"""raw.csv + labels.csv -> features.csv (one row per measurement).

    python experiments/make_features.py data/demo_calibration
    python experiments/make_features.py data/2026-11_E1 --heater-steps 5

The folder must contain raw.csv (written by record.py, can be several days
concatenated) and labels.csv (typed by hand, one row per sample_id).
"""
from __future__ import annotations

import argparse
import sys
from pathlib import Path

import pandas as pd

sys.path.insert(0, str(Path(__file__).resolve().parents[1]))

from enose.features import build_feature_table, odor_columns  # noqa: E402


def main() -> None:
    parser = argparse.ArgumentParser(description=__doc__, formatter_class=argparse.RawDescriptionHelpFormatter)
    parser.add_argument("folder", type=Path)
    parser.add_argument("--heater-steps", type=int, default=5)
    arguments = parser.parse_args()

    raw = pd.read_csv(arguments.folder / "raw.csv")
    labels = pd.read_csv(arguments.folder / "labels.csv")
    # A sample measured twice has more than one block of "purge" rows (two readings in the
    # same phase_second are normal: the device clock and the laptop clock drift a little).
    starts = raw["phase"].eq("purge") & raw["phase"].ne(raw["phase"].shift()) | raw["sample_id"].ne(
        raw["sample_id"].shift()) & raw["phase"].eq("purge")
    purge_blocks = starts.groupby(raw["sample_id"]).sum()
    merged = sorted(purge_blocks[purge_blocks > 1].index)
    if merged:
        print(f"注意：{len(merged)} 個樣本編號被量了不只一次，資料會混在一起，請改編號後重算：{merged[:5]}")
    short = raw[raw["phase"].isin(["purge", "sample"])].groupby(["sample_id", "phase"])["phase_second"].max()
    too_short = sorted(short[short < 60].index.get_level_values("sample_id").unique())
    if too_short:
        print(f"注意：{len(too_short)} 個樣本的換氣或量測階段不到 60 秒：{too_short[:5]}")
    table = build_feature_table(raw, labels, arguments.heater_steps)
    missing = sorted(set(labels["sample_id"]) - set(table["sample_id"]))
    if missing:
        print(f"注意：labels.csv 有 {len(missing)} 個樣本在 raw.csv 找不到：{missing[:5]}…")
    broken = table[table[odor_columns(table)].isna().any(axis=1)]
    if len(broken):
        print(f"注意：{len(broken)} 次量測有缺值（可能某一階段太短）：{list(broken['sample_id'][:5])}")
    table.to_csv(arguments.folder / "features.csv", index=False)
    print(f"已寫入 {arguments.folder / 'features.csv'}：{len(table)} 次量測 × {len(odor_columns(table))} 個氣味特徵")


if __name__ == "__main__":
    main()