/.*/"&"/g' foo2.txt | tr '\n' ',' "1.0","1.0","1.0","1.0","1.0","1.0","1.0","1.0","1.0","1.0","1.0","1.0","1.0","1.0","1.0","1.0","1.0","1.0","1.0","1.0","1.0","1.0","1.0","1.0","1.0","1.0","1.0","1.0","1.0","1.0","1.0","1.0","1.0","1.0","1.0","1.0","1.0","1.0","1.0","1.0","1.0","1.0","1.0","1.0","1.0","1.0","1.0","1.0","1.0","1.0","1.0","1.0","1.0","1.0","1.0","1.0","1.0","1.0","1.0","1.0","1.0","1.0","1.0","1.0","1.0","1.0","1.0","1.0","1.0","1.0","1.0","1.0","1.0","1.0","1.0","1.0","1.0","1.0","1.0","1.0","1.0","1.0","1.0","1.0","1.0","1.0","1.0","1.0","1.0","1.0","1.0","1.0","1.0","1.0","1.0","1.0","1.0","1.0","1.0","1.0","1.0","1.0","1.0","1.0","1.0","1.0","1.0","1.0","1.0","1.0","1.0","1.0","1.0","1.0","1.0","1.0","1.0","1.0","1.0","1.0","1.0","1.0","1.0","1.0","1.0","1.0","1.0","1.0","1.0","1.0","1.0","1.0","1.0","1.0","1.0","1.0","1.0","1.0","1.0","1.0","1.0","1.0","1.0","1.0","1.0","1.0","1.0","1.0","1.0","1.0","1.0","1.0","1.0","1.0","1.0","1.0", ---- \...but I learnt that we can actually do one better than this using 'paste' which allows you to http://unix.stackexchange.com/questions/114244/replace-all-newlines-to-space-except-the-last[replace new lines excluding the last one]. The only annoying thing about paste is that you can't pipe to it so we need to use process substitution instead: ~~~bash $ paste -s -d ',' <(sed 's/.*/"&"/g' foo2.txt) "1.0","1.0","1.0","1.0","1.0","1.0","1.0","1.0","1.0","1.0","1.0","1.0","1.0","1.0","1.0","1.0","1.0","1.0","1.0","1.0","1.0","1.0","1.0","1.0","1.0","1.0","1.0","1.0","1.0","1.0","1.0","1.0","1.0","1.0","1.0","1.0","1.0","1.0","1.0","1.0","1.0","1.0","1.0","1.0","1.0","1.0","1.0","1.0","1.0","1.0","1.0","1.0","1.0","1.0","1.0","1.0","1.0","1.0","1.0","1.0","1.0","1.0","1.0","1.0","1.0","1.0","1.0","1.0","1.0","1.0","1.0","1.0","1.0","1.0","1.0","1.0","1.0","1.0","1.0","1.0","1.0","1.0","1.0","1.0","1.0","1.0","1.0","1.0","1.0","1.0","1.0","1.0","1.0","1.0","1.0","1.0","1.0","1.0","1.0","1.0","1.0","1.0","1.0","1.0","1.0","1.0","1.0","1.0","1.0","1.0","1.0","1.0","1.0","1.0","1.0","1.0","1.0","1.0","1.0","1.0","1.0","1.0","1.0","1.0","1.0","1.0","1.0","1.0","1.0","1.0","1.0","1.0","1.0","1.0","1.0","1.0","1.0","1.0","1.0","1.0","1.0","1.0","1.0","1.0","1.0","1.0","1.0","1.0","1.0","1.0","1.0","1.0","1.0","1.0","1.0","1.0" ~~~ If we're only a Mac we could even automate the copy/paste step too by piping to 'pbcopy': ~~~bash $ paste -s -d ',' <(sed 's/.*/"&"/g' foo2.txt) | pbcopy ~~~

unix-converting-a-file-of-values-into-a-comma-separated-list

https://markhneedham.com/blog/2015/06/08/unix-converting-a-file-of-values-into-a-comma-separated-list

2015-06-23 22:34:47

R: Scraping the release dates of github projects

Continuing on from my blog post about http://www.markhneedham.com/blog/2015/06/21/r-scraping-neo4j-release-dates-with-rvest/[scraping Neo4j's release dates] I thought it'd be even more interesting to chart the release dates of some github projects. In theory the release dates should be accessible through the github API but the few that I looked at weren't returning any data so I scraped the data together. We'll be using rvest again and I first wrote the following function to extract the release versions and dates from a single page: [source,r] ---- library(dplyr) library(rvest) process_page = function(releases, session) { rows = session %>% html_nodes("ul.release-timeline-tags li") for(row in rows) { date = row %>% html_node("span.date") version = row %>% html_node("div.tag-info a") if(!is.null(version) && !is.null(date)) { date = date %>% html_text() %>% str_trim() version = version %>% html_text() %>% str_trim() releases = rbind(releases, data.frame(date = date, version = version)) } } return(releases) } ---- Let's try it out on the https://github.com/apache/cassandra/releases[Cassandra release page] and see what it comes back with: [source,r] ---- > r = process_page(data.frame(), html_session("https://github.com/apache/cassandra/releases")) > r date version 1 Jun 22, 2015 cassandra-2.1.7 2 Jun 22, 2015 cassandra-2.0.16 3 Jun 8, 2015 cassandra-2.1.6 4 Jun 8, 2015 cassandra-2.2.0-rc1 5 May 19, 2015 cassandra-2.2.0-beta1 6 May 18, 2015 cassandra-2.0.15 7 Apr 29, 2015 cassandra-2.1.5 8 Apr 1, 2015 cassandra-2.0.14 9 Apr 1, 2015 cassandra-2.1.4 10 Mar 16, 2015 cassandra-2.0.13 ---- That works pretty well but it's only one page! To get all the pages we can use the +++<cite>+++follow_link+++</cite>+++ function to follow the 'Next' link until there aren't anymore pages to process. We end up with the following function to do this: [source,r] ---- find_all_releases = function(starting_page) { s = html_session(starting_page) releases = data.frame() next_page = TRUE while(next_page) { possibleError = tryCatch({ releases = process_page(releases, s) s = s %>% follow_link("Next") }, error = function(e) { e }) if(inherits(possibleError, "error")){ next_page = FALSE } } return(releases) } ---- Let's try it out starting from the Cassandra page: [source,r] ---- > cassandra = find_all_releases("https://github.com/apache/cassandra/releases") Navigating to https://github.com/apache/cassandra/releases?after=cassandra-2.0.13 Navigating to https://github.com/apache/cassandra/releases?after=cassandra-2.0.10 Navigating to https://github.com/apache/cassandra/releases?after=cassandra-2.0.8 Navigating to https://github.com/apache/cassandra/releases?after=cassandra-1.2.13 Navigating to https://github.com/apache/cassandra/releases?after=cassandra-2.0.0-rc1 Navigating to https://github.com/apache/cassandra/releases?after=cassandra-1.2.3 Navigating to https://github.com/apache/cassandra/releases?after=cassandra-1.2.0-beta2 Navigating to https://github.com/apache/cassandra/releases?after=cassandra-1.0.10 Navigating to https://github.com/apache/cassandra/releases?after=cassandra-1.0.6 Navigating to https://github.com/apache/cassandra/releases?after=cassandra-1.0.0-rc2 Navigating to https://github.com/apache/cassandra/releases?after=cassandra-0.7.7 Navigating to https://github.com/apache/cassandra/releases?after=cassandra-0.7.4 Navigating to https://github.com/apache/cassandra/releases?after=cassandra-0.7.0-rc3 Navigating to https://github.com/apache/cassandra/releases?after=cassandra-0.6.4 Navigating to https://github.com/apache/cassandra/releases?after=cassandra-0.5.0-rc3 Navigating to https://github.com/apache/cassandra/releases?after=cassandra-0.4.0-final > cassandra %>% sample_n(10) date version 151 Mar 13, 2010 cassandra-0.5.0-rc2 25 Jul 3, 2014 cassandra-1.2.18 51 Jul 27, 2013 cassandra-1.2.8 21 Aug 19, 2014 cassandra-2.1.0-rc6 73 Sep 24, 2012 cassandra-1.2.0-beta1 158 Mar 13, 2010 cassandra-0.4.0-rc2 113 May 20, 2011 cassandra-0.7.6-2 15 Oct 24, 2014 cassandra-2.1.1 103 Sep 15, 2011 cassandra-1.0.0-beta1 93 Nov 29, 2011 cassandra-1.0.4 ---- I want to plot when the different releases happened in time and in order to do that we need to create an extra column containing the 'release series' which we can do with the following transformation: [source,r] ---- series = function(version) { parts = strsplit(as.character(version), "\\.") return(unlist(lapply(parts, function(p) paste(p %>% unlist %>% head(2), collapse = ".")))) } bySeries = cassandra %>% mutate(date2 = mdy(date), series = series(version), short_version = gsub("cassandra-", "", version), short_series = series(short_version)) > bySeries %>% sample_n(10) date version date2 series short_version short_series 3 Jun 8, 2015 cassandra-2.1.6 2015-06-08 cassandra-2.1 2.1.6 2.1 161 Mar 13, 2010 cassandra-0.4.0-beta1 2010-03-13 cassandra-0.4 0.4.0-beta1 0.4 62 Feb 15, 2013 cassandra-1.1.10 2013-02-15 cassandra-1.1 1.1.10 1.1 153 Mar 13, 2010 cassandra-0.5.0-beta2 2010-03-13 cassandra-0.5 0.5.0-beta2 0.5 37 Feb 7, 2014 cassandra-2.0.5 2014-02-07 cassandra-2.0 2.0.5 2.0 36 Feb 7, 2014 cassandra-1.2.15 2014-02-07 cassandra-1.2 1.2.15 1.2 29 Jun 2, 2014 cassandra-2.1.0-rc1 2014-06-02 cassandra-2.1 2.1.0-rc1 2.1 21 Aug 19, 2014 cassandra-2.1.0-rc6 2014-08-19 cassandra-2.1 2.1.0-rc6 2.1 123 Feb 16, 2011 cassandra-0.7.2 2011-02-16 cassandra-0.7 0.7.2 0.7 135 Nov 1, 2010 cassandra-0.7.0-beta3 2010-11-01 cassandra-0.7 0.7.0-beta3 0.7 ---- Now let's plot those releases and see what we get: [source,R] ---- ggplot(aes(x = date2, y = short_series), data = bySeries %>% filter(!grepl("beta|rc", short_version))) + geom_text(aes(label=short_version),hjust=0.5, vjust=0.5, size = 4, angle = 90) + theme_bw() ---- image::{{<siteurl>}}/uploads/2015/06/2015-06-23_22-59-19.png[2015 06 23 22 59 19,598] An interesting thing we can see from this visualisation is what overlap the various series of versions have. Most of the time there are only two series of versions overlapping but the 1.2, 2.0 and 2.1 series all overlap which is unusual. In this chart we excluded all beta and RC versions. Let's bring those back in and just show the last 3 versions: [source,r] ---- ggplot(aes(x = date2, y = short_series), data = bySeries %>% filter(grepl("2\\.[012]\\.|1\\.2\\.", short_version))) + geom_text(aes(label=short_version),hjust=0.5, vjust=0.5, size = 4, angle = 90) + theme_bw() ---- image::{{<siteurl>}}/uploads/2015/06/2015-06-23_23-08-04.png[2015 06 23 23 08 04,599] From this chart it's clearer that the 2.0 and 2.1 series have recent releases so there will probably be three overlapping versions when the 2.2 series is released as well. The chart is still a bit cluttered although less than before. I'm not sure of a better way of visualising this type of data so if you have any ideas do let me know!

r-scraping-the-release-dates-of-github-projects

https://markhneedham.com/blog/2015/06/23/r-scraping-the-release-dates-of-github-projects

2015-06-15 22:53:33

Northwind: Finding direct/transitive Reports in SQL and Neo4j's Cypher

Every few months we run a http://www.meetup.com/graphdb-london/events/222398828/[relational to graph meetup] at the Neo London office where we go through how to take your data from a relational database and into the graph. We use the https://code.google.com/p/northwindextended/downloads/detail?name=northwind.postgre.sql[Northwind dataset] which often comes as a demo dataset on relational databases and come up with some queries which seem graph in nature. My favourite query is one which finds out how employees are organised and who reports to whom. I thought it'd be quite interesting to see what it would look like in Postgres SQL as well, just for fun. We'll start off by getting a list of employees and the person they report to: [source,sql] ---- SELECT e."EmployeeID", e."ReportsTo" FROM employees AS e WHERE e."ReportsTo" is not null; EmployeeID | ReportsTo ------------+----------- 1 | 2 3 | 2 4 | 2 5 | 2 6 | 5 7 | 5 8 | 2 9 | 5 (8 rows) ---- In cypher we'd do this: [source,cypher] ---- MATCH (e:Employee)<-[:REPORTS_TO]-(sub) RETURN sub.EmployeeID, e.EmployeeID +-------------------------------+ | sub.EmployeeID | e.EmployeeID | +-------------------------------+ | "4" | "2" | | "5" | "2" | | "1" | "2" | | "3" | "2" | | "8" | "2" | | "9" | "5" | | "6" | "5" | | "7" | "5" | +-------------------------------+ 8 rows ---- Next let's find the big boss who doesn't report to anyone. First in SQL: [source,sql] ---- SELECT e."EmployeeID" AS bigBoss FROM employees AS e WHERE e."ReportsTo" is null bigboss --------- 2 (1 row) ---- And now cypher: [source,cypher] ---- MATCH (e:Employee) WHERE NOT (e)-[:REPORTS_TO]->() RETURN e.EmployeeID AS bigBoss +---------+ | bigBoss | +---------+ | "2" | +---------+ 1 row ---- We still don't need to join anything so the query isn't that interesting yet. Let's bring in some more properties from the manager record so we have to self join on the employees table: [source,sql] ---- SELECT e."FirstName", e."LastName", e."Title", manager."FirstName", manager."LastName", manager."Title" FROM employees AS e JOIN employees AS manager ON e."ReportsTo" = manager."EmployeeID" WHERE e."ReportsTo" is not null FirstName | LastName | Title | FirstName | LastName | Title -----------+-----------+--------------------------+-----------+----------+----------------------- Nancy | Davolio | Sales Representative | Andrew | Fuller | Vice President, Sales Janet | Leverling | Sales Representative | Andrew | Fuller | Vice President, Sales Margaret | Peacock | Sales Representative | Andrew | Fuller | Vice President, Sales Steven | Buchanan | Sales Manager | Andrew | Fuller | Vice President, Sales Michael | Suyama | Sales Representative | Steven | Buchanan | Sales Manager Robert | King | Sales Representative | Steven | Buchanan | Sales Manager Laura | Callahan | Inside Sales Coordinator | Andrew | Fuller | Vice President, Sales Anne | Dodsworth | Sales Representative | Steven | Buchanan | Sales Manager (8 rows) ---- [source,cypher] ---- MATCH (e:Employee)<-[:REPORTS_TO]-(sub) RETURN sub.FirstName, sub.LastName, sub.Title, e.FirstName, e.LastName, e.Title +----------------------------------------------------------------------------------------------------------------+ | sub.FirstName | sub.LastName | sub.Title | e.FirstName | e.LastName | e.Title | +----------------------------------------------------------------------------------------------------------------+ | "Margaret" | "Peacock" | "Sales Representative" | "Andrew" | "Fuller" | "Vice President, Sales" | | "Steven" | "Buchanan" | "Sales Manager" | "Andrew" | "Fuller" | "Vice President, Sales" | | "Nancy" | "Davolio" | "Sales Representative" | "Andrew" | "Fuller" | "Vice President, Sales" | | "Janet" | "Leverling" | "Sales Representative" | "Andrew" | "Fuller" | "Vice President, Sales" | | "Laura" | "Callahan" | "Inside Sales Coordinator" | "Andrew" | "Fuller" | "Vice President, Sales" | | "Anne" | "Dodsworth" | "Sales Representative" | "Steven" | "Buchanan" | "Sales Manager" | | "Michael" | "Suyama" | "Sales Representative" | "Steven" | "Buchanan" | "Sales Manager" | | "Robert" | "King" | "Sales Representative" | "Steven" | "Buchanan" | "Sales Manager" | +----------------------------------------------------------------------------------------------------------------+ 8 rows ---- Now let's see how many direct reports each manager has: [source,sql] ---- SELECT manager."EmployeeID" AS manager, COUNT(e."EmployeeID") AS reports FROM employees AS manager LEFT JOIN employees AS e ON e."ReportsTo" = manager."EmployeeID" GROUP BY manager ORDER BY reports DESC; manager | reports ---------+--------- 2 | 5 5 | 3 1 | 0 3 | 0 4 | 0 9 | 0 6 | 0 7 | 0 8 | 0 (9 rows) ---- [source,cypher] ---- MATCH (e:Employee) OPTIONAL MATCH (e)<-[rel:REPORTS_TO]-(report) RETURN e.EmployeeID AS employee, COUNT(rel) AS reports +--------------------+ | employee | reports | +--------------------+ | "2" | 5 | | "5" | 3 | | "8" | 0 | | "7" | 0 | | "1" | 0 | | "4" | 0 | | "6" | 0 | | "9" | 0 | | "3" | 0 | +--------------------+ 9 rows ---- Things start to get more interesting if we find the transitive reporting relationships that exist. I'm not an expert at Postgres but one way to achieve this is by writing http://www.postgresql.org/docs/current/static/queries-with.html[a recursive WITH query] like so: [source,sql] ---- WITH RECURSIVE recursive_employees("EmployeeID", "ReportsTo") AS ( SELECT e."EmployeeID", e."ReportsTo" FROM employees e UNION ALL SELECT e."EmployeeID", e."ReportsTo" FROM employees e, recursive_employees re WHERE e."EmployeeID" = re."ReportsTo" ) SELECT re."ReportsTo", COUNT(*) AS count FROM recursive_employees AS re WHERE re."ReportsTo" IS NOT NULL GROUP BY re."ReportsTo"; ReportsTo | count -----------+------- 2 | 8 5 | 3 (2 rows) ---- If there's a simpler way let me know in the comments. In cypher we only need to add one character, '*', after the 'REPORTS_TO' relationship to get it to recurse as far as it can. We'll also remove the 'OPTIONAL MATCH' so that we only get back people who have people reporting to them: [source,cypher] ---- MATCH (e:Employee)<-[rel:REPORTS_TO*]-(report) RETURN e.EmployeeID AS employee, COUNT(rel) AS reports +--------------------+ | employee | reports | +--------------------+ | "2" | 8 | | "5" | 3 | +--------------------+ 2 rows ---- Now I need to find some relational datasets with more complicated queries to play around with. If you have any ideas do let me know.

northwind-finding-directtransitive-reports-in-sql-and-neo4js-cypher

https://markhneedham.com/blog/2015/06/15/northwind-finding-directtransitive-reports-in-sql-and-neo4js-cypher

2015-06-12 23:12:32

The Willpower Instinct: Reducing time spent mindlessly scrolling for things to read

I recently finished reading Kelly McGonigal's excellent book 'http://www.amazon.co.uk/The-Willpower-Instinct-Kelly-McGonigal/dp/1583335080[The Willpower Instinct]' having previously watched her https://www.youtube.com/watch?v=V5BXuZL1HAg[Google talk of the same title] My main takeaway from the book is that there are things that we want to do (or not do) but doing them (or not as the case may be) isn't necessarily instinctive and so we need to develop some strategies to help ourselves out. In one of the early chapters she suggests picking a habit that you want to do less off and write down on a piece of paper every time you want to do it and how you're feeling at that point. After writing it down you're free to then follow through and do it but you don't have to if you change your mind. I was quite aware of the fact that I spend a lot of time idly scrolling from email to Twitter to Facebook to LinkedIn to news websites and back again so I thought it'd be interesting to track when/why I was doing this. The annoying thing about this habit is that it can *easily eat up 20-30 minutes at a time without you even noticing*. I've been tracking myself for about three weeks and in the first few days I noticed that the first thing I did as soon as I woke up was grab my phone and get into the cycle. It was quite frustrating to be lured in so early in the day but one of the suggestions in the book is that feeling guilty about something is actually detrimental to our progress. Instead we should note why it happened and then move on - the day isn't a write off because of one event! Kelly suggests that if we can work out the times when we're most likely to fall into our habits then we can pre-plan a mitigation strategy. From looking over my notes the following are the reasons why I want to start mindlessly scrolling: * I'm stuck on the problem I'm working on * I'm bored * I'm tired * I'm hungry * I'm getting distracted by notifications * I want to not think for a while The notifications bullet is easy to address - I turn off notifications on my phone for 4 hours at a time so I don't even know there's anything to read. I was intrigued to note that I got distracted when stuck on a problem - the main take away here is to check whether the urge to scroll mindlessly is being driven by having to think hard. If it is then I can choose to either get back to it or go for a short walk and then come back. But definitely don't start scrolling! I often find myself bored on my commute to work so I've addressed this by working out a book/paper I'm going to read the night before and then having that ready for the journey. Lunch time is prime time for mindless scrolling as well so I've filled that time with various https://www.youtube.com/watch?v=9MKY4KypBzg[computer] https://www.youtube.com/watch?v=KSdNYi55kjg[science]/http://www.thetalkingmachines.com/blog/2015/6/4/the-economic-impact-of-machine-learning-and-using-the-kernel-trick-to-dig-in-to-big-data[data] http://www.thetalkingmachines.com/blog/2015/3/13/how-machine-learning-got-where-it-is-and-the-future-of-the-field[science] http://www.partiallyderivative.com/news/2015/6/8/episode-24-can-killer-robots-marry-their-cousins[videos]. Since I started tracking my scrolling I've found myself sleeping earlier so my assumption is that the extra hours awake were being spent mindlessly scrolling which led to being more tired so a win all around in that respect. Something I've noticed is that I'm sometimes wasting time on other activities which I'm are not 'forbidden' but are equally unconstructive e.g. chat applications / watching music videos. The former are obviously useful for communicating with people so I've been trying to use them only when I actually want to chat to someone rather than mindlessly looking for messages to read. I also find myself not wanting to write down the times I've mindlessly scrolled when I'm doing it a lot on a given day. Being aware of this is helpful as I just write it down anyway and get on with the day. The summary of my experience so far is it seems beneficial - I don't think I've lost anything by not checking those mediums so often and I've definitely read a lot more than I usually do and been more focused as well. Now I need to go and try out some of the other exercises from the book - if you've read it / tried out any of the tips I'd love to hear what's worked well for you.

the-willpower-instinct-reducing-time-spent-mindlessly-scrolling-for-things-to-read

https://markhneedham.com/blog/2015/06/12/the-willpower-instinct-reducing-time-spent-mindlessly-scrolling-for-things-to-read

2015-06-25 23:14:51

R: Scraping Wimbledon draw data

Given Wimbledon starts next week I wanted to find a data set to explore before it gets underway. Having searched around and failed to find one I had to resort to scraping the http://www.atpworldtour.com/en/scores/archive/wimbledon/540/2013/results[ATP World Tour's event page] which displays the matches in an easy to access format. We'll be using the Wimbledon 2013 draw since Andy Murray won that year! This is what the page looks like: image::{{<siteurl>}}/uploads/2015/06/2015-06-25_23-47-16.png[2015 06 25 23 47 16,599] Each match is in its own row of a table and each column has a class attribute which makes it really easy to scrape. We'll be using R's rvest again. I wrote the following script which grabs the player names, seedings and score of the match and stores everything in a data frame: [source,r] ---- library(rvest) library(dplyr) library(stringr) s = html_session("http://www.atpworldtour.com/en/scores/archive/wimbledon/540/2013/results") rows = s %>% html_nodes("div#scoresResultsContent tr") matches = data.frame() for(row in rows) { players = row %>% html_nodes("td.day-table-name a") seedings = row %>% html_nodes("td.day-table-seed") score = row %>% html_node("td.day-table-score a") if(!is.null(score)) { player1 = players[1] %>% html_text() %>% str_trim() seeding1 = ifelse(!is.na(seedings[1]), seedings[1] %>% html_node("span") %>% html_text() %>% str_trim(), NA) player2 = players[2] %>% html_text() %>% str_trim() seeding2 = ifelse(!is.na(seedings[2]), seedings[2] %>% html_node("span") %>% html_text() %>% str_trim(), NA) matches = rbind(data.frame(winner = player1, winner_seeding = seeding1, loser = player2, loser_seeding = seeding2, score = score %>% html_text() %>% str_trim(), round = round), matches) } else { round = row %>% html_node("th") %>% html_text() } } ---- This is what the data frame looks like: [source,r] ---- > matches %>% sample_n(10) winner winner_seeding loser loser_seeding score round 61 Wayne Odesnik (4) Thiago Alves <NA> 61 64 1st Round Qualifying 4 Danai Udomchoke <NA> Marton Fucsovics <NA> 61 57 1210 1st Round Qualifying 233 Jerzy Janowicz (24) Lukasz Kubot <NA> 75 64 64 Quarter-Finals 90 Malek Jaziri <NA> Illya Marchenko (9) 674 75 64 2nd Round Qualifying 222 David Ferrer (4) Alexandr Dolgopolov (26) 676 762 26 61 62 Round of 32 54 Michal Przysiezny (11) Dusan Lojda <NA> 26 63 62 1st Round Qualifying 52 Go Soeda (13) Nikola Mektic <NA> 62 60 1st Round Qualifying 42 Ruben Bemelmans (23) Jonathan Dasnieres de Veigy <NA> 63 64 1st Round Qualifying 31 Mirza Basic <NA> Tsung-Hua Yang <NA> 674 33 (RET) 1st Round Qualifying 179 Jurgen Melzer <NA> Julian Reister (Q) 36 762 765 62 Round of 64 ---- It also contains qualifying matches which I'm not so interested in. Let's strip those out: [source,r] ---- main_matches = matches %>% filter(!grepl("Qualifying", round)) %>% mutate(year = 2013) ---- We'll also put a column in for 'year' so that we can handle the draws for multiple years later on. Next I wanted to clean up the data a bit. I'd like to be able to do some queries based on the seedings of the players but at the moment that column contains numeric brackets in values as well as some other values which indicate whether a player is a qualifier, lucky loser or wildcard entry. I started by adding a column to store this extra information: [source,r] ---- main_matches$winner_type = NA main_matches$winner_type[main_matches$winner_seeding == "(WC)"] = "wildcard" main_matches$winner_type[main_matches$winner_seeding == "(Q)"] = "qualifier" main_matches$winner_type[main_matches$winner_seeding == "(LL)"] = "lucky loser" main_matches$loser_type = NA main_matches$loser_type[main_matches$loser_seeding == "(WC)"] = "wildcard" main_matches$loser_type[main_matches$loser_seeding == "(Q)"] = "qualifier" main_matches$loser_type[main_matches$loser_seeding == "(LL)"] = "lucky loser" ---- And then I cleaned up the existing column: [source,r] ---- tidy_seeding = function(seeding) { no_brackets = gsub("\\(|\\)", "", seeding) return(gsub("WC|Q|L", NA, no_brackets)) } main_matches = main_matches %>% mutate(winner_seeding = as.numeric(tidy_seeding(winner_seeding)), loser_seeding = as.numeric(tidy_seeding(loser_seeding))) ---- Now we can write a query against the data frame to find out when the underdog won i.e. a player with no seeding beat a player with a seeding or a lower seeded player beat a higher seeded one: [source,r] ---- > main_matches %>% filter((winner_seeding > loser_seeding) | (is.na(winner_seeding) & !is.na(loser_seeding))) winner winner_seeding loser loser_seeding score round year 1 Jurgen Melzer NA Fabio Fognini 30 675 75 63 62 Round of 128 2013 2 Bernard Tomic NA Sam Querrey 21 766 763 36 26 63 Round of 128 2013 3 Feliciano Lopez NA Gilles Simon 19 62 64 7611 Round of 128 2013 4 Ivan Dodig NA Philipp Kohlschreiber 16 46 676 763 63 21 (RET) Round of 128 2013 5 Viktor Troicki NA Janko Tipsarevic 14 63 64 765 Round of 128 2013 6 Lleyton Hewitt NA Stan Wawrinka 11 64 75 63 Round of 128 2013 7 Steve Darcis NA Rafael Nadal 5 764 768 64 Round of 128 2013 8 Fernando Verdasco NA Julien Benneteau 31 761 764 64 Round of 64 2013 9 Grega Zemlja NA Grigor Dimitrov 29 36 764 36 64 119 Round of 64 2013 10 Adrian Mannarino NA John Isner 18 11 (RET) Round of 64 2013 11 Igor Sijsling NA Milos Raonic 17 75 64 764 Round of 64 2013 12 Kenny De Schepper NA Marin Cilic 10 (W/O) Round of 64 2013 13 Ernests Gulbis NA Jo-Wilfried Tsonga 6 36 63 63 (RET) Round of 64 2013 14 Sergiy Stakhovsky NA Roger Federer 3 675 765 75 765 Round of 64 2013 15 Lukasz Kubot NA Benoit Paire 25 61 63 64 Round of 32 2013 16 Kenny De Schepper NA Juan Monaco 22 64 768 64 Round of 32 2013 17 Jerzy Janowicz 24 Nicolas Almagro 15 766 63 64 Round of 32 2013 18 Andreas Seppi 23 Kei Nishikori 12 36 62 674 61 64 Round of 32 2013 19 Bernard Tomic NA Richard Gasquet 9 767 57 75 765 Round of 32 2013 20 Juan Martin Del Potro 8 David Ferrer 4 62 64 765 Quarter-Finals 2013 21 Andy Murray 2 Novak Djokovic 1 64 75 64 Finals 2013 ---- There are actually very few times when a lower seeded player beat a higher seeded one but there are quite a few instances of non seeds beating seeds. We've got 21 occurrences of underdogs winning out of a total of 127 matches. Let's filter that set of rows and see which seeds lost in the first round: [source,r] ---- > main_matches %>% filter(round == "Round of 128" & !is.na(loser_seeding)) winner winner_seeding loser loser_seeding score round year 1 Jurgen Melzer NA Fabio Fognini 30 675 75 63 62 Round of 128 2013 2 Bernard Tomic NA Sam Querrey 21 766 763 36 26 63 Round of 128 2013 3 Feliciano Lopez NA Gilles Simon 19 62 64 7611 Round of 128 2013 4 Ivan Dodig NA Philipp Kohlschreiber 16 46 676 763 63 21 (RET) Round of 128 2013 5 Viktor Troicki NA Janko Tipsarevic 14 63 64 765 Round of 128 2013 6 Lleyton Hewitt NA Stan Wawrinka 11 64 75 63 Round of 128 2013 7 Steve Darcis NA Rafael Nadal 5 764 768 64 Round of 128 2013 ---- Rafael Nadal is the most prominent but Stan Wawrinka also lost in the first round that year which I'd forgotten about! Next let's make the 'round' column an ordered factor one so that we can sort matches by round: [source,r] ---- main_matches$round = factor(main_matches$round, levels = c("Round of 128", "Round of 64", "Round of 32", "Round of 16", "Quarter-Finals", "Semi-Finals", "Finals")) > main_matches$round ... Levels: Round of 128 Round of 64 Round of 32 Round of 16 Quarter-Finals Semi-Finals Finals ---- We can now really easily work out which unseeded players went the furthest in the tournament: [source,r] ---- > main_matches %>% filter(is.na(loser_seeding)) %>% arrange(desc(round)) %>% head(5) winner winner_seeding loser loser_seeding score round year 1 Jerzy Janowicz 24 Lukasz Kubot NA 75 64 64 Quarter-Finals 2013 2 Andy Murray 2 Fernando Verdasco NA 46 36 61 64 75 Quarter-Finals 2013 3 Fernando Verdasco NA Kenny De Schepper NA 64 64 64 Round of 16 2013 4 Lukasz Kubot NA Adrian Mannarino NA 46 63 36 63 64 Round of 16 2013 5 Jerzy Janowicz 24 Jurgen Melzer NA 36 761 64 46 64 Round of 16 2013 ---- Next up I thought it'd be cool to write a function which showed which round each player exited in: [source,r] ---- round_reached = function(player, main_matches) { furthest_match = main_matches %>% filter(winner == player | loser == player) %>% arrange(desc(round)) %>% head(1) return(ifelse(furthest_match$winner == player, "Winner", as.character(furthest_match$round))) } ---- Our function isn't vectorisable - it only works if we pass in a single player at a time so we'll have to group the data frame by player before calling it. Let's check it works by seeing how far Andy Murray and Rafael Nadal got: [source,r] ---- > round_reached("Rafael Nadal", main_matches) [1] "Round of 128" > round_reached("Andy Murray", main_matches) [1] "Winner" ---- Great. What about if we try it against each of the top 8 seeds? [source,r] ---- > rbind(main_matches %>% filter(winner_seeding %in% 1:8) %>% mutate(name = winner, seeding = winner_seeding), main_matches %>% filter(loser_seeding %in% 1:8) %>% mutate(name = loser, seeding = loser_seeding)) %>% select(name, seeding) %>% distinct() %>% arrange(seeding) %>% group_by(name) %>% mutate(round_reached = round_reached(name, main_matches)) Source: local data frame [8 x 3] Groups: name name seeding round_reached 1 Novak Djokovic 1 Finals 2 Andy Murray 2 Winner 3 Roger Federer 3 Round of 64 4 David Ferrer 4 Quarter-Finals 5 Rafael Nadal 5 Round of 128 6 Jo-Wilfried Tsonga 6 Round of 64 7 Tomas Berdych 7 Quarter-Finals 8 Juan Martin Del Potro 8 Semi-Finals ---- Neat. Next up I want to do a comparison between the round they reached and the round you'd expect them to get to given their seeding but that's for the weekend! I've put a CSV file containing all the data in https://gist.github.com/mneedham/8c2b46d960c2208e9961[this gist] in case you want to play with it. I'm planning to scrape a few more years worth of data before Monday and add in some extra fields as well but in case I don't get around to it the https://gist.github.com/mneedham/8c2b46d960c2208e9961#file-wimbledon-r[full script in this blog post is included in the gist] as well so feel free to tweak it if tennis is your thing.

r-scraping-wimbledon-draw-data

https://markhneedham.com/blog/2015/06/25/r-scraping-wimbledon-draw-data

2015-12-04 07:52:34

Neo4j: Facts as nodes

On Tuesday I https://skillsmatter.com/skillscasts/7298-modelling-a-recommendation-engine-a-worked-example[spoke] at the Neo4j London user group about http://www.meetup.com/graphdb-london/events/226721630/[incrementally building a recommendation engine] and described the 'facts as nodes' modeling pattern, defined as follows in the http://graphdatabases.com/[Graph Databases book]: ____ When two or more domain entities interact for a period of time, a fact emerges. We represent a fact as a separate node with connections to each of the entities engaged in that fact. Modeling an action in terms of its product--that is, in terms of the thing that results from the action--produces a similar structure: an intermediate node that represents the outcome of an interaction between two or more entities. ____ We started with the following model describing a meetup member and the groups they've joined: image::{{<siteurl>}}/uploads/2015/12/2015-12-04_07-26-11.png[2015 12 04 07 26 11,597] This model works well for the query it was defined for - find groups similar to ones that I'm already a member of: [source,cypher] ---- MATCH (member:Member {name: "Mark Needham"})-[:MEMBER_OF]->(group)-[:HAS_TOPIC]->(topic) WITH member, topic, COUNT(*) AS score MATCH (topic)<-[:HAS_TOPIC]-(otherGroup) WHERE NOT (member)-[:MEMBER_OF]->(otherGroup) RETURN otherGroup.name, COLLECT(topic.name), SUM(score) as score ORDER BY score DESC ---- Prefixing that query with the 'PROFILE' keyword yields a query plan and the following summary text: [source,text] ---- Cypher version: CYPHER 2.3, planner: COST. 89100 total db hits in 113 ms. ---- In this model it feels like there is a +++<cite>+++membership+++</cite>+++ fact waiting to become a node. image::{{<siteurl>}}/uploads/2015/12/2015-12-04_07-35-38.png[2015 12 04 07 35 38,598] We can refactor towards that model with the following query: [source,cypher] ---- MATCH (member:Member)-[rel:MEMBER_OF]->(group) MERGE (membership:Membership {id: member.id + "_" + group.id}) SET membership.joined = rel.joined MERGE (member)-[:HAS_MEMBERSHIP]->(membership) MERGE (membership)-[:OF_GROUP]->(group); ---- We'd answer our initial question with the following query: [source,cypher] ---- MATCH (member:Member {name: "Mark Needham"})-[:HAS_MEMBERSHIP]->()-[:OF_GROUP]->(group:Group)-[:HAS_TOPIC]->(topic) WITH member, topic, COUNT(*) AS score MATCH (topic)<-[:HAS_TOPIC]-(otherGroup) WHERE NOT (member)-[:HAS_MEMBERSHIP]->(:Membership)-[:OF_GROUP]->(otherGroup:Group) RETURN otherGroup.name, COLLECT(topic.name), SUM(score) as score ORDER BY score DESC ---- at the following cost: [source,text] ---- Cypher version: CYPHER 2.3, planner: COST. 468201 total db hits in 346 ms. ---- The membership node hasn't proved its value yet - it does 4x more work to get the same result. However, the next question we want to answer is 'what group do people join after the Neo4j user group?' where it might come in handy. First we'll add a 'NEXT' relationship between a user's adjacent group memberships by writing the following query: [source,cypher] ---- MATCH (member:Member)-[:HAS_MEMBERSHIP]->(membership) WITH member, membership ORDER BY member.id, membership.joined WITH member, COLLECT(membership) AS memberships UNWIND RANGE(0,SIZE(memberships) - 2) as idx WITH memberships[idx] AS m1, memberships[idx+1] AS m2 MERGE (m1)-[:NEXT]->(m2); ---- And now for the query: [source,cypher] ---- MATCH (group:Group {name: "Neo4j - London User Group"})<-[:OF_GROUP]-(membership)-[:NEXT]->(nextMembership), (membership)<-[:HAS_MEMBERSHIP]-(member:Member)-[:HAS_MEMBERSHIP]->(nextMembership), (nextMembership)-[:OF_GROUP]->(nextGroup) RETURN nextGroup.name, COUNT(*) AS times ORDER BY times DESC ---- [source,text] ---- Cypher version: CYPHER 2.3, planner: COST. 23671 total db hits in 39 ms. ---- And for comparison - the same query using the initial model: [source,cypher] ---- MATCH (group:Group {name: "Neo4j - London User Group"})<-[membership:MEMBER_OF]-(member), (member)-[otherMembership:MEMBER_OF]->(otherGroup) WHERE membership.joined < otherMembership.joined WITH member, otherGroup ORDER BY otherMembership.joined WITH member, COLLECT(otherGroup)[0] AS nextGroup RETURN nextGroup.name, COUNT(*) AS times ORDER BY times DESC ---- [source,text] ---- Cypher version: CYPHER 2.3, planner: COST. 86179 total db hits in 138 ms. ---- This time the membership model does 3x less work, so depending on the question a different model works better. Given this observation we might choose to keep both models. The disadvantage of doing that is that we pay write and maintenance penalties to keep them both in sync. e.g. this is what queries to add a new membership or remove one would look like == Adding group membership [source,cypher] ---- WITH "Mark Needham" AS memberName, "Neo4j - London User Group" AS groupName, timestamp() AS now MATCH (group:Group {name: groupName}) MATCH (member:Member {name: memberName}) MERGE (member)-[memberOfRel:MEMBER_OF]->(group) ON CREATE SET memberOfRel.time = now MERGE (membership:Membership {id: member.id + "_" + group.id}) ON CREATE SET membership.joined = now MERGE (member)-[:HAS_MEMBERSHIP]->(membership) MERGE (membership)-[:OF_GROUP]->(group) ---- == Removing group membership [source,cypher] ---- WITH "Mark Needham" AS memberName, "Neo4j - London User Group" AS groupName, timestamp() AS now MATCH (group:Group {name: groupName}) MATCH (member:Member {name: memberName}) MATCH (member)-[memberOfRel:MEMBER_OF]->(group) MATCH (membership:Membership {id: member.id + "_" + group.id}) MATCH (member)-[hasMembershipRel:HAS_MEMBERSHIP]->(membership) MATCH (membership)-[ofGroupRel:OF_GROUP]->(group) DELETE memberOfRel, hasMembershipRel, ofGroupRel ---- The https://github.com/neo4j-meetups/modeling-worked-example[dataset is on github] so take a look at it and send any questions my way.

neo4j-facts-as-nodes

https://markhneedham.com/blog/2015/12/04/neo4j-facts-as-nodes

2015-12-20 12:12:46

Python: Squashing 'duplicate' pairs together

As part of a data cleaning pipeline I had pairs of ids of duplicate addresses that I wanted to group together. I couldn't work out how to solve the problem immediately so I simplified the problem into pairs of letters i.e. [source,text] ---- A B (A is the same as B) B C (B is the same as C) C D ... E F (E is the same as F) F G ... ---- The output that I want to get is: [source,text] ---- (A, B, C, D) (E, F, G) ---- I spent several hours trying to come up with a clever data structure to do this until https://reshmeeauckloo.wordpress.com/[Reshmee] suggested tracking the sets of duplicates using an array of arrays or list of lists since we're going to script this using Python. The actual data is in a CSV file but we'll create a list of tuples to save ourselves some work: [source,python] ---- pairs = [ ("A", "B"), ("B", "C"), ("C", "D"), ("E", "F"), ("F", "G") ] ---- We're going to iterate through the list of pairs and on each iteration we'll check if there's an entry in the list containing either of the values. There can be three outcomes from this check: . No entry - we'll add a new entry with our pair of values. . One entry - we'll add the other value to that entry. . Two entries - we'll merge them together replacing the existing entry. The first step is to write a function to check the list of lists for a matching pair: [source,python] ---- def find_matching_index(pair, dups): return [index for index, dup in enumerate(dups) if pair[0] in dup or pair[1] in dup] print find_matching_index(("A", "B"), [set(["D", "E"])]) [] print find_matching_index(("B", "C"), [set(["A", "B"])]) [0] print find_matching_index(("B", "C"), [set(["A", "B"]), set(["C", "D"])]) [0, 1] ---- Next we need to write a function which iterates over all our pairs of values and uses +++<cite>+++find_matching_index+++</cite>+++ to work out which decision to make: [source,python] ---- def extract_groups(items): dups = [] for pair in items: matching_index = find_matching_index(pair, dups) if len(matching_index) == 0: dups.append(set([pair[0], pair[1]])) elif len(matching_index) == 1: index = matching_index[0] matching_dup = dups[index] dups.pop(index) dups.append(matching_dup.union([pair[0], pair[1]])) else: index1, index2 = matching_index dup1 = dups[index1] dup2 = dups[index2] dups.pop(index1) dups.pop(index2 - 1) # the index decrements since we removed one entry on the previous line dups.append(dup1.union(dup2)) return dups ---- Now let's run this with a few test cases: [source,python] ---- test_cases = [ [ ("A", "B"), ("B", "C"), ("C", "D"), ("E", "F"), ("F", "G") ], [ ("A", "B"), ("B", "C"), ("C", "D"), ("E", "F"), ("F", "G"), ("G", "A"), ("G", "Z"), ("B", "D") ], [ ("A", "B"), ("B", "C"), ("C", "E"), ("E", "A") ], [ ("A", "B"), ("C", "D"), ("F", "G"), ("H", "I"), ("J", "A") ] ] for test_case in test_cases: print extract_groups(test_case) [set(['A', 'C', 'B', 'D']), set(['E', 'G', 'F'])] [set(['A', 'C', 'B', 'E', 'D', 'G', 'F', 'Z'])] [set(['A', 'C', 'B', 'E'])] [set(['C', 'D']), set(['G', 'F']), set(['I', 'H']), set(['A', 'J', 'B'])] ---- This certainly doesn't scale very well but since I only have a few hundred duplicate addresses it does the job for me. It feels like there should be a more functional way to write these functions without mutating all these lists but I haven't figured out what that is yet.

python-squashing-duplicate-pairs-together

https://markhneedham.com/blog/2015/12/20/python-squashing-duplicate-pairs-together

2015-12-27 12:24:05

R: Error in approxfun(x.values.1, y.values.1, method = "constant", f = 1, : zero non-NA points

I've been following http://www.r-bloggers.com/how-to-perform-a-logistic-regression-in-r/[Michy Alice's logistic regression tutorial] to create an attendance model for London dev meetups and ran into an interesting problem while doing so. Our dataset has a class imbalance i.e. most people RSVP 'no' to events which can lead to misleading accuracy score where predicting 'no' every time would lead to supposed high accuracy. [source,r] ---- Source: local data frame [2 x 2] attended n (dbl) (int) 1 0 1541 2 1 53 ---- I sampled the data using +++<cite>+++caret+++</cite>+++'s +++<cite>+++http://www.inside-r.org/packages/cran/caret/docs/upSample[upSample]+++</cite>+++ function to avoid this: [source,r] ---- attended = as.factor((df %>% dplyr::select(attended))$attended) upSampledDf = upSample(df %>% dplyr::select(-attended), attended) upSampledDf$attended = as.numeric(as.character(upSampledDf$Class)) ---- I then trained a logistic regression model but when I tried to plot the area under the curve I ran into trouble: [source,r] ---- p <- predict(model, newdata=test, type="response") pr <- prediction(p, test$attended) prf <- performance(pr, measure = "tpr", x.measure = "fpr") Error in approxfun(x.values.1, y.values.1, method = "constant", f = 1, : zero non-NA points ---- I don't have any NA values in my data frame so this message was a bit confusing to start with. As usual http://stackoverflow.com/questions/23836955/error-in-approxfunx-values-1-y-values-1-method-constant-f-1-zero-no/33028711#33028711[Stack Overflow came to the rescue] with the suggestion that I was probably missing positive/negative values for the independent variable i.e. 'approved'. A quick count on the test data frame using dplyr confirmed my mistake: [source,r] ---- > test %>% count(attended) Source: local data frame [1 x 2] attended n (dbl) (int) 1 1 582 ---- I'll have to http://stackoverflow.com/questions/9081498/the-correct-approach-of-randomly-re-ordering-a-matrix-in-r[randomly sort the data frame] and then reassign my training and test data frames to work around it.

r-error-in-approxfunx-values-1-y-values-1-method-constant-f-1-zero-non-na-points

https://markhneedham.com/blog/2015/12/27/r-error-in-approxfunx-values-1-y-values-1-method-constant-f-1-zero-non-na-points

2015-12-31 13:58:39

2015: A year in the life of the Neo4j London meetup group

Given we've only got a few more hours left of 2015 I thought it'd be fun to do a quick overview of how things have been going in the http://www.meetup.com/graphdb-london/events/227107130/[London chapter of the Neo4j meetup] https://github.com/neo4j-meetups/modeling-worked-example[using Neo4j] with a bit of R mixed in. We're going to be using the RNeo4j library to interact with the database along with a few other libraries which will help us out with different tasks: [source,r] ---- library(RNeo4j) library(ggplot2) library(dplyr) library(zoo) graph = startGraph("http://localhost:7474/db/data/", username = "neo4j", password = "myPassword") ---- Let's get to it: == Members [source,r] ---- query = "MATCH (:Group {name: {name}})<-[membership:MEMBER_OF]-() RETURN membership.joined AS timestamp" joinedDF = cypher(graph, query, name = "Neo4j - London User Group") joinedDF$joinDate = as.Date(as.POSIXct(joinedDF$timestamp / 1000, origin="1970-01-01")) joinedDF$joinDate = as.Date(as.POSIXct(joinedDF$timestamp / 1000, origin="1970-01-01")) ggplot(aes(x = year, y = n, label = n), data = joinedDF %>% mutate(year = format(joinDate, "%Y")) %>% count(year)) + geom_bar(stat = "identity", fill = "Dark Blue") + ggtitle("Number of new members by year") + geom_text(vjust=-0.5) ---- image::{{<siteurl>}}/uploads/2015/12/2015-12-31_12-23-06.png[2015 12 31 12 23 06,300] A bit down on 2014 but not too far away. We're still attracting new people who are interested in learning about graphs. Let's drill into those numbers a bit: [source,R] ---- byYearMon = joinedDF %>% filter(format(joinDate, "%Y") == 2015) %>% mutate(yearmon = as.Date(as.yearmon(joinDate))) %>% count(yearmon) ggplot(aes(x = yearmon, y = n, label = n), data = byYearMon) + geom_bar(stat = "identity", fill = "Dark Blue") + theme(axis.text.x = element_text(angle = 90, hjust = 1)) + scale_x_date(labels = date_format("%B"), breaks = "1 month") + ggtitle("Number of new members by month/year") ---- image::{{<siteurl>}}/uploads/2015/12/2015-12-31_12-39-04.png[2015 12 31 12 39 04,300] We had a bit of an end of year surge in October/November which was unexpected. December has been low in previous years, there was an April dip which I think is because we stopped doing events before http://graphconnect.com/gc2015-europe/[Graph Connect 2015]. I'm not sure about the September dip so let's have a look: [source,r] ---- eventsQuery = "MATCH (:Group {name: {name}})-[:HOSTED_EVENT]->(event) RETURN event.time + event.utcOffset AS timestamp" eventsDF = cypher(graph, eventsQuery, name = "Neo4j - London User Group") eventsDF$timestamp = as.Date(as.POSIXct(eventsDF$timestamp / 1000, origin="1970-01-01")) eventsByYearMon = eventsDF %>% filter(format(timestamp, "%Y") == 2015) %>% mutate(yearmon = as.Date(as.yearmon(timestamp))) %>% count(yearmon) merge(eventsByYearMon, byYearMon, by="yearmon") yearmon n.x n.y 1 2015-01-01 3 80 2 2015-02-01 6 76 3 2015-03-01 2 70 4 2015-04-01 2 53 5 2015-05-01 4 78 6 2015-06-01 5 83 7 2015-07-01 3 73 8 2015-08-01 5 73 9 2015-09-01 3 40 10 2015-10-01 3 94 11 2015-11-01 4 117 12 2015-12-01 3 48 ---- At first glance there doesn't seem to be any correlation between the number of events held and the number of new members so I think we'll have to look for another predictor of that variable! == Events Next let's have a look at the events we ran in 2015. We'll start with a quick chart showing the number of events we've run over the years: [source,r] ---- ggplot(aes(x = year, y = n, label = n), data = eventsDF %>% mutate(year = format(timestamp, "%Y")) %>% count(year)) + geom_bar(stat = "identity", fill = "Dark Blue") + theme(axis.text.x = element_text(angle = 90, hjust = 1)) + ggtitle("Number of events") ---- image::{{<siteurl>}}/uploads/2015/12/2015-12-31_13-43-15.png[2015 12 31 13 43 15,300] So less events than last year but how many people RSVPD 'yes' to the ones we did host? [source,r] ---- eventsQuery = "MATCH (:Group {name: {name}})-[:HOSTED_EVENT]->(event)<-[:RSVPD {response: 'yes'}]-() WHERE event.time + event.utcOffset < timestamp() WITH event, COUNT(*) AS rsvps RETURN event.time + event.utcOffset AS timestamp, rsvps" eventsDF = cypher(graph, eventsQuery, name = "Neo4j - London User Group") eventsDF$timestamp = as.Date(as.POSIXct(eventsDF$timestamp / 1000, origin="1970-01-01")) ggplot(aes(x = year, y = rsvps), data = eventsDF %>% mutate(year = format(timestamp, "%Y")) %>% group_by(year) %>% summarise(rsvps= sum(rsvps)) ) + geom_bar(stat = "identity", fill = "Dark Blue") + theme(axis.text.x = element_text(angle = 90, hjust = 1)) + ggtitle("Number of attendees") ---- image::{{<siteurl>}}/uploads/2015/12/2015-12-31_13-59-15.png[2015 12 31 13 54 50,300] Slightly more 'yes' RSVPs than last year. Now let's drill into the repeat events we ran this year: [source,r] ---- eventsQuery = "MATCH (:Group {name: {name}})-[:HOSTED_EVENT]->(event) WHERE {startYear} <= (event.time + event.utcOffset) < {endYear} RETURN event.name AS event, COUNT(*) AS times ORDER BY times DESC" eventsDF = cypher(graph, eventsQuery, name = "Neo4j - London User Group", startYear = as.numeric(as.POSIXct("2015-01-01", format="%Y-%m-%d")) * 1000, endYear = as.numeric(as.POSIXct("2015-12-31", format="%Y-%m-%d")) * 1000) eventsDF %>% filter(times > 1) event times 1 Relational to graph: A worked example 7 2 Intro to Graphs 6 3 Graph Modelling - Do's and Don'ts 5 4 Hands On Intro to Cypher - Neo4j's Query Language 3 5 Build your own recommendation engine with Neo4j in an hour 2 6 Fraud Detection using Neo4j 2 ---- I thought we'd run 'Intro to Graphs' most often but the data doesn't lie - it's all about relational to graph. And which were the most popular repeat events in terms of 'yes' RSVPs? [source,R] ---- eventsQuery = "MATCH (:Group {name: {name}})-[:HOSTED_EVENT]->(event) WHERE {startYear} <= (event.time + event.utcOffset) < {endYear} MATCH (event)<-[:RSVPD {response: 'yes'}]-() WITH event, COUNT(*) AS yesRSVPs WITH event.name AS event, COUNT(*) AS times, SUM(yesRSVPs) AS rsvps RETURN event, times, rsvps, rsvps / times AS rsvpsPerEvent ORDER BY rsvpsPerEvent DESC" eventsDF = cypher(graph, eventsQuery, name = "Neo4j - London User Group", startYear = as.numeric(as.POSIXct("2015-01-01", format="%Y-%m-%d")) * 1000, endYear = as.numeric(as.POSIXct("2015-12-31", format="%Y-%m-%d")) * 1000) eventsDF %>% filter(times > 1) event times rsvps rsvpsPerEvent 1 Fraud Detection using Neo4j 2 150 75 2 Intro to Graphs 6 352 58 3 Graph Modelling - Do's and Don'ts 5 281 56 4 Relational to graph: A worked example 7 367 52 5 Build your own recommendation engine with Neo4j in an hour 2 85 42 6 Hands On Intro to Cypher - Neo4j's Query Language 3 104 34 ---- It looks like fraud is a popular topic although we've only run it twice so perhaps best not to read too much into that. We're http://www.meetup.com/graphdb-london/events/227211069/[running that one again] in a couple of weeks if you're interested. Ignoring repeat events let's see which event drew the biggest crowd: [source,r] ---- eventsQuery = "MATCH (:Group {name: {name}})-[:HOSTED_EVENT]->(event) WHERE {startYear} <= (event.time + event.utcOffset) < {endYear} MATCH (event)<-[:RSVPD {response: 'yes'}]-() WITH event.id AS id, event.name AS event, COUNT(*) AS rsvps RETURN event, rsvps ORDER BY rsvps DESC" eventsDF = cypher(graph, eventsQuery, name = "Neo4j - London User Group", startYear = as.numeric(as.POSIXct("2015-01-01", format="%Y-%m-%d")) * 1000, endYear = as.numeric(as.POSIXct("2015-12-31", format="%Y-%m-%d")) * 1000) eventsDF %>% head(5) event rsvps 1 Neo4j Full Stack Applications + Python, R and Neo4j - The Data Science Stack 133 2 Modelling a recommendation engine: A worked example 118 3 Building a repository of biomedical ontologies with Neo4j 107 4 GraphHack @ Graph Connect: The night before Election Day 91 5 Bootstrapping a Recommendation Engine 88 ---- A http://www.meetup.com/graphdb-london/events/226184225/[double header] featuring https://twitter.com/_nicolemargaret[Nicole White] and https://twitter.com/mrmattwright[Matt Wright] proved to be the most popular event of the year and in fact the most popular in terms of 'yes' RSVPs so far: [source,r] ---- eventsQuery = "MATCH (:Group {name: {name}})-[:HOSTED_EVENT]->(event)<-[:RSVPD {response: 'yes'}]-() WITH event, COUNT(*) AS rsvps RETURN event.name AS event, event.time + event.utcOffset AS time, rsvps ORDER BY rsvps DESC" eventsDF = cypher(graph, eventsQuery, name = "Neo4j - London User Group") eventsDF$time = as.Date(as.POSIXct(eventsDF$time / 1000, origin="1970-01-01")) eventsDF %>% mutate(year = format(time, "%Y")) %>% dplyr::select(-time) %>% head(10) event rsvps year 1 Neo4j Full Stack Applications + Python, R and Neo4j - The Data Science Stack 133 2015 2 Modelling a recommendation engine: A worked example 118 2015 3 Building a repository of biomedical ontologies with Neo4j 107 2015 4 Real world Neo4j use cases 98 2014 5 The transport graph 94 2014 6 The Visualisation Special 93 2014 7 Impossible is Nothing by Jim Webber, Neo4j's Chief Scientist 93 2014 8 GraphHack @ Graph Connect: The night before Election Day 91 2015 9 Bootstrapping a Recommendation Engine 88 2015 10 Scraping and Graphing the Apple app store 88 2015 ---- 3 of the top 4 belong to 2015 and 6 of the top 10. Let's see what 2016 has in store. Thanks to everyone who's come along to one of our meetups and Happy New Year!

2015-a-year-in-the-life-of-the-neo4j-london-meetup-group

https://markhneedham.com/blog/2015/12/31/2015-a-year-in-the-life-of-the-neo4j-london-meetup-group

2015-12-31 10:47:01

Study until your mind wanders

I've previously found it very difficult to read math heavy content which has made it challenging to read http://www.amazon.co.uk/Distributed-Computing-2e-Fundamentals-Simulations/dp/0471453242/ref=sr_1_1?ie=UTF8&qid=1451556471&sr=8-1&keywords=distributed+computing+fundamentals[Distributed Computing] which I bought last May. After several false starts where I gave up after getting frustrated that I couldn't understand things the first time around and forgot everything if I left it a couple of days I decided to try again with a different approach. I've been trying a technique http://www.markhneedham.com/blog/2015/03/17/one-month-of-mini-habits/[I learned] from http://www.amazon.co.uk/Mini-Habits-Smaller-Bigger-Results-ebook/dp/B00HGKNBDK[Mini Habits] where every day I have a (very small) goal of *reading one page of the book*. Having such a small goal means that I can read the material as slowly as I like (repeating previous days if necessary). So far I've read 4 chapters (~100 pages) over the last month - some days I read 6 or 7 pages, other days I only manage one. The key is to keep the rhythm of reading something. I tried doing the reading at different times of the day - on the bus on the way to work, in the evening before going to sleep - and found that for me the best time is immediately when I wake up. To minimise my mind wandering I don't read any emails, chat messages or social media accounts before I start. Despite this, I've noticed that after a while *my mind starts to wander* while reading proofs and that's a signal for me to stop for the day. When I pick up again the next day I've often found that I understand what I was having difficulty with. I've read that meditation prior to studying is an effective way to quiet the mind and would be a more 'on demand' way of achieving the concentration required to read this type of material. I've never done any meditation so if anyone has any tips on where to start that'd be helpful.

study-until-your-mind-wanders

https://markhneedham.com/blog/2015/12/31/study-until-your-mind-wanders

2015-12-13 21:22:07

Neo4j: Specific relationship vs Generic relationship + property

For optimal traversal speed in Neo4j queries we should make our *relationship types as specific as possible*. Let's take a look at an example from the 'http://www.meetup.com/graphdb-london/events/226721630/[modelling a recommendations engine]' talk I https://skillsmatter.com/skillscasts/7298-modelling-a-recommendation-engine-a-worked-example[presented at Skillsmatter] a couple of weeks ago. I needed to decided how to model the 'RSVP' relationship between a +++<cite>+++Member+++</cite>+++ and an +++<cite>+++Event+++</cite>+++. A person can RSVP 'yes' or 'no' to an event and I'd like to capture both of these responses. i.e. we can choose between: image::{{<siteurl>}}/uploads/2015/12/2015-12-13_20-39-05.png[2015 12 13 20 39 05,300] and: image::{{<siteurl>}}/uploads/2015/12/2015-12-13_20-39-54.png[2015 12 13 20 39 54,300] When deciding on a model we mainly need to think about the types of queries that we want to write. We shouldn't forget about updating the model but in my experience more time is spent querying graphs than updating them. Let's take a look at each of those in turn: == What queries do we want to write? The first query was going to use previous 'yes' RSVPs as an indicator of interest for future events. We're not interested in 'no' RSVPs for this query. I started out with the generic RSVP relationship type with a 'response' property to distinguish between 'yes' and 'no': [source,cypher] ---- MATCH (member:Member {name: "Mark Needham"}) MATCH (futureEvent:Event) WHERE futureEvent.time >= timestamp() MATCH (futureEvent)<-[:HOSTED_EVENT]-(group) OPTIONAL MATCH (member)-[rsvp:RSVPD {response: "yes"}]->(pastEvent)<-[:HOSTED_EVENT]-(group) WHERE pastEvent.time < timestamp() RETURN group.name, futureEvent.name, COUNT(rsvp) AS previousEvents ORDER BY previousEvents DESC ---- This ran reasonably quickly but I was curious whether I could get the query to run any quicker by changing to the more specific model. Using the more specific relationship type our query reads: [source,cypher] ---- MATCH (member:Member {name: "Mark Needham"}) MATCH (futureEvent:Event) WHERE futureEvent.time >= timestamp() MATCH (futureEvent)<-[:HOSTED_EVENT]-(group) OPTIONAL MATCH (member)-[rsvp:RSVP_YES]->(pastEvent)<-[:HOSTED_EVENT]-(group) WHERE pastEvent.time < timestamp() RETURN group.name, futureEvent.name, COUNT(rsvp) AS previousEvents ORDER BY previousEvents DESC ---- We can now http://neo4j.com/docs/stable/how-do-i-profile-a-query.html[profile our query] and compare the db hits of both solutions: [source,text] ---- RSVPD {response: "yes"} Cypher version: CYPHER 2.3, planner: COST. 688635 total db hits in 232 ms. RSVP_YES Cypher version: CYPHER 2.3, planner: COST. 559866 total db hits in 207 ms. ---- So we get a slight gain by using the more specific relationship type. The reason the db hits is lower is partly because we've removed the need to lookup the 'response' property on every 'RSVP' property and check that it matches 'yes'. We're also evaluating fewer relationships since we only look at positive RSVPs, negative ones are ignored. Our next query might be to capture all the RSVPs made by a member and list them alongside the events: [source,cypher] ---- MATCH (member:Member {name: "Mark Needham"})-[rsvp:RSVPD]->(event) WHERE event.time < timestamp() RETURN event.name, event.time, rsvp.response ORDER BY event.time DESC ---- [source,cypher] ---- MATCH (member:Member {name: "Mark Needham"})-[rsvp:RSVP_YES|:RSVP_NO]->(event) WHERE event.time < timestamp() RETURN event.name, event.time, CASE TYPE(rsvp) WHEN "RSVP_YES" THEN "yes" ELSE "no" END AS response ORDER BY event.time DESC ---- Again we see a marginal db hits win for the more specific relationship type: [source,text] ---- RSVPD {response: "yes"} / RSVPD {response: "no"} Cypher version: CYPHER 2.3, planner: COST. 684 total db hits in 37 ms. RSVP_YES / RSVP_NO Cypher version: CYPHER 2.3, planner: COST. 541 total db hits in 24 ms. ---- However, the query is quite unwieldy and unless we store the response as a property on the relationship the code to return 'yes' or 'no' is a bit awkward. The more specific approach query would become even more painful to deal with if we introduced the 'waitlist' RSVP which we've chosen to exclude. == Will we need to update the relationship? Yes! Users are able to change their RSVP up until the event happens so we need to be able to handle that. Let's have a look at the queries we'd have to write to handle a change in RSVP using both models: == Generic relationship type [source,cypher] ---- MATCH (event:Event {id: {event_id}}) MATCH (member:Member {id: {member_id}}) MERGE (member)-[rsvpRel:RSVPD {id: {rsvp_id}}]->(event) ON CREATE SET rsvpRel.created = toint({mtime}) ON MATCH SET rsvpRel.lastModified = toint({mtime}) SET rsvpRel.response = {response} ---- == Specific relationship type [source,cypher] ---- MATCH (event:Event {id: {event_id}}) MATCH (member:Member {id: {member_id}}) FOREACH(ignoreMe IN CASE WHEN {response} = "yes" THEN [1] ELSE [] END | MERGE (member)-[rsvpYes:RSVP_YES {id: {rsvp_id}}]->(event) ON CREATE SET rsvpYes.created = toint({mtime}) ON MATCH SET rsvpYes.lastModified = toint({mtime}) MERGE (member)-[oldRSVP:RSVP_NO]->(event) DELETE oldRSVP ) FOREACH(ignoreMe IN CASE WHEN {response} = "no" THEN [1] ELSE [] END | MERGE (member)-[rsvpNo:RSVP_NO {id: {rsvp_id}}]->(event) ON CREATE SET rsvpNo.created = toint({mtime}) ON MATCH SET rsvpNo.lastModified = toint({mtime}) MERGE (member)-[oldRSVP:RSVP_YES]->(event) DELETE oldRSVP ) ---- As you can see, the code to update an RSVP is more complicated when using the specific relationship type due in part to Cypher not yet having first class support for conditionals. In summary, for our meetup.com model we gain speed improvements by using more specific relationship types but at the expense of some more complicated read queries and a significantly more convoluted update query. Depending on the cardinality of relationships in your model your mileage may vary but it's worth doing some profiling to compare all your options.

neo4j-specific-relationship-vs-generic-relationship-property

https://markhneedham.com/blog/2015/12/13/neo4j-specific-relationship-vs-generic-relationship-property

2012-03-20 06:53:18

Scala: Counting number of inversions (via merge sort) for an unsorted collection

The first programming questions of https://www.coursera.org/algo/class/index[algo-class] requires you to calculate the number of inversions it would take using merge sort to sort a collection in ascending order. I found quite http://www.cp.eng.chula.ac.th/~piak/teaching/algo/algo2008/count-inv.htm[a nice explanation here too]: ____ Finding "similarity" between two rankings. Given a sequence of n numbers 1..n (assume all numbers are distinct). Define a measure that tells us how far this list is from being in ascending order. The value should be 0 if a_1 < a_2 < \... < a_n and should be higher as the list is more "out of order". e.g. 2 4 1 3 5 1 2 3 4 5 The sequence 2, 4, 1, 3, 5 has three inversions (2,1), (4,1), (4,3). ____ The simple/naive way of solving this problem is to iterate through the collection in two loops and compare each value and its current index with the others, looking for ones which are not in the right order. I wrote the following Ruby code which seems to do the job: [source,ruby] ---- def number_of_inversions(arr) count = 0 arr.each_with_index do |item_a, index_a| arr.each_with_index do |item_b, index_b| if index_b >= index_a && item_a > item_b count +=1 end end end count end ---- [source,text] ---- >> number_of_inversions [6,5,4,3,2,1] => 15 ---- That works fine for small collections but since we're effectively doing nested for loops it actually takes O(n&sup2;) time which means that it pretty much kills my machine on the sample data set of 100,000 numbers. An O(n log(n)) solution is explained in the lectures which involves recursively splitting the collection in half (like in merge sort) and then counting how many elements remain in the left collection when we select an item from the right collection since this will tell us how many positions/inversions out of place that element is. e.g. if we're trying to work out how many inversions in the collection [1,3,5,2,4,6] we eventually end up merging [1,3,5] and [2,4,6] * Our first iteration puts '1' from the left collection into the new array. * The second iteration puts '2' from the right collection into the new array and there are two elements left in the left collection ('3' and '5') so we have 2 inversions (3,2 and 5,2). * Our third iteration puts '3' from the left collection into the new array. * Our fourth iteration puts '4' from the right collection into the new array and there is one element left in the left collection ('5') so we have 1 inversion (5,4) * Our fifth iteration puts '5' from the left collection and the sixth puts '6' from the right collection in. The number of inversions for that example is therefore 3. My colleague https://twitter.com/#!/anagri[Amir] and I decided to try and write some Scala code to do this and ended up with the following adapted merge sort: [source,scala] ---- import io.Source object MSort { def main(args:Array[String]) { val fileWithNumbers = "/Users/mneedham/Documents/algo-class/IntegerArray.txt" val inversions: BigInt = numberOfInversions(Source.fromFile(new java.io.File(fileWithNumbers)).getLines().map(Integer.parseInt).toList) println(inversions) } def numberOfInversions(collection: List[Int]): BigInt = { var count: BigInt = 0 def inversionsInner(innerCollection: List[Int]): List[Int] = { def merge(left: List[Int], right: List[Int]): Stream[Int] = (left, right) match { case (x :: xs, y :: ys) if x < y=> { Stream.cons(x, merge(xs, right)) } case (x :: xs, y :: ys) => { count = count + left.length; Stream.cons(y, merge(left, ys)) } case _ => if (left.isEmpty) right.toStream else left.toStream } val n = innerCollection.length / 2 if (n == 0) innerCollection else { val (lowerHalf, upperHalf) = innerCollection splitAt n merge(inversionsInner(lowerHalf), inversionsInner(upperHalf)).toList } } inversionsInner(collection) count } } ---- The interesting line is number 15 where we are going to select a value from the right collection and therefore increment our count by the number of items left in the left collection. It works but it's a bit annoying that we had to use a 'var' to keep track of the count since that's not really idiomatic Scala. I've been trying to find a way of passing the count around as an accumulator and returning it at the end but really struggled to get the code to compile when I started doing that and seemed to make the code a lot more complicated than it is now. I'm sure there is a way to do it but I can't see it at the moment! Since the mutation is reasonably well encapsulated I'm not sure whether it really matters that much but it's always interesting an interesting exercise to try and write code which doesn't mutate state so if anyone can see a good way to do it let me know.

scala-counting-number-of-inversions-via-merge-sort-for-an-unsorted-collection

https://markhneedham.com/blog/2012/03/20/scala-counting-number-of-inversions-via-merge-sort-for-an-unsorted-collection

2012-03-20 23:36:03

Haskell: Chaining functions to find the middle value in a collection

I've been playing around with writing merge sort in Haskell and eventually ended up with the following function: [source,haskell] ---- msort :: [Int] -> [Int] msort unsorted = let n = floor (fromIntegral(length unsorted) / 2) in if n == 0 then unsorted else let (left, right) = splitAt n unsorted in merge (msort left) (msort right) where merge [] right = right merge left [] = left merge left@(x:xs) right@(y:ys) = if x < y then x : merge xs right else y : merge left ys ---- The 3rd line was annoying me as it has way too many brackets on it and I was fairly sure that it should be possible to just combine the functions http://www.markhneedham.com/blog/2009/01/12/f-partial-function-application-with-the-function-composition-operator/[like I learnt to do in F# a few years ago]. It's pretty easy to do that for the first two functions 'length' and 'fromIntegral' which we can do like this: [source,haskell] ---- middle = fromIntegral . length ---- The third line now reads like this: [source,haskell] ---- let n = floor ((middle unsorted) / 2) ---- It's a slight improvement but still not that great. The problem with working out how to chain the division bit is that our value needs to be passed as the first argument to '/' so we can't do the following... [source,haskell] ---- middle = ((/) 2) . fromIntegral . length ---- ...since that divides 2 by the length of our collection rather than the other way around! [source,haskell] ---- > middle [1,2,3,4,5,6] 0.3333333333333333 ---- Instead we want to create an anonymous function around the '/' function and then apply floor: [source,haskell] ---- middle :: [Int] -> Int middle = floor . (\y -> y / 2) . fromIntegral . length ---- And merge sort now looks like this: [source,haskell] ---- msort :: [Int] -> [Int] msort unsorted = let n = middle unsorted in if n == 0 then unsorted else let (left, right) = splitAt n unsorted in merge (msort left) (msort right) where merge [] right = right merge left [] = left merge left@(x:xs) right@(y:ys) = if x < y then x : merge xs right else y : merge left ys ---- Which I think is pretty neat!

haskell-chaining-functions-to-find-the-middle-value-in-a-collection

https://markhneedham.com/blog/2012/03/20/haskell-chaining-functions-to-find-the-middle-value-in-a-collection

2012-03-20 23:55:51

Haskell: Newbie currying mistake

As I http://www.markhneedham.com/blog/2012/03/20/haskell-chaining-functions-to-find-the-middle-value-in-a-collection/[mentioned in my last post] I've spent a bit of this evening writing a merge sort function and one of the mistakes I made a few times was incorrectly passing arguments to the recursive calls of 'merge'. For example, this is one of the earlier versions of the function: [source,haskell] ---- middle :: [Int] -> Int middle = floor . (\y -> y / 2) . fromIntegral . length msort :: [Int] -> [Int] msort unsorted = let n = middle unsorted in if n == 0 then unsorted else let (left, right) = splitAt n unsorted in merge (msort left) (msort right) where merge [] right = right merge left [] = left merge left@(x:xs) right@(y:ys) = if x < y then x : merge(xs, right) else y : merge (left, ys) ---- Which doesn't actually compile: [source,text] ---- Couldn't match expected type `[a0]' with actual type `[a0] -> [a0]' In the return type of a call of `merge' In the second argument of `(:)', namely `merge (xs, right)' In the expression: x : merge (xs, right) ---- My defence for this mistake is that many of the other languages I've programmed in take function parameters separated by a comma but in this case I've actually only succeeded in currying the 'merge' function. i.e. it thinks I only wanted to pass one value to it and return a function, hence the error message! One correction would be to change the code to read like this so we can explicitly see that a function which takes 2 arguments can be called with each argument separately: [source,haskell] ---- msort :: [Int] -> [Int] msort unsorted = let n = middle unsorted in if n == 0 then unsorted else let (left, right) = splitAt n unsorted in merge (msort left) (msort right) where merge [] right = right merge left [] = left merge left@(x:xs) right@(y:ys) = if x < y then x : merge(xs)(right) else y : merge (left)(ys) ---- We can do exactly the same with the '/' function to use a simpler example: [source,haskell] ---- -- a long way of dividing 2 by 3 > ((/) 2) 3 0.6666666666666666 ---- The type of '/' is: [source,haskell] ---- > :t (/) (/) :: Fractional a => a -> a -> a ---- which means it takes in a 'Fractional' and then returns a function which takes in a 'Fractional' and returns a 'Fractional'. In Haskell http://stackoverflow.com/questions/4768453/type-signature-vs-function-equation-in-haskell[function application is left associative] which means that 'f x y' is the same as '(f x) y' so we can omit the parentheses and pass both arguments together: [source,haskell] ---- > (/) 2 3 0.6666666666666666 ---- And we can do the same thing in the merge sort of course: [source,haskell] ---- msort :: [Int] -> [Int] msort unsorted = let n = middle unsorted in if n == 0 then unsorted else let (left, right) = splitAt n unsorted in merge (msort left) (msort right) where merge [] right = right merge left [] = left merge left@(x:xs) right@(y:ys) = if x < y then x : merge xs right else y : merge left ys ---- A fairly simple mistake to make but it had me confused for a while! ''' Updated the explanation around how we can pass arguments to '/' after my colleague https://twitter.com/#!/gavri[Gavri] https://twitter.com/#!/gavri/status/182331266681671680[pointed out that the initial explanation was incorrect].

haskell-newbie-currying-mistake

https://markhneedham.com/blog/2012/03/20/haskell-newbie-currying-mistake

2012-03-29 06:45:59

Readability/Performance

I recently read the http://www.aosabook.org/en/graphite.html[Graphite chapter] of http://www.aosabook.org/en/index.html[The Architecture of Open Source Applications] book which mostly tells the story of how Chris Davis incrementally built out http://graphite.wikidot.com/[Graphite] - a pretty cool tool that can be used to do real time graphing of metrics. The whole chapter is a very good read but I found the design reflections especially interesting: ____ One of Graphite's greatest strengths and greatest weaknesses is the fact that very little of it was actually "designed" in the traditional sense. [It] evolved gradually, hurdle by hurdle, as problems arose. Many times the hurdles were foreseeable and various pre-emptive solutions seemed natural. However *it can be useful to avoid solving problems you do not actually have yet, even if it seems likely that you soon will*. ____ One of the main success criteria of the application that I'm currently working on is its performance - it doesn't have millisecond-ish latency requirements but it does need to do a lot of calculations and return within a reasonable amount of time. Several times we've been working on a bit of code and have written something which is *easy to read but runs in quadratic time* because we have a nested iteration over a collection. Having this performance requirement in mind has made us think twice about whether we should be looking for a better way of writing that code but that would seem to be a premature optimisation since we don't actually know that this bit of code is going to be the bottleneck. Instead we've started to put a performance testing infrastructure in place so that we can actually https://github.com/codahale/metrics[gather some data] that tells us where the problems in the code are. Davis goes on to state the following in his article: ____ The reason is that you can learn much more from closely studying actual failures than from theorizing about superior strategies. Problem solving is driven by both the empirical data we have at hand and our own knowledge and intuition. I've found that doubting your own wisdom sufficiently can force you to look at your empirical data more thoroughly. ____ As I http://www.markhneedham.com/blog/2012/03/28/testing-trying-not-to-overdo-it/[mentioned in my last post] the main logic in our application involves loading a bunch of data from Oracle and then processing various calculations on it. At the moment we're getting sub second responses from most requests and where that isn't the case the bottleneck has been in one of our database queries rather than in the code. Maybe eventually we'll want to optimise those bits of code which have quadratic running time and the good thing is that we've abstracted those bits of code so we could optimise that code with minimal impact to the rest of the code base. I've been working through http://www.algo-class.org[Algo Class] and one of the things that I've learnt is that you only really see problems with algorithms of quadratic running time if you are iterating through really big collections - otherwise it's barely noticeable. In our case the collections will probably be small so we unlikely to see that problem. On the other hand we are likely to be repeating the same calculations multiple times so we will probably look to add some caching if that starts to become a problem. Of course when we do that we'll need to check our performance tests before and after the change to ensure that we've actually addressed a real problem rather than theorised about it! For now though we're continuing to write our code in the way that is most readable/easy to understand and waiting for our performance tests to force us to change that approach. ''' I've not worked closely on anything which had proprietary trading style latency requirements so if your experience is different please let me know in the comments.

readabilityperformance

https://markhneedham.com/blog/2012/03/29/readabilityperformance

2012-03-28 00:10:46

Testing: Trying not to overdo it

The design of the code which contains the main logic of the application that I'm currently working on looks a bit like the diagram on the right hand side: image::{{<siteurl>}}/uploads/2012/03/orchestration-code.gif[Orchestration code,329] We load a bunch of stuff from an Oracle database, construct some objects from the data and then invoke a sequence of methods on those objects in order to execute our domain logic. Typically we might expect to see unit level test against all the classes described in this diagram but we've actually been trying out an approach where we don't test the orchestration code directly but rather only test it via the resource which makes use of it. We originally started off writing some tests around that code but they ended up being really similar to our database and resource tests. Having them around also made it difficult to change the way the orchestration worked since we'd end up breaking most of the tests when we tried to change anything. One disadvantage of not testing this code is that we end up using the debugger more when trying to work out why resource tests aren't working since we now have more code being directly tested. image::{{<siteurl>}}/uploads/2012/03/orchestration-tests2.gif[Orchestration tests2,427] On the other hand we've been forced to drive logic into the domain objects as a result since we don't have any other place to test that functionality from. Testing directly against the domain objects is much easier since everything's in memory and we can easily setup the data to be how we want it to be and inject it into the objects. Another approach we could have taken would be to mock out the dependencies of the orchestration code but since this code is mostly coordinating other classes there are a lot of dependencies and the tests ended up being quite complicated and brittle. Initially I was of the opinion that it wasn't a good idea to not test the orchestration code but looking back a month later I think it's working reasonably well and putting this constraint on ourselves has made the code easier to change while still being well tested.

testing-trying-not-to-overdo-it

https://markhneedham.com/blog/2012/03/28/testing-trying-not-to-overdo-it

2012-03-17 11:19:11

Coding: Wait for the abstractions to emerge

One of the things that I've learnt while developing code in an incremental way is that the way the code should be designed isn't going to be obvious straight away so we need to be patience and wait for it to emerge. There's often a tendency to pull out classes or methods but more recently I've been trying to follow an approach where I leave the code in one class/method and play around with/study it until I see a good abstraction to make. More often than not it takes a few times over multiple days before anything reveals itself. I recently watched https://twitter.com/#!/venkat_s[Venkat Subramaniam's] Agile India presentation 'http://www.youtube.com/watch?v=iGsPeR-SYYo&feature=youtu.be[How to Approach Refactoring]' in which amongst other bits of advice he suggests that we '*make it work then make it evolve*' which sounds like it's covering the same type of ground. Sometimes in the rush to abstract code to make it easier to test or to reduce the size of a class we end up creating an abstraction which doesn't really make sense in the business domain. Having big methods of inline code goes against what we've learnt but putting all the code together makes it much easier to see the real abstractions in the code than if some premature abstractions have been made and we have to remember those as well. It's still useful to be driven by the goal of pulling out small classes of behaviour/data that are easy to test but *spending a bit more time understanding the domain before doing so seems to work better*. We recently had some unwieldy code which we couldn't quite work out how to abstract so we left the logic inline until we knew better even though it looked quite messy. Eventually when talking to one of the subject matter experts they mentioned a term that they used which actually combined two things together which we'd originally thought were independent. A simple rule of thumb is that if there is no name in the domain for what you're abstracting then it might not make sense to do so. It doesn't always work but at least gets your thinking about whether you're improving the code with the abstraction you're about to make!

coding-wait-for-the-abstractions-to-emerge

https://markhneedham.com/blog/2012/03/17/coding-wait-for-the-abstractions-to-emerge

2012-03-19 23:25:47

Functional Programming: One function at a time

As I mentioned in an earlier post I got a bit stuck working out all the diagonals in the 20x20 grid of http://projecteuler.net/problem=11[Project Euler problem 11] and my http://www.markhneedham.com/blog/2012/03/13/functional-programming-shaping-the-data-to-fit-a-function/[colleague Uday ended up showing me how to do it]. I realised while watching him solve the problem that we'd been using quite different approaches to solving the problem and that his way worked way better than mine, at least in this context. My approach was to try and drive out the solution from the top down which meant in this case: * Get all of the diagonals left to right by iterating over a 2 dimensional array and accumulating sub arrays of the values 1 row and 1 column greater than the current position. I was stuck for ages trying to work out how to do that whereas Uday made it much easier by just focusing on one of the functions that we'd need, getting that to work and then moving onto the next one. After we'd iterated with that approach a few times we had a bunch of functions that we could compose together to solve the initial problem. We started with a simple function to find the value at position (x,y) in the array: [source,haskell] ---- findValue :: Int -> Int -> [[Int]] -> Int findValue x y grid = (grid !! x) !! y ---- We then worked on writing a function in the REPL which would find all the diagonals going down from the current position [source,haskell] ---- [findValue (0+z) (0+z) grid | z <- [0..19]] ---- We eventually realised that running that from anywhere but the top left hand corner would throw an exception since we'd gone out of bounds of the array: [source,text] ---- > [findValue (1+z) (1+z) grid | z <- [0..19]] [49,31,23,51,3,67,20,97,45,3,24,44,52,26,32,40,4,5,48,*** Exception: Prelude.(!!): index too large ---- We then wrote a function to check that a given point was actually in the array: [source,haskell] ---- hasItemSingle :: [a] -> Int -> Bool hasItemSingle array position = position >= 0 && (length array - 1) >= position hasItem :: [[a]] -> Int -> Int -> Bool hasItem array x y = hasItemSingle array x && hasItemSingle (array !! x) y ---- And our function to calculate the diagonals then used that: [source,text] ---- > [findValue (1+z) (1+z) grid | z <- [0..19], hasItem grid (1+z) (1+z) ] [49,31,23,51,3,67,20,97,45,3,24,44,52,26,32,40,4,5,48] ---- That function eventually ended up like this: [source,haskell] ---- diagonalAt :: Int -> Int -> [[Int]] -> [Int] diagonalAt x y grid = [findValue (x+z) (y+z) grid | z <- [0..(length grid)], hasItem grid (x + z) (y + z)] ---- The nice thing about taking this approach of only thinking about one function at a time is that you can actually make some progress since you only need to think about a small part of the problem as compared to the whole thing. I still start off trying to solve a problem from the outside in but if I get stuck then I start looking for some simple functions that I can write to get started and help me work towards an overall solution.

functional-programming-one-function-at-a-time

https://markhneedham.com/blog/2012/03/19/functional-programming-one-function-at-a-time

2012-03-21 00:50:37

Functional Programming: Handling the Options

A couple of weeks ago https://twitter.com/#!/channingwalton/status/177662437905010688[Channing Walton tweeted the following]: ____ Every time you call get on an Option a kitten dies. ____ As Channing points out in the comments he was referring to unguarded calls to 'get' which would lead to an exception if the Option was empty, therefore pretty much defeating the point of using an Option in the first place! We're using Dan Bodart's http://code.google.com/p/totallylazy/[totallylazy] library on the application I'm currently working on and in fact were calling 'get' on an Option so I wanted to see if we could get rid of it. The code in question is used to work out a price based on a base value and an increment value, both of which are optional. I've re-created some tests to show roughly how it works: [source,java] ---- @Test public void myOptionsExample() { Assert.assertThat(calculate(Option.<Double>none(), Option.<Double>none(), 20), is(Option.<Double>none())); Assert.assertThat(calculate(Option.<Double>none(), Option.some(12.0), 20), is(Option.some(240.0))); Assert.assertThat(calculate(Option.some(50.0), Option.some(12.0), 20), is(Option.some(290.0))); Assert.assertThat(calculate(Option.some(50.0), Option.<Double>none(), 20), is(Option.some(50.0))); } ---- [source,java] ---- private Option<Double> calculate(Option<Double> base, final Option<Double> increment, final int multiplier) { if(base.isEmpty() && increment.isEmpty()) { return Option.none(); } if(increment.isEmpty()){ return base; } double baseBit = base.getOrElse(0.0); double incrementBit = increment.get() * multiplier; return Option.some(baseBit + incrementBit); } ---- We can get rid of the 'increment.isEmpty()' line by making use of a 'getOrElse' a few lines later like so: [source,java] ---- private Option<Double> calculate(Option<Double> base, final Option<Double> increment, final int multiplier) { if(base.isEmpty() && increment.isEmpty()) { return Option.none(); } double baseBit = base.getOrElse(0.0); double incrementBit = increment.getOrElse(0.0) * multiplier; return Option.some(baseBit + incrementBit); } ---- In Scala we can often make use of a for expression when combining Options but I'm not sure it would help in this case because we want to return a 'Some' if either of the Options has a value and only return a 'None' if they're both empty. I can't figure out how to get rid of that opening if statement - I tried pushing the options into a list and then calling 'foldLeft' on that to sum up the components but we still have the problem of how to deal with returning a 'None' when the list is empty. If you know how to simplify that code either using totallylazy or another language please let me know in the comments!

functional-programming-handling-the-options

https://markhneedham.com/blog/2012/03/21/functional-programming-handling-the-options

2012-03-31 09:06:14

Micro Services: A simple example

In our code base we had the concept of a 'ProductSpeed' with two different constructors which initialised the object in different ways: [source,java] ---- public class ProductSpeed { public ProductSpeed(String name) { ... } public ProductSpeed(String name, int order)) { } } ---- In the cases where the first constructor was used the order of the product was irrelevant. When the second constructor was used we did care about it because we wanted to be able sort the products before showing them in a drop down list to the user. The reason for the discrepancy was that this object was being constructed from data which originated from two different systems and in one the concept of order existed and in the other it didn't. What we actually needed was to have two different versions of that object but we probably wouldn't want to name them 'ProductSpeedForSystem1' and 'ProductSpeedForSystem2'! In Domain Driven Design terms we actually have the concept of a 'ProductSpeed' but in two different bounded contexts which could just mean that they come under different packages if we're building everything in one (monolithic) application. However, we could see from looking at the way 'ProductSpeed' initialised from the second constructor was being used in the application that it didn't interact with anything else and so could easily be pulled out into its own mini application or http://2012.33degree.org/talk/show/67[micro service]. We're actually building an API for other systems to interact with and the initial design of the code described above was: image::{{<siteurl>}}/uploads/2012/03/api-before.gif[Api before,303] We get a product from the product list (which is sorted based on the ordering described!) and then post a request which includes the product amongst other things. After we'd pulled out a micro service it looked like this: image::{{<siteurl>}}/uploads/2012/03/api-after.gif[Api after,303] The choice of product is actually a step that you do before you make your request to the main API whereas we'd initially coupled them into the same deployable. These are the advantages I see from what we've done: * We can now *easily change the underlying data source* of the products micro service if we want to since it now has its own schema which we could switch out if necessary. * It takes about 5 minutes to populate all the products and we run the script to repopulate the main DB quite frequently. Now products can be loaded separately. * Our code is now *much simplified*! And some disadvantages: * We now have to *deploy two jars instead of one* so our deployment has become a bit more complicated. My colleague http://twitter.com/boicy[James Lewis] points out that we're effectively pushing the complexity from the application into the infrastructure when we design systems with lots of mini applications doing one thing. * Overall I think we have *more code* since there are some similarities between the objects in both contexts and we've now got two versions of each object since they're deployed separately. My experience is that sharing domain code generally leads to suffering so we're not doing that.

micro-services-a-simple-example

https://markhneedham.com/blog/2012/03/31/micro-services-a-simple-example

2012-03-31 22:49:18

Haskell: Pattern matching data types with named fields

One of my favourite things about coding in Haskell is that I often end up pattern matching against data types. I've been playing around with modelling cars coming into and out from a car park and changing the state of the car park accordingly. I started with these data type definitions: [source,haskell] ---- data CarParkState = Available Bool Int Int | AlmostFull Bool Int Int | Full Bool Int deriving (Show) data Action = Entering | Leaving deriving (Show) data Sticker = Handicap | None deriving (Show) ---- which were used in the following function: [source,haskell] ---- car_park_state :: CarParkState -> (Action, Sticker) -> CarParkState car_park_state (Available _ 8 handicap) (Entering, None) = AlmostFull True 9 handicap car_park_state (AlmostFull _ 11 handicap) (Entering, None) = Full True handicap car_park_state (AlmostFull _ 9 handicap) (Leaving, None) = Available True 8 handicap -- there are more states but I've cut them for brevity ---- This code isn't too bad but it's not obvious what the numbers '8', '11' and '9' are referring to without scrolling back up to the data type definition and checking. By coincidence I came across a http://en.wikibooks.org/wiki/Haskell/More_on_datatypes[Haskell data types wiki page] which introduced me to the concept of named fields. This means that rather than referring to the arguments of a data type by position we can give them a name instead: [source,haskell] ---- data CarParkState = Available { changed :: Bool, normalSpaces :: Int, handicapSpaces :: Int } | AlmostFull { changed :: Bool, normalSpaces :: Int, handicapSpaces :: Int } | Full { changed :: Bool, handicapSpaces :: Int } deriving (Show) ---- The +++<cite>+++car_park_state+++</cite>+++ function now reads like this: [source,haskell] ---- car_park_state :: CarParkState -> (Action, Sticker) -> CarParkState car_park_state state@(Available {normalSpaces=8}) (Entering, None) = AlmostFull True 9 (handicapSpaces state) car_park_state state@(AlmostFull {normalSpaces=11}) (Entering, None) = Full True (handicapSpaces state) car_park_state state@(AlmostFull {normalSpaces=9}) (Leaving, None) = Available True 8 (handicapSpaces state) ---- There's more code in this solution than the original although to me +++<cite>+++car_park_state+++</cite>+++ is clearer since you can tell what it's doing just by glancing at the code. I'm not sure which of these approaches is considered idiomatic Haskell but my feeling is that it's probably the former since it's more concise. On the other hand, the value of named fields goes up as the total number of parameters in a data types increases since people will find it increasingly difficult to remember which parameter is in each position.

haskell-pattern-matching-data-types-with-named-fields

https://markhneedham.com/blog/2012/03/31/haskell-pattern-matching-data-types-with-named-fields

2012-03-30 06:21:00

IntelliJ: Find/Replace using regular expressions with capture groups

Everyone now and then we end up having to write a bunch of mapping code and I quite like using IntelliJ's 'Replace' option to do it but always end up spending about 5 minutes trying to remember how to do capture groups so I thought I'd write it down this time. Given the following text in our file: [source,scala] ---- val mark = 0 val dave = 0 val john = 0 val alex = 0 ---- Let's say we wanted to prefix each of those names with 'cool' and had decided not to use Column mode for whatever reason. One way of doing that is to capture the names and then replace each of them with 'cool' appended on the beginning: A (very hacky) find regex could be this: [source,text] ---- \s([a-zA-Z]+)\s= ---- Where we capture all the letters in the variable name inside a group and then build our replacement string like so: [source,text] ---- cool$1 = ---- I always expect the capture group replacement syntax to be '\1', '\2' and so on but it actually uses a '$' instead. Hopefully now I shall remember!

intellij-findreplace-using-regular-expressions-with-capture-groups

https://markhneedham.com/blog/2012/03/30/intellij-findreplace-using-regular-expressions-with-capture-groups

2012-03-06 01:17:30

Choosing where to put the complexity

On the current application I'm working on we need to make use of some data which comes from another system so we've created an import script which creates a copy of that data so that we can use it in our application. In general we've been trying not to do too much manipulation of the data and keeping it close to the initial structure so that if something goes wrong with the import we can more easily trace the problem back to the original data source. image::{{<siteurl>}}/uploads/2012/03/complexity.gif[Complexity,474] While that approach has generally been fine we recently had a situation where the way the data was stored in the original database was quite de-normalised and recreating that structure made some of the code in the data access layer quite messy. We therefore decided to change the import script to normalise the data, thereby simplifying our database access code. For now the complexity trade off seems ok because we haven't had to change the schema that much so it'll still be reasonably easy to track the data back to the source. We currently don't have any tests specifically around the data import because there's not very much logic going on but if the complexity increases and we start to see problems with the import process then we'll need to change that. While we were having this conversation about where we should place the complexity I found it interesting that you could argue for each approach equally convincingly and neither really seemed definitively better - it was back to the world of http://www.markhneedham.com/blog/?s=trade+off[trade offs]!

choosing-where-to-put-the-complexity

https://markhneedham.com/blog/2012/03/06/choosing-where-to-put-the-complexity

2012-03-24 12:28:03

Haskell: Memoization using the power of laziness

I've been trying to solve http://projecteuler.net/problem=15[problem 15 of Project Euler] which requires you to find the number of routes that can be taken to navigate from the top corner of a grid down to the bottom right corner. For example there are six routes across a 2x2 grid: image::{{<siteurl>}}/uploads/2012/03/grid.gif[Grid,270] My initial solution looked like this: [source,haskell] ---- routes :: (Int, Int) -> Int -> Int routes origin size = inner origin size where inner origin@(x, y) size | x == size && y == size = 0 | x == size || y == size = 1 | otherwise = inner (x+1, y) size + inner (x, y+1) size ---- Which can be called like this: [source,haskell] ---- routes (0,0) 2 ---- Once we reach the edge of the grid i.e. 'x==size' or 'y==size' then we know there are no more routes from that path because you're not allowed to backtrack so we return a value of 1. At every other position we recurse twice, once going down the grid and once going across it to get all of the routes. This solution works fine for small sizes but it starts to take a serious amount of time to finish once 'size' gets above 11. Since the problem requires you to solve the problem for a grid size of 20 it doesn't suffice! From working through the problem on paper it was clear that a lot of the calculations were being repeated since there were many different ways of reaching each point on the grid. We therefore need to cache the calculations so that they wouldn't be repeated multiple times. The normal imperative language way of doing that would be to create a map or 2D array containing each grid position and then updating it once a grid position had been calculated. Unfortunately that doesn't really work with the normal array in Haskell because it's immutable and passing it around as a parameter doesn't seem to work either since we only have a new updated array when it's too late to be useful for other calculations. I came across a http://unspecified.wordpress.com/2011/11/04/lazy-dynamic-programming-with-arrays-in-haskell/[cool blog post by Matt Giuca] where he suggests that we need to think of things which have been computed and those which haven't rather than thinking about the problem in terms of which elements of the array have been mutated. He suggests creating an array where (in our case) the key represents a grid position and the value is a function call to work out how many routes we have from that position. Since Haskell is lazily evaluated that function won't be evaluated until that array position is accessed and once it's been calculated the value will be stored in the array for any future lookups. The code now ends up looking like this: [source,haskell] ---- routes :: Int -> Int routes size = arr ! (size, size) where arr = array ((0,0),(size,size)) [((x,y), inner (x,y) size) | x<-[0..size], y<-[0..size]] inner origin@(x, y) size | x == 0 && y == 0 = 0 | x == 0 || y == 0 = 1 | otherwise = arr ! (x-1, y) + arr ! (x, y-1) ---- We first create an array which contains entries for every position on the grid and a corresponding call to the function which works out the number of rotues from there. I had to change the way we navigate through the grid to be from the bottom right up to the top left corner since that made it much easier to just do a lookup on position (size, size) as the entry point. And now it returns instantly! [source,text] ---- > routes 20 137846528820 ----

haskell-memoization-using-the-power-of-laziness

https://markhneedham.com/blog/2012/03/24/haskell-memoization-using-the-power-of-laziness

2012-03-24 00:40:34

Saving the values of dynamically populated dropdown on back button

We wanted to be able to retain the value of a drop down menu that was being dynamically populated (via an AJAX call) when the user hit the back button but the AJAX request re-runs when we go hit back therefore losing our selection. Our initial thinking was that we might be able to store the value of the dropdown in a hidden field and then restore it into the dropdown using jQuery on page load but that approach didn't work since hidden fields don't seem to retain their values when you hit back. Another approach would be to update the URL with a new # value on each change of the dropdown and then parse the URL on page load to figure out which value should be selected. That would probably work but it seemed a bit too complicated. Eventually we realised that we could create an extra text box, 'hide' it by use of 'display:none' in our CSS and then copy the dropdown value into that every time it changed. We ended up with code something like this to capture the selected value: [source,javascript] ---- $("#dropdown-menu").live('change', function() { $("#hidden-field').val($(this).val()); }); ---- And then we populated the dropdown with the saved value on the callback from the AJAX call: [source,javascript] ---- $.ajax({ url: 'url/to/populate_dropdown', success: function(data) { $("#dropdown-menu").html(data); $("#dropdown-menu").val($("#hidden-field").val()); } }); ---- It seems to work pretty well and it's a simple technique as well so it worked in all the browsers that we tested it in.

saving-the-values-of-dynamically-populated-dropdown-on-back-button

https://markhneedham.com/blog/2012/03/24/saving-the-values-of-dynamically-populated-dropdown-on-back-button

2012-03-23 23:54:42

Oracle Spatial: Querying by a point/latitude/longitude

We're using Oracle Spatial on the application I'm working on and while most of the time any spatial queries we make are done from Java code we wanted to be able to run them directly from SQL as well to verify the code was working correctly. We normally end up forgetting how to construct a query so I thought I'd document it. Assuming we have a table +++<cite>+++table_with_shape+++</cite>+++ which has a column +++<cite>+++shape+++</cite>+++ which is a polygon, if we want to check whether a lat/long value interacts with that shape we can do that with the following query: [source,text] ---- SELECT * FROM table_with_shape tws WHERE SDO_ANYINTERACT(tws.shape, SDO_GEOMETRY(2001, 8307, SDO_POINT_TYPE(<LongValueHere>, <LatValueHere>, NULL), NULL, NULL) ) = 'TRUE' ---- The first parameter to +++<cite>+++SDO_GEOMETRY+++</cite>+++ http://docs.oracle.com/cd/B19306_01/appdev.102/b14255/sdo_objrelschema.htm#i1004087[defines the type of geometry] which in this case is a point. The second parameter is the coordinate system which is 8307 since we're using the 'WGS 84 longitude/latitude' system. The third parameter is our point and the rest of the parameters aren't interesting here so we pass null for them.

oracle-spatial-querying-by-a-pointlatitudelongitude

https://markhneedham.com/blog/2012/03/23/oracle-spatial-querying-by-a-pointlatitudelongitude