"""E1 sensor characteristics: rise time, steady response, recovery time, repeatability (CV).

    python experiments/e1_sensor_characteristics.py data/demo_calibration

Needs raw.csv and labels.csv (columns: sample_id, label, concentration; rh_level optional,
only rh_level == "mid" is used so humidity does not mix in). Writes to <folder>/results/:
    e1_per_measurement.csv   one row per measurement and channel
    e1_summary.csv           mean, SD and CV per gas x concentration x channel
    e1_curves.png            average response curve per gas
"""
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.common import output_dir, plt, save_figure  # noqa: E402
from enose.features import _tail_mean, recovery_time, rise_time_90  # noqa: E402

# (name in the report, raw column, heater step or None)
CHANNELS: tuple[tuple[str, str, int | None], ...] = (
    ("BME688 280°C", "gas_ohm", 2),
    ("SGP30 乙醇", "ethanol_raw", None),
    ("SCD40 CO2", "co2_ppm", None),
)


def channel_rows(raw: pd.DataFrame, column: str, step: int | None) -> pd.DataFrame:
    rows = raw if step is None else raw[raw["heater_step"] == step]
    return rows[rows[column] >= 0]


def main() -> None:
    parser = argparse.ArgumentParser(description=__doc__, formatter_class=argparse.RawDescriptionHelpFormatter)
    parser.add_argument("folder", type=Path)
    arguments = parser.parse_args()
    raw = pd.read_csv(arguments.folder / "raw.csv")
    labels = pd.read_csv(arguments.folder / "labels.csv")
    if "rh_level" in labels:
        labels = labels[labels["rh_level"] == "mid"]
    results = output_dir(arguments.folder / "results")

    records = []
    for sample in labels.itertuples():
        measurement = raw[raw["sample_id"] == sample.sample_id]
        for name, column, step in CHANNELS:
            rows = channel_rows(measurement, column, step)
            purge, during, recover = (rows[rows["phase"] == phase] for phase in ("purge", "sample", "recover"))
            baseline, steady = _tail_mean(purge, column), _tail_mean(during, column)
            records.append({
                "sample_id": sample.sample_id, "gas": sample.label, "concentration": sample.concentration,
                "channel": name, "baseline": baseline, "steady": steady,
                "relative_change": (steady - baseline) / abs(baseline) if "CO2" not in name else steady - baseline,
                "rise_time_90_s": rise_time_90(during, column, baseline),
                "recovery_time_s": recovery_time(recover, column, baseline),
            })
    table = pd.DataFrame(records)
    table.to_csv(results / "e1_per_measurement.csv", index=False)

    summary = table.groupby(["channel", "gas", "concentration"]).agg(
        n=("relative_change", "size"), mean_change=("relative_change", "mean"), sd_change=("relative_change", "std"),
        rise_time_90_s=("rise_time_90_s", "mean"), recovery_time_s=("recovery_time_s", "mean")).reset_index()
    summary["cv_pct"] = 100 * summary["sd_change"] / summary["mean_change"].abs()
    summary.to_csv(results / "e1_summary.csv", index=False)
    print(summary.round(3).to_string(index=False))
    print(f"\nCV < 10% 的條件：{(summary['cv_pct'] < 10).sum()} / {len(summary)}")

    figure, axes = plt.subplots(1, len(CHANNELS), figsize=(5 * len(CHANNELS), 3.6))
    middle = labels[labels["concentration"] == sorted(labels["concentration"].unique())[len(labels["concentration"].unique()) // 2]]
    for axis, (name, column, step) in zip(axes, CHANNELS):
        for gas, group in middle.groupby("label"):
            curves = []
            for sample_id in group["sample_id"]:
                measurement = raw[raw["sample_id"] == sample_id]
                rows = channel_rows(measurement, column, step).copy()
                # 10-second bins from the start of the measurement, so curves of different samples line up
                rows["t"] = (rows["device_ms"] - measurement["device_ms"].min()) // 10000 * 10
                curves.append(rows.groupby("t")[column].mean())
            mean_curve = pd.concat(curves, axis=1).mean(axis=1)
            axis.plot(mean_curve.index, mean_curve.to_numpy(), label=gas)
        axis.set_title(name)
        axis.set_xlabel("時間（秒）")
    axes[0].legend()
    save_figure(figure, results / "e1_curves.png")


if __name__ == "__main__":
    main()