searxng/searx/engines/wikidata.py
Marc Abonce Seguin 0d8970c8f2
only return one url per "type" in Wikidata (#2151)
i.e. only one official website, one Twitter, etc.
2020-08-27 21:44:48 +02:00

504 lines
18 KiB
Python

# -*- coding: utf-8 -*-
"""
Wikidata
@website https://wikidata.org
@provide-api yes (https://wikidata.org/w/api.php)
@using-api partially (most things require scraping)
@results JSON, HTML
@stable no (html can change)
@parse url, infobox
"""
from searx import logger
from searx.poolrequests import get
from searx.engines.xpath import extract_text
from searx.engines.wikipedia import _fetch_supported_languages, supported_languages_url
from searx.url_utils import urlencode
from searx.utils import match_language, eval_xpath
from json import loads
from lxml.html import fromstring
from lxml import etree
logger = logger.getChild('wikidata')
result_count = 1
# urls
wikidata_host = 'https://www.wikidata.org'
url_search = wikidata_host \
+ '/w/index.php?{query}&ns0=1'
wikidata_api = wikidata_host + '/w/api.php'
url_detail = wikidata_api\
+ '?action=parse&format=json&{query}'\
+ '&redirects=1&prop=text%7Cdisplaytitle%7Cparsewarnings'\
+ '&disableeditsection=1&preview=1&sectionpreview=1&disabletoc=1&utf8=1&formatversion=2'
url_map = 'https://www.openstreetmap.org/'\
+ '?lat={latitude}&lon={longitude}&zoom={zoom}&layers=M'
url_image = 'https://commons.wikimedia.org/wiki/Special:FilePath/{filename}?width=500&height=400'
# xpaths
div_ids_xpath = '//div[@id]'
wikidata_ids_xpath = '//ul[@class="mw-search-results"]/li//a/@href'
title_xpath = '//*[contains(@class,"wikibase-title-label")]'
description_xpath = '//div[contains(@class,"wikibase-entitytermsview-heading-description")]'
label_xpath = './/div[contains(@class,"wikibase-statementgroupview-property-label")]/a'
url_xpath = './/a[contains(@class,"external free") or contains(@class, "wb-external-id")]'
wikilink_xpath = './/ul[contains(@class,"wikibase-sitelinklistview-listview")]'\
+ '/li[contains(@data-wb-siteid,"{wikiid}")]//a/@href'
property_row_xpath = './/div[contains(@class,"wikibase-statementview")]'
preferred_rank_xpath = './/span[contains(@class,"wikibase-rankselector-preferred")]'
value_xpath = './/div[contains(@class,"wikibase-statementview-mainsnak")]'\
+ '/*/div[contains(@class,"wikibase-snakview-value")]'
language_fallback_xpath = '//sup[contains(@class,"wb-language-fallback-indicator")]'
calendar_name_xpath = './/sup[contains(@class,"wb-calendar-name")]'
media_xpath = value_xpath + '//div[contains(@class,"commons-media-caption")]//a'
def get_id_cache(result):
id_cache = {}
for e in eval_xpath(result, div_ids_xpath):
id = e.get('id')
if id.startswith('P'):
id_cache[id] = e
return id_cache
def request(query, params):
params['url'] = url_search.format(
query=urlencode({'search': query}))
return params
def response(resp):
results = []
htmlparser = etree.HTMLParser()
html = fromstring(resp.content.decode("utf-8"), parser=htmlparser)
search_results = eval_xpath(html, wikidata_ids_xpath)
if resp.search_params['language'].split('-')[0] == 'all':
language = 'en'
else:
language = match_language(resp.search_params['language'], supported_languages, language_aliases).split('-')[0]
# TODO: make requests asynchronous to avoid timeout when result_count > 1
for search_result in search_results[:result_count]:
wikidata_id = search_result.split('/')[-1]
url = url_detail.format(query=urlencode({'page': wikidata_id, 'uselang': language}))
htmlresponse = get(url)
jsonresponse = loads(htmlresponse.content.decode("utf-8"))
results += getDetail(jsonresponse, wikidata_id, language, resp.search_params['language'], htmlparser)
return results
def getDetail(jsonresponse, wikidata_id, language, locale, htmlparser):
results = []
urls = []
attributes = []
title = jsonresponse.get('parse', {}).get('displaytitle', {})
result = jsonresponse.get('parse', {}).get('text', {})
if not title or not result:
return results
title = fromstring(title, parser=htmlparser)
for elem in eval_xpath(title, language_fallback_xpath):
elem.getparent().remove(elem)
title = extract_text(eval_xpath(title, title_xpath))
result = fromstring(result, parser=htmlparser)
for elem in eval_xpath(result, language_fallback_xpath):
elem.getparent().remove(elem)
description = extract_text(eval_xpath(result, description_xpath))
id_cache = get_id_cache(result)
# URLS
# official website
add_url(urls, result, id_cache, 'P856', results=results)
# wikipedia
wikipedia_link_count = 0
wikipedia_link = get_wikilink(result, language + 'wiki')
if wikipedia_link:
wikipedia_link_count += 1
urls.append({'title': 'Wikipedia (' + language + ')',
'url': wikipedia_link})
if language != 'en':
wikipedia_en_link = get_wikilink(result, 'enwiki')
if wikipedia_en_link:
wikipedia_link_count += 1
urls.append({'title': 'Wikipedia (en)',
'url': wikipedia_en_link})
# TODO: get_wiki_firstlanguage
# if wikipedia_link_count == 0:
# more wikis
add_url(urls, result, id_cache, default_label='Wikivoyage (' + language + ')', link_type=language + 'wikivoyage')
add_url(urls, result, id_cache, default_label='Wikiquote (' + language + ')', link_type=language + 'wikiquote')
add_url(urls, result, id_cache, default_label='Wikimedia Commons', link_type='commonswiki')
add_url(urls, result, id_cache, 'P625', 'OpenStreetMap', link_type='geo')
# musicbrainz
add_url(urls, result, id_cache, 'P434', 'MusicBrainz', 'http://musicbrainz.org/artist/')
add_url(urls, result, id_cache, 'P435', 'MusicBrainz', 'http://musicbrainz.org/work/')
add_url(urls, result, id_cache, 'P436', 'MusicBrainz', 'http://musicbrainz.org/release-group/')
add_url(urls, result, id_cache, 'P966', 'MusicBrainz', 'http://musicbrainz.org/label/')
# IMDb
add_url(urls, result, id_cache, 'P345', 'IMDb', 'https://www.imdb.com/', link_type='imdb')
# source code repository
add_url(urls, result, id_cache, 'P1324')
# blog
add_url(urls, result, id_cache, 'P1581')
# social media links
add_url(urls, result, id_cache, 'P2397', 'YouTube', 'https://www.youtube.com/channel/')
add_url(urls, result, id_cache, 'P1651', 'YouTube', 'https://www.youtube.com/watch?v=')
add_url(urls, result, id_cache, 'P2002', 'Twitter', 'https://twitter.com/')
add_url(urls, result, id_cache, 'P2013', 'Facebook', 'https://facebook.com/')
add_url(urls, result, id_cache, 'P2003', 'Instagram', 'https://instagram.com/')
urls.append({'title': 'Wikidata',
'url': 'https://www.wikidata.org/wiki/'
+ wikidata_id + '?uselang=' + language})
# INFOBOX ATTRIBUTES (ROWS)
# DATES
# inception date
add_attribute(attributes, id_cache, 'P571', date=True)
# dissolution date
add_attribute(attributes, id_cache, 'P576', date=True)
# start date
add_attribute(attributes, id_cache, 'P580', date=True)
# end date
add_attribute(attributes, id_cache, 'P582', date=True)
# date of birth
add_attribute(attributes, id_cache, 'P569', date=True)
# date of death
add_attribute(attributes, id_cache, 'P570', date=True)
# date of spacecraft launch
add_attribute(attributes, id_cache, 'P619', date=True)
# date of spacecraft landing
add_attribute(attributes, id_cache, 'P620', date=True)
# nationality
add_attribute(attributes, id_cache, 'P27')
# country of origin
add_attribute(attributes, id_cache, 'P495')
# country
add_attribute(attributes, id_cache, 'P17')
# headquarters
add_attribute(attributes, id_cache, 'Q180')
# PLACES
# capital
add_attribute(attributes, id_cache, 'P36', trim=True)
# head of state
add_attribute(attributes, id_cache, 'P35', trim=True)
# head of government
add_attribute(attributes, id_cache, 'P6', trim=True)
# type of government
add_attribute(attributes, id_cache, 'P122')
# official language
add_attribute(attributes, id_cache, 'P37')
# population
add_attribute(attributes, id_cache, 'P1082', trim=True)
# area
add_attribute(attributes, id_cache, 'P2046')
# currency
add_attribute(attributes, id_cache, 'P38', trim=True)
# heigth (building)
add_attribute(attributes, id_cache, 'P2048')
# MEDIA
# platform (videogames)
add_attribute(attributes, id_cache, 'P400')
# author
add_attribute(attributes, id_cache, 'P50')
# creator
add_attribute(attributes, id_cache, 'P170')
# director
add_attribute(attributes, id_cache, 'P57')
# performer
add_attribute(attributes, id_cache, 'P175')
# developer
add_attribute(attributes, id_cache, 'P178')
# producer
add_attribute(attributes, id_cache, 'P162')
# manufacturer
add_attribute(attributes, id_cache, 'P176')
# screenwriter
add_attribute(attributes, id_cache, 'P58')
# production company
add_attribute(attributes, id_cache, 'P272')
# record label
add_attribute(attributes, id_cache, 'P264')
# publisher
add_attribute(attributes, id_cache, 'P123')
# original network
add_attribute(attributes, id_cache, 'P449')
# distributor
add_attribute(attributes, id_cache, 'P750')
# composer
add_attribute(attributes, id_cache, 'P86')
# publication date
add_attribute(attributes, id_cache, 'P577', date=True)
# genre
add_attribute(attributes, id_cache, 'P136')
# original language
add_attribute(attributes, id_cache, 'P364')
# isbn
add_attribute(attributes, id_cache, 'Q33057')
# software license
add_attribute(attributes, id_cache, 'P275')
# programming language
add_attribute(attributes, id_cache, 'P277')
# version
add_attribute(attributes, id_cache, 'P348', trim=True)
# narrative location
add_attribute(attributes, id_cache, 'P840')
# LANGUAGES
# number of speakers
add_attribute(attributes, id_cache, 'P1098')
# writing system
add_attribute(attributes, id_cache, 'P282')
# regulatory body
add_attribute(attributes, id_cache, 'P1018')
# language code
add_attribute(attributes, id_cache, 'P218')
# OTHER
# ceo
add_attribute(attributes, id_cache, 'P169', trim=True)
# founder
add_attribute(attributes, id_cache, 'P112')
# legal form (company/organization)
add_attribute(attributes, id_cache, 'P1454')
# operator
add_attribute(attributes, id_cache, 'P137')
# crew members (tripulation)
add_attribute(attributes, id_cache, 'P1029')
# taxon
add_attribute(attributes, id_cache, 'P225')
# chemical formula
add_attribute(attributes, id_cache, 'P274')
# winner (sports/contests)
add_attribute(attributes, id_cache, 'P1346')
# number of deaths
add_attribute(attributes, id_cache, 'P1120')
# currency code
add_attribute(attributes, id_cache, 'P498')
image = add_image(id_cache)
if len(attributes) == 0 and len(urls) == 2 and len(description) == 0:
results.append({
'url': urls[0]['url'],
'title': title,
'content': description
})
else:
results.append({
'infobox': title,
'id': wikipedia_link,
'content': description,
'img_src': image,
'attributes': attributes,
'urls': urls
})
return results
# only returns first match
def add_image(id_cache):
# P15: route map, P242: locator map, P154: logo, P18: image, P242: map, P41: flag, P2716: collage, P2910: icon
property_ids = ['P15', 'P242', 'P154', 'P18', 'P242', 'P41', 'P2716', 'P2910']
for property_id in property_ids:
image = id_cache.get(property_id, None)
if image is not None:
image_name = eval_xpath(image, media_xpath)
image_src = url_image.replace('{filename}', extract_text(image_name[0]))
return image_src
# setting trim will only returned high ranked rows OR the first row
def add_attribute(attributes, id_cache, property_id, default_label=None, date=False, trim=False):
attribute = id_cache.get(property_id, None)
if attribute is not None:
if default_label:
label = default_label
else:
label = extract_text(eval_xpath(attribute, label_xpath))
label = label[0].upper() + label[1:]
if date:
trim = True
# remove calendar name
calendar_name = eval_xpath(attribute, calendar_name_xpath)
for calendar in calendar_name:
calendar.getparent().remove(calendar)
concat_values = ""
values = []
first_value = None
for row in eval_xpath(attribute, property_row_xpath):
if not first_value or not trim or eval_xpath(row, preferred_rank_xpath):
value = eval_xpath(row, value_xpath)
if not value:
continue
value = extract_text(value)
# save first value in case no ranked row is found
if trim and not first_value:
first_value = value
else:
# to avoid duplicate values
if value not in values:
concat_values += value + ", "
values.append(value)
if trim and not values:
attributes.append({'label': label,
'value': first_value})
else:
attributes.append({'label': label,
'value': concat_values[:-2]})
# requires property_id unless it's a wiki link (defined in link_type)
def add_url(urls, result, id_cache, property_id=None, default_label=None, url_prefix=None, results=None,
link_type=None, only_first=True):
links = []
# wiki links don't have property in wikidata page
if link_type and 'wiki' in link_type:
links.append(get_wikilink(result, link_type))
else:
dom_element = id_cache.get(property_id, None)
if dom_element is not None:
if not default_label:
label = extract_text(eval_xpath(dom_element, label_xpath))
label = label[0].upper() + label[1:]
if link_type == 'geo':
links.append(get_geolink(dom_element))
elif link_type == 'imdb':
links.append(get_imdblink(dom_element, url_prefix))
else:
url_results = eval_xpath(dom_element, url_xpath)
for link in url_results:
if link is not None:
if url_prefix:
link = url_prefix + extract_text(link)
else:
link = extract_text(link)
links.append(link)
# append urls
for url in links:
if url is not None:
u = {'title': default_label or label, 'url': url}
if property_id == 'P856':
u['official'] = True
u['domain'] = url.split('/')[2]
urls.append(u)
if results is not None:
results.append(u)
if only_first:
break
def get_imdblink(result, url_prefix):
imdb_id = eval_xpath(result, value_xpath)
if imdb_id:
imdb_id = extract_text(imdb_id)
id_prefix = imdb_id[:2]
if id_prefix == 'tt':
url = url_prefix + 'title/' + imdb_id
elif id_prefix == 'nm':
url = url_prefix + 'name/' + imdb_id
elif id_prefix == 'ch':
url = url_prefix + 'character/' + imdb_id
elif id_prefix == 'co':
url = url_prefix + 'company/' + imdb_id
elif id_prefix == 'ev':
url = url_prefix + 'event/' + imdb_id
else:
url = None
return url
def get_geolink(result):
coordinates = eval_xpath(result, value_xpath)
if not coordinates:
return None
coordinates = extract_text(coordinates[0])
latitude, longitude = coordinates.split(',')
# convert to decimal
lat = int(latitude[:latitude.find(u'°')])
if latitude.find('\'') >= 0:
lat += int(latitude[latitude.find(u'°') + 1:latitude.find('\'')] or 0) / 60.0
if latitude.find('"') >= 0:
lat += float(latitude[latitude.find('\'') + 1:latitude.find('"')] or 0) / 3600.0
if latitude.find('S') >= 0:
lat *= -1
lon = int(longitude[:longitude.find(u'°')])
if longitude.find('\'') >= 0:
lon += int(longitude[longitude.find(u'°') + 1:longitude.find('\'')] or 0) / 60.0
if longitude.find('"') >= 0:
lon += float(longitude[longitude.find('\'') + 1:longitude.find('"')] or 0) / 3600.0
if longitude.find('W') >= 0:
lon *= -1
# TODO: get precision
precision = 0.0002
# there is no zoom information, deduce from precision (error prone)
# samples :
# 13 --> 5
# 1 --> 6
# 0.016666666666667 --> 9
# 0.00027777777777778 --> 19
# wolframalpha :
# quadratic fit { {13, 5}, {1, 6}, {0.0166666, 9}, {0.0002777777,19}}
# 14.1186-8.8322 x+0.625447 x^2
if precision < 0.0003:
zoom = 19
else:
zoom = int(15 - precision * 8.8322 + precision * precision * 0.625447)
url = url_map\
.replace('{latitude}', str(lat))\
.replace('{longitude}', str(lon))\
.replace('{zoom}', str(zoom))
return url
def get_wikilink(result, wikiid):
url = eval_xpath(result, wikilink_xpath.replace('{wikiid}', wikiid))
if not url:
return None
url = url[0]
if url.startswith('http://'):
url = url.replace('http://', 'https://')
elif url.startswith('//'):
url = 'https:' + url
return url