Merge pull request #2483 from return42/fix-google-news

[fix] revise of the google-News engine
This commit is contained in:
Alexandre Flament 2021-01-23 20:21:09 +01:00 committed by GitHub
commit 7d24850d49
No known key found for this signature in database
GPG key ID: 4AEE18F83AFDEB23
4 changed files with 167 additions and 68 deletions

View file

@ -177,7 +177,9 @@ PYLINT_FILES=\
searx/testing.py \
searx/engines/gigablast.py \
searx/engines/deviantart.py \
searx/engines/digg.py
searx/engines/digg.py \
searx/engines/google.py \
searx/engines/google_news.py
test.pylint: pyenvinstall
$(call cmd,pylint,$(PYLINT_FILES))

View file

@ -155,6 +155,11 @@ def get_lang_country(params, lang_list, custom_aliases):
return language, country, lang_country
def detect_google_sorry(resp):
resp_url = urlparse(resp.url)
if resp_url.netloc == 'sorry.google.com' or resp_url.path.startswith('/sorry'):
raise SearxEngineCaptchaException()
def request(query, params):
"""Google search request"""
@ -200,16 +205,10 @@ def request(query, params):
def response(resp):
"""Get response from google's search request"""
detect_google_sorry(resp)
results = []
# detect google sorry
resp_url = urlparse(resp.url)
if resp_url.netloc == 'sorry.google.com' or resp_url.path == '/sorry/IndexRedirect':
raise SearxEngineCaptchaException()
if resp_url.path.startswith('/sorry'):
raise SearxEngineCaptchaException()
# which subdomain ?
# subdomain = resp.search_params.get('google_subdomain')

View file

@ -12,10 +12,9 @@ Definitions`_.
Header set Content-Security-Policy "img-src 'self' data: ;"
"""
from urllib.parse import urlencode, urlparse, unquote
from urllib.parse import urlencode, unquote
from lxml import html
from searx import logger
from searx.exceptions import SearxEngineCaptchaException
from searx.utils import extract_text, eval_xpath
from searx.engines.google import _fetch_supported_languages, supported_languages_url # NOQA # pylint: disable=unused-import
@ -23,6 +22,7 @@ from searx.engines.google import (
get_lang_country,
google_domains,
time_range_dict,
detect_google_sorry,
)
logger = logger.getChild('google images')
@ -123,13 +123,7 @@ def response(resp):
"""Get response from google's search request"""
results = []
# detect google sorry
resp_url = urlparse(resp.url)
if resp_url.netloc == 'sorry.google.com' or resp_url.path == '/sorry/IndexRedirect':
raise SearxEngineCaptchaException()
if resp_url.path.startswith('/sorry'):
raise SearxEngineCaptchaException()
detect_google_sorry(resp)
# which subdomain ?
# subdomain = resp.search_params.get('google_subdomain')

View file

@ -1,12 +1,45 @@
# SPDX-License-Identifier: AGPL-3.0-or-later
"""
Google (News)
"""Google (News)
For detailed description of the *REST-full* API see: `Query Parameter
Definitions`_. Not all parameters can be appied, e.g. num_ (the number of
search results to return) is ignored.
.. _Query Parameter Definitions:
https://developers.google.com/custom-search/docs/xml_results#WebSearch_Query_Parameter_Definitions
.. _num: https://developers.google.com/custom-search/docs/xml_results#numsp
"""
# pylint: disable=invalid-name, missing-function-docstring
import binascii
import re
from urllib.parse import urlencode
from base64 import b64decode
from lxml import html
from searx.utils import match_language
from searx.engines.google import _fetch_supported_languages, supported_languages_url # NOQA # pylint: disable=unused-import
from searx import logger
from searx.utils import (
eval_xpath,
eval_xpath_list,
eval_xpath_getindex,
extract_text,
)
# pylint: disable=unused-import
from searx.engines.google import (
supported_languages_url,
_fetch_supported_languages,
detect_google_sorry,
)
# pylint: enable=unused-import
from searx.engines.google import (
get_lang_country,
filter_mapping,
)
# about
about = {
@ -18,72 +51,143 @@ about = {
"results": 'HTML',
}
# search-url
logger = logger.getChild('google news')
# compared to other google engines google-news has a different time range
# support. The time range is included in the search term.
time_range_dict = {
'day': 'when:1d',
'week': 'when:7d',
'month': 'when:1m',
'year': 'when:1y',
}
# engine dependent config
categories = ['news']
paging = True
paging = False
language_support = True
safesearch = True
use_locale_domain = True
time_range_support = True
number_of_results = 10
safesearch = True # not really, but it is not generated by google
search_url = 'https://www.google.com/search'\
'?{query}'\
'&tbm=nws'\
'&gws_rd=cr'\
'&{search_options}'
time_range_attr = "qdr:{range}"
time_range_dict = {'day': 'd',
'week': 'w',
'month': 'm',
'year': 'y'}
# do search-request
def request(query, params):
"""Google-News search request"""
search_options = {
'start': (params['pageno'] - 1) * number_of_results
}
language, country, lang_country = get_lang_country(
# pylint: disable=undefined-variable
params, supported_languages, language_aliases
)
subdomain = 'news.google.com'
if params['time_range'] in time_range_dict:
search_options['tbs'] = time_range_attr.format(range=time_range_dict[params['time_range']])
if params['time_range']: # in time_range_dict:
query += ' ' + time_range_dict[params['time_range']]
if safesearch and params['safesearch']:
search_options['safe'] = 'on'
query_url = 'https://'+ subdomain + '/search' + "?" + urlencode({
'q': query,
'hl': lang_country,
'lr': "lang_" + language,
'ie': "utf8",
'oe': "utf8",
'ceid' : "%s:%s" % (country, language),
'gl' : country,
})
params['url'] = search_url.format(query=urlencode({'q': query}),
search_options=urlencode(search_options))
if params['safesearch']:
query_url += '&' + urlencode({'safe': filter_mapping[params['safesearch']]})
if params['language'] != 'all':
language = match_language(params['language'], supported_languages, language_aliases).split('-')[0]
if language:
params['url'] += '&hl=' + language
params['url'] = query_url
logger.debug("query_url --> %s", query_url)
# en-US,en;q=0.8,en;q=0.5
params['headers']['Accept-Language'] = (
lang_country + ',' + language + ';q=0.8,' + language + ';q=0.5'
)
logger.debug("HTTP header Accept-Language --> %s",
params['headers']['Accept-Language'])
params['headers']['Accept'] = (
'text/html,application/xhtml+xml,application/xml;q=0.9,image/webp,*/*;q=0.8'
)
#params['google_subdomain'] = subdomain
return params
# get response from search-request
def response(resp):
"""Get response from google's search request"""
results = []
detect_google_sorry(resp)
# which subdomain ?
# subdomain = resp.search_params.get('google_subdomain')
# convert the text to dom
dom = html.fromstring(resp.text)
# parse results
for result in dom.xpath('//div[@class="g"]|//div[@class="g _cy"]'):
try:
r = {
'url': result.xpath('.//a[@class="l lLrAF"]')[0].attrib.get("href"),
'title': ''.join(result.xpath('.//a[@class="l lLrAF"]//text()')),
'content': ''.join(result.xpath('.//div[@class="st"]//text()')),
}
except:
continue
for result in eval_xpath_list(dom, '//div[@class="xrnccd"]'):
imgs = result.xpath('.//img/@src')
if len(imgs) and not imgs[0].startswith('data'):
r['img_src'] = imgs[0]
# The first <a> tag in the <article> contains the link to the
# article The href attribute of the <a> is a google internal link,
# we can't use. The real link is hidden in the jslog attribute:
#
# <a ...
# jslog="95014; 4:https://www.cnn.com/.../index.html; track:click"
# href="./articles/CAIiENu3nGS...?hl=en-US&amp;gl=US&amp;ceid=US%3Aen"
# ... />
results.append(r)
jslog = eval_xpath_getindex(result, './article/a/@jslog', 0)
url = re.findall('http[^;]*', jslog)
if url:
url = url[0]
else:
# The real URL is base64 encoded in the json attribute:
# jslog="95014; 5:W251bGwsbnVsbCxudW...giXQ==; track:click"
jslog = jslog.split(";")[1].split(':')[1].strip()
try:
padding = (4 -(len(jslog) % 4)) * "="
jslog = b64decode(jslog + padding)
except binascii.Error:
# URL cant be read, skip this result
continue
# now we have : b'[null, ... null,"https://www.cnn.com/.../index.html"]'
url = re.findall('http[^;"]*', str(jslog))[0]
# the first <h3> tag in the <article> contains the title of the link
title = extract_text(eval_xpath(result, './article/h3[1]'))
# the first <div> tag in the <article> contains the content of the link
content = extract_text(eval_xpath(result, './article/div[1]'))
# the second <div> tag contains origin publisher and the publishing date
pub_date = extract_text(eval_xpath(result, './article/div[2]//time'))
pub_origin = extract_text(eval_xpath(result, './article/div[2]//a'))
pub_info = []
if pub_origin:
pub_info.append(pub_origin)
if pub_date:
# The pub_date is mostly a string like 'yesertday', not a real
# timezone date or time. Therefore we can't use publishedDate.
pub_info.append(pub_date)
pub_info = ', '.join(pub_info)
if pub_info:
content = pub_info + ': ' + content
# The image URL is located in a preceding sibling <img> tag, e.g.:
# "https://lh3.googleusercontent.com/DjhQh7DMszk.....z=-p-h100-w100"
# These URL are long but not personalized (double checked via tor).
img_src = extract_text(result.xpath('preceding-sibling::a/figure/img/@src'))
results.append({
'url': url,
'title': title,
'content': content,
'img_src': img_src,
})
# return results
return results