searxng/searx/search.py

567 lines
19 KiB
Python
Raw Normal View History

'''
searx is free software: you can redistribute it and/or modify
it under the terms of the GNU Affero General Public License as published by
the Free Software Foundation, either version 3 of the License, or
(at your option) any later version.
searx is distributed in the hope that it will be useful,
but WITHOUT ANY WARRANTY; without even the implied warranty of
MERCHANTABILITY or FITNESS FOR A PARTICULAR PURPOSE. See the
GNU Affero General Public License for more details.
You should have received a copy of the GNU Affero General Public License
along with searx. If not, see < http://www.gnu.org/licenses/ >.
(C) 2013- by Adam Tauber, <asciimoo@gmail.com>
'''
import threading
import re
import searx.poolrequests as requests_lib
2014-07-07 11:59:27 +00:00
from itertools import izip_longest, chain
from operator import itemgetter
2014-12-14 00:18:01 +00:00
from Queue import Queue
from time import time
2014-07-07 11:59:27 +00:00
from urlparse import urlparse, unquote
from searx.engines import (
2014-10-19 10:41:04 +00:00
categories, engines
)
from searx.languages import language_codes
from searx.utils import gen_useragent, get_blocked_engines
from searx.query import Query
2015-01-09 03:13:05 +00:00
from searx import logger
2014-07-07 11:59:27 +00:00
2015-01-09 03:13:05 +00:00
logger = logger.getChild('search')
2014-07-07 11:59:27 +00:00
number_of_searches = 0
def search_request_wrapper(fn, url, engine_name, **kwargs):
try:
return fn(url, **kwargs)
2015-01-09 03:30:55 +00:00
except:
# increase errors stats
engines[engine_name].stats['errors'] += 1
# print engine name and specific error message
2015-01-09 03:30:55 +00:00
logger.exception('engine crash: {0}'.format(engine_name))
return
def threaded_requests(requests):
2014-12-14 00:18:01 +00:00
timeout_limit = max(r[2]['timeout'] for r in requests)
search_start = time()
2014-12-19 12:59:41 +00:00
for fn, url, request_args, engine_name in requests:
2014-12-18 09:11:56 +00:00
request_args['timeout'] = timeout_limit
th = threading.Thread(
target=search_request_wrapper,
args=(fn, url, engine_name),
2014-12-08 22:55:11 +00:00
kwargs=request_args,
name='search_request',
)
2014-12-19 12:59:41 +00:00
th._engine_name = engine_name
th.start()
for th in threading.enumerate():
2014-12-08 22:55:11 +00:00
if th.name == 'search_request':
2014-12-14 00:18:01 +00:00
remaining_time = max(0.0, timeout_limit - (time() - search_start))
th.join(remaining_time)
if th.isAlive():
2015-01-09 03:13:05 +00:00
logger.warning('engine timeout: {0}'.format(th._engine_name))
2014-12-14 00:18:01 +00:00
# get default reqest parameter
2014-07-07 11:59:27 +00:00
def default_request_params():
return {
'method': 'GET',
'headers': {},
'data': {},
'url': '',
'cookies': {},
'verify': True
}
2014-07-07 11:59:27 +00:00
# create a callback wrapper for the search engine results
2014-12-18 09:11:56 +00:00
def make_callback(engine_name, results_queue, callback, params):
2014-07-07 11:59:27 +00:00
# creating a callback wrapper for the search engine results
def process_callback(response, **kwargs):
2015-02-13 10:17:00 +00:00
# check if redirect comparing to the True value,
# because resp can be a Mock object, and any attribut name returns something.
if response.is_redirect is True:
2015-02-13 12:02:37 +00:00
logger.debug('{0} redirect on: {1}'.format(engine_name, response))
2015-02-13 10:17:00 +00:00
return
2014-07-07 11:59:27 +00:00
response.search_params = params
timeout_overhead = 0.2 # seconds
search_duration = time() - params['started']
timeout_limit = engines[engine_name].timeout + timeout_overhead
if search_duration > timeout_limit:
engines[engine_name].stats['page_load_time'] += timeout_limit
engines[engine_name].stats['errors'] += 1
return
2014-09-28 14:51:41 +00:00
# callback
search_results = callback(response)
2014-09-28 14:51:41 +00:00
# add results
2014-07-07 11:59:27 +00:00
for result in search_results:
result['engine'] = engine_name
results_queue.put_nowait((engine_name, search_results))
2014-09-28 14:51:41 +00:00
# update stats with current page-load-time
2014-12-18 09:11:56 +00:00
engines[engine_name].stats['page_load_time'] += search_duration
2014-09-28 14:51:41 +00:00
2014-07-07 11:59:27 +00:00
return process_callback
2014-09-28 14:51:41 +00:00
# return the meaningful length of the content for a result
2014-09-28 14:51:41 +00:00
def content_result_len(content):
if isinstance(content, basestring):
content = re.sub('[,;:!?\./\\\\ ()-_]', '', content)
2014-10-19 10:41:04 +00:00
return len(content)
else:
return 0
2014-07-07 11:59:27 +00:00
2014-09-28 14:51:41 +00:00
# score results and remove duplications
2014-07-07 11:59:27 +00:00
def score_results(results):
# calculate scoring parameters
2014-07-07 11:59:27 +00:00
flat_res = filter(
None, chain.from_iterable(izip_longest(*results.values())))
flat_len = len(flat_res)
engines_len = len(results)
2014-07-07 11:59:27 +00:00
results = []
# pass 1: deduplication + scoring
2014-07-07 11:59:27 +00:00
for i, res in enumerate(flat_res):
res['parsed_url'] = urlparse(res['url'])
res['host'] = res['parsed_url'].netloc
if res['host'].startswith('www.'):
res['host'] = res['host'].replace('www.', '', 1)
res['engines'] = [res['engine']]
2014-09-22 21:39:21 +00:00
2014-07-07 11:59:27 +00:00
weight = 1.0
2014-09-22 21:39:21 +00:00
# strip multiple spaces and cariage returns from content
if res.get('content'):
2014-10-19 10:41:04 +00:00
res['content'] = re.sub(' +', ' ',
res['content'].strip().replace('\n', ''))
2014-09-22 21:39:21 +00:00
# get weight of this engine if possible
2014-07-07 11:59:27 +00:00
if hasattr(engines[res['engine']], 'weight'):
weight = float(engines[res['engine']].weight)
# calculate score for that engine
2014-07-07 11:59:27 +00:00
score = int((flat_len - i) / engines_len) * weight + 1
# check for duplicates
2014-09-22 21:39:21 +00:00
duplicated = False
2014-07-07 11:59:27 +00:00
for new_res in results:
# remove / from the end of the url if required
2014-10-19 10:41:04 +00:00
p1 = res['parsed_url'].path[:-1]\
if res['parsed_url'].path.endswith('/')\
else res['parsed_url'].path
p2 = new_res['parsed_url'].path[:-1]\
if new_res['parsed_url'].path.endswith('/')\
else new_res['parsed_url'].path
# check if that result is a duplicate
2014-07-07 11:59:27 +00:00
if res['host'] == new_res['host'] and\
unquote(p1) == unquote(p2) and\
res['parsed_url'].query == new_res['parsed_url'].query and\
res.get('template') == new_res.get('template'):
duplicated = new_res
break
# merge duplicates together
2014-07-07 11:59:27 +00:00
if duplicated:
# using content with more text
2014-10-19 10:41:04 +00:00
if content_result_len(res.get('content', '')) >\
content_result_len(duplicated.get('content', '')):
2014-07-07 11:59:27 +00:00
duplicated['content'] = res['content']
# increase result-score
2014-07-07 11:59:27 +00:00
duplicated['score'] += score
# add engine to list of result-engines
2014-07-07 11:59:27 +00:00
duplicated['engines'].append(res['engine'])
# using https if possible
2014-07-07 11:59:27 +00:00
if duplicated['parsed_url'].scheme == 'https':
continue
elif res['parsed_url'].scheme == 'https':
duplicated['url'] = res['parsed_url'].geturl()
duplicated['parsed_url'] = res['parsed_url']
# if there is no duplicate found, append result
2014-07-07 11:59:27 +00:00
else:
res['score'] = score
results.append(res)
results = sorted(results, key=itemgetter('score'), reverse=True)
# pass 2 : group results by category and template
gresults = []
categoryPositions = {}
for i, res in enumerate(results):
# FIXME : handle more than one category per engine
2014-10-19 10:41:04 +00:00
category = engines[res['engine']].categories[0] + ':' + ''\
if 'template' not in res\
else res['template']
current = None if category not in categoryPositions\
else categoryPositions[category]
# group with previous results using the same category
# if the group can accept more result and is not too far
# from the current position
if current is not None and (current['count'] > 0)\
and (len(gresults) - current['index'] < 20):
# group with the previous results using
# the same category with this one
index = current['index']
gresults.insert(index, res)
2014-10-19 10:41:04 +00:00
# update every index after the current one
# (including the current one)
for k in categoryPositions:
v = categoryPositions[k]['index']
if v >= index:
categoryPositions[k]['index'] = v+1
# update this category
current['count'] -= 1
else:
# same category
gresults.append(res)
# update categoryIndex
2014-10-19 10:41:04 +00:00
categoryPositions[category] = {'index': len(gresults), 'count': 8}
# return gresults
return gresults
2014-09-28 14:51:41 +00:00
def merge_two_infoboxes(infobox1, infobox2):
if 'urls' in infobox2:
urls1 = infobox1.get('urls', None)
2014-10-19 10:41:04 +00:00
if urls1 is None:
2014-09-28 14:51:41 +00:00
urls1 = []
infobox1.set('urls', urls1)
urlSet = set()
for url in infobox1.get('urls', []):
urlSet.add(url.get('url', None))
2014-10-19 10:41:04 +00:00
2014-09-28 14:51:41 +00:00
for url in infobox2.get('urls', []):
if url.get('url', None) not in urlSet:
urls1.append(url)
if 'attributes' in infobox2:
attributes1 = infobox1.get('attributes', None)
2014-10-19 10:41:04 +00:00
if attributes1 is None:
2014-09-28 14:51:41 +00:00
attributes1 = []
infobox1.set('attributes', attributes1)
attributeSet = set()
for attribute in infobox1.get('attributes', []):
if attribute.get('label', None) not in attributeSet:
attributeSet.add(attribute.get('label', None))
2014-10-19 10:41:04 +00:00
2014-09-28 14:51:41 +00:00
for attribute in infobox2.get('attributes', []):
attributes1.append(attribute)
if 'content' in infobox2:
content1 = infobox1.get('content', None)
content2 = infobox2.get('content', '')
2014-10-19 10:41:04 +00:00
if content1 is not None:
2014-09-28 14:51:41 +00:00
if content_result_len(content2) > content_result_len(content1):
infobox1['content'] = content2
else:
infobox1.set('content', content2)
def merge_infoboxes(infoboxes):
results = []
infoboxes_id = {}
for infobox in infoboxes:
add_infobox = True
infobox_id = infobox.get('id', None)
2014-10-19 10:41:04 +00:00
if infobox_id is not None:
2014-09-28 14:51:41 +00:00
existingIndex = infoboxes_id.get(infobox_id, None)
2014-10-19 10:41:04 +00:00
if existingIndex is not None:
2014-09-28 14:51:41 +00:00
merge_two_infoboxes(results[existingIndex], infobox)
2014-10-19 10:41:04 +00:00
add_infobox = False
2014-09-28 14:51:41 +00:00
if add_infobox:
results.append(infobox)
infoboxes_id[infobox_id] = len(results)-1
return results
class Search(object):
"""Search information container"""
def __init__(self, request):
# init vars
super(Search, self).__init__()
self.query = None
self.engines = []
self.categories = []
self.paging = False
self.pageno = 1
self.lang = 'all'
# set blocked engines
self.blocked_engines = get_blocked_engines(engines, request.cookies)
self.results = []
2015-03-08 22:30:31 +00:00
self.suggestions = set()
self.answers = set()
2014-09-28 14:51:41 +00:00
self.infoboxes = []
self.request_data = {}
# set specific language if set
if request.cookies.get('language')\
and request.cookies['language'] in (x[0] for x in language_codes):
self.lang = request.cookies['language']
# set request method
if request.method == 'POST':
self.request_data = request.form
else:
self.request_data = request.args
# TODO better exceptions
if not self.request_data.get('q'):
raise Exception('noquery')
# set pagenumber
pageno_param = self.request_data.get('pageno', '1')
if not pageno_param.isdigit() or int(pageno_param) < 1:
raise Exception('wrong pagenumber')
self.pageno = int(pageno_param)
2014-10-19 10:41:04 +00:00
# parse query, if tags are set, which change
# the serch engine or search-language
query_obj = Query(self.request_data['q'], self.blocked_engines)
query_obj.parse_query()
# set query
self.query = query_obj.getSearchQuery()
# get last selected language in query, if possible
# TODO support search with multible languages
if len(query_obj.languages):
self.lang = query_obj.languages[-1]
self.engines = query_obj.engines
self.categories = []
2014-10-19 10:41:04 +00:00
# if engines are calculated from query,
# set categories by using that informations
if self.engines and query_obj.specific:
self.categories = list(set(engine['category']
2014-02-07 01:45:12 +00:00
for engine in self.engines))
2014-10-19 10:41:04 +00:00
# otherwise, using defined categories to
# calculate which engines should be used
else:
# set categories/engines
load_default_categories = True
for pd_name, pd in self.request_data.items():
if pd_name == 'categories':
self.categories.extend(categ for categ in pd.split(',') if categ in categories)
elif pd_name == 'engines':
pd_engines = [{'category': engines[engine].categories[0],
'name': engine}
for engine in pd.split(',') if engine in engines]
if pd_engines:
self.engines.extend(pd_engines)
load_default_categories = False
elif pd_name.startswith('category_'):
category = pd_name[9:]
# if category is not found in list, skip
2014-10-19 10:41:04 +00:00
if category not in categories:
continue
if pd != 'off':
# add category to list
self.categories.append(category)
elif category in self.categories:
# remove category from list if property is set to 'off'
self.categories.remove(category)
if not load_default_categories:
return
2014-10-19 10:41:04 +00:00
# if no category is specified for this search,
# using user-defined default-configuration which
# (is stored in cookie)
2014-02-11 12:13:51 +00:00
if not self.categories:
cookie_categories = request.cookies.get('categories', '')
cookie_categories = cookie_categories.split(',')
for ccateg in cookie_categories:
if ccateg in categories:
self.categories.append(ccateg)
2014-10-19 10:41:04 +00:00
# if still no category is specified, using general
# as default-category
2014-02-11 12:13:51 +00:00
if not self.categories:
self.categories = ['general']
2014-10-19 10:41:04 +00:00
# using all engines for that search, which are
# declared under the specific categories
for categ in self.categories:
self.engines.extend({'category': categ,
'name': engine.name}
for engine in categories[categ]
if (engine.name, categ) not in self.blocked_engines)
2014-02-09 00:07:18 +00:00
# do search-request
2014-07-07 11:59:27 +00:00
def search(self, request):
global number_of_searches
# init vars
2014-07-07 11:59:27 +00:00
requests = []
2014-12-14 00:18:01 +00:00
results_queue = Queue()
2014-12-19 18:40:40 +00:00
results = {}
# increase number of searches
2014-07-07 11:59:27 +00:00
number_of_searches += 1
# set default useragent
2014-10-19 10:41:04 +00:00
# user_agent = request.headers.get('User-Agent', '')
2014-07-07 11:59:27 +00:00
user_agent = gen_useragent()
# start search-reqest for all selected engines
2014-07-07 11:59:27 +00:00
for selected_engine in self.engines:
if selected_engine['name'] not in engines:
continue
engine = engines[selected_engine['name']]
# if paging is not supported, skip
2014-07-07 11:59:27 +00:00
if self.pageno > 1 and not engine.paging:
continue
2014-10-19 10:41:04 +00:00
# if search-language is set and engine does not
# provide language-support, skip
2014-07-07 11:59:27 +00:00
if self.lang != 'all' and not engine.language_support:
continue
# set default request parameters
2014-07-07 11:59:27 +00:00
request_params = default_request_params()
request_params['headers']['User-Agent'] = user_agent
request_params['category'] = selected_engine['category']
2014-12-18 09:11:56 +00:00
request_params['started'] = time()
2014-07-07 11:59:27 +00:00
request_params['pageno'] = self.pageno
request_params['language'] = self.lang
try:
# 0 = None, 1 = Moderate, 2 = Strict
request_params['safesearch'] = int(request.cookies.get('safesearch', 1))
except ValueError:
request_params['safesearch'] = 1
2014-10-19 10:41:04 +00:00
# update request parameters dependent on
# search-engine (contained in engines folder)
2014-12-18 09:11:56 +00:00
engine.request(self.query.encode('utf-8'), request_params)
2014-07-07 11:59:27 +00:00
if request_params['url'] is None:
# TODO add support of offline engines
pass
# create a callback wrapper for the search engine results
2014-07-07 11:59:27 +00:00
callback = make_callback(
selected_engine['name'],
2014-12-14 00:18:01 +00:00
results_queue,
2014-07-07 11:59:27 +00:00
engine.response,
2014-12-18 09:11:56 +00:00
request_params)
2014-07-07 11:59:27 +00:00
2014-10-19 10:41:04 +00:00
# create dictionary which contain all
# informations about the request
2014-07-07 11:59:27 +00:00
request_args = dict(
headers=request_params['headers'],
hooks=dict(response=callback),
cookies=request_params['cookies'],
timeout=engine.timeout,
verify=request_params['verify']
2014-07-07 11:59:27 +00:00
)
# specific type of request (GET or POST)
2014-07-07 11:59:27 +00:00
if request_params['method'] == 'GET':
req = requests_lib.get
2014-07-07 11:59:27 +00:00
else:
req = requests_lib.post
2014-07-07 11:59:27 +00:00
request_args['data'] = request_params['data']
# ignoring empty urls
if not request_params['url']:
continue
# append request to list
requests.append((req, request_params['url'],
request_args,
selected_engine['name']))
2014-12-19 18:40:40 +00:00
if not requests:
2015-03-08 22:30:31 +00:00
return self
# send all search-request
threaded_requests(requests)
2014-12-14 00:18:01 +00:00
while not results_queue.empty():
engine_name, engine_results = results_queue.get_nowait()
# TODO type checks
2015-03-08 22:30:31 +00:00
[self.suggestions.add(x['suggestion'])
for x in list(engine_results)
if 'suggestion' in x
and engine_results.remove(x) is None]
2015-03-08 22:30:31 +00:00
[self.answers.add(x['answer'])
for x in list(engine_results)
if 'answer' in x
and engine_results.remove(x) is None]
2015-03-08 22:30:31 +00:00
self.infoboxes.extend(x for x in list(engine_results)
if 'infobox' in x
and engine_results.remove(x) is None)
2014-12-14 00:18:01 +00:00
results[engine_name] = engine_results
# update engine-specific stats
2014-07-07 11:59:27 +00:00
for engine_name, engine_results in results.items():
engines[engine_name].stats['search_count'] += 1
engines[engine_name].stats['result_count'] += len(engine_results)
# score results and remove duplications
2015-03-08 22:30:31 +00:00
self.results = score_results(results)
2014-07-07 11:59:27 +00:00
2014-09-28 14:51:41 +00:00
# merge infoboxes according to their ids
2015-03-08 22:30:31 +00:00
self.infoboxes = merge_infoboxes(self.infoboxes)
2014-09-28 14:51:41 +00:00
# update engine stats, using calculated score
2015-03-08 22:30:31 +00:00
for result in self.results:
2014-07-07 11:59:27 +00:00
for res_engine in result['engines']:
engines[result['engine']]\
.stats['score_count'] += result['score']
2014-09-28 14:51:41 +00:00
# return results, suggestions, answers and infoboxes
2015-03-08 22:30:31 +00:00
return self