from pathlib import Path

# Replace hardcoded path with Hugging Face-aware fallback
from src.utils.paths import get_processed_path, _hf_download

def _load_defaults(dataset: str) -> Dict[str, Dict[str, Any]]:
    """
    Load defaults.json for a dataset.
    Try local path first; fall back to HF hub if needed.
    """
    try:
        fp = get_processed_path(dataset) / "index" / "defaults.json"
        if fp.exists():
            return json.loads(fp.read_text())
    except Exception:
        pass
    try:
        # fallback (root-level for HF structure)
        return json.loads(_hf_download("json/defaults.json").read_text())
    except Exception:
        return {}


# Likewise for these load functions:

def _load_user_vec(proc: Path, user_id: str) -> np.ndarray:
    try:
        dfu = _read_parquet(proc / "user_text_emb.parquet", ["user_id", "vector"])
    except FileNotFoundError:
        dfu = pd.read_parquet(_hf_download("parquet/user_text_emb.parquet"), columns=["user_id", "vector"])
    row = dfu[dfu["user_id"] == user_id]
    if row.empty:
        raise ValueError(f"user_id '{user_id}' not found. Run text embedding step.")
    v = np.asarray(row.iloc[0]["vector"], dtype=np.float32)
    return v / (np.linalg.norm(v) + 1e-12)


def _load_items_table(proc: Path) -> pd.DataFrame:
    try:
        items = _read_parquet(proc / "items_with_meta.parquet")
    except FileNotFoundError:
        items = pd.read_parquet(_hf_download("parquet/items_with_meta.parquet"))
    if ITEM_KEY not in items.columns:
        if items.index.name == ITEM_KEY:
            items = items.reset_index()
        else:
            raise KeyError(f"'{ITEM_KEY}' not found in items_with_meta.parquet")
    return items


def _user_seen_items(proc: Path, user_id: str) -> set:
    try:
        df = _read_parquet(proc / "reviews.parquet", ["user_id", ITEM_KEY])
    except FileNotFoundError:
        df = pd.read_parquet(_hf_download("parquet/reviews.parquet"), columns=["user_id", ITEM_KEY])
    return set(df[df["user_id"] == user_id][ITEM_KEY].tolist())