2021-01-05 10:24:39 +00:00
|
|
|
# SPDX-License-Identifier: AGPL-3.0-or-later
|
|
|
|
|
2021-09-22 11:36:34 +00:00
|
|
|
import gc
|
2020-12-24 08:28:16 +00:00
|
|
|
import typing
|
|
|
|
import types
|
|
|
|
import functools
|
|
|
|
import itertools
|
|
|
|
from time import time
|
2021-04-14 16:11:35 +00:00
|
|
|
from timeit import default_timer
|
2020-12-24 08:28:16 +00:00
|
|
|
from urllib.parse import urlparse
|
|
|
|
|
|
|
|
import re
|
2021-01-19 20:26:04 +00:00
|
|
|
from langdetect import detect_langs
|
|
|
|
from langdetect.lang_detect_exception import LangDetectException
|
2021-03-18 18:59:01 +00:00
|
|
|
import httpx
|
2020-12-24 08:28:16 +00:00
|
|
|
|
[httpx] replace searx.poolrequests by searx.network
settings.yml:
* outgoing.networks:
* can contains network definition
* propertiers: enable_http, verify, http2, max_connections, max_keepalive_connections,
keepalive_expiry, local_addresses, support_ipv4, support_ipv6, proxies, max_redirects, retries
* retries: 0 by default, number of times searx retries to send the HTTP request (using different IP & proxy each time)
* local_addresses can be "192.168.0.1/24" (it supports IPv6)
* support_ipv4 & support_ipv6: both True by default
see https://github.com/searx/searx/pull/1034
* each engine can define a "network" section:
* either a full network description
* either reference an existing network
* all HTTP requests of engine use the same HTTP configuration (it was not the case before, see proxy configuration in master)
2021-04-05 08:43:33 +00:00
|
|
|
from searx import network, logger
|
2021-09-22 11:36:34 +00:00
|
|
|
from searx.utils import gen_useragent
|
2020-12-24 08:28:16 +00:00
|
|
|
from searx.results import ResultContainer
|
2021-01-05 10:24:39 +00:00
|
|
|
from searx.search.models import SearchQuery, EngineRef
|
2020-12-24 08:28:16 +00:00
|
|
|
from searx.search.processors import EngineProcessor
|
2021-04-14 15:23:15 +00:00
|
|
|
from searx.metrics import counter_inc
|
2020-12-24 08:28:16 +00:00
|
|
|
|
|
|
|
|
2021-01-08 18:04:04 +00:00
|
|
|
logger = logger.getChild('searx.search.checker')
|
|
|
|
|
2020-12-24 08:28:16 +00:00
|
|
|
HTML_TAGS = [
|
2021-12-27 08:16:03 +00:00
|
|
|
# fmt: off
|
2020-12-24 08:28:16 +00:00
|
|
|
'embed', 'iframe', 'object', 'param', 'picture', 'source', 'svg', 'math', 'canvas', 'noscript', 'script',
|
|
|
|
'del', 'ins', 'area', 'audio', 'img', 'map', 'track', 'video', 'a', 'abbr', 'b', 'bdi', 'bdo', 'br', 'cite',
|
|
|
|
'code', 'data', 'dfn', 'em', 'i', 'kdb', 'mark', 'q', 'rb', 'rp', 'rt', 'rtc', 'ruby', 's', 'samp', 'small',
|
|
|
|
'span', 'strong', 'sub', 'sup', 'time', 'u', 'var', 'wbr', 'style', 'blockquote', 'dd', 'div', 'dl', 'dt',
|
|
|
|
'figcaption', 'figure', 'hr', 'li', 'ol', 'p', 'pre', 'ul', 'button', 'datalist', 'fieldset', 'form', 'input',
|
|
|
|
'label', 'legend', 'meter', 'optgroup', 'option', 'output', 'progress', 'select', 'textarea', 'applet',
|
|
|
|
'frame', 'frameset'
|
2021-12-27 08:16:03 +00:00
|
|
|
# fmt: on
|
2020-12-24 08:28:16 +00:00
|
|
|
]
|
|
|
|
|
|
|
|
|
|
|
|
def get_check_no_html():
|
|
|
|
rep = ['<' + tag + '[^\>]*>' for tag in HTML_TAGS]
|
|
|
|
rep += ['</' + tag + '>' for tag in HTML_TAGS]
|
|
|
|
pattern = re.compile('|'.join(rep))
|
|
|
|
|
|
|
|
def f(text):
|
|
|
|
return pattern.search(text.lower()) is None
|
|
|
|
|
|
|
|
return f
|
|
|
|
|
|
|
|
|
|
|
|
_check_no_html = get_check_no_html()
|
|
|
|
|
|
|
|
|
|
|
|
def _is_url(url):
|
|
|
|
try:
|
|
|
|
result = urlparse(url)
|
|
|
|
except ValueError:
|
|
|
|
return False
|
|
|
|
if result.scheme not in ('http', 'https'):
|
|
|
|
return False
|
|
|
|
return True
|
|
|
|
|
|
|
|
|
|
|
|
@functools.lru_cache(maxsize=8192)
|
2021-09-22 11:36:34 +00:00
|
|
|
def _download_and_check_if_image(image_url: str) -> bool:
|
|
|
|
"""Download an URL and check if the Content-Type starts with "image/"
|
|
|
|
This function should not be called directly: use _is_url_image
|
|
|
|
otherwise the cache of functools.lru_cache contains data: URL which might be huge.
|
|
|
|
"""
|
2020-12-24 08:28:16 +00:00
|
|
|
retry = 2
|
|
|
|
|
|
|
|
while retry > 0:
|
|
|
|
a = time()
|
|
|
|
try:
|
2021-09-22 11:36:34 +00:00
|
|
|
# use "image_proxy" (avoid HTTP/2)
|
|
|
|
network.set_context_network_name('image_proxy')
|
2022-01-22 15:11:42 +00:00
|
|
|
r, stream = network.stream(
|
2021-12-27 08:26:22 +00:00
|
|
|
'GET',
|
|
|
|
image_url,
|
|
|
|
timeout=10.0,
|
|
|
|
allow_redirects=True,
|
|
|
|
headers={
|
|
|
|
'User-Agent': gen_useragent(),
|
|
|
|
'Accept': 'text/html,application/xhtml+xml,application/xml;q=0.9,image/webp,*/*;q=0.8',
|
|
|
|
'Accept-Language': 'en-US;q=0.5,en;q=0.3',
|
|
|
|
'Accept-Encoding': 'gzip, deflate, br',
|
|
|
|
'DNT': '1',
|
|
|
|
'Connection': 'keep-alive',
|
|
|
|
'Upgrade-Insecure-Requests': '1',
|
|
|
|
'Sec-GPC': '1',
|
|
|
|
'Cache-Control': 'max-age=0',
|
|
|
|
},
|
|
|
|
)
|
2021-09-22 11:36:34 +00:00
|
|
|
r.close()
|
2021-09-28 13:26:34 +00:00
|
|
|
if r.status_code == 200:
|
|
|
|
is_image = r.headers.get('content-type', '').startswith('image/')
|
|
|
|
else:
|
|
|
|
is_image = False
|
2021-09-22 11:36:34 +00:00
|
|
|
del r
|
|
|
|
del stream
|
|
|
|
return is_image
|
2021-03-18 18:59:01 +00:00
|
|
|
except httpx.TimeoutException:
|
2020-12-24 08:28:16 +00:00
|
|
|
logger.error('Timeout for %s: %i', image_url, int(time() - a))
|
|
|
|
retry -= 1
|
2021-03-18 18:59:01 +00:00
|
|
|
except httpx.HTTPError:
|
2020-12-24 08:28:16 +00:00
|
|
|
logger.exception('Exception for %s', image_url)
|
|
|
|
return False
|
2021-09-22 11:36:34 +00:00
|
|
|
return False
|
|
|
|
|
|
|
|
|
|
|
|
def _is_url_image(image_url) -> bool:
|
2021-12-27 08:26:22 +00:00
|
|
|
"""Normalize image_url"""
|
2021-09-22 11:36:34 +00:00
|
|
|
if not isinstance(image_url, str):
|
|
|
|
return False
|
|
|
|
|
|
|
|
if image_url.startswith('//'):
|
|
|
|
image_url = 'https:' + image_url
|
|
|
|
|
|
|
|
if image_url.startswith('data:'):
|
|
|
|
return image_url.startswith('data:image/')
|
|
|
|
|
|
|
|
if not _is_url(image_url):
|
|
|
|
return False
|
|
|
|
|
|
|
|
return _download_and_check_if_image(image_url)
|
2020-12-24 08:28:16 +00:00
|
|
|
|
|
|
|
|
|
|
|
def _search_query_to_dict(search_query: SearchQuery) -> typing.Dict[str, typing.Any]:
|
|
|
|
return {
|
|
|
|
'query': search_query.query,
|
|
|
|
'lang': search_query.lang,
|
|
|
|
'pageno': search_query.pageno,
|
|
|
|
'safesearch': search_query.safesearch,
|
|
|
|
'time_range': search_query.time_range,
|
|
|
|
}
|
|
|
|
|
|
|
|
|
2021-12-27 08:26:22 +00:00
|
|
|
def _search_query_diff(
|
|
|
|
sq1: SearchQuery, sq2: SearchQuery
|
|
|
|
) -> typing.Tuple[typing.Dict[str, typing.Any], typing.Dict[str, typing.Any]]:
|
2020-12-24 08:28:16 +00:00
|
|
|
param1 = _search_query_to_dict(sq1)
|
|
|
|
param2 = _search_query_to_dict(sq2)
|
|
|
|
common = {}
|
|
|
|
diff = {}
|
|
|
|
for k, value1 in param1.items():
|
|
|
|
value2 = param2[k]
|
|
|
|
if value1 == value2:
|
|
|
|
common[k] = value1
|
|
|
|
else:
|
|
|
|
diff[k] = (value1, value2)
|
|
|
|
return (common, diff)
|
|
|
|
|
|
|
|
|
|
|
|
class TestResults:
|
|
|
|
|
2021-01-08 18:04:04 +00:00
|
|
|
__slots__ = 'errors', 'logs', 'languages'
|
2020-12-24 08:28:16 +00:00
|
|
|
|
|
|
|
def __init__(self):
|
|
|
|
self.errors: typing.Dict[str, typing.List[str]] = {}
|
2021-01-08 18:04:04 +00:00
|
|
|
self.logs: typing.Dict[str, typing.List[typing.Any]] = {}
|
|
|
|
self.languages: typing.Set[str] = set()
|
2020-12-24 08:28:16 +00:00
|
|
|
|
2021-01-08 18:04:04 +00:00
|
|
|
def add_error(self, test, message, *args):
|
|
|
|
# message to self.errors
|
2020-12-24 08:28:16 +00:00
|
|
|
errors_for_test = self.errors.setdefault(test, [])
|
|
|
|
if message not in errors_for_test:
|
|
|
|
errors_for_test.append(message)
|
2021-01-08 18:04:04 +00:00
|
|
|
# (message, *args) to self.logs
|
|
|
|
logs_for_test = self.logs.setdefault(test, [])
|
|
|
|
if (message, *args) not in logs_for_test:
|
|
|
|
logs_for_test.append((message, *args))
|
2020-12-24 08:28:16 +00:00
|
|
|
|
2021-01-08 18:04:04 +00:00
|
|
|
def add_language(self, language):
|
|
|
|
self.languages.add(language)
|
2020-12-24 08:28:16 +00:00
|
|
|
|
|
|
|
@property
|
|
|
|
def succesfull(self):
|
|
|
|
return len(self.errors) == 0
|
|
|
|
|
|
|
|
def __iter__(self):
|
|
|
|
for test_name, errors in self.errors.items():
|
|
|
|
for error in sorted(errors):
|
|
|
|
yield (test_name, error)
|
|
|
|
|
|
|
|
|
|
|
|
class ResultContainerTests:
|
|
|
|
|
|
|
|
__slots__ = 'test_name', 'search_query', 'result_container', 'languages', 'stop_test', 'test_results'
|
|
|
|
|
2021-12-27 08:26:22 +00:00
|
|
|
def __init__(
|
|
|
|
self, test_results: TestResults, test_name: str, search_query: SearchQuery, result_container: ResultContainer
|
|
|
|
):
|
2020-12-24 08:28:16 +00:00
|
|
|
self.test_name = test_name
|
|
|
|
self.search_query = search_query
|
|
|
|
self.result_container = result_container
|
|
|
|
self.languages: typing.Set[str] = set()
|
|
|
|
self.test_results = test_results
|
|
|
|
self.stop_test = False
|
|
|
|
|
|
|
|
@property
|
|
|
|
def result_urls(self):
|
|
|
|
results = self.result_container.get_ordered_results()
|
2021-03-25 08:37:37 +00:00
|
|
|
return [result['url'] for result in results if 'url' in result]
|
2020-12-24 08:28:16 +00:00
|
|
|
|
2021-01-08 18:04:04 +00:00
|
|
|
def _record_error(self, message: str, *args) -> None:
|
|
|
|
sq = _search_query_to_dict(self.search_query)
|
|
|
|
sqstr = ' '.join(['{}={!r}'.format(k, v) for k, v in sq.items()])
|
|
|
|
self.test_results.add_error(self.test_name, message, *args, '(' + sqstr + ')')
|
2020-12-24 08:28:16 +00:00
|
|
|
|
|
|
|
def _add_language(self, text: str) -> typing.Optional[str]:
|
2021-01-19 20:26:04 +00:00
|
|
|
try:
|
|
|
|
r = detect_langs(str(text)) # pylint: disable=E1101
|
|
|
|
except LangDetectException:
|
|
|
|
return None
|
|
|
|
|
|
|
|
if len(r) > 0 and r[0].prob > 0.95:
|
|
|
|
self.languages.add(r[0].lang)
|
|
|
|
self.test_results.add_language(r[0].lang)
|
2020-12-24 08:28:16 +00:00
|
|
|
return None
|
|
|
|
|
|
|
|
def _check_result(self, result):
|
|
|
|
if not _check_no_html(result.get('title', '')):
|
2021-01-08 18:04:04 +00:00
|
|
|
self._record_error('HTML in title', repr(result.get('title', '')))
|
2020-12-24 08:28:16 +00:00
|
|
|
if not _check_no_html(result.get('content', '')):
|
2021-01-08 18:04:04 +00:00
|
|
|
self._record_error('HTML in content', repr(result.get('content', '')))
|
2021-03-25 08:37:37 +00:00
|
|
|
if result.get('url') is None:
|
|
|
|
self._record_error('url is None')
|
2020-12-24 08:28:16 +00:00
|
|
|
|
|
|
|
self._add_language(result.get('title', ''))
|
|
|
|
self._add_language(result.get('content', ''))
|
|
|
|
|
|
|
|
template = result.get('template', 'default.html')
|
|
|
|
if template == 'default.html':
|
|
|
|
return
|
|
|
|
if template == 'code.html':
|
|
|
|
return
|
|
|
|
if template == 'torrent.html':
|
|
|
|
return
|
|
|
|
if template == 'map.html':
|
|
|
|
return
|
|
|
|
if template == 'images.html':
|
|
|
|
thumbnail_src = result.get('thumbnail_src')
|
|
|
|
if thumbnail_src is not None:
|
|
|
|
if not _is_url_image(thumbnail_src):
|
2021-01-08 18:04:04 +00:00
|
|
|
self._record_error('thumbnail_src URL is invalid', thumbnail_src)
|
2020-12-24 08:28:16 +00:00
|
|
|
elif not _is_url_image(result.get('img_src')):
|
2021-01-08 18:04:04 +00:00
|
|
|
self._record_error('img_src URL is invalid', result.get('img_src'))
|
2020-12-24 08:28:16 +00:00
|
|
|
if template == 'videos.html' and not _is_url_image(result.get('thumbnail')):
|
2021-01-08 18:04:04 +00:00
|
|
|
self._record_error('thumbnail URL is invalid', result.get('img_src'))
|
2020-12-24 08:28:16 +00:00
|
|
|
|
|
|
|
def _check_results(self, results: list):
|
|
|
|
for result in results:
|
|
|
|
self._check_result(result)
|
|
|
|
|
|
|
|
def _check_answers(self, answers):
|
|
|
|
for answer in answers:
|
|
|
|
if not _check_no_html(answer):
|
2021-01-08 18:04:04 +00:00
|
|
|
self._record_error('HTML in answer', answer)
|
2020-12-24 08:28:16 +00:00
|
|
|
|
|
|
|
def _check_infoboxes(self, infoboxes):
|
|
|
|
for infobox in infoboxes:
|
|
|
|
if not _check_no_html(infobox.get('content', '')):
|
2021-01-08 18:04:04 +00:00
|
|
|
self._record_error('HTML in infobox content', infobox.get('content', ''))
|
2020-12-24 08:28:16 +00:00
|
|
|
self._add_language(infobox.get('content', ''))
|
|
|
|
for attribute in infobox.get('attributes', {}):
|
|
|
|
if not _check_no_html(attribute.get('value', '')):
|
2021-01-08 18:04:04 +00:00
|
|
|
self._record_error('HTML in infobox attribute value', attribute.get('value', ''))
|
2020-12-24 08:28:16 +00:00
|
|
|
|
|
|
|
def check_basic(self):
|
|
|
|
if len(self.result_container.unresponsive_engines) > 0:
|
|
|
|
for message in self.result_container.unresponsive_engines:
|
|
|
|
self._record_error(message[1] + ' ' + (message[2] or ''))
|
|
|
|
self.stop_test = True
|
|
|
|
return
|
|
|
|
|
|
|
|
results = self.result_container.get_ordered_results()
|
|
|
|
if len(results) > 0:
|
|
|
|
self._check_results(results)
|
|
|
|
|
|
|
|
if len(self.result_container.answers) > 0:
|
|
|
|
self._check_answers(self.result_container.answers)
|
|
|
|
|
|
|
|
if len(self.result_container.infoboxes) > 0:
|
|
|
|
self._check_infoboxes(self.result_container.infoboxes)
|
|
|
|
|
|
|
|
def has_infobox(self):
|
2021-01-05 10:24:39 +00:00
|
|
|
"""Check the ResultContainer has at least one infobox"""
|
2020-12-24 08:28:16 +00:00
|
|
|
if len(self.result_container.infoboxes) == 0:
|
|
|
|
self._record_error('No infobox')
|
|
|
|
|
|
|
|
def has_answer(self):
|
2021-01-05 10:24:39 +00:00
|
|
|
"""Check the ResultContainer has at least one answer"""
|
2020-12-24 08:28:16 +00:00
|
|
|
if len(self.result_container.answers) == 0:
|
|
|
|
self._record_error('No answer')
|
|
|
|
|
|
|
|
def has_language(self, lang):
|
2021-01-05 10:24:39 +00:00
|
|
|
"""Check at least one title or content of the results is written in the `lang`.
|
|
|
|
|
|
|
|
Detected using pycld3, may be not accurate"""
|
2020-12-24 08:28:16 +00:00
|
|
|
if lang not in self.languages:
|
|
|
|
self._record_error(lang + ' not found')
|
|
|
|
|
|
|
|
def not_empty(self):
|
2021-01-05 10:24:39 +00:00
|
|
|
"""Check the ResultContainer has at least one answer or infobox or result"""
|
2020-12-24 08:28:16 +00:00
|
|
|
result_types = set()
|
|
|
|
results = self.result_container.get_ordered_results()
|
|
|
|
if len(results) > 0:
|
|
|
|
result_types.add('results')
|
|
|
|
|
|
|
|
if len(self.result_container.answers) > 0:
|
|
|
|
result_types.add('answers')
|
|
|
|
|
|
|
|
if len(self.result_container.infoboxes) > 0:
|
|
|
|
result_types.add('infoboxes')
|
|
|
|
|
|
|
|
if len(result_types) == 0:
|
|
|
|
self._record_error('No result')
|
|
|
|
|
|
|
|
def one_title_contains(self, title: str):
|
2021-01-05 10:24:39 +00:00
|
|
|
"""Check one of the title contains `title` (case insensitive comparaison)"""
|
2020-12-24 08:28:16 +00:00
|
|
|
title = title.lower()
|
|
|
|
for result in self.result_container.get_ordered_results():
|
|
|
|
if title in result['title'].lower():
|
|
|
|
return
|
|
|
|
self._record_error(('{!r} not found in the title'.format(title)))
|
|
|
|
|
|
|
|
|
|
|
|
class CheckerTests:
|
|
|
|
|
|
|
|
__slots__ = 'test_results', 'test_name', 'result_container_tests_list'
|
|
|
|
|
2021-12-27 08:26:22 +00:00
|
|
|
def __init__(
|
|
|
|
self, test_results: TestResults, test_name: str, result_container_tests_list: typing.List[ResultContainerTests]
|
|
|
|
):
|
2020-12-24 08:28:16 +00:00
|
|
|
self.test_results = test_results
|
|
|
|
self.test_name = test_name
|
|
|
|
self.result_container_tests_list = result_container_tests_list
|
|
|
|
|
|
|
|
def unique_results(self):
|
2021-03-25 08:37:37 +00:00
|
|
|
"""Check the results of each ResultContainer is unique"""
|
2020-12-24 08:28:16 +00:00
|
|
|
urls_list = [rct.result_urls for rct in self.result_container_tests_list]
|
|
|
|
if len(urls_list[0]) > 0:
|
|
|
|
# results on the first page
|
|
|
|
for i, urls_i in enumerate(urls_list):
|
|
|
|
for j, urls_j in enumerate(urls_list):
|
|
|
|
if i < j and urls_i == urls_j:
|
2021-12-27 08:26:22 +00:00
|
|
|
common, diff = _search_query_diff(
|
|
|
|
self.result_container_tests_list[i].search_query,
|
|
|
|
self.result_container_tests_list[j].search_query,
|
|
|
|
)
|
2020-12-24 08:28:16 +00:00
|
|
|
common_str = ' '.join(['{}={!r}'.format(k, v) for k, v in common.items()])
|
2021-12-27 08:26:22 +00:00
|
|
|
diff1_str = ', '.join(['{}={!r}'.format(k, v1) for (k, (v1, v2)) in diff.items()])
|
|
|
|
diff2_str = ', '.join(['{}={!r}'.format(k, v2) for (k, (v1, v2)) in diff.items()])
|
|
|
|
self.test_results.add_error(
|
|
|
|
self.test_name,
|
|
|
|
'results are identitical for {} and {} ({})'.format(diff1_str, diff2_str, common_str),
|
|
|
|
)
|
2020-12-24 08:28:16 +00:00
|
|
|
|
|
|
|
|
|
|
|
class Checker:
|
|
|
|
|
|
|
|
__slots__ = 'processor', 'tests', 'test_results'
|
|
|
|
|
|
|
|
def __init__(self, processor: EngineProcessor):
|
|
|
|
self.processor = processor
|
|
|
|
self.tests = self.processor.get_tests()
|
|
|
|
self.test_results = TestResults()
|
|
|
|
|
|
|
|
@property
|
|
|
|
def engineref_list(self):
|
|
|
|
engine_name = self.processor.engine_name
|
|
|
|
engine_category = self.processor.engine.categories[0]
|
|
|
|
return [EngineRef(engine_name, engine_category)]
|
|
|
|
|
|
|
|
@staticmethod
|
|
|
|
def search_query_matrix_iterator(engineref_list, matrix):
|
|
|
|
p = []
|
|
|
|
for name, values in matrix.items():
|
|
|
|
if isinstance(values, (tuple, list)):
|
|
|
|
l = [(name, value) for value in values]
|
|
|
|
else:
|
|
|
|
l = [(name, values)]
|
|
|
|
p.append(l)
|
|
|
|
|
|
|
|
for kwargs in itertools.product(*p):
|
|
|
|
kwargs = {k: v for k, v in kwargs}
|
|
|
|
query = kwargs['query']
|
|
|
|
params = dict(kwargs)
|
|
|
|
del params['query']
|
|
|
|
yield SearchQuery(query, engineref_list, **params)
|
|
|
|
|
|
|
|
def call_test(self, obj, test_description):
|
|
|
|
if isinstance(test_description, (tuple, list)):
|
|
|
|
method, args = test_description[0], test_description[1:]
|
|
|
|
else:
|
|
|
|
method = test_description
|
|
|
|
args = ()
|
|
|
|
if isinstance(method, str) and hasattr(obj, method):
|
|
|
|
getattr(obj, method)(*args)
|
|
|
|
elif isinstance(method, types.FunctionType):
|
|
|
|
method(*args)
|
|
|
|
else:
|
2021-12-27 08:26:22 +00:00
|
|
|
self.test_results.add_error(
|
|
|
|
obj.test_name,
|
|
|
|
'method {!r} ({}) not found for {}'.format(method, method.__class__.__name__, obj.__class__.__name__),
|
|
|
|
)
|
2020-12-24 08:28:16 +00:00
|
|
|
|
|
|
|
def call_tests(self, obj, test_descriptions):
|
|
|
|
for test_description in test_descriptions:
|
|
|
|
self.call_test(obj, test_description)
|
|
|
|
|
|
|
|
def search(self, search_query: SearchQuery) -> ResultContainer:
|
|
|
|
result_container = ResultContainer()
|
|
|
|
engineref_category = search_query.engineref_list[0].category
|
|
|
|
params = self.processor.get_params(search_query, engineref_category)
|
|
|
|
if params is not None:
|
2021-04-14 15:23:15 +00:00
|
|
|
counter_inc('engine', search_query.engineref_list[0].name, 'search', 'count', 'sent')
|
2021-04-14 16:11:35 +00:00
|
|
|
self.processor.search(search_query.query, params, result_container, default_timer(), 5)
|
2020-12-24 08:28:16 +00:00
|
|
|
return result_container
|
|
|
|
|
|
|
|
def get_result_container_tests(self, test_name: str, search_query: SearchQuery) -> ResultContainerTests:
|
|
|
|
result_container = self.search(search_query)
|
|
|
|
result_container_check = ResultContainerTests(self.test_results, test_name, search_query, result_container)
|
|
|
|
result_container_check.check_basic()
|
|
|
|
return result_container_check
|
|
|
|
|
|
|
|
def run_test(self, test_name):
|
|
|
|
test_parameters = self.tests[test_name]
|
|
|
|
search_query_list = list(Checker.search_query_matrix_iterator(self.engineref_list, test_parameters['matrix']))
|
|
|
|
rct_list = [self.get_result_container_tests(test_name, search_query) for search_query in search_query_list]
|
|
|
|
stop_test = False
|
|
|
|
if 'result_container' in test_parameters:
|
|
|
|
for rct in rct_list:
|
|
|
|
stop_test = stop_test or rct.stop_test
|
|
|
|
if not rct.stop_test:
|
|
|
|
self.call_tests(rct, test_parameters['result_container'])
|
|
|
|
if not stop_test:
|
|
|
|
if 'test' in test_parameters:
|
|
|
|
checker_tests = CheckerTests(self.test_results, test_name, rct_list)
|
|
|
|
self.call_tests(checker_tests, test_parameters['test'])
|
|
|
|
|
|
|
|
def run(self):
|
|
|
|
for test_name in self.tests:
|
|
|
|
self.run_test(test_name)
|
2021-09-22 11:36:34 +00:00
|
|
|
# clear cache
|
|
|
|
_download_and_check_if_image.cache_clear()
|
|
|
|
# force a garbage collector
|
|
|
|
gc.collect()
|