[fix] revise google images engine

this commit is picked from #1985

searx/engines/google_images.py

@@ -1,97 +1,205 @@
-"""
+# SPDX-License-Identifier: AGPL-3.0-or-later
- Google (Images)
+"""Google (Images)
 
- @website     https://www.google.com
+:website:     https://images.google.com (redirected to subdomain www.)
- @provide-api yes (https://developers.google.com/custom-search/)
+:provide-api: yes (https://developers.google.com/custom-search/)
+:using-api:   not the offical, since it needs registration to another service
+:results:     HTML
+:stable:      no
+:template:    images.html
+:parse:       url, title, content, source, thumbnail_src, img_src
+
+For detailed description of the *REST-full* API see: `Query Parameter
+Definitions`_.
+
+.. _admonition:: Content-Security-Policy (CSP)
+
+   This engine needs to allow images from the `data URLs`_ (prefixed with the
+   ``data:` scheme).::
+
+     Header set Content-Security-Policy "img-src 'self' data: ;"
+
+.. _Query Parameter Definitions:
+   https://developers.google.com/custom-search/docs/xml_results#WebSearch_Query_Parameter_Definitions
 
- @using-api   no
- @results     HTML chunks with JSON inside
- @stable      no
- @parse       url, title, img_src
 """
 
-from datetime import date, timedelta
-from json import loads
 from lxml import html
-from searx.url_utils import urlencode
+from flask_babel import gettext
+from searx import logger
+from searx.url_utils import urlencode, urlparse
+from searx.utils import eval_xpath
+from searx.engines.xpath import extract_text
+
+# pylint: disable=unused-import
+from searx.engines.google import (
+    supported_languages_url
+    ,  _fetch_supported_languages
+)
+# pylint: enable=unused-import
+
+from searx.engines.google import (
+    get_lang_country
+    , google_domains
+    , time_range_dict
+)
+
+logger = logger.getChild('google images')
 
 # engine dependent config
+
 categories = ['images']
-paging = True
+paging = False
-safesearch = True
+language_support = True
+use_locale_domain = True
 time_range_support = True
-number_of_results = 100
+safesearch = True
 
-search_url = 'https://www.google.com/search'\
+filter_mapping = {
-    '?{query}'\
+    0 : 'images',
-    '&tbm=isch'\
+    1 : 'active',
-    '&yv=2'\
+    2 : 'active'
-    '&{search_options}'
+}
-time_range_attr = "qdr:{range}"
-time_range_custom_attr = "cdr:1,cd_min:{start},cd_max{end}"
-time_range_dict = {'day': 'd',
-                   'week': 'w',
-                   'month': 'm'}
 
+def scrap_out_thumbs(dom):
+    """Scrap out thumbnail data from <script> tags.
+    """
+    ret_val = dict()
+    for script in eval_xpath(dom, '//script[contains(., "_setImgSrc(")]'):
+        _script = script.text
+        # _setImgSrc('0','data:image\/jpeg;base64,\/9j\/4AAQSkZJR ....');
+        _thumb_no, _img_data = _script[len("_setImgSrc("):-2].split(",",1)
+        _thumb_no = _thumb_no.replace("'","")
+        _img_data = _img_data.replace("'","")
+        _img_data = _img_data.replace(r"\/", r"/")
+        ret_val[_thumb_no] = _img_data.replace(r"\x3d", "=")
+    return ret_val
 
-# do search-request
 def request(query, params):
-    search_options = {
+    """Google-Video search request"""
-        'ijn': params['pageno'] - 1,
+
-        'start': (params['pageno'] - 1) * number_of_results
+    language, country, lang_country = get_lang_country(
-    }
+        # pylint: disable=undefined-variable
+        params, supported_languages, language_aliases
+    )
+    subdomain = 'www.' + google_domains.get(country.upper(), 'google.com')
+
+    query_url = 'https://'+ subdomain + '/search' + "?" + urlencode({
+        'q':   query,
+        'tbm': "isch",
+        'hl':  lang_country,
+        'lr': "lang_" + language,
+        'ie': "utf8",
+        'oe': "utf8",
+        'num': 30,
+    })
 
     if params['time_range'] in time_range_dict:
-        search_options['tbs'] = time_range_attr.format(range=time_range_dict[params['time_range']])
+        query_url += '&' + urlencode({'tbs': 'qdr:' + time_range_dict[params['time_range']]})
-    elif params['time_range'] == 'year':
+    if params['safesearch']:
-        now = date.today()
+        query_url += '&' + urlencode({'safe': filter_mapping[params['safesearch']]})
-        then = now - timedelta(days=365)
-        start = then.strftime('%m/%d/%Y')
-        end = now.strftime('%m/%d/%Y')
-        search_options['tbs'] = time_range_custom_attr.format(start=start, end=end)
 
-    if safesearch and params['safesearch']:
+    params['url'] = query_url
-        search_options['safe'] = 'on'
+    logger.debug("query_url --> %s", query_url)
-
-    params['url'] = search_url.format(query=urlencode({'q': query}),
-                                      search_options=urlencode(search_options))
 
+    params['headers']['Accept-Language'] = (
+        "%s,%s;q=0.8,%s;q=0.5" % (lang_country, language, language))
+    logger.debug(
+        "HTTP 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_url = urlparse(resp.url)
+    if resp_url.netloc == 'sorry.google.com' or resp_url.path == '/sorry/IndexRedirect':
+        raise RuntimeWarning('sorry.google.com')
+
+    if resp_url.path.startswith('/sorry'):
+        raise RuntimeWarning(gettext('CAPTCHA required'))
+
+    # which subdomain ?
+    # subdomain = resp.search_params.get('google_subdomain')
+
+    # convert the text to dom
     dom = html.fromstring(resp.text)
+    img_bas64_map = scrap_out_thumbs(dom)
 
     # parse results
-    for result in dom.xpath('//div[contains(@class, "rg_meta")]/text()'):
+    #
+    # root element::
+    #     <div id="islmp" ..>
+    # result div per image::
+    #     <div jsmodel="tTXmib"> / <div jsaction="..." data-id="..."
+    #     The data-id matches to a item in a json-data structure in::
+    #         <script nonce="I+vqelcy/01CKiBJi5Z1Ow">AF_initDataCallback({key: 'ds:1', ... data:function(){return [ ...
+    #     In this structure the ling to the origin PNG, JPG or whatever is given
+    #     (we do not blow out the link there, you could still implement that)
+    # first link per image-div contains a <img> with the data-iid for bas64 encoded image data::
+    #      <img class="rg_i Q4LuWd" data-iid="0"
+    # second link per image-div is the target link::
+    #      <a class="VFACy kGQAp" href="https://en.wikipedia.org/wiki/The_Sacrament_of_the_Last_Supper">
+    # the second link also contains two div tags with the *description* and *publisher*::
+    #      <div class="WGvvNb">The Sacrament of the Last Supper ...</div>
+    #      <div class="fxgdke">en.wikipedia.org</div>
+
+    root = eval_xpath(dom, '//div[@id="islmp"]')
+    if not root:
+        logger.error("did not find root element id='islmp'")
+        return results
+
+    root = root[0]
+    for img_node in eval_xpath(root, './/img[contains(@class, "rg_i")]'):
 
         try:
-            metadata = loads(result)
+            img_alt = eval_xpath(img_node, '@alt')[0]
 
-            img_format = metadata.get('ity', '')
+            img_base64_id = eval_xpath(img_node, '@data-iid')
-            img_width = metadata.get('ow', '')
+            if img_base64_id:
-            img_height = metadata.get('oh', '')
+                img_base64_id = img_base64_id[0]
-            if img_width and img_height:
+                thumbnail_src = img_bas64_map[img_base64_id]
-                img_format += " {0}x{1}".format(img_width, img_height)
+            else:
+                thumbnail_src = eval_xpath(img_node, '@src')
+                if not thumbnail_src:
+                    thumbnail_src = eval_xpath(img_node, '@data-src')
+                if thumbnail_src:
+                    thumbnail_src = thumbnail_src[0]
+                else:
+                    thumbnail_src = ''
 
-            source = metadata.get('st', '')
+            link_node = eval_xpath(img_node, '../../../a[2]')[0]
-            source_url = metadata.get('isu', '')
+            url = eval_xpath(link_node, '@href')[0]
-            if source_url:
-                source += " ({0})".format(source_url)
 
-            results.append({'url': metadata['ru'],
+            pub_nodes = eval_xpath(link_node, './div/div')
-                            'title': metadata['pt'],
+            pub_descr = img_alt
-                            'content': metadata.get('s', ''),
+            pub_source = ''
-                            'source': source,
+            if pub_nodes:
-                            'img_format': img_format,
+                pub_descr = extract_text(pub_nodes[0])
-                            'thumbnail_src': metadata['tu'],
+                pub_source = extract_text(pub_nodes[1])
-                            'img_src': metadata['ou'],
-                            'template': 'images.html'})
 
-        except:
+            results.append({
+                'url': url,
+                'title': img_alt,
+                'content': pub_descr,
+                'source': pub_source,
+                'img_src': url,
+                # 'img_format': img_format,
+                'thumbnail_src': thumbnail_src,
+                'template': 'images.html'
+            })
+        except Exception as e:  # pylint: disable=broad-except
+            logger.error(e, exc_info=True)
+            #from lxml import etree
+            #logger.debug(etree.tostring(img_node, pretty_print=True))
+            #import pdb
+            #pdb.set_trace()
             continue
 
     return results