fix(site-finder): address PR #116 auto-review M1-M5

M1 (mutation risk): replace EMPTY_DUMP_RESULT direct refs with
_make_empty_result() factory. dict(...) shallow copy left nested
nspd_dump shared by reference across concurrent requests — single
mutation pollutes module sentinel for all subsequent calls. Now
каждый caller gets independent dict.

M2 (O(N) spatial scan): SELECT extended denormalized counts
(territorial_zones_count, zouit_count, engineering_count). Each
spatial helper accepts layer_counts and early-returns when count=0
— skips heavy jsonb_array_elements + ST_Transform/ST_Intersects
scan entirely. Critical для quarters с 2000+ features.

M4 (documentation): _trigger_harvest docstring describes known
burst/no-dedup limitation + TODO Redis SETNX (отдельный PR).

M5 (test fragility): _make_db_mock_with_spatial docstring describes
positional-call contract — db.execute order (0=dump, 1=zoning, 2=zouit,
3=engineering) и зависимость от count-values.

+4 new tests (17 total, all pass):
- test_make_empty_result_returns_independent_copies (mutation safety)
- test_make_empty_result_overrides
- test_early_exit_all_counts_zero_no_spatial_queries
- test_early_exit_partial_counts

Per auto-review on 3068a9c.
This commit is contained in:
lekss361 2026-05-13 09:03:35 +03:00
parent 3068a9c531
commit aef8308463
2 changed files with 193 additions and 42 deletions

View file

@ -29,7 +29,8 @@ _DUMP_MAX_AGE_DAYS = 180
# Радиус поиска инженерных сооружений (метры) — договорённость #44 I3.
_ENGINEERING_RADIUS_M = 200
# Пустой результат — возвращается когда дамп недоступен, устарел или при ошибке.
# Sentinel для isinstance-проверок и read-only fallback в parcels.py try/except.
# НИКОГДА не мутировать — использовать _make_empty_result() для новых dict.
EMPTY_DUMP_RESULT: dict[str, Any] = {
"nspd_zoning": None,
"nspd_zouit_overlaps": [],
@ -44,6 +45,31 @@ EMPTY_DUMP_RESULT: dict[str, Any] = {
}
def _make_empty_result(
*,
fetched_at_utc: str | None = None,
stale: bool = False,
harvest_triggered: bool = False,
total_features: int | None = None,
) -> dict[str, Any]:
"""Создаёт свежую копию empty-dump result с возможностью переопределить поля.
Вызывать вместо EMPTY_DUMP_RESULT напрямую чтобы не мутировать singleton.
"""
return {
"nspd_zoning": None,
"nspd_zouit_overlaps": [],
"nspd_engineering_nearby": [],
"nspd_dump": {
"available": False,
"fetched_at_utc": fetched_at_utc,
"stale": stale,
"harvest_triggered": harvest_triggered,
"total_features": total_features,
},
}
def derive_quarter_cad(cad_num: str) -> str | None:
"""3-сегментный кадастровый квартал из любого кадастрового номера.
@ -80,21 +106,23 @@ def get_quarter_dump_data(
Если parcel_wkt=None возвращает только freshness metadata (нет geom для
spatial queries).
"""
empty_result = EMPTY_DUMP_RESULT
quarter = derive_quarter_cad(cad_num)
if quarter is None:
logger.warning("quarter_dump_lookup: cannot derive quarter from cad=%s", cad_num)
return empty_result
return _make_empty_result()
# Читаем строку дампа из БД
# Читаем строку дампа из БД. Денормализованные счётчики слоёв используются
# для early-exit в spatial helpers (M2 mitigation).
row = db.execute(
text(
"""
SELECT quarter_cad,
fetched_at_utc,
total_features,
harvest_error
harvest_error,
territorial_zones_count,
zouit_count,
engineering_count
FROM nspd_quarter_dumps
WHERE quarter_cad = :q
"""
@ -108,15 +136,7 @@ def get_quarter_dump_data(
if row is None:
# Дампа нет — ставим harvest в очередь
harvest_triggered = _trigger_harvest(quarter)
result = dict(empty_result)
result["nspd_dump"] = {
"available": False,
"fetched_at_utc": None,
"stale": False,
"harvest_triggered": harvest_triggered,
"total_features": None,
}
return result
return _make_empty_result(harvest_triggered=harvest_triggered)
fetched_at: datetime = row[1]
# Убедимся что timezone-aware для корректного сравнения
@ -125,6 +145,9 @@ def get_quarter_dump_data(
total_features: int | None = row[2]
harvest_error: str | None = row[3]
territorial_zones_count: int = row[4] or 0
zouit_count: int = row[5] or 0
engineering_count: int = row[6] or 0
is_stale = (now - fetched_at) > max_age
has_error = harvest_error is not None
@ -132,15 +155,12 @@ def get_quarter_dump_data(
# Устаревший или с ошибкой — триггерим повторный harvest
if is_stale or has_error:
harvest_triggered = _trigger_harvest(quarter)
result = dict(empty_result)
result["nspd_dump"] = {
"available": False,
"fetched_at_utc": fetched_at.isoformat(),
"stale": is_stale,
"harvest_triggered": harvest_triggered,
"total_features": total_features,
}
return result
return _make_empty_result(
fetched_at_utc=fetched_at.isoformat(),
stale=is_stale,
harvest_triggered=harvest_triggered,
total_features=total_features,
)
# Свежий дамп без ошибок — извлекаем spatial данные
dump_meta: dict[str, Any] = {
@ -160,9 +180,14 @@ def get_quarter_dump_data(
"nspd_dump": dump_meta,
}
nspd_zoning = _get_zoning(db, quarter, parcel_wkt)
nspd_zouit = _get_zouit_overlaps(db, quarter, parcel_wkt)
nspd_engineering = _get_engineering_nearby(db, quarter, parcel_wkt)
layer_counts = {
"territorial_zones_count": territorial_zones_count,
"zouit_count": zouit_count,
"engineering_count": engineering_count,
}
nspd_zoning = _get_zoning(db, quarter, parcel_wkt, layer_counts)
nspd_zouit = _get_zouit_overlaps(db, quarter, parcel_wkt, layer_counts)
nspd_engineering = _get_engineering_nearby(db, quarter, parcel_wkt, layer_counts)
return {
"nspd_zoning": nspd_zoning,
@ -175,11 +200,20 @@ def get_quarter_dump_data(
# ── Spatial helpers ───────────────────────────────────────────────────────────
def _get_zoning(db: Session, quarter: str, parcel_wkt: str) -> dict[str, Any] | None:
def _get_zoning(
db: Session,
quarter: str,
parcel_wkt: str,
layer_counts: dict[str, int] | None = None,
) -> dict[str, Any] | None:
"""G1: ПЗЗ территориальная зона по centroid участка из dump.
Geometry в features_json в EPSG:3857 ST_Transform на чтении 4326.
layer_counts денормализованные счётчики из строки дампа. Если
territorial_zones_count == 0 пропускаем heavy jsonb_array_elements scan.
"""
if layer_counts is not None and layer_counts.get("territorial_zones_count", 1) == 0:
return None
try:
row = db.execute(
text(
@ -228,12 +262,17 @@ def _get_zouit_overlaps(
db: Session,
quarter: str,
parcel_wkt: str,
layer_counts: dict[str, int] | None = None,
) -> list[dict[str, Any]]:
"""G3: ЗОУИТ которые пересекают участок.
Проверяем 5 групп: zouit_okn, zouit_engineering, zouit_natural,
zouit_protected, zouit_other.
layer_counts денормализованные счётчики. Если zouit_count == 0
пропускаем heavy jsonb_array_elements scan.
"""
if layer_counts is not None and layer_counts.get("zouit_count", 1) == 0:
return []
try:
rows = db.execute(
text(
@ -285,8 +324,15 @@ def _get_engineering_nearby(
db: Session,
quarter: str,
parcel_wkt: str,
layer_counts: dict[str, int] | None = None,
) -> list[dict[str, Any]]:
"""I3: Инженерные сооружения в радиусе _ENGINEERING_RADIUS_M от centroid участка."""
"""I3: Инженерные сооружения в радиусе _ENGINEERING_RADIUS_M от centroid участка.
layer_counts денормализованные счётчики. Если engineering_count == 0
пропускаем heavy jsonb_array_elements + ST_DWithin scan.
"""
if layer_counts is not None and layer_counts.get("engineering_count", 1) == 0:
return []
try:
rows = db.execute(
text(
@ -353,6 +399,12 @@ def _trigger_harvest(quarter: str) -> bool:
"""Fire-and-forget harvest_quarter.apply_async(). Возвращает True если enqueue OK.
Ленивый импорт чтобы избежать circular import (tasks services tasks).
Known limitation: burst из N concurrent analyze_parcel запросов на один
и тот же ещё не закешированный квартал может поставить N одинаковых задач
в очередь (нет дедупликации). UPSERT в harvest_quarter идемпотентен, поэтому
данные не портятся, но WAF traffic тратится впустую.
TODO: добавить Redis SETNX lock с TTL перед apply_async отдельная issue/PR.
"""
try:
from app.workers.tasks.nspd_sync import harvest_quarter

View file

@ -7,6 +7,8 @@ Mock-based — без реальной БД и Celery. Проверяет:
- get_quarter_dump_data: свежий дамп с territorial_zones nspd_zoning populated
- get_quarter_dump_data: свежий дамп с 2 zouit nspd_zouit_overlaps length=2
- get_quarter_dump_data: parcel_wkt=None metadata only, no spatial queries
- _make_empty_result: mutation-safety каждый вызов возвращает независимый dict
- early-exit: layer_counts == 0 spatial query не вызывается
"""
from __future__ import annotations
@ -16,6 +18,8 @@ from typing import Any
from unittest.mock import MagicMock, patch
from app.services.site_finder.quarter_dump_lookup import (
EMPTY_DUMP_RESULT,
_make_empty_result,
derive_quarter_cad,
get_quarter_dump_data,
)
@ -87,8 +91,9 @@ def test_get_quarter_dump_no_row() -> None:
def test_get_quarter_dump_stale() -> None:
"""Дамп старше 180 дней → stale=True, harvest triggered, available=False."""
old_date = datetime.now(UTC) - timedelta(days=200)
# row: (quarter_cad, fetched_at_utc, total_features, harvest_error)
db = _make_db_mock(row=("66:41:0204016", old_date, 50, None))
# row: (quarter_cad, fetched_at_utc, total_features, harvest_error,
# territorial_zones_count, zouit_count, engineering_count)
db = _make_db_mock(row=("66:41:0204016", old_date, 50, None, 5, 2, 1))
with patch(
"app.services.site_finder.quarter_dump_lookup._trigger_harvest",
@ -106,7 +111,7 @@ def test_get_quarter_dump_stale() -> None:
def test_get_quarter_dump_with_harvest_error() -> None:
"""Дамп с harvest_error → harvest_triggered=True, available=False."""
fresh_date = datetime.now(UTC) - timedelta(days=10)
db = _make_db_mock(row=("66:41:0204016", fresh_date, 0, "WAF 429"))
db = _make_db_mock(row=("66:41:0204016", fresh_date, 0, "WAF 429", 0, 0, 0))
with patch(
"app.services.site_finder.quarter_dump_lookup._trigger_harvest",
@ -125,7 +130,7 @@ def test_get_quarter_dump_with_harvest_error() -> None:
def test_get_quarter_dump_no_parcel_geometry() -> None:
"""parcel_wkt=None → metadata only, нет spatial queries."""
fresh_date = datetime.now(UTC) - timedelta(days=5)
db = _make_db_mock(row=("66:41:0204016", fresh_date, 120, None))
db = _make_db_mock(row=("66:41:0204016", fresh_date, 120, None, 10, 3, 2))
result = get_quarter_dump_data(db, "66:41:0204016", parcel_wkt=None)
@ -149,11 +154,17 @@ def _make_db_mock_with_spatial(
) -> MagicMock:
"""Mock DB с разными результатами для разных запросов.
Порядок вызовов:
1. read dump row
2. zoning (first())
3. zouit (fetchall())
4. engineering (fetchall())
Порядок вызовов db.execute жёстко привязан к реализации get_quarter_dump_data:
0. dump lookup .first()
1. zoning .first() (пропускается если territorial_zones_count == 0)
2. zouit .fetchall() (пропускается если zouit_count == 0)
3. engineering .fetchall() (пропускается если engineering_count == 0)
Хрупко при реорганизации порядка запросов в production-коде. При изменении
порядка в get_quarter_dump_data обновить индексы call_count здесь.
dump_row должен быть 7-элементным tuple:
(quarter_cad, fetched_at_utc, total_features, harvest_error,
territorial_zones_count, zouit_count, engineering_count).
"""
db = MagicMock()
call_count = 0
@ -185,7 +196,7 @@ def _make_db_mock_with_spatial(
def test_get_quarter_dump_fresh_zoning() -> None:
"""Свежий дамп с territorial_zone → nspd_zoning populated."""
fresh_date = datetime.now(UTC) - timedelta(days=30)
dump_row = ("66:41:0204016", fresh_date, 100, None)
dump_row = ("66:41:0204016", fresh_date, 100, None, 5, 0, 0)
# zone_props JSONB в виде dict (SQLAlchemy вернёт dict)
zone_props: dict[str, Any] = {
"reg_numb_border": "Ж-3",
@ -214,7 +225,7 @@ def test_get_quarter_dump_fresh_zoning() -> None:
def test_get_quarter_dump_fresh_zoning_no_match() -> None:
"""Свежий дамп, centroid вне всех territorial_zones → nspd_zoning=None."""
fresh_date = datetime.now(UTC) - timedelta(days=30)
dump_row = ("66:41:0204016", fresh_date, 100, None)
dump_row = ("66:41:0204016", fresh_date, 100, None, 3, 0, 0)
db = _make_db_mock_with_spatial(
dump_row=dump_row,
@ -234,7 +245,9 @@ def test_get_quarter_dump_fresh_zoning_no_match() -> None:
def test_get_quarter_dump_fresh_zouit_overlaps() -> None:
"""Свежий дамп с 2 zouit → nspd_zouit_overlaps length=2."""
fresh_date = datetime.now(UTC) - timedelta(days=10)
dump_row = ("66:41:0204016", fresh_date, 200, None)
# territorial_zones_count > 0 чтобы все 3 spatial запроса вызывались
# (иначе call_count в mock смещается и zouit попадает не в тот слот)
dump_row = ("66:41:0204016", fresh_date, 200, None, 1, 5, 0)
props1: dict[str, Any] = {"name": "ОКН Дом купца", "subcategory": "okn"}
props2: dict[str, Any] = {"name": "ОЗ газопровода", "subcategory": "engineering"}
@ -267,7 +280,7 @@ def test_get_quarter_dump_fresh_zouit_overlaps() -> None:
def test_get_quarter_dump_includes_nspd_fields() -> None:
"""Базовая проверка что все 4 nspd-ключа присутствуют в возвращаемом dict."""
fresh_date = datetime.now(UTC) - timedelta(days=1)
dump_row = ("66:41:0204016", fresh_date, 55, None)
dump_row = ("66:41:0204016", fresh_date, 55, None, 2, 1, 0)
db = _make_db_mock_with_spatial(
dump_row=dump_row,
@ -288,3 +301,89 @@ def test_get_quarter_dump_includes_nspd_fields() -> None:
assert "stale" in dump
assert "harvest_triggered" in dump
assert "total_features" in dump
# ── _make_empty_result — mutation safety ─────────────────────────────────────
def test_make_empty_result_returns_independent_copies() -> None:
"""Каждый вызов _make_empty_result() возвращает независимый dict.
Мутация nspd_dump в одном результате не должна влиять на другой
и на модульный sentinel EMPTY_DUMP_RESULT.
"""
a = _make_empty_result()
b = _make_empty_result()
# Вложенный dict — независимый объект
assert a["nspd_dump"] is not b["nspd_dump"]
assert a["nspd_dump"] is not EMPTY_DUMP_RESULT["nspd_dump"]
# Мутируем один — другой не трогается
a["nspd_dump"]["harvest_triggered"] = True
a["nspd_zouit_overlaps"].append({"group_key": "test"})
assert b["nspd_dump"]["harvest_triggered"] is False
assert b["nspd_zouit_overlaps"] == []
assert EMPTY_DUMP_RESULT["nspd_dump"]["harvest_triggered"] is False
assert EMPTY_DUMP_RESULT["nspd_zouit_overlaps"] == []
def test_make_empty_result_overrides() -> None:
"""_make_empty_result принимает override-аргументы."""
r = _make_empty_result(harvest_triggered=True, stale=True, total_features=42)
assert r["nspd_dump"]["harvest_triggered"] is True
assert r["nspd_dump"]["stale"] is True
assert r["nspd_dump"]["total_features"] == 42
assert r["nspd_dump"]["available"] is False
# ── early-exit when layer_counts == 0 ────────────────────────────────────────
def test_early_exit_all_counts_zero_no_spatial_queries() -> None:
"""Если все denormalized counts == 0 — spatial queries не вызываются."""
fresh_date = datetime.now(UTC) - timedelta(days=5)
# territorial_zones_count=0, zouit_count=0, engineering_count=0
dump_row = ("66:41:0204016", fresh_date, 50, None, 0, 0, 0)
db = _make_db_mock_with_spatial(dump_row=dump_row)
result = get_quarter_dump_data(
db, "66:41:0204016", parcel_wkt="POLYGON ((60.6 56.8, 60.7 56.8, 60.7 56.9, 60.6 56.8))"
)
# Только 1 execute-вызов — чтение dump row; spatial helpers пропущены
assert db.execute.call_count == 1
assert result["nspd_zoning"] is None
assert result["nspd_zouit_overlaps"] == []
assert result["nspd_engineering_nearby"] == []
assert result["nspd_dump"]["available"] is True
def test_early_exit_partial_counts() -> None:
"""territorial_zones_count > 0, zouit_count == 0 → zoning query вызывается,
zouit и engineering нет."""
fresh_date = datetime.now(UTC) - timedelta(days=5)
dump_row = ("66:41:0204016", fresh_date, 30, None, 3, 0, 0)
zone_props: dict[str, Any] = {"reg_numb_border": "Ж-1", "type_zone": "Зона жилая"}
zoning_row_mock = MagicMock()
zoning_row_mock.__getitem__ = lambda self, i: zone_props if i == 0 else None
db = _make_db_mock_with_spatial(
dump_row=dump_row,
zoning_row=zoning_row_mock,
zouit_rows=[],
engineering_rows=[],
)
result = get_quarter_dump_data(
db, "66:41:0204016", parcel_wkt="POLYGON ((60.6 56.8, 60.7 56.8, 60.7 56.9, 60.6 56.8))"
)
# dump lookup (0) + zoning (1) = 2 вызова; zouit и engineering — пропущены
assert db.execute.call_count == 2
assert result["nspd_zoning"] is not None
assert result["nspd_zoning"]["zone_code"] == "Ж-1"
assert result["nspd_zouit_overlaps"] == []
assert result["nspd_engineering_nearby"] == []