"""Build the per-plot modeling dataset from per-tuber morphometrics and ground truth.

QC-filter the per-tuber dataframe, aggregate to (trial, Plot_std), merge
with the cleaned ground truth, and compute the regression target
obs_sum_B_A1_A2_OV.

Usage:
    uv run python scripts/build_plot_model_dataset.py \\
        --tubers outputs/df_class_real_scale_ALL.csv \\
        --out    outputs/df_model_yield_by_plot.csv
"""

from __future__ import annotations

import argparse
from pathlib import Path

import pandas as pd

from potatopheno.aggregation import aggregate_by_plot
from potatopheno.filtering import (
    ADC_QC_DROP_PLOTS,
    derive_axis_ratio_pca,
    drop_qc_plots,
    filter_axis_ratio,
)
from potatopheno.harvest import add_plot_std, merge_with_ground_truth


def _parse_args() -> argparse.Namespace:
    p = argparse.ArgumentParser(description=__doc__.splitlines()[0])
    p.add_argument(
        "--tubers",
        type=Path,
        required=True,
        help="Per-tuber morphometrics CSV produced by extract_morphometrics.py.",
    )
    p.add_argument(
        "--gt",
        type=Path,
        required=True,
        help="Cleaned ground-truth CSV.",
    )
    p.add_argument(
        "--out",
        type=Path,
        required=True,
        help="Path of the plot-level model CSV to write.",
    )
    p.add_argument(
        "--qc-drop-plots",
        type=lambda s: tuple(t.strip() for t in s.split(",") if t.strip()),
        default=ADC_QC_DROP_PLOTS,
        help="Comma-separated list of Plot IDs (image_id[-8:]) to drop as QC. "
        f"Default: {','.join(ADC_QC_DROP_PLOTS)}.",
    )
    p.add_argument(
        "--no-ar-filter",
        action="store_true",
        help="Skip the axis-ratio-PCA filter entirely.",
    )
    return p.parse_args()


def main() -> None:
    args = _parse_args()

    df_tubers = pd.read_csv(args.tubers)
    n0 = len(df_tubers)

    df = drop_qc_plots(df_tubers, drop_plots=args.qc_drop_plots)
    n1 = len(df)

    df = derive_axis_ratio_pca(df)
    if not args.no_ar_filter:
        df = filter_axis_ratio(df)
    n2 = len(df)

    df = add_plot_std(df)

    df_agg = aggregate_by_plot(df)
    df_gt = pd.read_csv(args.gt)
    df_model = merge_with_ground_truth(df_agg, df_gt)

    args.out.parent.mkdir(parents=True, exist_ok=True)
    df_model.to_csv(args.out, index=False)

    print(f"qc_drop_plots={list(args.qc_drop_plots)}  ar_filter={'OFF' if args.no_ar_filter else 'ON'}")
    print(
        f"tubers: {n0} → after qc-drop: {n1} → "
        f"after AR filter: {n2}"
    )
    print(
        f"plots: {df_model['Plot_std'].nunique()} across "
        f"{df_model['trial'].nunique()} trials"
    )
    print(f"[DONE] wrote {len(df_model)} rows → {args.out}")


if __name__ == "__main__":
    main()