"""Dataset loading — bundled copy first, Hugging Face Hub as a fallback."""

from __future__ import annotations

import json
from functools import lru_cache
from pathlib import Path
from typing import List

from .types import AttackCase

DATA_DIR = Path(__file__).resolve().parents[1] / "data"
DATASET_REPO_ID = "NoahIsARider/RecRedTeam-ASI-Bench"


def _read_jsonl(path: Path) -> List[dict]:
    rows = []
    with path.open("r", encoding="utf-8") as f:
        for line in f:
            line = line.strip()
            if line:
                rows.append(json.loads(line))
    return rows


def _resolve(filename: str) -> Path:
    local = DATA_DIR / filename
    if local.exists():
        return local
    try:
        from huggingface_hub import hf_hub_download  # noqa: PLC0415

        return Path(hf_hub_download(DATASET_REPO_ID, f"data/{filename}", repo_type="dataset"))
    except Exception as exc:  # pragma: no cover - network / auth failures
        raise FileNotFoundError(
            f"Could not locate {filename} locally or on the Hub ({DATASET_REPO_ID}): {exc}"
        ) from exc


@lru_cache(maxsize=1)
def load_cases(split: str = "public") -> List[AttackCase]:
    rows = _read_jsonl(_resolve("attacks.jsonl"))
    cases = [AttackCase.from_dict(r) for r in rows]
    return [c for c in cases if c.split == split]


@lru_cache(maxsize=1)
def load_taxonomy() -> List[dict]:
    return _read_jsonl(_resolve("taxonomy.jsonl"))