data/util/prepare_mosaec_partial_structure.py

#!/usr/bin/env python3
from __future__ import annotations

import argparse
import shutil
from pathlib import Path

import pandas as pd

DATASET_DIRS = ("charged", "neutral")

COLUMN_MAP = {
    "density_g/cm3": "density_g_cm3",
    "lcd_ang_H2": "lcd_ang_H2",
    "pld_ang_H2": "pld_ang_H2",
    "asa_m2/cm3_H2": "asa_m2_cm3_H2",
    "asa_m2/g_H2": "asa_m2_g_H2",
    "void_fraction_H2": "void_fraction_H2",
    "av_ang3_H2": "av_ang3_H2",
    "av_cm3/g_H2": "av_cm3_g_H2",
    "void_fraction_probe-occupiable_H2": "void_fraction_probe-occupiable_H2",
    "av_probe-occupiable_ang3_H2": "av_probe-occupiable_ang3_H2",
    "av_probe-occupiable_cm3/g_H2": "av_probe-occupiable_cm3_g_H2",
}

TARGET_COLUMNS = [
    "cif",
    "density_g_cm3",
    "lcd_ang_H2",
    "pld_ang_H2",
    "asa_m2_cm3_H2",
    "asa_m2_g_H2",
    "void_fraction_H2",
    "av_ang3_H2",
    "av_cm3_g_H2",
    "void_fraction_probe-occupiable_H2",
    "av_probe-occupiable_ang3_H2",
    "av_probe-occupiable_cm3_g_H2",
]


def normalize_cif(series: pd.Series) -> pd.Series:
    cif = series.astype(str).str.strip()
    return cif.str.replace(r"\\.cif$", "", regex=True)


def load_properties_table(csv_path: Path) -> tuple[pd.DataFrame, set[str]]:
    required = ["cif", *COLUMN_MAP.keys()]
    df = pd.read_csv(csv_path)

    missing = [c for c in required if c not in df.columns]
    if missing:
        raise ValueError(f"Missing required columns in CSV: {', '.join(missing)}")

    df["cif"] = normalize_cif(df["cif"])

    duplicated = df["cif"].duplicated(keep=False)
    if duplicated.any():
        dup_count = int(duplicated.sum())
        print(
            f"[WARN] Found {dup_count} duplicated cif rows in CSV. "
            "Keeping the first occurrence for each cif."
        )

    props = df[["cif", *COLUMN_MAP.keys()]].copy()
    props = props.rename(columns=COLUMN_MAP)
    props = props.drop_duplicates(subset=["cif"], keep="first")
    props = props[TARGET_COLUMNS]

    return props, set(props["cif"])


def process_subset(subset_dir: Path, valid_cif: set[str], props: pd.DataFrame) -> dict[str, int]:
    raw_dir = subset_dir / "raw"
    raw_dir.mkdir(exist_ok=True)

    moved = 0
    deleted = 0

    # Move/delete CIFs from subset root
    for cif_path in subset_dir.glob("*.cif"):
        if cif_path.stem in valid_cif:
            destination = raw_dir / cif_path.name
            if destination.exists():
                cif_path.unlink()
            else:
                shutil.move(str(cif_path), str(destination))
                moved += 1
        else:
            cif_path.unlink()
            deleted += 1

    # Clean invalid CIFs inside raw/
    for cif_path in raw_dir.glob("*.cif"):
        if cif_path.stem not in valid_cif:
            cif_path.unlink()
            deleted += 1

    raw_cif = sorted(p.stem for p in raw_dir.glob("*.cif") if p.stem in valid_cif)

    props_indexed = props.set_index("cif")
    available = [c for c in raw_cif if c in props_indexed.index]
    missing_in_csv = [c for c in raw_cif if c not in props_indexed.index]

    if missing_in_csv:
        print(
            f"[WARN] {subset_dir.name}: {len(missing_in_csv)} files in raw/ "
            "not found in CSV; skipped in id_prop.csv"
        )

    id_prop = props_indexed.loc[available].reset_index()
    id_prop_path = subset_dir / "id_prop.csv"
    id_prop.to_csv(id_prop_path, index=False)

    return {
        "moved": moved,
        "deleted": deleted,
        "raw_count": len(raw_cif),
        "id_prop_rows": len(id_prop),
    }


def parse_args() -> argparse.Namespace:
    parser = argparse.ArgumentParser(
        description=(
            "Prepare MOSAEC-DB full/charged and full/neutral datasets: "
            "move valid .cif files into raw/ and create id_prop.csv"
        )
    )
    parser.add_argument(
        "--csv",
        default="mosaec-db.csv",
        help="Path to source CSV (default: mosaec-db.csv)",
    )
    parser.add_argument(
        "--root",
        default=".",
        help="Root directory containing charged and neutral folders (default: current directory)",
    )
    return parser.parse_args()


def main() -> None:
    args = parse_args()
    root = Path(args.root).resolve()
    csv_path = (root / args.csv).resolve() if not Path(args.csv).is_absolute() else Path(args.csv)

    if not csv_path.exists():
        raise FileNotFoundError(f"CSV file not found: {csv_path}")

    props, valid_cif = load_properties_table(csv_path)

    print(f"Loaded {len(valid_cif)} unique cif values from: {csv_path}")
    print()

    for subset_name in DATASET_DIRS:
        subset_dir = root / subset_name
        if not subset_dir.exists() or not subset_dir.is_dir():
            print(f"[WARN] Skip {subset_name}: folder not found at {subset_dir}")
            continue

        stats = process_subset(subset_dir, valid_cif, props)
        print(
            f"{subset_name}: moved={stats['moved']}, deleted={stats['deleted']}, "
            f"raw_files={stats['raw_count']}, id_prop_rows={stats['id_prop_rows']}"
        )


if __name__ == "__main__":
    main()