paper_list.py

from __future__ import annotations

import numpy as np
import pandas as pd
import requests
from huggingface_hub.hf_api import SpaceInfo


class PaperList:
    def __init__(self):
        self.organization_name = 'ICML2022'
        self.table = pd.read_csv('papers.csv')
        self._preprcess_table()

        self.table_header = '''
            <tr>
                <td width="50%">Paper</td>
                <td width="26%">Authors</td>
                <td width="4%">pdf</td>
                <td width="4%">arXiv</td>
                <td width="4%">GitHub</td>
                <td width="4%">HF Spaces</td>
                <td width="4%">HF Models</td>
                <td width="4%">HF Datasets</td>
            </tr>'''

    @staticmethod
    def load_space_info(author: str) -> list[SpaceInfo]:
        path = 'https://huggingface.co/api/spaces'
        r = requests.get(path, params={'author': author})
        d = r.json()
        return [SpaceInfo(**x) for x in d]

    def add_spaces_to_table(self, organization_name: str,
                            df: pd.DataFrame) -> pd.DataFrame:
        spaces = self.load_space_info(organization_name)
        name2space = {
            s.id.split('/')[1].lower(): f'https://huggingface.co/spaces/{s.id}'
            for s in spaces
        }
        df['hf_space'] = df.loc[:, ['hf_space', 'github']].apply(
            lambda x: x[0] if isinstance(x[0], str) else name2space.get(
                x[1].split('/')[-1].lower()
                if isinstance(x[1], str) else '', np.nan),
            axis=1)
        return df

    def _preprcess_table(self) -> None:
        self.table = self.add_spaces_to_table(self.organization_name,
                                              self.table)
        self.table['title_lowercase'] = self.table.title.str.lower()

        rows = []
        for row in self.table.itertuples():
            paper = f'<a href="{row.url}" target="_blank">{row.title}</a>'
            pdf = f'<a href="{row.pdf}" target="_blank">pdf</a>'
            arxiv = f'<a href="{row.arxiv}" target="_blank">arXiv</a>' if isinstance(
                row.arxiv, str) else ''
            github = f'<a href="{row.github}" target="_blank">GitHub</a>' if isinstance(
                row.github, str) else ''
            hf_space = f'<a href="{row.hf_space}" target="_blank">Space</a>' if isinstance(
                row.hf_space, str) else ''
            hf_model = f'<a href="{row.hf_model}" target="_blank">Model</a>' if isinstance(
                row.hf_model, str) else ''
            hf_dataset = f'<a href="{row.hf_dataset}" target="_blank">Dataset</a>' if isinstance(
                row.hf_dataset, str) else ''
            row = f'''
                <tr>
                    <td>{paper}</td>
                    <td>{row.authors}</td>
                    <td>{pdf}</td>
                    <td>{arxiv}</td>
                    <td>{github}</td>
                    <td>{hf_space}</td>
                    <td>{hf_model}</td>
                    <td>{hf_dataset}</td>
                </tr>'''
            rows.append(row)
        self.table['html_table_content'] = rows

    def render(self, search_query: str, case_sensitive: bool,
               filter_names: list[str]) -> tuple[int, str]:
        df = self.add_spaces_to_table(self.organization_name, self.table)
        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 = 'HF Space' in filter_names
        has_hf_model = 'HF Model' in filter_names
        has_hf_dataset = 'HF Dataset' in filter_names
        df = self.filter_table(df, has_arxiv, has_github, has_hf_space,
                               has_hf_model, has_hf_dataset)
        return len(df), 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.isna()]
        if has_github:
            df = df[~df.github.isna()]
        if has_hf_space:
            df = df[~df.hf_space.isna()]
        if has_hf_model:
            df = df[~df.hf_model.isna()]
        if has_hf_dataset:
            df = df[~df.hf_dataset.isna()]
        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