Merge pull request #1061 from a01200356/bing

[fix] Language support for Bing Images and Videos

.travis.yml

@@ -13,7 +13,7 @@ python:
 before_install:
   - "export DISPLAY=:99.0"
   - "sh -e /etc/init.d/xvfb start"
-  - npm install less less-plugin-clean-css grunt-cli
+  - npm install less@2.7 less-plugin-clean-css grunt-cli
   - export PATH=`pwd`/node_modules/.bin:$PATH
   - ./manage.sh install_geckodriver ~/drivers
   - export PATH=~/drivers:$PATH

searx/engines/bing_images.py

@@ -18,7 +18,6 @@
 from lxml import html
 from json import loads
 import re
-from searx.engines.bing import _fetch_supported_languages, supported_languages_url
 from searx.url_utils import urlencode
 
 # engine dependent config
@@ -26,6 +25,8 @@ categories = ['images']
 paging = True
 safesearch = True
 time_range_support = True
+language_support = True
+supported_languages_url = 'https://www.bing.com/account/general'
 
 # search-url
 base_url = 'https://www.bing.com/'
@@ -45,23 +46,41 @@ safesearch_types = {2: 'STRICT',
 _quote_keys_regex = re.compile('({|,)([a-z][a-z0-9]*):(")', re.I | re.U)
 
 
+# get supported region code
+def get_region_code(lang, lang_list=None):
+    region = None
+    if lang in (lang_list or supported_languages):
+        region = lang
+    elif lang.startswith('no'):
+        region = 'nb-NO'
+    else:
+        # try to get a supported country code with language
+        lang = lang.split('-')[0]
+        for lc in (lang_list or supported_languages):
+            if lang == lc.split('-')[0]:
+                region = lc
+                break
+    if region:
+        return region.lower()
+    else:
+        return 'en-us'
+
+
 # do search-request
 def request(query, params):
     offset = (params['pageno'] - 1) * 10 + 1
 
-    # required for cookie
-    if params['language'] == 'all':
-        language = 'en-US'
-    else:
-        language = params['language']
-
     search_path = search_string.format(
         query=urlencode({'q': query}),
         offset=offset)
 
+    language = get_region_code(params['language'])
+
     params['cookies']['SRCHHPGUSR'] = \
-        'NEWWND=0&NRSLT=-1&SRCHLANG=' + language.split('-')[0] +\
+        'ADLT=' + safesearch_types.get(params['safesearch'], 'DEMOTE')
-        '&ADLT=' + safesearch_types.get(params['safesearch'], 'DEMOTE')
+
+    params['cookies']['_EDGE_S'] = 'mkt=' + language +\
+        '&ui=' + language + '&F=1'
 
     params['url'] = base_url + search_path
     if params['time_range'] in time_range_dict:
@@ -106,3 +125,22 @@ def response(resp):
 
     # return results
     return results
+
+
+# get supported languages from their site
+def _fetch_supported_languages(resp):
+    supported_languages = []
+    dom = html.fromstring(resp.text)
+
+    regions_xpath = '//div[@id="region-section-content"]' \
+                    + '//ul[@class="b_vList"]/li/a/@href'
+
+    regions = dom.xpath(regions_xpath)
+    for region in regions:
+        code = re.search('setmkt=[^\&]+', region).group()[7:]
+        if code == 'nb-NO':
+            code = 'no-NO'
+
+        supported_languages.append(code)
+
+    return supported_languages

searx/engines/bing_videos.py

@@ -12,6 +12,7 @@
 
 from json import loads
 from lxml import html
+from searx.engines.bing_images import _fetch_supported_languages, supported_languages_url, get_region_code
 from searx.engines.xpath import extract_text
 from searx.url_utils import urlencode
 
@@ -21,6 +22,7 @@ paging = True
 safesearch = True
 time_range_support = True
 number_of_results = 10
+language_support = True
 
 search_url = 'https://www.bing.com/videos/asyncv2?{query}&async=content&'\
              'first={offset}&count={number_of_results}&CW=1366&CH=25&FORM=R5VR5'
@@ -45,7 +47,8 @@ def request(query, params):
         'ADLT=' + safesearch_types.get(params['safesearch'], 'DEMOTE')
 
     # language cookie
-    params['cookies']['_EDGE_S'] = 'mkt=' + params['language'].lower() + '&F=1'
+    region = get_region_code(params['language'], lang_list=supported_languages)
+    params['cookies']['_EDGE_S'] = 'mkt=' + region + '&F=1'
 
     # query and paging
     params['url'] = search_url.format(query=urlencode({'q': query}),

searx/engines/duckduckgo.py

@@ -134,4 +134,4 @@ def _fetch_supported_languages(resp):
     regions_json = loads(response_page)
     supported_languages = map((lambda x: x[3:] + '-' + x[:2].upper()), regions_json.keys())
 
-    return supported_languages
+    return list(supported_languages)

searx/engines/swisscows.py

@@ -118,7 +118,7 @@ def _fetch_supported_languages(resp):
     dom = fromstring(resp.text)
     options = dom.xpath('//div[@id="regions-popup"]//ul/li/a')
     for option in options:
-        code = option.xpath('./@data-val')[0]
+        code = option.xpath('./@data-search-language')[0]
         if code.startswith('nb-'):
             code = code.replace('nb', 'no', 1)
         supported_languages.append(code)

searx/languages.py

@@ -5,6 +5,11 @@
 language_codes = (
     (u"ar-SA", u"العربية", u"", u"Arabic"),
     (u"bg-BG", u"Български", u"", u"Bulgarian"),
+    (u"ca", u"Català", u"", u"Catalan"),
+    (u"ca-AD", u"Català", u"Andorra", u"Catalan"),
+    (u"ca-CT", u"Català", u"", u"Catalan"),
+    (u"ca-ES", u"Català", u"Espanya", u"Catalan"),
+    (u"ca-FR", u"Català", u"França", u"Catalan"),
     (u"cs-CZ", u"Čeština", u"", u"Czech"),
     (u"da-DK", u"Dansk", u"", u"Danish"),
     (u"de", u"Deutsch", u"", u"German"),
@@ -15,9 +20,7 @@ language_codes = (
     (u"en", u"English", u"", u"English"),
     (u"en-AU", u"English", u"Australia", u"English"),
     (u"en-CA", u"English", u"Canada", u"English"),
-    (u"en-CY", u"English", u"Cyprus", u"English"),
     (u"en-GB", u"English", u"United Kingdom", u"English"),
-    (u"en-GD", u"English", u"Grenada", u"English"),
     (u"en-ID", u"English", u"Indonesia", u"English"),
     (u"en-IE", u"English", u"Ireland", u"English"),
     (u"en-IN", u"English", u"India", u"English"),
@@ -28,6 +31,7 @@ language_codes = (
     (u"en-US", u"English", u"United States", u"English"),
     (u"en-ZA", u"English", u"South Africa", u"English"),
     (u"es", u"Español", u"", u"Spanish"),
+    (u"es-AD", u"Español", u"Andorra", u"Spanish"),
     (u"es-AR", u"Español", u"Argentina", u"Spanish"),
     (u"es-CL", u"Español", u"Chile", u"Spanish"),
     (u"es-CO", u"Español", u"Colombia", u"Spanish"),
@@ -38,38 +42,32 @@ language_codes = (
     (u"et-EE", u"Eesti", u"", u"Estonian"),
     (u"fi-FI", u"Suomi", u"", u"Finnish"),
     (u"fr", u"Français", u"", u"French"),
+    (u"fr-AD", u"Français", u"Andorre", u"French"),
     (u"fr-BE", u"Français", u"Belgique", u"French"),
     (u"fr-CA", u"Français", u"Canada", u"French"),
     (u"fr-CH", u"Français", u"Suisse", u"French"),
     (u"fr-FR", u"Français", u"France", u"French"),
     (u"he-IL", u"עברית", u"", u"Hebrew"),
-    (u"hr-HR", u"Hrvatski", u"", u"Croatian"),
     (u"hu-HU", u"Magyar", u"", u"Hungarian"),
-    (u"id-ID", u"Bahasa Indonesia", u"", u"Indonesian"),
     (u"it", u"Italiano", u"", u"Italian"),
     (u"it-CH", u"Italiano", u"Svizzera", u"Italian"),
     (u"it-IT", u"Italiano", u"Italia", u"Italian"),
     (u"ja-JP", u"日本語", u"", u"Japanese"),
     (u"ko-KR", u"한국어", u"", u"Korean"),
-    (u"lt-LT", u"Lietuvių", u"", u"Lithuanian"),
-    (u"lv-LV", u"Latviešu", u"", u"Latvian"),
-    (u"ms-MY", u"Bahasa Melayu", u"", u"Malay"),
     (u"nl", u"Nederlands", u"", u"Dutch"),
     (u"nl-BE", u"Nederlands", u"België", u"Dutch"),
     (u"nl-NL", u"Nederlands", u"Nederland", u"Dutch"),
     (u"no-NO", u"Norsk", u"", u"Norwegian"),
     (u"pl-PL", u"Polski", u"", u"Polish"),
     (u"pt", u"Português", u"", u"Portuguese"),
+    (u"pt-AD", u"Português", u"Andorra", u"Portuguese"),
     (u"pt-BR", u"Português", u"Brasil", u"Portuguese"),
     (u"pt-PT", u"Português", u"Portugal", u"Portuguese"),
     (u"ro-RO", u"Română", u"", u"Romanian"),
     (u"ru-RU", u"Русский", u"", u"Russian"),
-    (u"sk-SK", u"Slovenčina", u"", u"Slovak"),
-    (u"sl", u"Slovenščina", u"", u"Slovenian"),
     (u"sv-SE", u"Svenska", u"", u"Swedish"),
     (u"th-TH", u"ไทย", u"", u"Thai"),
     (u"tr-TR", u"Türkçe", u"", u"Turkish"),
-    (u"vi-VN", u"Tiếng Việt", u"", u"Vietnamese"),
     (u"zh", u"中文", u"", u"Chinese"),
     (u"zh-CN", u"中文", u"中国", u"Chinese"),
     (u"zh-HK", u"中文", u"香港", u"Chinese"),

tests/unit/engines/test_bing_images.py

@@ -8,10 +8,12 @@ from searx.testing import SearxTestCase
 class TestBingImagesEngine(SearxTestCase):
 
     def test_request(self):
+        bing_images.supported_languages = ['fr-FR', 'en-US']
+
         query = 'test_query'
         dicto = defaultdict(dict)
         dicto['pageno'] = 1
-        dicto['language'] = 'fr_FR'
+        dicto['language'] = 'fr-FR'
         dicto['safesearch'] = 1
         dicto['time_range'] = ''
         params = bing_images.request(query, dicto)
@@ -19,12 +21,19 @@ class TestBingImagesEngine(SearxTestCase):
         self.assertTrue(query in params['url'])
         self.assertTrue('bing.com' in params['url'])
         self.assertTrue('SRCHHPGUSR' in params['cookies'])
-        self.assertTrue('fr' in params['cookies']['SRCHHPGUSR'])
+        self.assertTrue('DEMOTE' in params['cookies']['SRCHHPGUSR'])
+        self.assertTrue('_EDGE_S' in params['cookies'])
+        self.assertTrue('fr-fr' in params['cookies']['_EDGE_S'])
+
+        dicto['language'] = 'fr'
+        params = bing_images.request(query, dicto)
+        self.assertTrue('_EDGE_S' in params['cookies'])
+        self.assertTrue('fr-fr' in params['cookies']['_EDGE_S'])
 
         dicto['language'] = 'all'
         params = bing_images.request(query, dicto)
-        self.assertIn('SRCHHPGUSR', params['cookies'])
+        self.assertTrue('_EDGE_S' in params['cookies'])
-        self.assertIn('en', params['cookies']['SRCHHPGUSR'])
+        self.assertTrue('en-us' in params['cookies']['_EDGE_S'])
 
     def test_response(self):
         self.assertRaises(AttributeError, bing_images.response, None)
@@ -82,3 +91,28 @@ class TestBingImagesEngine(SearxTestCase):
         self.assertEqual(results[0]['content'], '')
         self.assertEqual(results[0]['thumbnail_src'], 'thumb_url')
         self.assertEqual(results[0]['img_src'], 'img_url')
+
+    def test_fetch_supported_languages(self):
+        html = """
+        <div>
+            <div id="region-section-content">
+                <ul class="b_vList">
+                    <li>
+                        <a href="https://bing...&setmkt=de-DE&s...">Germany</a>
+                        <a href="https://bing...&setmkt=nb-NO&s...">Norway</a>
+                    </li>
+                </ul>
+                <ul class="b_vList">
+                    <li>
+                        <a href="https://bing...&setmkt=es-AR&s...">Argentina</a>
+                    </li>
+                </ul>
+            </div>
+        </div>
+        """
+        response = mock.Mock(text=html)
+        languages = list(bing_images._fetch_supported_languages(response))
+        self.assertEqual(len(languages), 3)
+        self.assertIn('de-DE', languages)
+        self.assertIn('no-NO', languages)
+        self.assertIn('es-AR', languages)

tests/unit/engines/test_bing_videos.py

@@ -8,6 +8,8 @@ from searx.testing import SearxTestCase
 class TestBingVideosEngine(SearxTestCase):
 
     def test_request(self):
+        bing_videos.supported_languages = ['fr-FR', 'en-US']
+
         query = 'test_query'
         dicto = defaultdict(dict)
         dicto['pageno'] = 1

tests/unit/engines/test_swisscows.py

@@ -139,9 +139,9 @@ class TestSwisscowsEngine(SearxTestCase):
             <div id="regions-popup">
                 <div>
                     <ul>
-                        <li><a data-val="browser"></a></li>
+                        <li><a data-search-language="browser"></a></li>
-                        <li><a data-val="de-CH"></a></li>
+                        <li><a data-search-language="de-CH"></a></li>
-                        <li><a data-val="fr-CH"></a></li>
+                        <li><a data-search-language="fr-CH"></a></li>
                     </ul>
                 </div>
             </div>

utils/fetch_languages.py

@@ -8,13 +8,13 @@
 # are written in current directory to avoid overwriting in case something goes wrong.
 
 from requests import get
-from urllib import urlencode
 from lxml.html import fromstring
-from json import loads, dumps
+from json import loads, dump
 import io
 from sys import path
 path.append('../searx')  # noqa
 from searx import settings
+from searx.url_utils import urlencode
 from searx.engines import initialize_engines, engines
 
 # Geonames API for country names.
@@ -70,7 +70,7 @@ def get_country_name(locale):
     json = loads(response.text)
     content = json.get('geonames', None)
     if content is None or len(content) != 1:
-        print "No country name found for " + locale[0] + "-" + locale[1]
+        print("No country name found for " + locale[0] + "-" + locale[1])
         return ''
 
     return content[0].get('countryName', '')
@@ -84,11 +84,11 @@ def fetch_supported_languages():
             try:
                 engines_languages[engine_name] = engines[engine_name].fetch_supported_languages()
             except Exception as e:
-                print e
+                print(e)
 
     # write json file
     with io.open(engines_languages_file, "w", encoding="utf-8") as f:
-        f.write(unicode(dumps(engines_languages, ensure_ascii=False, encoding="utf-8")))
+        dump(engines_languages, f, ensure_ascii=False)
 
 
 # Join all language lists.
@@ -97,7 +97,7 @@ def join_language_lists():
     global languages
     # include wikipedia first for more accurate language names
     languages = {code: lang for code, lang
-                 in engines_languages['wikipedia'].iteritems()
+                 in engines_languages['wikipedia'].items()
                  if valid_code(code)}
 
     for engine_name in engines_languages:
@@ -121,7 +121,7 @@ def join_language_lists():
     # filter list to include only languages supported by most engines
     min_supported_engines = int(0.70 * len(engines_languages))
     languages = {code: lang for code, lang
-                 in languages.iteritems()
+                 in languages.items()
                  if len(lang.get('counter', [])) >= min_supported_engines or
                  len(languages.get(code.split('-')[0], {}).get('counter', [])) >= min_supported_engines}
 
@@ -165,7 +165,7 @@ def filter_single_country_languages():
 
 # Write languages.py.
 def write_languages_file():
-    new_file = open(languages_file, 'w')
+    new_file = open(languages_file, 'wb')
     file_content = '# -*- coding: utf-8 -*-\n'\
                    + '# list of language codes\n'\
                    + '# this file is generated automatically by utils/update_search_languages.py\n'\