[fix] yahoo_news engine

This commit is contained in:
Alexandre Flament 2015-05-02 21:08:56 +02:00
parent bbd83f5a51
commit 44ed4424f6
2 changed files with 89 additions and 89 deletions

View File

@ -23,15 +23,15 @@ paging = True
language_support = True language_support = True
# search-url # search-url
search_url = 'https://news.search.yahoo.com/search?{query}&b={offset}&fl=1&vl=lang_{lang}' # noqa 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
# specific xpath variables # specific xpath variables
results_xpath = '//div[@class="res"]' results_xpath = '//ol[contains(@class,"searchCenterMiddle")]//li'
url_xpath = './/h3/a/@href' url_xpath = './/h3/a/@href'
title_xpath = './/h3/a' title_xpath = './/h3/a'
content_xpath = './/div[@class="abstr"]' content_xpath = './/div[@class="compText"]'
publishedDate_xpath = './/span[@class="timestamp"]' publishedDate_xpath = './/span[contains(@class,"tri")]'
suggestion_xpath = '//div[@id="satat"]//a' suggestion_xpath = '//div[contains(@class,"VerALSOTRY")]//a'
# do search-request # do search-request
@ -48,11 +48,18 @@ def request(query, params):
lang=language) lang=language)
# TODO required? # TODO required?
params['cookies']['sB'] = 'fl=1&vl=lang_{lang}&sh=1&rw=new&v=1'\ params['cookies']['sB'] = '"v=1&vm=p&fl=1&vl=lang_{lang}&sh=1&pn=10&rw=new'\
.format(lang=language) .format(lang=language)
return params return params
def sanitize_url(url):
if ".yahoo.com/" in url:
return re.sub(u"\;\_ylt\=.+$", "", url)
else:
return url
# get response from search-request # get response from search-request
def response(resp): def response(resp):
results = [] results = []
@ -61,13 +68,17 @@ def response(resp):
# parse results # parse results
for result in dom.xpath(results_xpath): for result in dom.xpath(results_xpath):
url = parse_url(extract_url(result.xpath(url_xpath), search_url)) urls = result.xpath(url_xpath)
if len(urls) != 1:
continue
url = sanitize_url(parse_url(extract_url(urls, search_url)))
title = extract_text(result.xpath(title_xpath)[0]) title = extract_text(result.xpath(title_xpath)[0])
content = extract_text(result.xpath(content_xpath)[0]) content = extract_text(result.xpath(content_xpath)[0])
# parse publishedDate # parse publishedDate
publishedDate = extract_text(result.xpath(publishedDate_xpath)[0]) publishedDate = extract_text(result.xpath(publishedDate_xpath)[0])
# still useful ?
if re.match("^[0-9]+ minute(s|) ago$", publishedDate): if re.match("^[0-9]+ minute(s|) ago$", publishedDate):
publishedDate = datetime.now() - timedelta(minutes=int(re.match(r'\d+', publishedDate).group())) # noqa publishedDate = datetime.now() - timedelta(minutes=int(re.match(r'\d+', publishedDate).group())) # noqa
else: else:

View File

@ -39,19 +39,24 @@ class TestYahooNewsEngine(SearxTestCase):
self.assertEqual(yahoo_news.response(response), []) self.assertEqual(yahoo_news.response(response), [])
html = """ html = """
<div class="res"> <ol class=" reg searchCenterMiddle">
<div> <li class="first">
<h3> <div class="compTitle">
<a class="yschttl spt" href="http://this.is.the.url" target="_blank"> <h3>
This is <a class="yschttl spt" href="http://this.is.the.url" target="_blank">
the <b>title</b>... This is
</a> the <b>title</b>...
</h3> </a>
</div> </h3>
<span class="url">Business via Yahoo! Finance</span> &nbsp; <span class="timestamp">Feb 03 09:45am</span> </div>
<div class="abstr"> <div>
This is the content <span class="cite">Business via Yahoo!</span>
</div> <span class="tri fc-2nd ml-10">May 01 10:00 AM</span>
</div>
<div class="compText">
This is the content
</div>
</li>
</div> </div>
""" """
response = mock.Mock(text=html) response = mock.Mock(text=html)
@ -63,48 +68,59 @@ class TestYahooNewsEngine(SearxTestCase):
self.assertEqual(results[0]['content'], 'This is the content') self.assertEqual(results[0]['content'], 'This is the content')
html = """ html = """
<div class="res"> <ol class=" reg searchCenterMiddle">
<div> <li class="first">
<h3> <div class="compTitle">
<a class="yschttl spt" href="http://this.is.the.url" target="_blank"> <h3>
This is <a class="yschttl spt" href="http://this.is.the.url" target="_blank">
the <b>title</b>... This is
</a> the <b>title</b>...
</h3> </a>
</div> </h3>
<span class="url">Business via Yahoo!</span> &nbsp; <span class="timestamp">2 hours, 22 minutes ago</span> </div>
<div class="abstr"> <div>
This is the content <span class="cite">Business via Yahoo!</span>
</div> <span class="tri fc-2nd ml-10">2 hours, 22 minutes ago</span>
</div> </div>
<div class="res"> <div class="compText">
<div> This is the content
<h3> </div>
<a class="yschttl spt" href="http://this.is.the.url" target="_blank"> </li>
This is <li>
the <b>title</b>... <div class="compTitle">
</a> <h3>
</h3> <a class="yschttl spt" href="http://this.is.the.url" target="_blank">
</div> This is
<span class="url">Business via Yahoo!</span> &nbsp; <span class="timestamp">22 minutes ago</span> the <b>title</b>...
<div class="abstr"> </a>
This is the content </h3>
</div> </div>
</div> <div>
<div class="res"> <span class="cite">Business via Yahoo!</span>
<div> <span class="tri fc-2nd ml-10">22 minutes ago</span>
<h3> </div>
<a class="yschttl spt" href="http://this.is.the.url" target="_blank"> <div class="compText">
This is This is the content
the <b>title</b>... </div>
</a> </li>
</h3> <li>
</div> <div class="compTitle">
<span class="url">Business via Yahoo!</span> &nbsp; <span class="timestamp">Feb 03 09:45am 1900</span> <h3>
<div class="abstr"> <a class="yschttl spt" href="http://this.is.the.url" target="_blank">
This is the content This is
</div> the <b>title</b>...
</div> </a>
</h3>
</div>
<div>
<span class="cite">Business via Yahoo!</span>
<span class="tri fc-2nd ml-10">Feb 03 09:45AM 1900</span>
</div>
<div class="compText">
This is the content
</div>
</li>
</ol>
""" """
response = mock.Mock(text=html) response = mock.Mock(text=html)
results = yahoo_news.response(response) results = yahoo_news.response(response)
@ -114,30 +130,3 @@ class TestYahooNewsEngine(SearxTestCase):
self.assertEqual(results[0]['url'], 'http://this.is.the.url/') self.assertEqual(results[0]['url'], 'http://this.is.the.url/')
self.assertEqual(results[0]['content'], 'This is the content') self.assertEqual(results[0]['content'], 'This is the content')
self.assertEqual(results[2]['publishedDate'].year, datetime.now().year) self.assertEqual(results[2]['publishedDate'].year, datetime.now().year)
html = """
<li class="b_algo" u="0|5109|4755453613245655|UAGjXgIrPH5yh-o5oNHRx_3Zta87f_QO">
<div Class="sa_mc">
<div class="sb_tlst">
<h2>
<a href="http://this.should.be.the.link/" h="ID=SERP,5124.1">
<strong>This</strong> should be the title</a>
</h2>
</div>
<div class="sb_meta">
<cite>
<strong>this</strong>.meta.com</cite>
<span class="c_tlbxTrg">
<span class="c_tlbxH" H="BASE:CACHEDPAGEDEFAULT" K="SERP,5125.1">
</span>
</span>
</div>
<p>
<strong>This</strong> should be the content.</p>
</div>
</li>
"""
response = mock.Mock(text=html)
results = yahoo_news.response(response)
self.assertEqual(type(results), list)
self.assertEqual(len(results), 0)