1
0
mirror of https://github.com/searxng/searxng.git synced 2024-11-22 20:17:45 +01:00
searxng/searx/engines/wikidata.py

797 lines
27 KiB
Python
Raw Permalink Normal View History

# SPDX-License-Identifier: AGPL-3.0-or-later
"""This module implements the Wikidata engine. Some implementations are shared
from :ref:`wikipedia engine`.
"""
# pylint: disable=missing-class-docstring
2014-09-28 16:51:41 +02:00
from typing import TYPE_CHECKING
from hashlib import md5
from urllib.parse import urlencode, unquote
from json import loads
from dateutil.parser import isoparse
from babel.dates import format_datetime, format_date, format_time, get_datetime_format
from searx.data import WIKIDATA_UNITS
from searx.network import post, get
from searx.utils import searx_useragent, get_string_replaces_function
from searx.external_urls import get_external_url, get_earth_coordinates_url, area_to_osm_zoom
[fix] searxng_extra/update/update_engine_descriptions.py (part 1) Follow up of #2269 The script to update the descriptions of the engines does no longer work since PR #2269 has been merged. searx/engines/wikipedia.py ========================== 1. There was a misusage of zh-classical.wikipedia.org: - `zh-classical` is dedicate to classical Chinese [1] which is not traditional Chinese [2]. - zh.wikipedia.org has LanguageConverter enabled [3] and is going to dynamically show simplified or traditional Chinese according to the HTTP Accept-Language header. 2. The update_engine_descriptions.py needs a list of all wikipedias. The implementation from #2269 included only a reduced list: - https://meta.wikimedia.org/wiki/Wikipedia_article_depth - https://meta.wikimedia.org/wiki/List_of_Wikipedias searxng_extra/update/update_engine_descriptions.py ================================================== Before PR #2269 there was a match_language() function that did an approximation using various methods. With PR #2269 there are only the types in the data model of the languages, which can be recognized by babel. The approximation methods, which are needed (only here) in the determination of the descriptions, must be replaced by other methods. [1] https://en.wikipedia.org/wiki/Classical_Chinese [2] https://en.wikipedia.org/wiki/Traditional_Chinese_characters [3] https://www.mediawiki.org/wiki/Writing_systems#LanguageConverter Closes: https://github.com/searxng/searxng/issues/2330 Signed-off-by: Markus Heiser <markus.heiser@darmarit.de>
2023-04-04 15:17:12 +02:00
from searx.engines.wikipedia import (
fetch_wikimedia_traits,
get_wiki_params,
)
from searx.enginelib.traits import EngineTraits
if TYPE_CHECKING:
import logging
logger: logging.Logger
traits: EngineTraits
# about
about = {
"website": 'https://wikidata.org/',
"wikidata_id": 'Q2013',
"official_api_documentation": 'https://query.wikidata.org/',
"use_official_api": True,
"require_api_key": False,
"results": 'JSON',
}
display_type = ["infobox"]
"""A list of display types composed from ``infobox`` and ``list``. The latter
one will add a hit to the result list. The first one will show a hit in the
info box. Both values can be set, or one of the two can be set."""
# SPARQL
SPARQL_ENDPOINT_URL = 'https://query.wikidata.org/sparql'
SPARQL_EXPLAIN_URL = 'https://query.wikidata.org/bigdata/namespace/wdq/sparql?explain'
WIKIDATA_PROPERTIES = {
'P434': 'MusicBrainz',
'P435': 'MusicBrainz',
'P436': 'MusicBrainz',
'P966': 'MusicBrainz',
'P345': 'IMDb',
'P2397': 'YouTube',
'P1651': 'YouTube',
'P2002': 'Twitter',
'P2013': 'Facebook',
'P2003': 'Instagram',
}
# SERVICE wikibase:mwapi : https://www.mediawiki.org/wiki/Wikidata_Query_Service/User_Manual/MWAPI
# SERVICE wikibase:label: https://en.wikibooks.org/wiki/SPARQL/SERVICE_-_Label#Manual_Label_SERVICE
# https://en.wikibooks.org/wiki/SPARQL/WIKIDATA_Precision,_Units_and_Coordinates
# https://www.mediawiki.org/wiki/Wikibase/Indexing/RDF_Dump_Format#Data_model
# optimization:
# * https://www.wikidata.org/wiki/Wikidata:SPARQL_query_service/query_optimization
# * https://github.com/blazegraph/database/wiki/QueryHints
QUERY_TEMPLATE = """
SELECT ?item ?itemLabel ?itemDescription ?lat ?long %SELECT%
WHERE
{
SERVICE wikibase:mwapi {
bd:serviceParam wikibase:endpoint "www.wikidata.org";
wikibase:api "EntitySearch";
wikibase:limit 1;
mwapi:search "%QUERY%";
mwapi:language "%LANGUAGE%".
?item wikibase:apiOutputItem mwapi:item.
}
hint:Prior hint:runFirst "true".
%WHERE%
SERVICE wikibase:label {
bd:serviceParam wikibase:language "%LANGUAGE%,en".
?item rdfs:label ?itemLabel .
?item schema:description ?itemDescription .
%WIKIBASE_LABELS%
}
}
GROUP BY ?item ?itemLabel ?itemDescription ?lat ?long %GROUP_BY%
"""
2018-07-06 10:31:01 +02:00
# Get the calendar names and the property names
QUERY_PROPERTY_NAMES = """
SELECT ?item ?name
WHERE {
{
SELECT ?item
WHERE { ?item wdt:P279* wd:Q12132 }
} UNION {
VALUES ?item { %ATTRIBUTES% }
}
OPTIONAL { ?item rdfs:label ?name. }
}
"""
2014-09-28 16:51:41 +02:00
# see the property "dummy value" of https://www.wikidata.org/wiki/Q2013 (Wikidata)
# hard coded here to avoid to an additional SPARQL request when the server starts
DUMMY_ENTITY_URLS = set(
"http://www.wikidata.org/entity/" + wid for wid in ("Q4115189", "Q13406268", "Q15397819", "Q17339402")
)
2014-09-28 16:51:41 +02:00
# https://www.w3.org/TR/sparql11-query/#rSTRING_LITERAL1
# https://lists.w3.org/Archives/Public/public-rdf-dawg/2011OctDec/0175.html
sparql_string_escape = get_string_replaces_function(
# fmt: off
{
'\t': '\\\t',
'\n': '\\\n',
'\r': '\\\r',
'\b': '\\\b',
'\f': '\\\f',
'\"': '\\\"',
'\'': '\\\'',
'\\': '\\\\'
}
# fmt: on
)
replace_http_by_https = get_string_replaces_function({'http:': 'https:'})
def get_headers():
# user agent: https://www.mediawiki.org/wiki/Wikidata_Query_Service/User_Manual#Query_limits
return {'Accept': 'application/sparql-results+json', 'User-Agent': searx_useragent()}
def get_label_for_entity(entity_id, language):
name = WIKIDATA_PROPERTIES.get(entity_id)
if name is None:
name = WIKIDATA_PROPERTIES.get((entity_id, language))
if name is None:
name = WIKIDATA_PROPERTIES.get((entity_id, language.split('-')[0]))
if name is None:
name = WIKIDATA_PROPERTIES.get((entity_id, 'en'))
if name is None:
name = entity_id
return name
def send_wikidata_query(query, method='GET'):
if method == 'GET':
# query will be cached by wikidata
http_response = get(SPARQL_ENDPOINT_URL + '?' + urlencode({'query': query}), headers=get_headers())
else:
# query won't be cached by wikidata
http_response = post(SPARQL_ENDPOINT_URL, data={'query': query}, headers=get_headers())
if http_response.status_code != 200:
logger.debug('SPARQL endpoint error %s', http_response.content.decode())
logger.debug('request time %s', str(http_response.elapsed))
http_response.raise_for_status()
return loads(http_response.content.decode())
2014-09-28 16:51:41 +02:00
def request(query, params):
[fix] searxng_extra/update/update_engine_descriptions.py (part 1) Follow up of #2269 The script to update the descriptions of the engines does no longer work since PR #2269 has been merged. searx/engines/wikipedia.py ========================== 1. There was a misusage of zh-classical.wikipedia.org: - `zh-classical` is dedicate to classical Chinese [1] which is not traditional Chinese [2]. - zh.wikipedia.org has LanguageConverter enabled [3] and is going to dynamically show simplified or traditional Chinese according to the HTTP Accept-Language header. 2. The update_engine_descriptions.py needs a list of all wikipedias. The implementation from #2269 included only a reduced list: - https://meta.wikimedia.org/wiki/Wikipedia_article_depth - https://meta.wikimedia.org/wiki/List_of_Wikipedias searxng_extra/update/update_engine_descriptions.py ================================================== Before PR #2269 there was a match_language() function that did an approximation using various methods. With PR #2269 there are only the types in the data model of the languages, which can be recognized by babel. The approximation methods, which are needed (only here) in the determination of the descriptions, must be replaced by other methods. [1] https://en.wikipedia.org/wiki/Classical_Chinese [2] https://en.wikipedia.org/wiki/Traditional_Chinese_characters [3] https://www.mediawiki.org/wiki/Writing_systems#LanguageConverter Closes: https://github.com/searxng/searxng/issues/2330 Signed-off-by: Markus Heiser <markus.heiser@darmarit.de>
2023-04-04 15:17:12 +02:00
eng_tag, _wiki_netloc = get_wiki_params(params['searxng_locale'], traits)
query, attributes = get_query(query, eng_tag)
logger.debug("request --> language %s // len(attributes): %s", eng_tag, len(attributes))
2015-01-02 12:33:40 +01:00
params['method'] = 'POST'
params['url'] = SPARQL_ENDPOINT_URL
params['data'] = {'query': query}
params['headers'] = get_headers()
[fix] searxng_extra/update/update_engine_descriptions.py (part 1) Follow up of #2269 The script to update the descriptions of the engines does no longer work since PR #2269 has been merged. searx/engines/wikipedia.py ========================== 1. There was a misusage of zh-classical.wikipedia.org: - `zh-classical` is dedicate to classical Chinese [1] which is not traditional Chinese [2]. - zh.wikipedia.org has LanguageConverter enabled [3] and is going to dynamically show simplified or traditional Chinese according to the HTTP Accept-Language header. 2. The update_engine_descriptions.py needs a list of all wikipedias. The implementation from #2269 included only a reduced list: - https://meta.wikimedia.org/wiki/Wikipedia_article_depth - https://meta.wikimedia.org/wiki/List_of_Wikipedias searxng_extra/update/update_engine_descriptions.py ================================================== Before PR #2269 there was a match_language() function that did an approximation using various methods. With PR #2269 there are only the types in the data model of the languages, which can be recognized by babel. The approximation methods, which are needed (only here) in the determination of the descriptions, must be replaced by other methods. [1] https://en.wikipedia.org/wiki/Classical_Chinese [2] https://en.wikipedia.org/wiki/Traditional_Chinese_characters [3] https://www.mediawiki.org/wiki/Writing_systems#LanguageConverter Closes: https://github.com/searxng/searxng/issues/2330 Signed-off-by: Markus Heiser <markus.heiser@darmarit.de>
2023-04-04 15:17:12 +02:00
params['language'] = eng_tag
params['attributes'] = attributes
return params
def response(resp):
results = []
jsonresponse = loads(resp.content.decode())
language = resp.search_params['language']
attributes = resp.search_params['attributes']
logger.debug("request --> language %s // len(attributes): %s", language, len(attributes))
seen_entities = set()
for result in jsonresponse.get('results', {}).get('bindings', []):
attribute_result = {key: value['value'] for key, value in result.items()}
entity_url = attribute_result['item']
if entity_url not in seen_entities and entity_url not in DUMMY_ENTITY_URLS:
seen_entities.add(entity_url)
results += get_results(attribute_result, attributes, language)
else:
logger.debug('The SPARQL request returns duplicate entities: %s', str(attribute_result))
2014-09-28 16:51:41 +02:00
return results
2014-09-28 16:51:41 +02:00
_IMG_SRC_DEFAULT_URL_PREFIX = "https://commons.wikimedia.org/wiki/Special:FilePath/"
_IMG_SRC_NEW_URL_PREFIX = "https://upload.wikimedia.org/wikipedia/commons/thumb/"
def get_thumbnail(img_src):
"""Get Thumbnail image from wikimedia commons
Images from commons.wikimedia.org are (HTTP) redirected to
upload.wikimedia.org. The redirected URL can be calculated by this
function.
- https://stackoverflow.com/a/33691240
"""
logger.debug('get_thumbnail(): %s', img_src)
if not img_src is None and _IMG_SRC_DEFAULT_URL_PREFIX in img_src.split()[0]:
img_src_name = unquote(img_src.replace(_IMG_SRC_DEFAULT_URL_PREFIX, "").split("?", 1)[0].replace("%20", "_"))
img_src_name_first = img_src_name
img_src_name_second = img_src_name
if ".svg" in img_src_name.split()[0]:
img_src_name_second = img_src_name + ".png"
img_src_size = img_src.replace(_IMG_SRC_DEFAULT_URL_PREFIX, "").split("?", 1)[1]
img_src_size = img_src_size[img_src_size.index("=") + 1 : img_src_size.index("&")]
img_src_name_md5 = md5(img_src_name.encode("utf-8")).hexdigest()
img_src = (
_IMG_SRC_NEW_URL_PREFIX
+ img_src_name_md5[0]
+ "/"
+ img_src_name_md5[0:2]
+ "/"
+ img_src_name_first
+ "/"
+ img_src_size
+ "px-"
+ img_src_name_second
)
logger.debug('get_thumbnail() redirected: %s', img_src)
return img_src
def get_results(attribute_result, attributes, language):
# pylint: disable=too-many-branches
results = []
infobox_title = attribute_result.get('itemLabel')
infobox_id = attribute_result['item']
infobox_id_lang = None
infobox_urls = []
infobox_attributes = []
infobox_content = attribute_result.get('itemDescription', [])
img_src = None
img_src_priority = 0
for attribute in attributes:
value = attribute.get_str(attribute_result, language)
if value is not None and value != '':
attribute_type = type(attribute)
if attribute_type in (WDURLAttribute, WDArticle):
# get_select() method : there is group_concat(distinct ...;separator=", ")
# split the value here
for url in value.split(', '):
infobox_urls.append({'title': attribute.get_label(language), 'url': url, **attribute.kwargs})
# "normal" results (not infobox) include official website and Wikipedia links.
if "list" in display_type and (attribute.kwargs.get('official') or attribute_type == WDArticle):
2023-03-21 00:41:36 +01:00
results.append({'title': infobox_title, 'url': url, "content": infobox_content})
# update the infobox_id with the wikipedia URL
# first the local wikipedia URL, and as fallback the english wikipedia URL
if attribute_type == WDArticle and (
(attribute.language == 'en' and infobox_id_lang is None) or attribute.language != 'en'
):
infobox_id_lang = attribute.language
infobox_id = url
elif attribute_type == WDImageAttribute:
# this attribute is an image.
# replace the current image only the priority is lower
# (the infobox contain only one image).
if attribute.priority > img_src_priority:
img_src = get_thumbnail(value)
img_src_priority = attribute.priority
elif attribute_type == WDGeoAttribute:
# geocoordinate link
# use the area to get the OSM zoom
2023-09-15 09:53:03 +02:00
# Note: ignore the unit (must be km² otherwise the calculation is wrong)
# Should use normalized value p:P2046/psn:P2046/wikibase:quantityAmount
area = attribute_result.get('P2046')
osm_zoom = area_to_osm_zoom(area) if area else 19
url = attribute.get_geo_url(attribute_result, osm_zoom=osm_zoom)
if url:
infobox_urls.append({'title': attribute.get_label(language), 'url': url, 'entity': attribute.name})
else:
infobox_attributes.append(
{'label': attribute.get_label(language), 'value': value, 'entity': attribute.name}
)
if infobox_id:
infobox_id = replace_http_by_https(infobox_id)
# add the wikidata URL at the end
infobox_urls.append({'title': 'Wikidata', 'url': attribute_result['item']})
if (
"list" in display_type
and img_src is None
and len(infobox_attributes) == 0
and len(infobox_urls) == 1
and len(infobox_content) == 0
):
results.append({'url': infobox_urls[0]['url'], 'title': infobox_title, 'content': infobox_content})
elif "infobox" in display_type:
results.append(
{
'infobox': infobox_title,
'id': infobox_id,
'content': infobox_content,
'img_src': img_src,
'urls': infobox_urls,
'attributes': infobox_attributes,
}
)
2014-09-28 16:51:41 +02:00
return results
def get_query(query, language):
attributes = get_attributes(language)
select = [a.get_select() for a in attributes]
where = list(filter(lambda s: len(s) > 0, [a.get_where() for a in attributes]))
wikibase_label = list(filter(lambda s: len(s) > 0, [a.get_wikibase_label() for a in attributes]))
group_by = list(filter(lambda s: len(s) > 0, [a.get_group_by() for a in attributes]))
query = (
QUERY_TEMPLATE.replace('%QUERY%', sparql_string_escape(query))
.replace('%SELECT%', ' '.join(select))
.replace('%WHERE%', '\n '.join(where))
.replace('%WIKIBASE_LABELS%', '\n '.join(wikibase_label))
.replace('%GROUP_BY%', ' '.join(group_by))
.replace('%LANGUAGE%', language)
)
return query, attributes
def get_attributes(language):
# pylint: disable=too-many-statements
attributes = []
def add_value(name):
attributes.append(WDAttribute(name))
def add_amount(name):
attributes.append(WDAmountAttribute(name))
def add_label(name):
attributes.append(WDLabelAttribute(name))
def add_url(name, url_id=None, **kwargs):
attributes.append(WDURLAttribute(name, url_id, kwargs))
def add_image(name, url_id=None, priority=1):
attributes.append(WDImageAttribute(name, url_id, priority))
def add_date(name):
attributes.append(WDDateAttribute(name))
# Dates
for p in [
'P571', # inception date
'P576', # dissolution date
'P580', # start date
'P582', # end date
'P569', # date of birth
'P570', # date of death
'P619', # date of spacecraft launch
'P620',
]: # date of spacecraft landing
add_date(p)
for p in [
'P27', # country of citizenship
'P495', # country of origin
'P17', # country
'P159',
]: # headquarters location
add_label(p)
# Places
for p in [
'P36', # capital
'P35', # head of state
'P6', # head of government
'P122', # basic form of government
'P37',
]: # official language
add_label(p)
add_value('P1082') # population
add_amount('P2046') # area
add_amount('P281') # postal code
add_label('P38') # currency
add_amount('P2048') # height (building)
# Media
for p in [
'P400', # platform (videogames, computing)
'P50', # author
'P170', # creator
'P57', # director
'P175', # performer
'P178', # developer
'P162', # producer
'P176', # manufacturer
'P58', # screenwriter
'P272', # production company
'P264', # record label
'P123', # publisher
'P449', # original network
'P750', # distributed by
'P86',
]: # composer
add_label(p)
add_date('P577') # publication date
add_label('P136') # genre (music, film, artistic...)
add_label('P364') # original language
add_value('P212') # ISBN-13
add_value('P957') # ISBN-10
add_label('P275') # copyright license
add_label('P277') # programming language
add_value('P348') # version
add_label('P840') # narrative location
# Languages
add_value('P1098') # number of speakers
add_label('P282') # writing system
add_label('P1018') # language regulatory body
add_value('P218') # language code (ISO 639-1)
# Other
add_label('P169') # ceo
add_label('P112') # founded by
add_label('P1454') # legal form (company, organization)
add_label('P137') # operator (service, facility, ...)
add_label('P1029') # crew members (tripulation)
add_label('P225') # taxon name
add_value('P274') # chemical formula
add_label('P1346') # winner (sports, contests, ...)
add_value('P1120') # number of deaths
add_value('P498') # currency code (ISO 4217)
# URL
add_url('P856', official=True) # official website
attributes.append(WDArticle(language)) # wikipedia (user language)
if not language.startswith('en'):
attributes.append(WDArticle('en')) # wikipedia (english)
add_url('P1324') # source code repository
add_url('P1581') # blog
add_url('P434', url_id='musicbrainz_artist')
add_url('P435', url_id='musicbrainz_work')
add_url('P436', url_id='musicbrainz_release_group')
add_url('P966', url_id='musicbrainz_label')
add_url('P345', url_id='imdb_id')
add_url('P2397', url_id='youtube_channel')
add_url('P1651', url_id='youtube_video')
add_url('P2002', url_id='twitter_profile')
add_url('P2013', url_id='facebook_profile')
add_url('P2003', url_id='instagram_profile')
# Map
attributes.append(WDGeoAttribute('P625'))
# Image
add_image('P15', priority=1, url_id='wikimedia_image') # route map
add_image('P242', priority=2, url_id='wikimedia_image') # locator map
add_image('P154', priority=3, url_id='wikimedia_image') # logo
add_image('P18', priority=4, url_id='wikimedia_image') # image
add_image('P41', priority=5, url_id='wikimedia_image') # flag
add_image('P2716', priority=6, url_id='wikimedia_image') # collage
add_image('P2910', priority=7, url_id='wikimedia_image') # icon
return attributes
class WDAttribute:
__slots__ = ('name',)
def __init__(self, name):
self.name = name
def get_select(self):
return '(group_concat(distinct ?{name};separator=", ") as ?{name}s)'.replace('{name}', self.name)
def get_label(self, language):
return get_label_for_entity(self.name, language)
def get_where(self):
return "OPTIONAL { ?item wdt:{name} ?{name} . }".replace('{name}', self.name)
def get_wikibase_label(self):
return ""
def get_group_by(self):
return ""
def get_str(self, result, language): # pylint: disable=unused-argument
return result.get(self.name + 's')
def __repr__(self):
return '<' + str(type(self).__name__) + ':' + self.name + '>'
class WDAmountAttribute(WDAttribute):
def get_select(self):
return '?{name} ?{name}Unit'.replace('{name}', self.name)
def get_where(self):
return """ OPTIONAL { ?item p:{name} ?{name}Node .
?{name}Node rdf:type wikibase:BestRank ; ps:{name} ?{name} .
OPTIONAL { ?{name}Node psv:{name}/wikibase:quantityUnit ?{name}Unit. } }""".replace(
'{name}', self.name
)
def get_group_by(self):
return self.get_select()
def get_str(self, result, language):
value = result.get(self.name)
unit = result.get(self.name + "Unit")
if unit is not None:
unit = unit.replace('http://www.wikidata.org/entity/', '')
return value + " " + get_label_for_entity(unit, language)
return value
class WDArticle(WDAttribute):
__slots__ = 'language', 'kwargs'
def __init__(self, language, kwargs=None):
super().__init__('wikipedia')
self.language = language
self.kwargs = kwargs or {}
def get_label(self, language):
# language parameter is ignored
return "Wikipedia ({language})".replace('{language}', self.language)
def get_select(self):
return "?article{language} ?articleName{language}".replace('{language}', self.language)
def get_where(self):
return """OPTIONAL { ?article{language} schema:about ?item ;
schema:inLanguage "{language}" ;
schema:isPartOf <https://{language}.wikipedia.org/> ;
schema:name ?articleName{language} . }""".replace(
'{language}', self.language
)
def get_group_by(self):
return self.get_select()
def get_str(self, result, language):
key = 'article{language}'.replace('{language}', self.language)
return result.get(key)
class WDLabelAttribute(WDAttribute):
def get_select(self):
return '(group_concat(distinct ?{name}Label;separator=", ") as ?{name}Labels)'.replace('{name}', self.name)
2014-09-28 16:51:41 +02:00
def get_where(self):
return "OPTIONAL { ?item wdt:{name} ?{name} . }".replace('{name}', self.name)
2014-09-28 16:51:41 +02:00
def get_wikibase_label(self):
return "?{name} rdfs:label ?{name}Label .".replace('{name}', self.name)
2014-09-28 16:51:41 +02:00
def get_str(self, result, language):
return result.get(self.name + 'Labels')
class WDURLAttribute(WDAttribute):
HTTP_WIKIMEDIA_IMAGE = 'http://commons.wikimedia.org/wiki/Special:FilePath/'
__slots__ = 'url_id', 'kwargs'
def __init__(self, name, url_id=None, kwargs=None):
super().__init__(name)
self.url_id = url_id
self.kwargs = kwargs
def get_str(self, result, language):
value = result.get(self.name + 's')
if self.url_id and value is not None and value != '':
value = value.split(',')[0]
url_id = self.url_id
if value.startswith(WDURLAttribute.HTTP_WIKIMEDIA_IMAGE):
value = value[len(WDURLAttribute.HTTP_WIKIMEDIA_IMAGE) :]
url_id = 'wikimedia_image'
return get_external_url(url_id, value)
return value
class WDGeoAttribute(WDAttribute):
def get_label(self, language):
return "OpenStreetMap"
def get_select(self):
return "?{name}Lat ?{name}Long".replace('{name}', self.name)
def get_where(self):
return """OPTIONAL { ?item p:{name}/psv:{name} [
wikibase:geoLatitude ?{name}Lat ;
wikibase:geoLongitude ?{name}Long ] }""".replace(
'{name}', self.name
)
def get_group_by(self):
return self.get_select()
def get_str(self, result, language):
latitude = result.get(self.name + 'Lat')
longitude = result.get(self.name + 'Long')
if latitude and longitude:
return latitude + ' ' + longitude
return None
def get_geo_url(self, result, osm_zoom=19):
latitude = result.get(self.name + 'Lat')
longitude = result.get(self.name + 'Long')
if latitude and longitude:
return get_earth_coordinates_url(latitude, longitude, osm_zoom)
return None
class WDImageAttribute(WDURLAttribute):
__slots__ = ('priority',)
def __init__(self, name, url_id=None, priority=100):
super().__init__(name, url_id)
self.priority = priority
class WDDateAttribute(WDAttribute):
def get_select(self):
return '?{name} ?{name}timePrecision ?{name}timeZone ?{name}timeCalendar'.replace('{name}', self.name)
def get_where(self):
# To remove duplicate, add
# FILTER NOT EXISTS { ?item p:{name}/psv:{name}/wikibase:timeValue ?{name}bis FILTER (?{name}bis < ?{name}) }
# this filter is too slow, so the response function ignore duplicate results
# (see the seen_entities variable)
return """OPTIONAL { ?item p:{name}/psv:{name} [
wikibase:timeValue ?{name} ;
wikibase:timePrecision ?{name}timePrecision ;
wikibase:timeTimezone ?{name}timeZone ;
wikibase:timeCalendarModel ?{name}timeCalendar ] . }
hint:Prior hint:rangeSafe true;""".replace(
'{name}', self.name
)
def get_group_by(self):
return self.get_select()
def format_8(self, value, locale): # pylint: disable=unused-argument
# precision: less than a year
return value
def format_9(self, value, locale):
year = int(value)
# precision: year
if year < 1584:
if year < 0:
return str(year - 1)
return str(year)
timestamp = isoparse(value)
return format_date(timestamp, format='yyyy', locale=locale)
def format_10(self, value, locale):
# precision: month
timestamp = isoparse(value)
return format_date(timestamp, format='MMMM y', locale=locale)
def format_11(self, value, locale):
# precision: day
timestamp = isoparse(value)
return format_date(timestamp, format='full', locale=locale)
def format_13(self, value, locale):
timestamp = isoparse(value)
# precision: minute
return (
get_datetime_format(format, locale=locale)
.replace("'", "")
.replace('{0}', format_time(timestamp, 'full', tzinfo=None, locale=locale))
.replace('{1}', format_date(timestamp, 'short', locale=locale))
)
def format_14(self, value, locale):
# precision: second.
return format_datetime(isoparse(value), format='full', locale=locale)
DATE_FORMAT = {
'0': ('format_8', 1000000000),
'1': ('format_8', 100000000),
'2': ('format_8', 10000000),
'3': ('format_8', 1000000),
'4': ('format_8', 100000),
'5': ('format_8', 10000),
'6': ('format_8', 1000),
'7': ('format_8', 100),
'8': ('format_8', 10),
'9': ('format_9', 1), # year
'10': ('format_10', 1), # month
'11': ('format_11', 0), # day
'12': ('format_13', 0), # hour (not supported by babel, display minute)
'13': ('format_13', 0), # minute
'14': ('format_14', 0), # second
}
def get_str(self, result, language):
value = result.get(self.name)
if value == '' or value is None:
return None
precision = result.get(self.name + 'timePrecision')
date_format = WDDateAttribute.DATE_FORMAT.get(precision)
if date_format is not None:
format_method = getattr(self, date_format[0])
precision = date_format[1]
try:
if precision >= 1:
t = value.split('-')
if value.startswith('-'):
value = '-' + t[1]
else:
value = t[0]
return format_method(value, language)
except Exception: # pylint: disable=broad-except
return value
return value
def debug_explain_wikidata_query(query, method='GET'):
if method == 'GET':
http_response = get(SPARQL_EXPLAIN_URL + '&' + urlencode({'query': query}), headers=get_headers())
else:
http_response = post(SPARQL_EXPLAIN_URL, data={'query': query}, headers=get_headers())
http_response.raise_for_status()
return http_response.content
def init(engine_settings=None): # pylint: disable=unused-argument
# WIKIDATA_PROPERTIES : add unit symbols
for k, v in WIKIDATA_UNITS.items():
WIKIDATA_PROPERTIES[k] = v['symbol']
# WIKIDATA_PROPERTIES : add property labels
wikidata_property_names = []
for attribute in get_attributes('en'):
if type(attribute) in (WDAttribute, WDAmountAttribute, WDURLAttribute, WDDateAttribute, WDLabelAttribute):
if attribute.name not in WIKIDATA_PROPERTIES:
wikidata_property_names.append("wd:" + attribute.name)
query = QUERY_PROPERTY_NAMES.replace('%ATTRIBUTES%', " ".join(wikidata_property_names))
jsonresponse = send_wikidata_query(query)
for result in jsonresponse.get('results', {}).get('bindings', {}):
name = result['name']['value']
lang = result['name']['xml:lang']
entity_id = result['item']['value'].replace('http://www.wikidata.org/entity/', '')
WIKIDATA_PROPERTIES[(entity_id, lang)] = name.capitalize()
def fetch_traits(engine_traits: EngineTraits):
[fix] searxng_extra/update/update_engine_descriptions.py (part 1) Follow up of #2269 The script to update the descriptions of the engines does no longer work since PR #2269 has been merged. searx/engines/wikipedia.py ========================== 1. There was a misusage of zh-classical.wikipedia.org: - `zh-classical` is dedicate to classical Chinese [1] which is not traditional Chinese [2]. - zh.wikipedia.org has LanguageConverter enabled [3] and is going to dynamically show simplified or traditional Chinese according to the HTTP Accept-Language header. 2. The update_engine_descriptions.py needs a list of all wikipedias. The implementation from #2269 included only a reduced list: - https://meta.wikimedia.org/wiki/Wikipedia_article_depth - https://meta.wikimedia.org/wiki/List_of_Wikipedias searxng_extra/update/update_engine_descriptions.py ================================================== Before PR #2269 there was a match_language() function that did an approximation using various methods. With PR #2269 there are only the types in the data model of the languages, which can be recognized by babel. The approximation methods, which are needed (only here) in the determination of the descriptions, must be replaced by other methods. [1] https://en.wikipedia.org/wiki/Classical_Chinese [2] https://en.wikipedia.org/wiki/Traditional_Chinese_characters [3] https://www.mediawiki.org/wiki/Writing_systems#LanguageConverter Closes: https://github.com/searxng/searxng/issues/2330 Signed-off-by: Markus Heiser <markus.heiser@darmarit.de>
2023-04-04 15:17:12 +02:00
"""Uses languages evaluated from :py:obj:`wikipedia.fetch_wikimedia_traits
<searx.engines.wikipedia.fetch_wikimedia_traits>` and removes
- ``traits.custom['wiki_netloc']``: wikidata does not have net-locations for
the languages and the list of all
- ``traits.custom['WIKIPEDIA_LANGUAGES']``: not used in the wikipedia engine
"""
fetch_wikimedia_traits(engine_traits)
engine_traits.custom['wiki_netloc'] = {}
[fix] searxng_extra/update/update_engine_descriptions.py (part 1) Follow up of #2269 The script to update the descriptions of the engines does no longer work since PR #2269 has been merged. searx/engines/wikipedia.py ========================== 1. There was a misusage of zh-classical.wikipedia.org: - `zh-classical` is dedicate to classical Chinese [1] which is not traditional Chinese [2]. - zh.wikipedia.org has LanguageConverter enabled [3] and is going to dynamically show simplified or traditional Chinese according to the HTTP Accept-Language header. 2. The update_engine_descriptions.py needs a list of all wikipedias. The implementation from #2269 included only a reduced list: - https://meta.wikimedia.org/wiki/Wikipedia_article_depth - https://meta.wikimedia.org/wiki/List_of_Wikipedias searxng_extra/update/update_engine_descriptions.py ================================================== Before PR #2269 there was a match_language() function that did an approximation using various methods. With PR #2269 there are only the types in the data model of the languages, which can be recognized by babel. The approximation methods, which are needed (only here) in the determination of the descriptions, must be replaced by other methods. [1] https://en.wikipedia.org/wiki/Classical_Chinese [2] https://en.wikipedia.org/wiki/Traditional_Chinese_characters [3] https://www.mediawiki.org/wiki/Writing_systems#LanguageConverter Closes: https://github.com/searxng/searxng/issues/2330 Signed-off-by: Markus Heiser <markus.heiser@darmarit.de>
2023-04-04 15:17:12 +02:00
engine_traits.custom['WIKIPEDIA_LANGUAGES'] = []