Lou·Ka — tous les logements à louer du Québec, un seul endroit.
- 264commits
- 1branches
- 0releases
- 180.7 MBon disk
- maindefault branch
- 5 days agolast push
1# -----------------------------------------------------------------------------2# Lou-Ka — Agrégateur de logements à louer (province de Québec)3# Auteur : Simon-Pierre Boucher — contact@spboucher.ai4# db.py : persistance SQLite — upsert avec détection de changements,5# cycle de vie avec délai de grâce (2 syncs), détection de dérive,6# cache des pages détail, cache de géocodage.7# -----------------------------------------------------------------------------8from __future__ import annotations910import json11import sqlite312import statistics13import time14from pathlib import Path1516from .schema import Listing1718DB_PATH = Path(__file__).resolve().parent.parent / "data" / "louka.db"1920# Nombre d'exécutions consécutives où une annonce doit être absente de la21# source avant d'être désactivée (délai de grâce contre les ratés ponctuels).22MISS_GRACE = 22324# Dérive : si une source retourne <= DRIFT_RATIO × sa médiane historique25# (médiane >= DRIFT_MIN_BASE annonces), on alerte et on suspend les retraits.26DRIFT_RATIO = 0.2527DRIFT_MIN_BASE = 828DRIFT_HISTORY = 52930_SCHEMA = """31CREATE TABLE IF NOT EXISTS listings (32 uid TEXT PRIMARY KEY,33 source TEXT NOT NULL,34 external_id TEXT NOT NULL,35 url TEXT,36 title TEXT,37 address TEXT,38 sector TEXT,39 city TEXT,40 unit_type TEXT,41 bedrooms REAL,42 bathrooms REAL,43 price REAL,44 price_label TEXT,45 availability TEXT,46 availability_date TEXT,47 area_sqft REAL,48 pets TEXT,49 furnished INTEGER,50 description TEXT,51 amenities TEXT, -- JSON (liste de textes source)52 details TEXT, -- JSON (champs structurés : inclusions, parking…)53 images TEXT, -- JSON54 lat REAL,55 lng REAL,56 geocode_failed INTEGER DEFAULT 0,57 content_hash TEXT,58 first_seen REAL,59 last_seen REAL,60 updated_at REAL,61 miss_count INTEGER DEFAULT 0,62 active INTEGER DEFAULT 163);64CREATE INDEX IF NOT EXISTS idx_listings_source ON listings(source);65CREATE INDEX IF NOT EXISTS idx_listings_city ON listings(city);66CREATE INDEX IF NOT EXISTS idx_listings_active ON listings(active);67CREATE INDEX IF NOT EXISTS idx_listings_geo ON listings(lat, lng);6869CREATE TABLE IF NOT EXISTS sync_log (70 id INTEGER PRIMARY KEY AUTOINCREMENT,71 source TEXT,72 ts REAL,73 found INTEGER,74 added INTEGER,75 updated INTEGER,76 removed INTEGER,77 ok INTEGER,78 message TEXT,79 stats TEXT -- JSON : taux de champs null, missed, alerte…80);8182CREATE TABLE IF NOT EXISTS detail_cache (83 source TEXT NOT NULL,84 external_id TEXT NOT NULL,85 key TEXT, -- hash du contenu « liste » de l'annonce86 payload TEXT, -- JSON opaque propre au connecteur87 fetched_at REAL,88 PRIMARY KEY (source, external_id)89);9091CREATE TABLE IF NOT EXISTS price_log (92 uid TEXT NOT NULL,93 ts REAL NOT NULL,94 price REAL -- prix observé (NULL = retiré de l'affichage)95);96CREATE INDEX IF NOT EXISTS idx_price_log_uid ON price_log(uid);9798CREATE TABLE IF NOT EXISTS poi_cache (99 coord_key TEXT PRIMARY KEY, -- "lat,lng" arrondi à 4 décimales (~11 m)100 lat REAL,101 lng REAL,102 pois TEXT, -- JSON : [{cat, name, dist_m}] (plus proche/catégorie)103 ts REAL104);105106CREATE TABLE IF NOT EXISTS geocode_cache (107 address TEXT PRIMARY KEY, -- adresse normalisée (clé de cache)108 lat REAL,109 lng REAL,110 provider TEXT,111 failed INTEGER DEFAULT 0,112 ts REAL113);114115CREATE TABLE IF NOT EXISTS users (116 id INTEGER PRIMARY KEY AUTOINCREMENT,117 google_sub TEXT UNIQUE, -- identifiant stable Google (OpenID « sub »)118 ka_id TEXT, -- identifiant membre « ka-0123456789 » (unique)119 email TEXT,120 name TEXT,121 picture TEXT, -- URL de l'avatar Google122 public INTEGER DEFAULT 0, -- profil public /u/{ka_id} (opt-in)123 role TEXT, -- « locataire » | « gestionnaire » | NULL124 org_source TEXT, -- source réclamée (gestionnaire)125 created_at REAL,126 last_login REAL127);128129CREATE TABLE IF NOT EXISTS favorites (130 user_id INTEGER NOT NULL,131 uid TEXT NOT NULL, -- listings.uid132 ts REAL,133 PRIMARY KEY (user_id, uid)134);135136CREATE TABLE IF NOT EXISTS image_checks (137 url TEXT PRIMARY KEY, -- URL d'image vérifiée (imgcheck.py)138 status TEXT, -- ok | dead | small | error139 width INTEGER,140 height INTEGER,141 bytes INTEGER,142 sha1 TEXT, -- hash du contenu (doublons / placeholders)143 checked_at REAL144);145146CREATE TABLE IF NOT EXISTS env_tiles (147 tile_key TEXT PRIMARY KEY, -- « ty,tx » (tuiles 0,5° — voir poi.py/environment.py)148 data TEXT, -- JSON : routes/rails/aéro/industriel/bars/cyclable/POI149 fetched_at REAL150);151152CREATE TABLE IF NOT EXISTS kascores (153 coord_key TEXT PRIMARY KEY, -- « lat,lng » arrondi à 4 décimales (immeuble)154 lat REAL, lng REAL,155 walk REAL, -- KA Walk Score 0-100 (NULL = données insuffisantes)156 transit REAL, -- KA Transit Score (NULL = non desservi/inconnu)157 bike REAL,158 calme REAL,159 services REAL,160 global REAL, -- moyenne pondérée par défaut (voir kascores.py)161 details TEXT, -- JSON : détail par score (fiche + méthodologie)162 version TEXT, -- version du barème (recalcul si changement)163 computed_at REAL164);165166CREATE TABLE IF NOT EXISTS source_profiles (167 source_id TEXT PRIMARY KEY, -- id de la source (data/sources.json)168 owner_user_id INTEGER, -- gestionnaire qui a réclamé la page169 tagline TEXT, -- accroche personnalisée170 description TEXT,171 website TEXT,172 phone TEXT,173 email TEXT,174 logo_url TEXT,175 updated_at REAL176);177"""178179# Colonnes ajoutées après la v1 — migration automatique des bases existantes.180_MIGRATIONS = {181 "listings": {182 "availability_date": "TEXT",183 "area_sqft": "REAL",184 "pets": "TEXT",185 "furnished": "INTEGER",186 "details": "TEXT",187 "geocode_failed": "INTEGER DEFAULT 0",188 "miss_count": "INTEGER DEFAULT 0",189 "dauid": "TEXT", # aire de diffusion 2021 (stats de quartier)190 "digest": "TEXT", # JSON louka/textmine.py (description structurée)191 "dup_of": "TEXT", # uid de l'annonce canonique si doublon inter-sources192 "dup_sources": "TEXT", # JSON : autres sources où l'annonce est publiée193 "bedrooms": "REAL", # chambres fermées (convention QC : 4½ = 2)194 "bathrooms": "REAL", # salles de bain (1.5 = salle d'eau en plus)195 "completeness": "REAL", # score de complétude 0–100 (quality.py)196 "published": "INTEGER DEFAULT 1", # 0 = quarantaine (sous le seuil)197 "quality_reasons": "TEXT", # JSON : raisons de la quarantaine198 "images_ok": "TEXT", # JSON : galerie nettoyée (imgcheck.py)199 "img_audit": "TEXT", # JSON : traçabilité du contrôle images200 "coord_key": "TEXT", # clé immeuble « lat,lng » 4 déc. (jointure kascores)201 },202 "sync_log": {203 "stats": "TEXT",204 },205 "users": {206 "ka_id": "TEXT",207 "public": "INTEGER DEFAULT 0",208 "role": "TEXT",209 "org_source": "TEXT",210 "display_name": "TEXT", # nom choisi (prime sur le nom Google)211 "avatar_url": "TEXT", # photo téléversée (prime sur picture)212 "bio": "TEXT",213 "city": "TEXT",214 "phone": "TEXT",215 "website": "TEXT",216 "socials": "TEXT", # JSON {instagram, facebook, x, linkedin…}217 },218}219220221def connect() -> sqlite3.Connection:222 DB_PATH.parent.mkdir(parents=True, exist_ok=True)223 con = sqlite3.connect(DB_PATH)224 con.row_factory = sqlite3.Row225 con.executescript(_SCHEMA)226 for table, cols in _MIGRATIONS.items():227 existing = {r["name"] for r in con.execute(f"PRAGMA table_info({table})")}228 for col, decl in cols.items():229 if col not in existing:230 con.execute(f"ALTER TABLE {table} ADD COLUMN {col} {decl}")231 # index sur des colonnes de migration : APRÈS les ALTER TABLE (une base232 # antérieure n'a pas encore la colonne au moment du CREATE INDEX)233 con.execute("CREATE UNIQUE INDEX IF NOT EXISTS users_ka_id ON users(ka_id)")234 con.execute("CREATE INDEX IF NOT EXISTS idx_listings_pub"235 " ON listings(active, published)")236 con.execute("CREATE INDEX IF NOT EXISTS idx_listings_coord"237 " ON listings(coord_key)")238 con.commit()239 # WAL : lectures (web) et écritures (sync, imgcheck) concurrentes sans240 # verrou global ; busy_timeout évite les « database is locked » ponctuels.241 con.execute("PRAGMA journal_mode=WAL")242 con.execute("PRAGMA busy_timeout=30000")243 return con244245246# ---------------------------------------------------------------------------247# Synchronisation d'une source248# ---------------------------------------------------------------------------249250def _drift_alert(con: sqlite3.Connection, source: str, found: int,251 null_price_rate: float) -> str | None:252 """Détecte une dérive du connecteur (chute du volume ou des prix extraits).253254 Retourne un message d'alerte, ou None si tout est normal.255 """256 hist = con.execute(257 "SELECT found, stats FROM sync_log WHERE source=? AND ok=1"258 " ORDER BY ts DESC LIMIT ?", (source, DRIFT_HISTORY)).fetchall()259 if len(hist) < 3:260 return None261 med_found = statistics.median(r["found"] for r in hist)262 if med_found >= DRIFT_MIN_BASE and found <= DRIFT_RATIO * med_found:263 return (f"dérive: {found} annonce(s) trouvée(s) contre une médiane de "264 f"{med_found:.0f} — retraits suspendus, vérifier le connecteur")265 if found >= DRIFT_MIN_BASE and null_price_rate >= 0.8:266 rates = []267 for r in hist:268 try:269 rates.append(json.loads(r["stats"] or "{}")["null_price_rate"])270 except (KeyError, ValueError, TypeError):271 continue272 if rates and statistics.median(rates) <= 0.3:273 return (f"dérive: {null_price_rate:.0%} des annonces sans prix "274 f"(habituellement {statistics.median(rates):.0%}) — "275 "le format de la source a probablement changé")276 return None277278279def sync_source(con: sqlite3.Connection, source: str,280 listings: list[Listing]) -> dict:281 """Synchronise les annonces d'une source.282283 - nouvelle annonce -> insertion284 - annonce modifiée -> mise à jour (comparaison de content_hash)285 - annonce disparue -> miss_count += 1, puis active=0 après MISS_GRACE286 exécutions consécutives (délai de grâce)287 - dérive détectée -> alerte consignée, retraits suspendus288 """289 now = time.time()290 added = updated = 0291 seen_uids = set()292293 n = len(listings)294 null_price = sum(1 for l in listings if l.price is None)295 null_addr = sum(1 for l in listings if not l.address)296 null_price_rate = round(null_price / n, 3) if n else 0.0297298 alert = _drift_alert(con, source, n, null_price_rate)299300 for lst in listings:301 seen_uids.add(lst.uid)302 h = lst.content_hash()303 row = con.execute(304 "SELECT content_hash, price, images FROM listings WHERE uid=?",305 (lst.uid,)).fetchone()306 # score de complétude + seuil de publication (quarantaine sous le seuil)307 from .quality import evaluate308 q_score, q_pub, q_reasons = evaluate({309 "title": lst.title,310 "price": lst.price, "address": lst.address, "city": lst.city,311 "sector": lst.sector, "lat": lst.lat, "lng": lst.lng,312 "description": lst.description, "images": lst.images,313 "unit_type": lst.unit_type, "bedrooms": lst.bedrooms,314 "bathrooms": lst.bathrooms, "availability": lst.availability,315 "availability_date": lst.availability_date,316 "area_sqft": lst.area_sqft, "amenities": lst.amenities,317 "details": lst.details, "url": lst.url})318 params = dict(319 uid=lst.uid, source=lst.source, external_id=lst.external_id,320 url=lst.url, title=lst.title, address=lst.address,321 sector=lst.sector, city=lst.city, unit_type=lst.unit_type,322 bedrooms=lst.bedrooms, bathrooms=lst.bathrooms,323 price=lst.price, price_label=lst.price_label,324 availability=lst.availability,325 availability_date=lst.availability_date,326 area_sqft=lst.area_sqft, pets=lst.pets,327 furnished=(None if lst.furnished is None else int(lst.furnished)),328 description=lst.description,329 digest=(json.dumps(lst.digest, ensure_ascii=False)330 if getattr(lst, "digest", None) else None),331 amenities=json.dumps(lst.amenities, ensure_ascii=False),332 details=json.dumps(lst.details, ensure_ascii=False),333 images=json.dumps(lst.images, ensure_ascii=False),334 lat=lst.lat, lng=lst.lng, content_hash=h, now=now,335 completeness=q_score, published=int(q_pub),336 quality_reasons=(json.dumps(q_reasons, ensure_ascii=False)337 if q_reasons else None),338 )339 if row is None:340 con.execute(341 """INSERT INTO listings (uid, source, external_id, url, title,342 address, sector, city, unit_type, bedrooms, bathrooms,343 price, price_label,344 availability, availability_date, area_sqft, pets, furnished,345 description, digest, amenities, details, images, lat, lng,346 content_hash, first_seen, last_seen, updated_at,347 miss_count, active, completeness, published,348 quality_reasons)349 VALUES (:uid,:source,:external_id,:url,:title,:address,350 :sector,:city,:unit_type,:bedrooms,:bathrooms,351 :price,:price_label,:availability,352 :availability_date,:area_sqft,:pets,:furnished,353 :description,:digest,:amenities,:details,:images,:lat,:lng,354 :content_hash,:now,:now,:now,0,1,355 :completeness,:published,:quality_reasons)""", params)356 if lst.price is not None: # prix initial = point de départ de l'historique357 con.execute("INSERT INTO price_log (uid, ts, price) VALUES (?,?,?)",358 (lst.uid, now, lst.price))359 added += 1360 elif row["content_hash"] != h:361 # COALESCE : ne jamais écraser des coordonnées géocodées par null362 con.execute(363 """UPDATE listings SET url=:url, title=:title,364 address=:address, sector=:sector, city=:city,365 unit_type=:unit_type, bedrooms=:bedrooms,366 bathrooms=:bathrooms, price=:price,367 price_label=:price_label, availability=:availability,368 availability_date=:availability_date,369 area_sqft=:area_sqft, pets=:pets, furnished=:furnished,370 description=:description, digest=:digest,371 amenities=:amenities, details=:details, images=:images,372 lat=COALESCE(:lat, lat), lng=COALESCE(:lng, lng),373 content_hash=:content_hash, last_seen=:now,374 updated_at=:now, miss_count=0, active=1,375 completeness=:completeness, published=:published,376 quality_reasons=:quality_reasons377 WHERE uid=:uid""", params)378 if (row["images"] or "[]") != params["images"]:379 # galerie modifiée : invalider l'audit d'images (imgcheck.py)380 con.execute("UPDATE listings SET images_ok=NULL, img_audit=NULL"381 " WHERE uid=?", (lst.uid,))382 if lst.price != row["price"]: # changement de prix -> historique383 con.execute("INSERT INTO price_log (uid, ts, price) VALUES (?,?,?)",384 (lst.uid, now, lst.price))385 updated += 1386 else:387 con.execute(388 "UPDATE listings SET last_seen=?, miss_count=0, active=1 WHERE uid=?",389 (now, lst.uid))390391 # Annonces de cette source qui n'apparaissent plus : délai de grâce,392 # puis désactivation. Suspendu si une dérive est détectée.393 removed = missed = 0394 if not alert:395 for r in con.execute(396 "SELECT uid, miss_count FROM listings WHERE source=? AND active=1",397 (source,)).fetchall():398 if r["uid"] in seen_uids:399 continue400 missed += 1401 if r["miss_count"] + 1 >= MISS_GRACE:402 con.execute(403 "UPDATE listings SET active=0, miss_count=?, updated_at=?"404 " WHERE uid=?", (r["miss_count"] + 1, now, r["uid"]))405 removed += 1406 else:407 con.execute("UPDATE listings SET miss_count=miss_count+1 WHERE uid=?",408 (r["uid"],))409410 stats = {411 "null_price_rate": null_price_rate,412 "null_address_rate": round(null_addr / n, 3) if n else 0.0,413 "missed": missed,414 }415 if alert:416 stats["alert"] = alert417 con.execute(418 "INSERT INTO sync_log (source, ts, found, added, updated, removed, ok,"419 " message, stats) VALUES (?,?,?,?,?,?,1,?,?)",420 (source, now, n, added, updated, removed, alert or "ok",421 json.dumps(stats, ensure_ascii=False)))422 con.commit()423 out = {"source": source, "found": n, "added": added,424 "updated": updated, "removed": removed}425 if alert:426 out["alert"] = alert427 return out428429430def log_failure(con: sqlite3.Connection, source: str, message: str) -> None:431 con.execute(432 "INSERT INTO sync_log (source, ts, found, added, updated, removed, ok, message)"433 " VALUES (?,?,0,0,0,0,0,?)", (source, time.time(), message))434 con.commit()435436437# ---------------------------------------------------------------------------438# Cache des pages détail (« détail si nouveau/modifié »)439# ---------------------------------------------------------------------------440441def get_cached_detail(con: sqlite3.Connection, source: str,442 external_id: str, key: str) -> dict | None:443 """Payload détail mis en cache si la clé (hash liste) n'a pas changé."""444 row = con.execute(445 "SELECT key, payload FROM detail_cache WHERE source=? AND external_id=?",446 (source, external_id)).fetchone()447 if row and row["key"] == key and row["payload"]:448 try:449 return json.loads(row["payload"])450 except ValueError:451 return None452 return None453454455def put_cached_detail(con: sqlite3.Connection, source: str,456 external_id: str, key: str, payload: dict) -> None:457 con.execute(458 "INSERT INTO detail_cache (source, external_id, key, payload, fetched_at)"459 " VALUES (?,?,?,?,?)"460 " ON CONFLICT(source, external_id) DO UPDATE SET"461 " key=excluded.key, payload=excluded.payload, fetched_at=excluded.fetched_at",462 (source, external_id, key, json.dumps(payload, ensure_ascii=False),463 time.time()))464 con.commit()465