From e63b21eccb139127de57d3ed0178d4ba8a1bf654 Mon Sep 17 00:00:00 2001
From: bot-backend <bot-backend@gendsgn.local>
Date: Wed, 17 Jun 2026 23:07:22 +0300
Subject: [PATCH] fix(scrapers): write resolved house_id to
 listings.house_id_fk in realtime link hook

---
 .../backend/app/services/scrapers/base.py     | 18 ++++++
 .../backend/tests/test_base_save_listings.py  | 64 +++++++++++++++++++
 2 files changed, 82 insertions(+)

diff --git a/tradein-mvp/backend/app/services/scrapers/base.py b/tradein-mvp/backend/app/services/scrapers/base.py
index 9fe05390..5c9fa1db 100644
--- a/tradein-mvp/backend/app/services/scrapers/base.py
+++ b/tradein-mvp/backend/app/services/scrapers/base.py
@@ -521,6 +521,24 @@ def _link_listing_to_house(db: Session, listing_id: int, lot: ScrapedLot) -> Non
             source_url=lot.house_url or lot.source_url,
         )
 
+        # Mirror the resolved house into listings.house_id_fk so direct
+        # `JOIN houses ON house_id_fk = id` keeps working on freshly scraped /
+        # re-scraped active rows. Without this the FK is populated only by the
+        # offline backfill (063_*.sql / backfill_listing_sources.py), leaving
+        # new active listings unlinked (avito 0/444, cian ~7%). Mirrors the
+        # backfill script's column/param style. IS DISTINCT FROM avoids a no-op
+        # write when the linkage is already correct.
+        if house_id is not None:
+            db.execute(
+                text(
+                    "UPDATE listings "
+                    "   SET house_id_fk = CAST(:hid AS bigint) "
+                    " WHERE id = CAST(:lid AS bigint) "
+                    "   AND house_id_fk IS DISTINCT FROM CAST(:hid AS bigint)"
+                ),
+                {"hid": house_id, "lid": listing_id},
+            )
+
     upsert_listing_source(
         db,
         listing_id=listing_id,
diff --git a/tradein-mvp/backend/tests/test_base_save_listings.py b/tradein-mvp/backend/tests/test_base_save_listings.py
index 1a38b7e8..30a924e4 100644
--- a/tradein-mvp/backend/tests/test_base_save_listings.py
+++ b/tradein-mvp/backend/tests/test_base_save_listings.py
@@ -596,6 +596,70 @@ def test_save_listings_null_area_and_date_pass_none_in_params():
     assert params["listing_date"] is None
 
 
+# ── Tests: house_id_fk realtime link write (avito/cian unlinked fix) ─────────
+
+
+def _find_house_fk_update(db: MagicMock):
+    """Return (sql, params) of the UPDATE listings SET house_id_fk call, or None.
+
+    match_or_create_house / upsert_listing_source are patched (no db.execute), so
+    the only db.execute calls are the listings INSERT (call 0) and — when a house
+    is resolved — the house_id_fk UPDATE added by the realtime link hook.
+    """
+    for call in db.execute.call_args_list:
+        sql = str(call.args[0])
+        if "house_id_fk" in sql and "UPDATE listings" in sql:
+            return sql, call.args[1]
+    return None
+
+
+def test_save_listings_writes_house_id_fk_when_house_resolved(_patch_matching):
+    """House resolved (address/coords present) → UPDATE listings.house_id_fk runs
+    with the resolved house_id and the saved listing id."""
+    _patch_matching["house"].return_value = (909, 1.0, "geo")
+
+    db = _mock_db(inserted=True, listing_id=4242)
+    lot = _cian_lot(source_id="c-fk-1")  # has address + lat/lon → house resolved
+
+    save_listings(db, [lot])
+
+    found = _find_house_fk_update(db)
+    assert found is not None, "UPDATE listings SET house_id_fk was not issued"
+    _sql, params = found
+    assert params["hid"] == 909
+    assert params["lid"] == 4242
+
+
+def test_save_listings_no_house_fk_update_when_house_not_resolved(_patch_matching):
+    """No address and no coords → match_or_create_house skipped → no FK UPDATE."""
+    db = _mock_db(inserted=True, listing_id=4243)
+    # base lot has no address / lat / lon → house resolution skipped
+    lot = _base_lot(source="avito", source_id="a-fk-no", address=None, lat=None, lon=None)
+
+    save_listings(db, [lot])
+
+    assert _patch_matching["house"].call_count == 0
+    assert _find_house_fk_update(db) is None
+
+
+def test_save_listings_house_fk_update_guards_no_op_write(_patch_matching):
+    """The UPDATE is idempotency-guarded (IS DISTINCT FROM) and casts to bigint
+    (psycopg v3 CAST, never ::bigint)."""
+    _patch_matching["house"].return_value = (55, 1.0, "cadastr")
+
+    db = _mock_db(inserted=True, listing_id=1)
+    lot = _cian_lot(source_id="c-fk-guard")
+
+    save_listings(db, [lot])
+
+    found = _find_house_fk_update(db)
+    assert found is not None
+    sql, _params = found
+    assert "IS DISTINCT FROM" in sql
+    assert "CAST(:hid AS bigint)" in sql
+    assert "::bigint" not in sql
+
+
 # ── Tests 18-20: Yandex rich fields (predictedPrice, trend, description, agency) ──
 
 
-- 
2.45.3