paper_list.py

from __future__ import annotations

import pandas as pd


class PaperList:
    def __init__(self):
        self.organization_name = "ICML2023"
        self.table = pd.read_json("papers.json").fillna("")

        claim_info = pd.read_csv("claim_info.csv", dtype={"arxiv_id": str, "n_authors": int, "n_linked_authors": int})
        self.table = pd.merge(self.table, claim_info, on="arxiv_id", how="left")
        self.table[["n_authors", "n_linked_authors"]] = (
            self.table[["n_authors", "n_linked_authors"]].fillna(-1).astype(int)
        )

        self._preprocess_table()

        self.table_header = """
            <tr>
                <td width="38%">Title</td>
                <td width="20%">Authors</td>
                <td width="5%">Type</td>
                <td width="5%">arXiv</td>
                <td width="5%">GitHub</td>
                <td width="7%">Paper pages</td>
                <td width="5%">Spaces</td>
                <td width="5%">Models</td>
                <td width="5%">Datasets</td>
                <td width="5%">Claimed</td>
            </tr>"""

    def _preprocess_table(self) -> None:
        self.table["title_lowercase"] = self.table.title.str.lower()
        self.table["arxiv"] = self.table.arxiv_id.apply(lambda x: f"https://arxiv.org/abs/{x}" if x else "")
        self.table["hf_paper"] = self.table.arxiv_id.apply(lambda x: f"https://huggingface.co/papers/{x}" if x else "")
        self.table["authors"] = self.table.authors.apply(lambda x: ", ".join(x))

        rows = []
        for row in self.table.itertuples():
            title = f'<a href="{row.url}" target="_blank">{row.title}</a>'
            arxiv = f'<a href="{row.arxiv}" target="_blank">arXiv</a>' if row.arxiv else ""
            github = f'<a href="{row.github}" target="_blank">GitHub</a>' if row.github else ""
            hf_paper = f'<a href="{row.hf_paper}" target="_blank">Paper page</a>' if row.hf_paper else ""
            hf_space = f'<a href="{row.hf_space}" target="_blank">Space</a>' if row.hf_space else ""
            hf_model = f'<a href="{row.hf_model}" target="_blank">Model</a>' if row.hf_model else ""
            hf_dataset = f'<a href="{row.hf_dataset}" target="_blank">Dataset</a>' if row.hf_dataset else ""
            author_linked = "✅" if row.n_linked_authors > 0 else ""
            n_linked_authors = "" if row.n_linked_authors == -1 else row.n_linked_authors
            n_authors = "" if row.n_authors == -1 else row.n_authors
            claimed_paper = "" if n_linked_authors == "" else f"{n_linked_authors}/{n_authors} {author_linked}"
            row = f"""
                <tr>
                    <td>{title}</td>
                    <td>{row.authors}</td>
                    <td>{row.type}</td>
                    <td>{arxiv}</td>
                    <td>{github}</td>
                    <td>{hf_paper}</td>
                    <td>{hf_space}</td>
                    <td>{hf_model}</td>
                    <td>{hf_dataset}</td>
                    <td>{claimed_paper}</td>
                </tr>"""
            rows.append(row)
        self.table["html_table_content"] = rows

    def render(
        self,
        search_query: str,
        case_sensitive: bool,
        filter_names: list[str],
        presentation_type: str,
    ) -> tuple[str, str]:
        df = self.table
        if presentation_type != "(ALL)":
            df = df[df.type == presentation_type.lower()]
        if search_query:
            if case_sensitive:
                df = df[df.title.str.contains(search_query)]
            else:
                df = df[df.title_lowercase.str.contains(search_query.lower())]
        has_arxiv = "arXiv" in filter_names
        has_github = "GitHub" in filter_names
        has_hf_space = "Space" in filter_names
        has_hf_model = "Model" in filter_names
        has_hf_dataset = "Dataset" in filter_names
        df = self.filter_table(df, has_arxiv, has_github, has_hf_space, has_hf_model, has_hf_dataset)
        n_claimed = len(df[df.n_linked_authors > 0])
        return f"{len(df)} ({n_claimed} claimed)", self.to_html(df, self.table_header)

    @staticmethod
    def filter_table(
        df: pd.DataFrame,
        has_arxiv: bool,
        has_github: bool,
        has_hf_space: bool,
        has_hf_model: bool,
        has_hf_dataset: bool,
    ) -> pd.DataFrame:
        if has_arxiv:
            df = df[df.arxiv != ""]
        if has_github:
            df = df[df.github != ""]
        if has_hf_space:
            df = df[df.hf_space != ""]
        if has_hf_model:
            df = df[df.hf_model != ""]
        if has_hf_dataset:
            df = df[df.hf_dataset != ""]
        return df

    @staticmethod
    def to_html(df: pd.DataFrame, table_header: str) -> str:
        table_data = "".join(df.html_table_content)
        html = f"""
        <table>
            {table_header}
            {table_data}
        </table>"""
        return html