submissions.py

"""Submissions queue backed by the existing hugging-science/feedback HF Dataset.

`requests.jsonl` holds every non-"feedback" type (dataset/model/organization/
blog/challenge/collaboration); `feedback.jsonl` holds "feedback". Both are
written by the feedback-api Space's /submit endpoint. Ingestion happens
there, not here — this module only reads those files and updates a row's
status when the reviewer approves/rejects/acknowledges it.
"""

import json
import threading
from datetime import datetime, timezone

from huggingface_hub import HfApi
from huggingface_hub.utils import EntryNotFoundError

from about import DATASET_REPO, FEEDBACK_FILE, REQUESTS_FILE, TOKEN

_api = HfApi(token=TOKEN)
_lock = threading.Lock()

ALL_FILES = (REQUESTS_FILE, FEEDBACK_FILE)


def _now() -> str:
    return datetime.now(timezone.utc).isoformat()


def _read_file(filename: str) -> list[dict]:
    try:
        path = _api.hf_hub_download(
            repo_id=DATASET_REPO, repo_type="dataset", filename=filename
        )
    except EntryNotFoundError:
        return []
    with open(path, encoding="utf-8") as f:
        return [json.loads(line) for line in f if line.strip()]


def _write_file(filename: str, rows: list[dict]) -> None:
    content = "\n".join(json.dumps(row) for row in rows) + "\n"
    _api.upload_file(
        path_or_fileobj=content.encode("utf-8"),
        path_in_repo=filename,
        repo_id=DATASET_REPO,
        repo_type="dataset",
        commit_message="Update requests queue",
    )


def _read_all() -> list[dict]:
    return [row for filename in ALL_FILES for row in _read_file(filename)]


def list_submissions(status: str | None = None, type_: str | None = None) -> list[dict]:
    rows = _read_all()
    if status is not None:
        rows = [r for r in rows if r["status"] == status]
    if type_ is not None:
        rows = [r for r in rows if r["type"] == type_]
    return rows


def get_submission(submission_id: str) -> dict | None:
    for row in _read_all():
        if row["id"] == submission_id:
            return row
    return None


def _update_row(submission_id: str, mutate_fn) -> None:
    """Find `submission_id` in whichever file its type lives in, mutate it in
    place, and rewrite only that file."""
    with _lock:
        for filename in ALL_FILES:
            rows = _read_file(filename)
            for row in rows:
                if row["id"] == submission_id:
                    mutate_fn(row)
                    _write_file(filename, rows)
                    return


def mark_approved(submission_id: str, pr_url: str) -> None:
    def mutate(row):
        row["status"] = "approved"
        row["reviewed_at"] = _now()
        row["pr_url"] = pr_url

    _update_row(submission_id, mutate)


def mark_acknowledged(submission_id: str) -> None:
    """For non-PR types (feedback, collaboration, challenge): close out a
    submission without opening a PR — e.g. once it's been read or the
    requester has been emailed."""

    def mutate(row):
        row["status"] = "acknowledged"
        row["reviewed_at"] = _now()

    _update_row(submission_id, mutate)


def mark_rejected(submission_id: str, reason: str | None) -> None:
    def mutate(row):
        row["status"] = "rejected"
        row["reviewed_at"] = _now()
        row["reject_reason"] = reason

    _update_row(submission_id, mutate)