searxng/searx/engines/google.py

368 lines
12 KiB
Python
Raw Normal View History

2014-12-05 19:03:16 +00:00
# Google (Web)
#
2014-09-01 13:10:05 +00:00
# @website https://www.google.com
# @provide-api yes (https://developers.google.com/custom-search/)
2014-12-05 19:03:16 +00:00
#
# @using-api no
# @results HTML
# @stable no (HTML can change)
# @parse url, title, content, suggestion
2014-01-29 18:28:38 +00:00
import re
from cgi import escape
2014-01-29 18:28:38 +00:00
from urllib import urlencode
2014-12-05 19:03:16 +00:00
from urlparse import urlparse, parse_qsl
from lxml import html
from searx.poolrequests import get
from searx.engines.xpath import extract_text, extract_url
2014-01-29 18:28:38 +00:00
2014-09-01 13:10:05 +00:00
# engine dependent config
2014-01-29 18:28:38 +00:00
categories = ['general']
2014-09-01 13:10:05 +00:00
paging = True
language_support = True
use_locale_domain = True
# based on https://en.wikipedia.org/wiki/List_of_Google_domains and tests
default_hostname = 'www.google.com'
country_to_hostname = {
'BG': 'www.google.bg', # Bulgaria
'CZ': 'www.google.cz', # Czech Republic
'DE': 'www.google.de', # Germany
'DK': 'www.google.dk', # Denmark
'AT': 'www.google.at', # Austria
'CH': 'www.google.ch', # Switzerland
'GR': 'www.google.gr', # Greece
'AU': 'www.google.com.au', # Australia
'CA': 'www.google.ca', # Canada
'GB': 'www.google.co.uk', # United Kingdom
'ID': 'www.google.co.id', # Indonesia
'IE': 'www.google.ie', # Ireland
'IN': 'www.google.co.in', # India
'MY': 'www.google.com.my', # Malaysia
'NZ': 'www.google.co.nz', # New Zealand
'PH': 'www.google.com.ph', # Philippines
'SG': 'www.google.com.sg', # Singapore
# 'US': 'www.google.us', # United State, redirect to .com
'ZA': 'www.google.co.za', # South Africa
'AR': 'www.google.com.ar', # Argentina
'CL': 'www.google.cl', # Chile
'ES': 'www.google.es', # Span
'MX': 'www.google.com.mx', # Mexico
'EE': 'www.google.ee', # Estonia
'FI': 'www.google.fi', # Finland
'BE': 'www.google.be', # Belgium
'FR': 'www.google.fr', # France
'IL': 'www.google.co.il', # Israel
'HR': 'www.google.hr', # Croatia
'HU': 'www.google.hu', # Hungary
'IT': 'www.google.it', # Italy
'JP': 'www.google.co.jp', # Japan
'KR': 'www.google.co.kr', # South Korean
'LT': 'www.google.lt', # Lithuania
'LV': 'www.google.lv', # Latvia
'NO': 'www.google.no', # Norway
'NL': 'www.google.nl', # Netherlands
'PL': 'www.google.pl', # Poland
'BR': 'www.google.com.br', # Brazil
'PT': 'www.google.pt', # Portugal
'RO': 'www.google.ro', # Romania
'RU': 'www.google.ru', # Russia
'SK': 'www.google.sk', # Slovakia
'SL': 'www.google.si', # Slovenia (SL -> si)
'SE': 'www.google.se', # Sweden
'TH': 'www.google.co.th', # Thailand
'TR': 'www.google.com.tr', # Turkey
'UA': 'www.google.com.ua', # Ikraine
# 'CN': 'www.google.cn', # China, only from china ?
'HK': 'www.google.com.hk', # Hong kong
'TW': 'www.google.com.tw' # Taiwan
}
2014-01-29 18:28:38 +00:00
# osm
url_map = 'https://www.openstreetmap.org/'\
+ '?lat={latitude}&lon={longitude}&zoom={zoom}&layers=M'
2014-09-01 13:10:05 +00:00
# search-url
search_path = '/search'
search_url = ('https://{hostname}' +
2014-12-05 19:03:16 +00:00
search_path +
'?{query}&start={offset}&gbv=1')
2014-01-31 03:35:23 +00:00
# other URLs
map_hostname_start = 'maps.google.'
maps_path = '/maps'
redirect_path = '/url'
images_path = '/images'
# specific xpath variables
2014-12-05 19:03:16 +00:00
results_xpath = '//li[@class="g"]'
url_xpath = './/h3/a/@href'
title_xpath = './/h3'
content_xpath = './/span[@class="st"]'
content_misc_xpath = './/div[@class="f slp"]'
suggestion_xpath = '//p[@class="_Bmc"]'
# map : detail location
map_address_xpath = './/div[@class="s"]//table//td[2]/span/text()'
map_phone_xpath = './/div[@class="s"]//table//td[2]/span/span'
map_website_url_xpath = 'h3[2]/a/@href'
map_website_title_xpath = 'h3[2]'
# map : near the location
map_near = 'table[@class="ts"]//tr'
map_near_title = './/h4'
map_near_url = './/h4/a/@href'
map_near_phone = './/span[@class="nobr"]'
# images
images_xpath = './/div/a'
image_url_xpath = './@href'
image_img_src_xpath = './img/@src'
# property names
# FIXME : no translation
property_address = "Address"
property_phone = "Phone number"
# cookies
pref_cookie = ''
nid_cookie = {}
# see https://support.google.com/websearch/answer/873?hl=en
def get_google_pref_cookie():
global pref_cookie
if pref_cookie == '':
resp = get('https://www.google.com/ncr', allow_redirects=False)
pref_cookie = resp.cookies["PREF"]
return pref_cookie
2014-12-05 19:03:16 +00:00
def get_google_nid_cookie(google_hostname):
global nid_cookie
if google_hostname not in nid_cookie:
resp = get('https://' + google_hostname)
nid_cookie[google_hostname] = resp.cookies.get("NID", None)
return nid_cookie[google_hostname]
# remove google-specific tracking-url
def parse_url(url_string, google_hostname):
# sanity check
if url_string is None:
return url_string
# normal case
parsed_url = urlparse(url_string)
2014-12-05 19:03:16 +00:00
if (parsed_url.netloc in [google_hostname, '']
and parsed_url.path == redirect_path):
query = dict(parse_qsl(parsed_url.query))
return query['q']
else:
return url_string
2014-01-29 18:28:38 +00:00
2014-12-05 19:03:16 +00:00
# returns extract_text on the first result selected by the xpath or None
def extract_text_from_dom(result, xpath):
r = result.xpath(xpath)
if len(r) > 0:
return escape(extract_text(r[0]))
return None
2014-09-01 13:10:05 +00:00
# do search-request
2014-01-29 18:28:38 +00:00
def request(query, params):
offset = (params['pageno'] - 1) * 10
2014-09-01 13:10:05 +00:00
if params['language'] == 'all':
language = 'en'
country = 'US'
else:
language_array = params['language'].lower().split('_')
if len(language_array) == 2:
country = language_array[1]
else:
country = 'US'
language = language_array[0] + ',' + language_array[0] + '-' + country
if use_locale_domain:
google_hostname = country_to_hostname.get(country.upper(), default_hostname)
else:
google_hostname = default_hostname
2014-09-01 13:10:05 +00:00
2014-01-29 20:14:38 +00:00
params['url'] = search_url.format(offset=offset,
query=urlencode({'q': query}),
hostname=google_hostname)
params['headers']['Accept-Language'] = language
params['headers']['Accept'] = 'text/html,application/xhtml+xml,application/xml;q=0.9,*/*;q=0.8'
if google_hostname == default_hostname:
params['cookies']['PREF'] = get_google_pref_cookie()
params['cookies']['NID'] = get_google_nid_cookie(google_hostname)
params['google_hostname'] = google_hostname
2014-09-01 13:10:05 +00:00
2014-01-29 18:28:38 +00:00
return params
2014-09-01 13:10:05 +00:00
# get response from search-request
2014-01-29 18:28:38 +00:00
def response(resp):
results = []
2014-09-01 13:10:05 +00:00
# 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')
# which hostname ?
google_hostname = resp.search_params.get('google_hostname')
google_url = "https://" + google_hostname
# convert the text to dom
dom = html.fromstring(resp.text)
2014-01-29 18:28:38 +00:00
2014-09-01 13:10:05 +00:00
# parse results
for result in dom.xpath(results_xpath):
title = extract_text(result.xpath(title_xpath)[0])
try:
url = parse_url(extract_url(result.xpath(url_xpath), google_url), google_hostname)
parsed_url = urlparse(url, google_hostname)
# map result
if ((parsed_url.netloc == google_hostname and parsed_url.path.startswith(maps_path))
or (parsed_url.netloc.startswith(map_hostname_start))):
x = result.xpath(map_near)
if len(x) > 0:
# map : near the location
results = results + parse_map_near(parsed_url, x, google_hostname)
else:
# map : detail about a location
results = results + parse_map_detail(parsed_url, result, google_hostname)
# google news
elif (parsed_url.netloc == google_hostname
and parsed_url.path == search_path):
# skipping news results
pass
# images result
elif (parsed_url.netloc == google_hostname
and parsed_url.path == images_path):
# only thumbnail image provided,
# so skipping image results
# results = results + parse_images(result, google_hostname)
pass
else:
# normal result
content = extract_text_from_dom(result, content_xpath)
if content is None:
continue
content_misc = extract_text_from_dom(result, content_misc_xpath)
if content_misc is not None:
content = content_misc + "<br />" + content
# append result
2014-12-05 19:03:16 +00:00
results.append({'url': url,
'title': title,
2015-06-05 22:18:00 +00:00
'content': content
})
except:
continue
# parse suggestion
for suggestion in dom.xpath(suggestion_xpath):
# append suggestion
results.append({'suggestion': escape(extract_text(suggestion))})
2014-09-01 13:10:05 +00:00
# return results
2014-01-29 18:28:38 +00:00
return results
2014-12-05 19:03:16 +00:00
def parse_images(result, google_hostname):
results = []
for image in result.xpath(images_xpath):
url = parse_url(extract_text(image.xpath(image_url_xpath)[0]), google_hostname)
img_src = extract_text(image.xpath(image_img_src_xpath)[0])
2014-12-05 19:03:16 +00:00
# append result
results.append({'url': url,
'title': '',
'content': '',
'img_src': img_src,
2015-06-05 21:56:23 +00:00
'template': 'images.html'
})
return results
def parse_map_near(parsed_url, x, google_hostname):
results = []
for result in x:
title = extract_text_from_dom(result, map_near_title)
url = parse_url(extract_text_from_dom(result, map_near_url), google_hostname)
2015-06-05 21:56:23 +00:00
attributes = []
phone = extract_text_from_dom(result, map_near_phone)
2015-06-05 21:56:23 +00:00
add_attributes(attributes, property_phone, phone, 'tel:' + phone)
results.append({'title': title,
'url': url,
'content': attributes_to_html(attributes)
})
return results
def parse_map_detail(parsed_url, result, google_hostname):
results = []
# try to parse the geoloc
m = re.search('@([0-9\.]+),([0-9\.]+),([0-9]+)', parsed_url.path)
if m is None:
m = re.search('ll\=([0-9\.]+),([0-9\.]+)\&z\=([0-9]+)', parsed_url.query)
if m is not None:
2015-06-05 21:56:23 +00:00
# geoloc found (ignored)
lon = float(m.group(2)) # noqa
lat = float(m.group(1)) # noqa
zoom = int(m.group(3)) # noqa
# attributes
attributes = []
2015-06-05 21:56:23 +00:00
address = extract_text_from_dom(result, map_address_xpath)
phone = extract_text_from_dom(result, map_phone_xpath)
add_attributes(attributes, property_address, address, 'geo:' + str(lat) + ',' + str(lon))
add_attributes(attributes, property_phone, phone, 'tel:' + phone)
# title / content / url
website_title = extract_text_from_dom(result, map_website_title_xpath)
content = extract_text_from_dom(result, content_xpath)
website_url = parse_url(extract_text_from_dom(result, map_website_url_xpath), google_hostname)
2015-06-05 21:56:23 +00:00
# add a result if there is a website
if website_url is not None:
2015-06-05 21:56:23 +00:00
results.append({'title': website_title,
'content': (content + '<br />' if content is not None else '')
+ attributes_to_html(attributes),
'url': website_url
})
return results
2015-06-05 21:56:23 +00:00
def add_attributes(attributes, name, value, url):
if value is not None and len(value) > 0:
2015-06-05 21:56:23 +00:00
attributes.append({'label': name, 'value': value, 'url': url})
def attributes_to_html(attributes):
retval = '<table class="table table-striped">'
for a in attributes:
2015-06-05 21:56:23 +00:00
value = a.get('value')
if 'url' in a:
value = '<a href="' + a.get('url') + '">' + value + '</a>'
retval = retval + '<tr><th>' + a.get('label') + '</th><td>' + value + '</td></tr>'
retval = retval + '</table>'
return retval