app.py

#!/usr/bin/env python

import copy
import tempfile

import gradio as gr
from huggingface_hub import CommitOperationAdd, HfApi
import pandas as pd
from papers import PaperList

REPO_ID = "CVPR2024/CVPR2024-papers"
FILENAME = "data.csv"

api = HfApi()


paper_list = PaperList()

path = api.hf_hub_download(repo_id=REPO_ID, filename=FILENAME, repo_type="dataset")
actual_df = pd.read_csv(path)
paper_id_to_index = {str(row["id"]): i for i, row in actual_df.iterrows()}


with gr.Blocks() as demo_search:
    with gr.Group():
        search_title = gr.Textbox(label="Search title")
        search_author = gr.Textbox(label="Search author")
    df = gr.Dataframe(
        value=paper_list.df_prettified,
        datatype=paper_list.get_column_datatypes(paper_list.get_column_names()),
        type="pandas",
        row_count=(0, "dynamic"),
        interactive=False,
        height=1000,
        elem_id="table",
        wrap=True,
    )

    inputs = [
        search_title,
        search_author,
    ]
    gr.on(
        triggers=[
            search_title.submit,
            search_author.submit,
        ],
        fn=paper_list.search,
        inputs=inputs,
        outputs=df,
        queue=False,
        api_name=False,
    )
    demo_search.load(
        fn=paper_list.search,
        inputs=inputs,
        outputs=df,
        queue=False,
        api_name=False,
    )


def load_data(paper_id: str) -> tuple[str, str, str, str, str, str, str, str, str]:
    try:
        index = paper_id_to_index[paper_id]
    except KeyError:
        raise gr.Error(f"Paper ID {paper_id} not found.")
    
    paper = actual_df.iloc[index]

    return (
        paper["id"],
        paper["title"],
        paper["authors"],
        paper["arxiv_id"],
         "\n".join([PaperList.create_link("GitHub", url) for url in paper["GitHub"]] if paper["GitHub"]!="[]" else " "),
        "\n".join([PaperList.create_link(repo_id, f"https://huggingface.co/spaces/{repo_id}")
                    for repo_id in paper["Space"]
                    ] if paper["Space"] != "[]" else [" "]),
        "\n".join([PaperList.create_link(repo_id, f"https://huggingface.co/{repo_id}") for repo_id in paper["Model"]]
                    if paper["Model"] != "[]" else [" "]),
        "\n".join([PaperList.create_link(repo_id, f"https://huggingface.co/datasets/{repo_id}") for repo_id in paper["Dataset"]
                    ] if paper["Dataset"] != "[]" else [" "]
                )
    )


def split_and_strip(s: str) -> list[str]:
    return [x.strip() for x in s.split("\n") if x.strip()]


def create_pr(
    paper_id: str,
    title: str,
    authors: str,
    arxiv_id: str,
    project_page: str,
    github_links: str,
    space_ids: str,
    model_ids: str,
    dataset_ids: str,
    oauth_token: gr.OAuthToken | None,
) -> str:
    if oauth_token is None:
        return "Please log in first."

    try:
        index = paper_id_to_index[paper_id]
    except KeyError:
        raise gr.Error(f"Paper ID {paper_id} not found.")

    data = copy.deepcopy(df)
    data[index]["title"] = title.strip()
    data[index]["authors"] = authors
    data[index]["arxiv_id"] = arxiv_id.strip()
    data[index]["GitHub"] = github_links
    data[index]["Space"] = space_ids
    data[index]["Model"] = model_ids
    data[index]["Dataset"] = dataset_ids
    with tempfile.NamedTemporaryFile(mode="w", delete=False) as f:
        data.to_csv(f)

    commit = CommitOperationAdd(FILENAME, f.name)
    res = api.create_commit(
        repo_id=REPO_ID,
        operations=[commit],
        commit_message=f"Update {paper_id}",
        repo_type="dataset",
        create_pr=True,
        token=oauth_token.token,
    )
    return res.pr_url


with gr.Blocks() as demo_edit:
    with gr.Group():
        paper_id = gr.Textbox(label="ID", max_lines=1)
        load_button = gr.Button("Load")
    with gr.Group():
        title = gr.Textbox(label="Title", max_lines=1)
        authors = gr.Textbox(label="Authors", lines=5)
        arxiv_id = gr.Textbox(label="arXiv ID", max_lines=1, placeholder="2404.00000")
        github_links = gr.Textbox(
            label="GitHub links",
            lines=5,
            placeholder="https://github.com/aaa/bbb\nhttps://github.com/ccc/ddd",
        )
        space_ids = gr.Textbox(label="Space IDs", lines=5, placeholder="org_name1/repo_name1\norg_name2/repo_name2")
        model_ids = gr.Textbox(label="Model IDs", lines=5, placeholder="org_name1/repo_name1\norg_name2/repo_name2")
        dataset_ids = gr.Textbox(
            label="Dataset IDs", lines=5, placeholder="org_name1/repo_name1\norg_name2/repo_name2"
        )
    create_pr_button = gr.Button("Create PR")
    result = gr.Textbox(label="Result", max_lines=1)

    gr.on(
        triggers=[
            paper_id.submit,
            load_button.click,
        ],
        fn=load_data,
        inputs=paper_id,
        outputs=[
            paper_id,
            title,
            authors,
            arxiv_id,
            github_links,
            space_ids,
            model_ids,
            dataset_ids,
        ],
        queue=False,
        api_name=False,
    )
    create_pr_button.click(
        fn=create_pr,
        inputs=[
            paper_id,
            title,
            authors,
            arxiv_id,
            github_links,
            space_ids,
            model_ids,
            dataset_ids,
        ],
        outputs=result,
        queue=False,
        api_name=False,
    )

with gr.Blocks(css="style.css") as demo:
    gr.Markdown(
        "You can create PRs to update the CSV files in the [CVPR2024-papers repo](https://huggingface.co/datasets/CVPR2024/CVPR2024-papers) with this Space."
    )
    with gr.Tabs():
        with gr.Tab(label="Step 1: Login"):
            gr.Markdown("To create a PR, you first need to log in. Please press the login button below.")
            gr.LoginButton()
        with gr.Tab(label="Step 2: Search for paper ID"):
            gr.Markdown("Search for the paper you would like to update and find its paper ID.")
            demo_search.render()
        with gr.Tab(label="Step 3: Edit and create PR"):
            gr.Markdown("Enter the paper ID in the field below and press the Load button.")
            gr.Markdown("After making the necessary changes, press the Create PR button.")
            demo_edit.render()

if __name__ == "__main__":
    demo.queue(api_open=False).launch(show_api=False, debug=True)