on master help sign in
Merge pull request #613 from mmuman/ina

Add INA search engine

commit678b87f9d5969668a07c36f9b80be5a5b2d08587 authorAdam Tauber <asciimoo@gmail.com> dateSat 9 Jul 2016, 22:16 (10 years ago) parentfcc3264b538de5
3 files changed+153
searx/engines/ina.py+83 searx/settings.yml+6 tests/unit/engines/test_ina.py+64
searx/engines/ina.pynew file+83
@@ -0,0 +1,83 @@
1+# INA (Videos)
2+#
3+# @website https://www.ina.fr/
4+# @provide-api no
5+#
6+# @using-api no
7+# @results HTML (using search portal)
8+# @stable no (HTML can change)
9+# @parse url, title, content, publishedDate, thumbnail
10+#
11+# @todo set content-parameter with correct data
12+# @todo embedded (needs some md5 from video page)
13+
14+from json import loads
15+from urllib import urlencode
16+from lxml import html
17+from HTMLParser import HTMLParser
18+from searx.engines.xpath import extract_text
19+from dateutil import parser
20+
21+# engine dependent config
22+categories = ['videos']
23+paging = True
24+page_size = 48
25+
26+# search-url
27+base_url = 'https://www.ina.fr'
28+search_url = base_url + '/layout/set/ajax/recherche/result?autopromote=&hf={ps}&b={start}&type=Video&r=&{query}'
29+
30+# specific xpath variables
31+results_xpath = '//div[contains(@class,"search-results--list")]/div[@class="media"]'
32+url_xpath = './/a/@href'
33+title_xpath = './/h3[@class="h3--title media-heading"]'
34+thumbnail_xpath = './/img/@src'
35+publishedDate_xpath = './/span[@class="broadcast"]'
36+content_xpath = './/p[@class="media-body__summary"]'
37+
38+
39+# do search-request
40+def request(query, params):
41+ params['url'] = search_url.format(ps=page_size,
42+ start=params['pageno'] * page_size,
43+ query=urlencode({'q': query}))
44+
45+ return params
46+
47+
48+# get response from search-request
49+def response(resp):
50+ results = []
51+
52+ # we get html in a JSON container...
53+ response = loads(resp.text)
54+ if "content" not in response:
55+ return []
56+ dom = html.fromstring(response["content"])
57+ p = HTMLParser()
58+
59+ # parse results
60+ for result in dom.xpath(results_xpath):
61+ videoid = result.xpath(url_xpath)[0]
62+ url = base_url + videoid
63+ title = p.unescape(extract_text(result.xpath(title_xpath)))
64+ thumbnail = extract_text(result.xpath(thumbnail_xpath)[0])
65+ if thumbnail[0] == '/':
66+ thumbnail = base_url + thumbnail
67+ d = extract_text(result.xpath(publishedDate_xpath)[0])
68+ d = d.split('/')
69+ # force ISO date to avoid wrong parsing
70+ d = "%s-%s-%s" % (d[2], d[1], d[0])
71+ publishedDate = parser.parse(d)
72+ content = extract_text(result.xpath(content_xpath))
73+
74+ # append result
75+ results.append({'url': url,
76+ 'title': title,
77+ 'content': content,
78+ 'template': 'videos.html',
79+ 'publishedDate': publishedDate,
80+ 'thumbnail': thumbnail})
81+
82+ # return results
83+ return results
searx/settings.yml+6
@@ -254,6 +254,12 @@ engines:
254254 disabled : True
255255 shortcut : habr
256256
257+ - name : ina
258+ engine : ina
259+ shortcut : in
260+ timeout : 6.0
261+ disabled : True
262+
257263 - name : mixcloud
258264 engine : mixcloud
259265 shortcut : mc
tests/unit/engines/test_ina.pynew file+64
@@ -0,0 +1,64 @@
1+from collections import defaultdict
2+import mock
3+from searx.engines import ina
4+from searx.testing import SearxTestCase
5+
6+
7+class TestInaEngine(SearxTestCase):
8+
9+ def test_request(self):
10+ query = 'test_query'
11+ dicto = defaultdict(dict)
12+ dicto['pageno'] = 0
13+ params = ina.request(query, dicto)
14+ self.assertTrue('url' in params)
15+ self.assertTrue(query in params['url'])
16+ self.assertTrue('ina.fr' in params['url'])
17+
18+ def test_response(self):
19+ self.assertRaises(AttributeError, ina.response, None)
20+ self.assertRaises(AttributeError, ina.response, [])
21+ self.assertRaises(AttributeError, ina.response, '')
22+ self.assertRaises(AttributeError, ina.response, '[]')
23+
24+ response = mock.Mock(text='{}')
25+ self.assertEqual(ina.response(response), [])
26+
27+ response = mock.Mock(text='{"data": []}')
28+ self.assertEqual(ina.response(response), [])
29+
30+ json = """
31+ {"content":"\\t<div class=\\"container\\">\\n\\t\\n\
32+ <!-- DEBUT CONTENU PRINCIPAL -->\\n<div class=\\"row\\">\\n\
33+ <div class=\\"search-results--list\\"><div class=\\"media\\">\\n\
34+ \\t\\t\\t\\t<a class=\\"media-left media-video premium xiti_click_action\\" \
35+ data-xiti-params=\\"recherche_v4::resultats_conference_de_presse_du_general_de_gaulle::N\\" \
36+ href=\\"\/video\/CAF89035682\/conference-de-presse-du-general-de-gaulle-video.html\\">\\n\
37+ <img src=\\"https:\/\/www.ina.fr\/images_v2\/140x105\/CAF89035682.jpeg\\" \
38+ alt=\\"Conf\\u00e9rence de presse du G\\u00e9n\\u00e9ral de Gaulle \\">\\n\
39+ \\t\\t\\t\\t\\t<\/a>\\n\
40+ \\t\\t\\t\\t\\t<div class=\\"media-body\\">\\n\\t\\t\\t\\t\\t\\t<h3 class=\\"h3--title media-heading\\">\\n\
41+ \\t\\t\\t\\t\\t\\t\\t<a class=\\"xiti_click_action\\" \
42+ data-xiti-params=\\"recherche_v4::resultats_conference_de_presse_du_general_de_gaulle::N\\" \
43+ href=\\"\/video\/CAF89035682\/conference-de-presse-du-general-de-gaulle-video.html\\">\
44+ Conf\\u00e9rence de presse du G\\u00e9n\\u00e9ral de Gaulle <\/a>\\n\
45+ <\/h3>\\n\
46+ <div class=\\"media-body__info\\">\\n<span class=\\"broadcast\\">27\/11\/1967<\/span>\\n\
47+ <span class=\\"views\\">29321 vues<\/span>\\n\
48+ <span class=\\"duration\\">01h 33m 07s<\/span>\\n\
49+ <\/div>\\n\
50+ <p class=\\"media-body__summary\\">VERSION INTEGRALE DE LA CONFERENCE DE PRESSE DU GENERAL DE GAULLE . \
51+ - PA le Pr\\u00e9sident DE GAULLE : il ouvre les bras et s'assied. DP journalis...<\/p>\\n\
52+ <\/div>\\n<\/div><!-- \/.media -->\\n"
53+ }
54+ """
55+ response = mock.Mock(text=json)
56+ results = ina.response(response)
57+ self.assertEqual(type(results), list)
58+ self.assertEqual(len(results), 1)
59+ self.assertEqual(results[0]['title'], u'Conf\xe9rence de presse du G\xe9n\xe9ral de Gaulle')
60+ self.assertEqual(results[0]['url'],
61+ 'https://www.ina.fr/video/CAF89035682/conference-de-presse-du-general-de-gaulle-video.html')
62+ self.assertEqual(results[0]['content'],
63+ u"VERSION INTEGRALE DE LA CONFERENCE DE PRESSE DU GENERAL DE GAULLE ."
64+ u" - PA le Pr\u00e9sident DE GAULLE : il ouvre les bras et s'assied. DP journalis...")