merge_euhydro.py

"""Convert per-basin EU-Hydro GeoPackages into curated GeoParquet shards.

For each `euhydro_*_v013_GPKG.zip` in the current directory: extract the
package to a temp dir once, pull the layers that matter for water analysis,
drop Z/M dimensions and admin-only columns, write one GeoParquet per basin
per layer under `eu_hydro_master_skeleton_geoparquet/<layer_dir>/`, and
delete the extracted temp dir.
"""

from __future__ import annotations

import glob
import shutil
import sys
import tempfile
import time
import zipfile
from pathlib import Path

import geopandas as gpd
import pandas as pd


EXCLUDED_BASINS = ("guiana", "iceland", "islands")

LAYER_OUTPUT_DIR = {
    "River_Net_l": "river_lines",
    "River_Net_p": "river_polygons",
    "InlandWater": "inland_water",
    "RiverBasins": "river_basins",
}

DROP_COLS = {"BEGLIFEVER", "ENDLIFEVER", "UPDAT_BY", "UPDAT_WHEN"}

OUTPUT_DIR = Path("eu_hydro_master_skeleton_geoparquet")


def basin_from_zip(zip_path: str) -> str:
    return Path(zip_path).stem.removeprefix("euhydro_").removesuffix("_v013_GPKG")


def extract_gpkg(zip_path: str, dest_dir: Path) -> Path | None:
    """Extract only the primary basin .gpkg from the zip into dest_dir and return its path."""
    with zipfile.ZipFile(zip_path) as zf:
        names = zf.namelist()
        target = next(
            (n for n in names if n.lower().endswith(".gpkg") and "drainage_network" not in n.lower()),
            None,
        )
        if target is None:
            return None
        zf.extract(target, dest_dir)
        return dest_dir / target


def process_basin(zip_path: str, target_crs) -> tuple[list[dict], object]:
    basin = basin_from_zip(zip_path)
    rows: list[dict] = []
    tmp_root = Path(tempfile.mkdtemp(prefix=f"euhydro_{basin}_"))
    try:
        t0 = time.perf_counter()
        gpkg = extract_gpkg(zip_path, tmp_root)
        if gpkg is None:
            print(f"  skip {basin}: no matching .gpkg member in zip", flush=True)
            return [], target_crs
        print(f"  extracted in {time.perf_counter() - t0:.1f}s", flush=True)

        for layer, subdir in LAYER_OUTPUT_DIR.items():
            t1 = time.perf_counter()
            try:
                gdf = gpd.read_file(gpkg, layer=layer, force_2d=True)
            except Exception as e:
                print(f"  {basin}/{layer}: read failed ({e})", flush=True)
                continue

            if gdf.empty:
                print(f"  {basin}/{layer}: empty", flush=True)
                continue

            if target_crs is None:
                target_crs = gdf.crs
            elif gdf.crs != target_crs:
                gdf = gdf.to_crs(target_crs)

            gdf = gdf.drop(columns=[c for c in DROP_COLS if c in gdf.columns])
            gdf["source_basin"] = basin

            out_dir = OUTPUT_DIR / subdir
            out_dir.mkdir(parents=True, exist_ok=True)
            out_file = out_dir / f"euhydro_{basin}_v013.geoparquet"
            gdf.to_parquet(out_file, index=False, compression="zstd")

            n = len(gdf)
            rows.append({"layer": layer, "file": f"{subdir}/{out_file.name}", "source_basin": basin, "features": n})
            print(f"  {basin}/{layer}: {n} features in {time.perf_counter() - t1:.1f}s", flush=True)
    finally:
        shutil.rmtree(tmp_root, ignore_errors=True)

    return rows, target_crs


def main() -> None:
    zips = sorted(glob.glob("euhydro_*_v013_GPKG.zip"))
    if not zips:
        raise RuntimeError("No euhydro_*_v013_GPKG.zip files found in the current directory.")

    OUTPUT_DIR.mkdir(parents=True, exist_ok=True)

    target_crs = None
    manifest_rows: list[dict] = []

    for zp in zips:
        basin = basin_from_zip(zp)
        if any(x in basin.lower() for x in EXCLUDED_BASINS):
            print(f"skipping excluded basin: {basin}", flush=True)
            continue
        print(f"[{basin}] processing {zp}", flush=True)
        rows, target_crs = process_basin(zp, target_crs)
        manifest_rows.extend(rows)

    if not manifest_rows:
        raise RuntimeError("No output files were written.")

    manifest_path = OUTPUT_DIR / "manifest.csv"
    pd.DataFrame(manifest_rows).sort_values(["layer", "file"]).to_csv(manifest_path, index=False)

    print("Done.")
    print(f"Wrote {len(manifest_rows)} shards across {len(LAYER_OUTPUT_DIR)} layers to {OUTPUT_DIR}")
    print(f"Manifest: {manifest_path}")


if __name__ == "__main__":
    main()