#!/usr/bin/env python3
"""
Offline-Extraktion der Ban-Yaro-POIs aus einer OSM-.pbf-Datei.

Ersetzt das Live-Scannen gegen overpass-api.de (backend/routes/osm.py) durch
einen einmaligen/periodischen Batch-Lauf — keine OSM-Live-Last, kein Bann mehr.

Das Kategorie→Tag-Mapping ist 1:1 aus OSM_QUERIES in backend/routes/osm.py
übernommen. Schreibt ins selbe Schema wie die Produktiv-Tabelle `osm_pois`
(osm_id, type, lat, lon, name, opening_hours, phone, website, cached_at).

Aufruf:
    python3 extract_osm_pois.py <input.osm.pbf> <output.sqlite>
"""
import sys
import sqlite3
import osmium


# --- Kategorie-Klassifikation (1:1 aus OSM_QUERIES, backend/routes/osm.py) ---
def classify(t) -> list[str]:
    """Gibt alle Ban-Yaro-Typen zurück, die auf dieses OSM-Objekt passen."""
    types: list[str] = []
    a = t.get("amenity")
    shop = t.get("shop")
    craft = t.get("craft")
    leisure = t.get("leisure")
    tourism = t.get("tourism")
    dog = t.get("dog")
    outdoor = t.get("outdoor_seating")
    # "hundefreundlich, breiter gefasst": explizit erlaubt ODER Terrasse
    dog_ok = dog in ("yes", "allowed", "leashed")

    if a == "waste_basket":
        types.append("waste_basket")
    if a == "drinking_water":
        types.append("drinking_water")
    if a == "veterinary":
        types.append("tierarzt")
    if a == "bench":
        types.append("bank")
    if leisure == "dog_park" or (leisure == "park" and dog == "yes"):
        types.append("dog_park")
    if shop == "pet":
        types.append("shop")
    if shop == "pet_grooming" or craft == "pet_grooming":
        types.append("hundesalon")
    if (a in ("restaurant", "cafe") and (dog_ok or outdoor == "yes")) or a == "biergarten":
        types.append("restaurant")
    if tourism in ("hotel", "guest_house", "hostel") and dog_ok:
        types.append("hotel")
    return types


SCHEMA = """
CREATE TABLE IF NOT EXISTS osm_pois (
    osm_id        INTEGER NOT NULL,
    type          TEXT    NOT NULL,
    lat           REAL    NOT NULL,
    lon           REAL    NOT NULL,
    name          TEXT,
    opening_hours TEXT,
    phone         TEXT,
    website       TEXT,
    user_edited   INTEGER NOT NULL DEFAULT 0,
    cached_at     TEXT    NOT NULL DEFAULT (datetime('now')),
    PRIMARY KEY (osm_id, type)
);
CREATE INDEX IF NOT EXISTS idx_osm_pois_loc ON osm_pois(type, lat, lon);
"""


class PoiHandler(osmium.SimpleHandler):
    def __init__(self, conn):
        super().__init__()
        self.conn = conn
        self.rows = 0
        self.objs = 0

    def _save(self, osm_id, tags, lat, lon):
        types = classify(tags)
        if not types:
            return
        self.objs += 1
        name = tags.get("name")
        oh = tags.get("opening_hours")
        phone = tags.get("phone") or tags.get("contact:phone")
        web = tags.get("website") or tags.get("contact:website")
        for ty in types:
            self.conn.execute(
                "INSERT OR REPLACE INTO osm_pois "
                "(osm_id, type, lat, lon, name, opening_hours, phone, website, cached_at) "
                "VALUES (?,?,?,?,?,?,?,?, datetime('now'))",
                (osm_id, ty, lat, lon, name, oh, phone, web),
            )
            self.rows += 1

    def node(self, n):
        if n.tags:
            self._save(n.id, n.tags, n.location.lat, n.location.lon)

    def way(self, w):
        # Wege (z. B. Tierarzt im Gebäude) → Schwerpunkt aus Knoten ("out center")
        if not w.tags:
            return
        lats, lons = [], []
        for nd in w.nodes:
            if nd.location.valid():
                lats.append(nd.location.lat)
                lons.append(nd.location.lon)
        if lats:
            self._save(w.id, w.tags, sum(lats) / len(lats), sum(lons) / len(lons))


def main():
    if len(sys.argv) != 3:
        print(__doc__)
        sys.exit(1)
    src, dst = sys.argv[1], sys.argv[2]

    conn = sqlite3.connect(dst)
    conn.executescript(SCHEMA)

    h = PoiHandler(conn)
    # locations=True: Knoten-Koordinaten im Speicher halten, damit Wege einen
    # Schwerpunkt bekommen. flex_mem skaliert bis Länder-Extrakte.
    h.apply_file(src, locations=True, idx="flex_mem")

    conn.commit()
    print(f"\nObjekte mit Treffer: {h.objs:,}   eingefügte Zeilen: {h.rows:,}")
    print("\nPro Typ:")
    for ty, cnt in conn.execute(
        "SELECT type, COUNT(*) FROM osm_pois GROUP BY type ORDER BY 2 DESC"
    ):
        print(f"  {ty:16s} {cnt:>8,}")
    conn.close()


if __name__ == "__main__":
    main()