"""Load the master config (config/config.yaml) as a plain dict.

The YAML file is the single source of truth for paths, grid, index definition, lead
blocks and the cross-validation scheme.
"""
from __future__ import annotations
import os
import yaml

_DEFAULT = os.path.join(
    os.path.dirname(os.path.dirname(os.path.dirname(os.path.abspath(__file__)))),
    "config", "config.yaml",
)


def load_config(path: str | None = None) -> dict:
    with open(path or _DEFAULT, "r") as f:
        return yaml.safe_load(f)


def lead_blocks(cfg: dict) -> dict[str, tuple[int, int]]:
    """Return {block_name: (day_start, day_end)} inclusive, both ends."""
    return {k: (int(v[0]), int(v[1])) for k, v in cfg["lead_blocks"].items()}


def block_predictors(cfg: dict, block: str) -> list[str]:
    """Lead-dependent predictor NAME list for one block: backbone + the block's extras.

    Falls back to the backbone alone if `predictors` is absent or the block has no entry.
    Order = backbone first, then the block extras in config order (deduplicated).
    """
    spec = cfg.get("predictors")
    if not spec:
        from .features import FEATURE_NAMES
        return list(FEATURE_NAMES)
    names = list(spec.get("backbone", []))
    for n in spec.get("by_block", {}).get(block, []):
        if n not in names:
            names.append(n)
    return names


def block_predictor_index(cfg: dict, block: str, full_names) -> list[int]:
    """Column indices of the block's lead-dependent predictors into `full_names`.

    `full_names` is the ordered candidate-matrix column list (e.g.
    features.FEATURE_NAMES + features_ext.EXT_FEATURE_NAMES + features_ext.GRID_FEATURE_NAMES).
    Raises if a requested predictor is not present, so a config typo fails loudly.
    """
    idx = {n: i for i, n in enumerate(full_names)}
    want = block_predictors(cfg, block)
    missing = [n for n in want if n not in idx]
    if missing:
        raise KeyError(f"block {block}: predictors not in candidate matrix: {missing}")
    return [idx[n] for n in want]