searxng/searx/engines/yahoo_news.py

105 lines
3.3 KiB
Python
Raw Normal View History

2014-10-19 10:18:21 +00:00
# Yahoo (News)
#
2014-09-01 14:17:29 +00:00
# @website https://news.yahoo.com
2014-10-19 10:18:21 +00:00
# @provide-api yes (https://developer.yahoo.com/boss/search/)
# $0.80/1000 queries
#
2014-09-01 14:17:29 +00:00
# @using-api no (because pricing)
# @results HTML (using search portal)
# @stable no (HTML can change)
# @parse url, title, content, publishedDate
2014-03-04 12:11:53 +00:00
from urllib import urlencode
from lxml import html
from searx.engines.xpath import extract_text, extract_url
2014-03-08 18:09:03 +00:00
from searx.engines.yahoo import parse_url
from datetime import datetime, timedelta
import re
2014-03-18 12:19:50 +00:00
from dateutil import parser
2014-03-04 12:11:53 +00:00
2014-09-01 14:17:29 +00:00
# engine dependent config
2014-03-04 12:11:53 +00:00
categories = ['news']
2014-09-01 14:17:29 +00:00
paging = True
language_support = True
# search-url
2015-05-02 19:08:56 +00:00
search_url = 'https://news.search.yahoo.com/search?{query}&b={offset}&{lang}=uh3_news_web_gs_1&pz=10&xargs=0&vl=lang_{lang}' # noqa
2014-09-01 14:17:29 +00:00
# specific xpath variables
2015-05-02 19:08:56 +00:00
results_xpath = '//ol[contains(@class,"searchCenterMiddle")]//li'
2014-03-04 12:11:53 +00:00
url_xpath = './/h3/a/@href'
title_xpath = './/h3/a'
2015-05-02 19:08:56 +00:00
content_xpath = './/div[@class="compText"]'
publishedDate_xpath = './/span[contains(@class,"tri")]'
suggestion_xpath = '//div[contains(@class,"VerALSOTRY")]//a'
2014-03-04 12:11:53 +00:00
2014-09-01 14:17:29 +00:00
# do search-request
2014-03-04 12:11:53 +00:00
def request(query, params):
offset = (params['pageno'] - 1) * 10 + 1
2014-09-01 14:17:29 +00:00
2014-03-04 12:11:53 +00:00
if params['language'] == 'all':
language = 'en'
else:
language = params['language'].split('_')[0]
2014-10-19 10:18:21 +00:00
2014-03-04 12:11:53 +00:00
params['url'] = search_url.format(offset=offset,
2014-09-01 14:17:29 +00:00
query=urlencode({'p': query}),
lang=language)
# TODO required?
2015-05-02 19:08:56 +00:00
params['cookies']['sB'] = '"v=1&vm=p&fl=1&vl=lang_{lang}&sh=1&pn=10&rw=new'\
2014-03-04 12:11:53 +00:00
.format(lang=language)
return params
2015-05-02 19:08:56 +00:00
def sanitize_url(url):
if ".yahoo.com/" in url:
2016-07-11 13:29:47 +00:00
return re.sub(u"\\;\\_ylt\\=.+$", "", url)
2015-05-02 19:08:56 +00:00
else:
return url
2014-09-01 14:17:29 +00:00
# get response from search-request
2014-03-04 12:11:53 +00:00
def response(resp):
results = []
2014-09-01 14:17:29 +00:00
2014-03-04 12:11:53 +00:00
dom = html.fromstring(resp.text)
2014-09-01 14:17:29 +00:00
# parse results
2014-03-04 12:11:53 +00:00
for result in dom.xpath(results_xpath):
2015-05-02 19:08:56 +00:00
urls = result.xpath(url_xpath)
if len(urls) != 1:
continue
url = sanitize_url(parse_url(extract_url(urls, search_url)))
2014-03-04 12:11:53 +00:00
title = extract_text(result.xpath(title_xpath)[0])
content = extract_text(result.xpath(content_xpath)[0])
2014-09-01 14:17:29 +00:00
# parse publishedDate
publishedDate = extract_text(result.xpath(publishedDate_xpath)[0])
2015-05-02 19:08:56 +00:00
# still useful ?
if re.match("^[0-9]+ minute(s|) ago$", publishedDate):
2014-03-15 19:20:41 +00:00
publishedDate = datetime.now() - timedelta(minutes=int(re.match(r'\d+', publishedDate).group())) # noqa
else:
2014-03-15 19:20:41 +00:00
if re.match("^[0-9]+ hour(s|), [0-9]+ minute(s|) ago$",
publishedDate):
timeNumbers = re.findall(r'\d+', publishedDate)
2014-03-15 19:20:41 +00:00
publishedDate = datetime.now()\
- timedelta(hours=int(timeNumbers[0]))\
- timedelta(minutes=int(timeNumbers[1]))
else:
2014-03-18 14:56:22 +00:00
publishedDate = parser.parse(publishedDate)
if publishedDate.year == 1900:
publishedDate = publishedDate.replace(year=datetime.now().year)
2014-09-01 14:17:29 +00:00
# append result
2014-03-15 19:20:41 +00:00
results.append({'url': url,
'title': title,
'content': content,
'publishedDate': publishedDate})
2014-03-04 12:11:53 +00:00
2014-09-01 14:17:29 +00:00
# return results
2014-03-04 12:11:53 +00:00
return results