feat(tradein): avito_houses.py — Houses Catalog parser (state + reviews + history + recs) #449

Merged
lekss361 merged 2 commits from feat/tradein-avito-houses-parser into main 2026-05-23 12:37:25 +00:00
5 changed files with 1692 additions and 0 deletions

File diff suppressed because it is too large Load diff

View file

@ -93,6 +93,21 @@ class ScrapedLot(BaseModel):
photo_urls: list[str] = Field(default_factory=list)
raw_payload: dict[str, Any] | None = None
# Cian-specific extensions (Stage 2 of CianScraper v1)
living_area_m2: float | None = None
bedrooms_count: int | None = None
balconies_count: int | None = None
loggias_count: int | None = None
description_minhash: str | None = None
cadastral_number: str | None = None
building_cadastral_number: str | None = None
phones: list[dict] = Field(default_factory=list)
is_homeowner: bool | None = None
is_pro_seller: bool | None = None
bargain_allowed: bool | None = None
sale_type: str | None = None
metro_stations: list[dict] = Field(default_factory=list)
def compute_dedup_hash(self) -> str:
"""SHA256(source + (source_id или source_url) + price_rub) — стабильный uniqueness key.

View file

@ -0,0 +1,93 @@
"""Shared utility для извлечения Cian Redux state из _cianConfig.
Pattern: window._cianConfig['<mfe>'].push({key:..., value:..., priority:..., filter:...})
Used by SERP / detail / newbuilding / valuation scrapers.
"""
import json
import logging
import re
from typing import Any
logger = logging.getLogger(__name__)
# Match both patterns Cian uses:
# 1. window._cianConfig['mfe'].push({key:..., value:..., priority:..., filter:...})
# 2. window._cianConfig['mfe'] = window._cianConfig['mfe'] || [];
# window._cianConfig['mfe'].push({...})
_RE_CIAN_PUSH = re.compile(
r"window\._cianConfig\['(?P<mfe>[\w-]+)'\]"
r"(?:\.push\(\s*|"
r"\s*=\s*window\._cianConfig\['[\w-]+'\]\s*\|\|\s*\[\];\s*"
r"window\._cianConfig\['[\w-]+'\]\.push\(\s*)"
r"\{\s*key:\s*['\"](?P<key>\w+)['\"]\s*,\s*"
r"value:\s*(?P<value>.+?)\s*,\s*"
r"priority:",
re.DOTALL,
)
def extract_state(html: str, mfe: str, key: str = "initialState") -> dict[str, Any] | None:
"""Extract state.value where _cianConfig[<mfe>] entry.key matches.
Returns parsed JSON dict, or None if not found / parse failed.
Logs warning at debug level on parse failure for diagnostics.
"""
for match in _RE_CIAN_PUSH.finditer(html):
if match.group("mfe") != mfe or match.group("key") != key:
continue
value_str = match.group("value").strip()
# Strategy 1: direct JSON parse (value уже JSON-like)
try:
return json.loads(value_str)
except json.JSONDecodeError:
pass
# Strategy 2: JSON.parse("escaped string")
m_jp = re.match(r'JSON\.parse\(\s*"(.+?)"\s*\)\s*$', value_str, re.DOTALL)
if m_jp:
escaped = m_jp.group(1)
unescaped = (
escaped
.replace('\\"', '"')
.replace('\\\\', '\\')
.replace('\\n', '\n')
.replace('\\/', '/')
)
try:
return json.loads(unescaped)
except json.JSONDecodeError as exc:
logger.debug(
"JSON.parse() unescape failed for mfe=%s key=%s: %s", mfe, key, exc
)
continue
# Strategy 3: raw JS object literal — try demjson3 (optional dep)
try:
import demjson3 # type: ignore[import-untyped]
return demjson3.decode(value_str)
except ImportError:
logger.debug(
"demjson3 not installed; cannot parse raw JS object for mfe=%s key=%s", mfe, key
)
except Exception as exc:
logger.debug("demjson3 parse failed for mfe=%s key=%s: %s", mfe, key, exc)
return None
def extract_all_states(html: str) -> dict[str, dict[str, Any]]:
"""Returns {mfe_name: {key: parsed_value}} for all push() entries.
Useful for discovery / debugging during scraper development.
"""
result: dict[str, dict[str, Any]] = {}
for match in _RE_CIAN_PUSH.finditer(html):
mfe = match.group("mfe")
key = match.group("key")
parsed = extract_state(html, mfe, key)
if parsed is not None:
result.setdefault(mfe, {})[key] = parsed
return result

View file

@ -0,0 +1,397 @@
"""Offline smoke for avito_houses parser.
Uses minimal JSON fixture mimicking Avito __preloadedState__ structure.
Не требует сети, БД или curl_cffi тестирует только чистый парсер.
"""
from datetime import date
import pytest
from app.services.scrapers.avito_houses import (
HouseCatalogEnrichment,
HouseInfo,
HouseReview,
MiniSerpListing,
PlacementHistoryItem,
RecommendationItem,
SellerInfo,
_parse_ru_date,
parse_houses_state,
)
# ---------------------------------------------------------------------------
# Minimal __preloadedState__ fixture (Python dict — не raw HTML)
# ---------------------------------------------------------------------------
MINIMAL_STATE: dict = {
"data": {"data": {"page": {"placeholders": [
{"type": "breadcrumbs"},
{"type": "title"},
{"type": "gallery", "props": {}},
{
"type": "housePage",
"props": {"developmentData": {
"avitoId": 3171365,
"id": "MTE3LjM3MDg5MA",
"title": "Akademika Postovskogo 17a",
"address": "ул. Постовского, 17а",
"fullAddress": "Свердловская обл., Екатеринбург, ул. Постовского, 17а",
"coords": {"lat": 56.790699, "lng": 60.580191},
"aboutDevelopment": {"expandParams": {"items": [
{"title": "Параметры дома", "params": [
{"type": "Год постройки", "value": "2019"},
{"type": "Этажей", "value": "25"},
{"type": "Тип дома", "value": "Монолитный"},
{"type": "Пассажирский лифт", "value": "2"},
{"type": "Консьерж", "value": "Да"},
{"type": "Закрытый двор", "value": "Да"},
{"type": "Детская площадка", "value": "Да"},
]},
]}},
"developer": {"name": "Атомстройкомплекс", "key": "atomstroy"},
"mapPreview": {
"distance": "в 5 минутах ходьбы",
"objects": "школа, парк, аптека",
"pins": [{"lat": 56.79, "lng": 60.58, "type": "school"}],
},
"ratingBadge": {"info": {"score": 4.666, "scoreString": "4,7"}},
"ratingSummaryStat": {
"ratingStat": [{"score": 5, "count": 4}, {"score": 4, "count": 2}],
"reviewCount": 6,
},
}},
},
{
"type": "reviews",
"props": {"entries": [
{"type": "rating", "value": {
"id": 12345,
"author": {"title": "Иван"},
"reviewTitle": "Хорошо",
"score": 5,
"modelExperience": "Живу в своей квартире",
"rated": "9 октября 2025",
"textSections": [
{"title": "", "text": "Общий отзыв"},
{"title": "Преимущества", "text": "Парк рядом"},
{"title": "Недостатки", "text": "Парковки мало"},
],
}},
{"type": "pages", "value": {"nextPageUrl": "/reviews?page=2"}},
]},
},
{
"type": "miniSerp",
"props": {"items": [
{
"id": 7986882804,
"title": "3-к. квартира",
"description": "Описание",
"url": "/ekaterinburg/kvartiry/test_7986882804",
"price": "11 990 000 ₽",
"geo": {"content": "Чкаловская, 2,7 км", "colors": ["#cf2734"]},
"seller": {
"name": "DOMRF66",
"type": "Компания",
"from": "На Авито с июля 2012",
"url": "/brands/domrf66",
"logos": [],
},
},
]},
},
{
"type": "housePlacementHistory",
"props": {"items": [
{
"id": 7000000001,
"title": "1-к. квартира",
"startPrice": 5000000,
"startPriceDate": 1697000000,
"lastPrice": 4900000,
"lastPriceDate": 1715000000,
"exposure": 210,
"itemImage": {},
},
]},
},
{"type": "uxFeedback"},
{
"type": "recommendations",
"props": {"items": [
{
"title": "Похожая",
"priceRange": "от 6 млн",
"additionalInfo": ["Адрес 1"],
"url": "/rec1",
"images": [{"208x156": "https://avito.ru/img1.jpg"}],
},
]},
},
{"type": "padding"},
]}}}
}
# ---------------------------------------------------------------------------
# Helpers
# ---------------------------------------------------------------------------
def test_ru_date_parse() -> None:
assert _parse_ru_date("9 октября 2025") == date(2025, 10, 9)
assert _parse_ru_date("31 декабря 2020") == date(2020, 12, 31)
assert _parse_ru_date("1 января 2000") == date(2000, 1, 1)
assert _parse_ru_date(None) is None
assert _parse_ru_date("bogus") is None
assert _parse_ru_date("") is None
# ---------------------------------------------------------------------------
# Main parse test
# ---------------------------------------------------------------------------
def test_parse_minimal_state() -> None:
e = parse_houses_state(MINIMAL_STATE, "/catalog/houses/ekaterinburg/test/3171365")
assert isinstance(e, HouseCatalogEnrichment)
assert e.house_url == "/catalog/houses/ekaterinburg/test/3171365"
def test_house_info() -> None:
e = parse_houses_state(MINIMAL_STATE, "/catalog/houses/ekaterinburg/test/3171365")
h = e.house
assert isinstance(h, HouseInfo)
assert h.ext_id == 3171365
assert h.ext_id_hash == "MTE3LjM3MDg5MA"
assert h.title == "Akademika Postovskogo 17a"
assert h.short_address == "ул. Постовского, 17а"
assert h.full_address == "Свердловская обл., Екатеринбург, ул. Постовского, 17а"
assert h.lat == 56.790699
assert h.lon == 60.580191
def test_house_expand_params() -> None:
e = parse_houses_state(MINIMAL_STATE, "/catalog/houses/ekaterinburg/test/3171365")
h = e.house
assert h.year_built == 2019
assert h.total_floors == 25
assert h.house_type == "monolith"
assert h.passenger_elevators == 2
assert h.has_concierge is True
assert h.closed_yard is True
assert h.has_playground is True
def test_house_developer_and_rating() -> None:
e = parse_houses_state(MINIMAL_STATE, "/catalog/houses/ekaterinburg/test/3171365")
h = e.house
assert h.developer_name == "Атомстройкомплекс"
assert h.developer_key == "atomstroy"
assert h.rating_score == 4.666
assert h.rating_string == "4,7"
assert h.reviews_count == 6
assert len(h.rating_distribution) == 2
def test_house_map_and_infra() -> None:
e = parse_houses_state(MINIMAL_STATE, "/catalog/houses/ekaterinburg/test/3171365")
h = e.house
assert h.infrastructure_walk_distance == "в 5 минутах ходьбы"
assert h.infrastructure_summary == "школа, парк, аптека"
assert len(h.map_pins) == 1
assert h.map_pins[0]["type"] == "school"
def test_house_raw_characteristics() -> None:
e = parse_houses_state(MINIMAL_STATE, "/catalog/houses/ekaterinburg/test/3171365")
# raw_characteristics = полный expandParams items (1 категория)
assert len(e.house.raw_characteristics) == 1
assert e.house.raw_characteristics[0]["title"] == "Параметры дома"
def test_reviews() -> None:
e = parse_houses_state(MINIMAL_STATE, "/catalog/houses/ekaterinburg/test/3171365")
assert len(e.reviews) == 1
assert e.reviews_next_page_url == "/reviews?page=2"
r = e.reviews[0]
assert isinstance(r, HouseReview)
assert r.ext_review_id == 12345
assert r.author_name == "Иван"
assert r.review_title == "Хорошо"
assert r.score == 5
assert r.model_experience == "Живу в своей квартире"
assert r.rated_date == date(2025, 10, 9)
assert r.text_main == "Общий отзыв"
assert r.text_pros == "Парк рядом"
assert r.text_cons == "Парковки мало"
assert r.raw_payload is not None
def test_mini_serp() -> None:
e = parse_houses_state(MINIMAL_STATE, "/catalog/houses/ekaterinburg/test/3171365")
assert len(e.mini_serp) == 1
m = e.mini_serp[0]
assert isinstance(m, MiniSerpListing)
assert m.ext_item_id == 7986882804
assert m.price_rub == 11990000
assert m.metro_text == "Чкаловская, 2,7 км"
assert m.metro_color == "#cf2734"
assert m.description == "Описание"
def test_seller_extraction() -> None:
e = parse_houses_state(MINIMAL_STATE, "/catalog/houses/ekaterinburg/test/3171365")
m = e.mini_serp[0]
assert m.seller is not None
s = m.seller
assert isinstance(s, SellerInfo)
assert s.ext_seller_id == "domrf66"
assert s.name == "DOMRF66"
assert s.seller_type == "Компания"
assert s.on_platform_since == "На Авито с июля 2012"
assert s.profile_url == "/brands/domrf66"
def test_placement_history() -> None:
e = parse_houses_state(MINIMAL_STATE, "/catalog/houses/ekaterinburg/test/3171365")
assert len(e.placement_history) == 1
p = e.placement_history[0]
assert isinstance(p, PlacementHistoryItem)
assert p.ext_item_id == "7000000001"
assert p.title == "1-к. квартира"
assert p.start_price == 5000000
assert p.last_price == 4900000
assert p.exposure_days == 210
# removed_date ВСЕГДА None — виджет не отдаёт!
assert p.removed_date is None
# raw_payload не содержит itemImage (скипается)
assert p.raw_payload is not None
assert "itemImage" not in p.raw_payload
# start_price_date и last_price_date — из unix timestamp
assert isinstance(p.start_price_date, date)
assert isinstance(p.last_price_date, date)
def test_recommendations() -> None:
e = parse_houses_state(MINIMAL_STATE, "/catalog/houses/ekaterinburg/test/3171365")
assert len(e.recommendations) == 1
rec = e.recommendations[0]
assert isinstance(rec, RecommendationItem)
assert rec.title == "Похожая"
assert rec.price_text == "от 6 млн"
assert rec.address == "Адрес 1"
assert rec.url == "/rec1"
assert rec.image_url == "https://avito.ru/img1.jpg"
def test_missing_optional_widgets() -> None:
"""Парсер не падает если опциональные виджеты отсутствуют."""
state_minimal: dict = {
"data": {"data": {"page": {"placeholders": [
{
"type": "housePage",
"props": {"developmentData": {
"avitoId": 9999,
"coords": {"lat": 56.0, "lng": 60.0},
"aboutDevelopment": {"expandParams": {"items": []}},
"developer": {},
"mapPreview": {},
"ratingBadge": {"info": {}},
"ratingSummaryStat": {},
}},
},
]}}}
}
e = parse_houses_state(state_minimal, "/catalog/houses/test/9999")
assert e.house.ext_id == 9999
assert e.reviews == []
assert e.mini_serp == []
assert e.placement_history == []
assert e.recommendations == []
assert e.reviews_next_page_url is None
def test_invalid_state_raises() -> None:
"""ValueError если placeholders не найден."""
with pytest.raises(ValueError, match="placeholders"):
parse_houses_state({}, "/catalog/houses/test/1")
def test_missing_house_page_raises() -> None:
"""ValueError если housePage виджет отсутствует."""
state: dict = {
"data": {"data": {"page": {"placeholders": [
{"type": "breadcrumbs"},
]}}}
}
with pytest.raises(ValueError, match="housePage"):
parse_houses_state(state, "/catalog/houses/test/1")
def test_house_type_normalization() -> None:
"""Все варианты типа дома нормализуются корректно."""
from app.services.scrapers.avito_houses import _normalize_house_type
assert _normalize_house_type("Монолитный") == "monolith"
assert _normalize_house_type("Панельный") == "panel"
assert _normalize_house_type("Кирпичный") == "brick"
assert _normalize_house_type("Монолитно-кирпичный") == "monolith_brick"
assert _normalize_house_type("Блочный") == "block"
assert _normalize_house_type("Деревянный") == "wood"
assert _normalize_house_type("Неизвестный") == "other"
assert _normalize_house_type(None) is None
def test_strip_price() -> None:
"""_strip_price корректно извлекает числа из строк Avito."""
from app.services.scrapers.avito_houses import _strip_price
assert _strip_price("11 990 000 ₽") == 11990000
assert _strip_price("5 000 000 ₽") == 5000000
assert _strip_price(None) is None
assert _strip_price("") is None
def test_mini_serp_no_seller() -> None:
"""Объявление без продавца не вызывает ошибку."""
state: dict = {
"data": {"data": {"page": {"placeholders": [
{
"type": "housePage",
"props": {"developmentData": {
"avitoId": 111,
"coords": {"lat": 56.0, "lng": 60.0},
"aboutDevelopment": {"expandParams": {"items": []}},
"developer": {},
"mapPreview": {},
"ratingBadge": {"info": {}},
"ratingSummaryStat": {},
}},
},
{
"type": "miniSerp",
"props": {"items": [
{"id": 12345, "title": "1-к.", "url": "/test", "price": "3 000 000 ₽",
"geo": {"content": "Центр", "colors": []}},
]},
},
]}}}
}
e = parse_houses_state(state, "/catalog/houses/test/111")
assert len(e.mini_serp) == 1
assert e.mini_serp[0].seller is None
assert e.mini_serp[0].price_rub == 3000000

View file

@ -0,0 +1,46 @@
"""Tests for cian_state_parser."""
import pytest
from app.services.scrapers.cian_state_parser import extract_state
def test_extract_state_simple_json():
html = '''
<html><body>
<script>
window._cianConfig['frontend-serp'] = window._cianConfig['frontend-serp'] || [];
window._cianConfig['frontend-serp'].push({
key: 'initialState',
value: {"results": {"offers": [{"id": 1}]}},
priority: 1
});
</script>
</body></html>
'''
state = extract_state(html, mfe='frontend-serp', key='initialState')
assert state is not None
assert state['results']['offers'][0]['id'] == 1
def test_extract_state_json_parse_form():
html = r'''
window._cianConfig['frontend-offer-card'].push({
key: 'defaultState',
value: JSON.parse("{\"offerData\":{\"offer\":{\"cianId\":123}}}"),
priority: 1
});
'''
state = extract_state(html, mfe='frontend-offer-card', key='defaultState')
assert state is not None
assert state['offerData']['offer']['cianId'] == 123
def test_extract_state_missing_returns_none():
html = "<html></html>"
assert extract_state(html, mfe='frontend-serp') is None
def test_extract_state_wrong_mfe_returns_none():
html = '''
window._cianConfig['other-mfe'].push({key: 'initialState', value: {}, priority: 1});
'''
assert extract_state(html, mfe='frontend-serp', key='initialState') is None