import shutil
from rocksdict import Rdict, Options, SliceTransform, PlainTableFactoryOptions, DBCompressionType
import os
import argparse
import numpy as np
import pandas as pd
import smart_open
import json 
import hashlib
import tqdm
from indexed_zstd import IndexedZstdFile

def db_options():
    opt = Options()
    # create table
    opt.create_if_missing(True)
    # config to more jobs
    opt.set_max_background_jobs(8)
    # configure mem-table to a large value (1024 MB)
    opt.set_write_buffer_size(1024*1024*1024)
    # configure l0 and l1 size, let them have the same size (1 GB)
    opt.set_max_bytes_for_level_base(1024*1024*1024)
    # 1024 MB file size
    opt.set_target_file_size_base(1024*1024*1024)
    # use a smaller compaction multiplier
    opt.set_max_bytes_for_level_multiplier(4.0)
    # use 8-byte prefix (2 ^ 64 is far enough for transaction counts)
    opt.set_prefix_extractor(SliceTransform.create_max_len_prefix(8))
    opt.set_compression_type(DBCompressionType.none())
    return opt


# create DB
emb_db = Rdict("emb_rdict/", db_options())

existing_keys = set()
for key in tqdm.tqdm(emb_db.keys(), desc="add keys in memory"):
    existing_keys.add(key)


parser = argparse.ArgumentParser(description='Add data to Rdict database.')
parser.add_argument('--input', type=str, nargs="+")
parser.add_argument('--emb', type=str, nargs="+")
args = parser.parse_args()

input_paths = sorted(args.input)
emb_paths = sorted(args.emb)

assert len(input_paths) == len(emb_paths)

for input_path, emb_path in tqdm.tqdm(zip(input_paths, emb_paths), total=len(input_paths), desc="process files"):
    print(f"Open input {input_path} with {emb_path}")

    emb = np.load(emb_path)
    print("Embeddings:", len(emb))

    #Load the texts
    if input_path.endswith(".parquet"):
        print("Read parquet")
        df = pd.read_parquet(input_path)
        texts = [t.strip() for t in df['text']]
    elif input_path.endswith(".jsonl") or input_path.endswith(".jsonl.gz"):
        texts = []
        with smart_open.open(input_path, "rt") as fIn:
            for line in fIn:
                texts.append(json.loads(line)['text'].strip())
    elif input_path.endswith(".zst"):
        texts = []
        with IndexedZstdFile(input_path) as fIn:
            for line in fIn:
                texts.append(json.loads(line)['text'].strip())
    else:
        raise ValueError("Unknown input format")

    assert len(texts) == len(emb)
   
    #Add embeddings to the embeddings DB
    for idx in tqdm.trange(len(texts), desc="add to rdict"):
        emb_id = hashlib.md5(texts[idx].encode()).digest()
        if emb_id not in existing_keys:
            emb_db[emb_id] = emb[idx]
            existing_keys.add(emb_id)

print("DONE")