Аддитивный Foundation-модуль (эпик #2277 Group F4): scraper_kit/providers/_base.py извлекает 3 реально повторяющихся паттерна, найденных чтением avito/cian/yandex/ domclick providers (~13k строк): - build_browser_fetcher(config: ScraperConfig, source, ...) — config MANDATORY, устраняет footgun-класс #2322/#2330 (BrowserFetcher конструировался вручную в 6+ местах, endpoint/proxy-pool wiring легко забыть/потерять). - build_curl_cffi_session / build_document_session + http_proxies() — общий curl_cffi AsyncSession(impersonate=chrome120, proxies=..., headers=...) boilerplate, дублированный в avito/serp.py, avito/detail.py, avito/imv.py, cian/detail.py, cian/newbuilding.py (x2), cian/session.py. - DOCUMENT_HEADERS — идентичный "document" headers dict, дублированный трижды. Retry/backoff-циклы (yandex tarpit-retry, avito page-retry, tenacity в BaseScraper) сознательно НЕ обобщены — разная семантика ошибок/действий между попытками, обобщение дало бы leaky abstraction (см. docstring модуля). Existing providers/{avito,cian,yandex,domclick}/*.py НЕ тронуты — миграция call-site'ов на новый модуль это отдельные future issues (F4a-d). Заодно: актуализирован устаревший докстринг app/services/scraper_adapters.py ("Пока НИКТО их не использует") — адаптеры wired в прод с #2192 через app.scheduler_main._run_kit_scheduler.
168 lines
6.3 KiB
Python
168 lines
6.3 KiB
Python
"""Тесты shared-фундамента `scraper_kit.providers._base` (#2358, Foundation).
|
||
|
||
Покрывает:
|
||
- `http_proxies()` — url → dict / None passthrough.
|
||
- `build_curl_cffi_session()` / `build_document_session()` — корректные kwargs
|
||
в curl_cffi `AsyncSession` (impersonate/timeout/proxies/headers/cookies).
|
||
- `build_browser_fetcher()` — `config` MANDATORY (нет default в сигнатуре);
|
||
endpoint/use_pool/proxy_provider корректно прокинуты; `fetch_timeout_s`
|
||
опционален (дефолт `BrowserFetcher`, если не передан явно).
|
||
|
||
Изолированные — никакой интеграции с existing providers/avito|cian|yandex|domclick.
|
||
"""
|
||
|
||
from __future__ import annotations
|
||
|
||
import inspect
|
||
from typing import Any
|
||
|
||
from scraper_kit.browser_fetcher import BrowserFetcher
|
||
from scraper_kit.providers._base import (
|
||
DEFAULT_IMPERSONATE,
|
||
DOCUMENT_HEADERS,
|
||
build_browser_fetcher,
|
||
build_curl_cffi_session,
|
||
build_document_session,
|
||
http_proxies,
|
||
)
|
||
|
||
|
||
class _FakeConfig:
|
||
"""Минимальная ScraperConfig-заглушка — только поля, читаемые build_browser_fetcher."""
|
||
|
||
def __init__(self, endpoint: str, use_pool: bool = False) -> None:
|
||
self.browser_http_endpoint = endpoint
|
||
self.use_proxy_pool_browser = use_pool
|
||
|
||
|
||
class _FakeProxyProvider:
|
||
"""ProxyProvider-заглушка — сюда достаточно структурного соответствия, вызовы не нужны."""
|
||
|
||
def acquire(self, provider: str) -> Any:
|
||
raise NotImplementedError
|
||
|
||
def release(self, lease: Any) -> None:
|
||
raise NotImplementedError
|
||
|
||
def mark_health(self, lease: Any, ok: bool, **_: Any) -> None:
|
||
raise NotImplementedError
|
||
|
||
|
||
# ── http_proxies ──────────────────────────────────────────────────────────────
|
||
|
||
|
||
def test_http_proxies_none_passthrough() -> None:
|
||
assert http_proxies(None) is None
|
||
|
||
|
||
def test_http_proxies_empty_string_is_none() -> None:
|
||
assert http_proxies("") is None
|
||
|
||
|
||
def test_http_proxies_builds_dict_for_both_schemes() -> None:
|
||
result = http_proxies("http://user:pass@proxy:8080")
|
||
assert result == {
|
||
"http": "http://user:pass@proxy:8080",
|
||
"https": "http://user:pass@proxy:8080",
|
||
}
|
||
|
||
|
||
# ── build_curl_cffi_session / build_document_session ──────────────────────────
|
||
|
||
|
||
async def test_build_curl_cffi_session_wires_impersonate_timeout_proxies() -> None:
|
||
session = build_curl_cffi_session(proxy_url="http://proxy:8080", timeout=15.0)
|
||
try:
|
||
assert session.impersonate == DEFAULT_IMPERSONATE
|
||
assert session.timeout == 15.0
|
||
assert session.proxies == {"http": "http://proxy:8080", "https": "http://proxy:8080"}
|
||
finally:
|
||
await session.close()
|
||
|
||
|
||
async def test_build_curl_cffi_session_no_proxy_direct_connection() -> None:
|
||
session = build_curl_cffi_session(proxy_url=None)
|
||
try:
|
||
# curl_cffi нормализует proxies=None в {} (falsy) — прямое подключение.
|
||
assert not session.proxies
|
||
finally:
|
||
await session.close()
|
||
|
||
|
||
async def test_build_curl_cffi_session_passes_headers_and_cookies() -> None:
|
||
session = build_curl_cffi_session(
|
||
proxy_url=None, headers={"X-Test": "1"}, cookies={"sid": "abc"}
|
||
)
|
||
try:
|
||
assert dict(session.headers).get("x-test") == "1"
|
||
assert session.cookies.get("sid") == "abc"
|
||
finally:
|
||
await session.close()
|
||
|
||
|
||
async def test_build_document_session_uses_document_headers() -> None:
|
||
session = build_document_session(proxy_url=None)
|
||
try:
|
||
headers = dict(session.headers)
|
||
for key, value in DOCUMENT_HEADERS.items():
|
||
assert headers.get(key.lower()) == value
|
||
finally:
|
||
await session.close()
|
||
|
||
|
||
async def test_build_document_session_custom_timeout_and_proxy() -> None:
|
||
session = build_document_session(proxy_url="http://proxy:8080", timeout=10.0)
|
||
try:
|
||
assert session.timeout == 10.0
|
||
assert session.proxies == {"http": "http://proxy:8080", "https": "http://proxy:8080"}
|
||
finally:
|
||
await session.close()
|
||
|
||
|
||
# ── build_browser_fetcher ──────────────────────────────────────────────────────
|
||
|
||
|
||
def test_build_browser_fetcher_config_is_mandatory_in_signature() -> None:
|
||
"""config — параметр БЕЗ default в сигнатуре: пропустить его невозможно (footgun-guard)."""
|
||
sig = inspect.signature(build_browser_fetcher)
|
||
config_param = sig.parameters["config"]
|
||
assert config_param.default is inspect.Parameter.empty
|
||
|
||
|
||
def test_build_browser_fetcher_wires_endpoint_and_use_pool() -> None:
|
||
config = _FakeConfig(endpoint="http://browser:3000", use_pool=True)
|
||
provider = _FakeProxyProvider()
|
||
|
||
fetcher = build_browser_fetcher(config, "avito", proxy_provider=provider) # type: ignore[arg-type]
|
||
|
||
assert isinstance(fetcher, BrowserFetcher)
|
||
assert fetcher._endpoint == "http://browser:3000"
|
||
assert fetcher._use_pool is True
|
||
assert fetcher._proxy_provider is provider
|
||
assert fetcher._source == "avito"
|
||
|
||
|
||
def test_build_browser_fetcher_use_pool_false_by_default_config() -> None:
|
||
config = _FakeConfig(endpoint="http://browser:3000", use_pool=False)
|
||
|
||
fetcher = build_browser_fetcher(config, "cian") # type: ignore[arg-type]
|
||
|
||
assert fetcher._use_pool is False
|
||
assert fetcher._proxy_provider is None
|
||
|
||
|
||
def test_build_browser_fetcher_default_timeout_when_not_specified() -> None:
|
||
config = _FakeConfig(endpoint="http://browser:3000")
|
||
|
||
fetcher = build_browser_fetcher(config, "yandex") # type: ignore[arg-type]
|
||
|
||
# Дефолт BrowserFetcher (_HTTP_TIMEOUT_S = 120.0) — не переопределён.
|
||
assert fetcher._fetch_timeout_s == 120.0
|
||
|
||
|
||
def test_build_browser_fetcher_explicit_timeout_override() -> None:
|
||
config = _FakeConfig(endpoint="http://browser:3000")
|
||
|
||
fetcher = build_browser_fetcher(config, "yandex", fetch_timeout_s=30.0) # type: ignore[arg-type]
|
||
|
||
assert fetcher._fetch_timeout_s == 30.0
|