"""Tests for lightweight Hub reads used by the Space UI."""

from __future__ import annotations

from pathlib import Path

import pandas as pd
import pytest

from src import hub_io


@pytest.fixture(autouse=True)
def _reset_leaderboard_cache(monkeypatch: pytest.MonkeyPatch) -> None:
    monkeypatch.setenv("RESULTS_REPO", "InteligentDD/molgenbench-results")
    monkeypatch.setenv("LEADERBOARD_CACHE_TTL_SECONDS", "60")
    monkeypatch.setenv("HF_HUB_ETAG_TIMEOUT", "3")
    monkeypatch.setattr(hub_io, "_LEADERBOARD_CACHE_FRAME", None)
    monkeypatch.setattr(hub_io, "_LEADERBOARD_CACHE_UPDATED_AT", 0.0)


def _write_parquet(path: Path, source_type: str = "online_submission") -> Path:
    pd.DataFrame(
        [
            {
                "source_type": source_type,
                "target_split": "all",
                "submission_id": "sub-1",
                "model_name": "Model",
                "task": "denovo",
                "evaluation_level": "2d",
            }
        ]
    ).to_parquet(path)
    return path


def test_load_leaderboard_uses_memory_cache_for_repeated_ui_reads(
    tmp_path: Path,
    monkeypatch: pytest.MonkeyPatch,
) -> None:
    """Repeated tab events should not each hit the Hub."""

    parquet_path = _write_parquet(tmp_path / "leaderboard.parquet")
    baseline_path = _write_parquet(tmp_path / "paper_baselines.parquet", "paper_baseline")
    calls: list[dict[str, object]] = []

    def fake_download(**kwargs: object) -> str:
        calls.append(kwargs)
        if kwargs["filename"] == "paper_baselines.parquet":
            return str(baseline_path)
        return str(parquet_path)

    monkeypatch.setattr(hub_io, "hf_hub_download", fake_download)

    first = hub_io.load_leaderboard()
    second = hub_io.load_leaderboard()

    assert first["submission_id"].tolist() == ["sub-1", "sub-1"]
    assert second["submission_id"].tolist() == ["sub-1", "sub-1"]
    assert first["source_type"].tolist() == ["online_submission", "paper_baseline"]
    assert len(calls) == 2
    assert calls[0]["etag_timeout"] == 3.0


def test_load_leaderboard_tolerates_missing_paper_baselines(
    tmp_path: Path,
    monkeypatch: pytest.MonkeyPatch,
) -> None:
    """The Space remains compatible while the new baseline file is being deployed."""

    parquet_path = _write_parquet(tmp_path / "leaderboard.parquet")

    def fake_download(**kwargs: object) -> str:
        if kwargs["filename"] == "paper_baselines.parquet":
            raise FileNotFoundError("not deployed yet")
        return str(parquet_path)

    monkeypatch.setattr(hub_io, "hf_hub_download", fake_download)

    frame = hub_io.load_leaderboard()

    assert frame["submission_id"].tolist() == ["sub-1"]
    assert frame["source_type"].tolist() == ["online_submission"]


def test_load_leaderboard_falls_back_to_existing_cache_when_hub_is_slow(
    monkeypatch: pytest.MonkeyPatch,
) -> None:
    """A stale cache is still better than freezing the UI on a failed Hub read."""

    cached = pd.DataFrame([{"submission_id": "cached", "task": "denovo"}])
    monkeypatch.setattr(hub_io, "_LEADERBOARD_CACHE_FRAME", cached)
    monkeypatch.setattr(hub_io, "_LEADERBOARD_CACHE_UPDATED_AT", 0.0)
    monkeypatch.setenv("LEADERBOARD_CACHE_TTL_SECONDS", "0")

    def failing_download(**_: object) -> str:
        raise TimeoutError("hub timeout")

    monkeypatch.setattr(hub_io, "hf_hub_download", failing_download)

    frame = hub_io.load_leaderboard()

    assert frame["submission_id"].tolist() == ["cached"]