Merge pull request #540 from a01200356/wikipedia_infobox

[enh] wikipedia infobox
This commit is contained in:
Adam Tauber 2016-04-19 20:40:48 +02:00
commit f46057feb2
9 changed files with 308 additions and 18 deletions

View File

@ -43,3 +43,4 @@ generally made searx better:
- Kang-min Liu
- Kirill Isakov
- Guilhem Bonnefille
- Marc Abonce Seguin

View File

@ -1,5 +1,6 @@
import json
from urllib import urlencode
from re import compile, sub
from lxml import html
from searx.utils import html_to_text
from searx.engines.xpath import extract_text
@ -7,6 +8,8 @@ from searx.engines.xpath import extract_text
url = 'https://api.duckduckgo.com/'\
+ '?{query}&format=json&pretty=0&no_redirect=1&d=1'
http_regex = compile(r'^http:')
def result_to_text(url, text, htmlResult):
# TODO : remove result ending with "Meaning" or "Category"
@ -19,8 +22,8 @@ def result_to_text(url, text, htmlResult):
def request(query, params):
# TODO add kl={locale}
params['url'] = url.format(query=urlencode({'q': query}))
params['headers']['Accept-Language'] = params['language']
return params
@ -103,6 +106,10 @@ def response(resp):
urls.append({'title': search_res.get('DefinitionSource'),
'url': definitionURL})
# to merge with wikidata's infobox
if infobox_id:
infobox_id = http_regex.sub('https:', infobox_id)
# entity
entity = search_res.get('Entity', None)
# TODO continent / country / department / location / waterfall /

View File

@ -86,11 +86,11 @@ def getDetail(jsonresponse, wikidata_id, language, locale):
results.append({'title': title, 'url': official_website})
wikipedia_link_count = 0
if language != 'en':
wikipedia_link = get_wikilink(result, language + 'wiki')
wikipedia_link_count += add_url(urls,
'Wikipedia (' + language + ')',
get_wikilink(result, language +
'wiki'))
wikipedia_link)
if language != 'en':
wikipedia_en_link = get_wikilink(result, 'enwiki')
wikipedia_link_count += add_url(urls,
'Wikipedia (en)',
@ -188,7 +188,7 @@ def getDetail(jsonresponse, wikidata_id, language, locale):
else:
results.append({
'infobox': title,
'id': wikipedia_en_link,
'id': wikipedia_link,
'content': description,
'attributes': attributes,
'urls': urls

114
searx/engines/wikipedia.py Normal file
View File

@ -0,0 +1,114 @@
"""
Wikipedia (Web)
@website https://{language}.wikipedia.org
@provide-api yes
@using-api yes
@results JSON
@stable yes
@parse url, infobox
"""
from json import loads
from urllib import urlencode, quote
# search-url
base_url = 'https://{language}.wikipedia.org/'
search_postfix = 'w/api.php?'\
'action=query'\
'&format=json'\
'&{query}'\
'&prop=extracts|pageimages'\
'&exintro'\
'&explaintext'\
'&pithumbsize=300'\
'&redirects'
# set language in base_url
def url_lang(lang):
if lang == 'all':
language = 'en'
else:
language = lang.split('_')[0]
return base_url.format(language=language)
# do search-request
def request(query, params):
if query.islower():
query += '|' + query.title()
params['url'] = url_lang(params['language']) \
+ search_postfix.format(query=urlencode({'titles': query}))
return params
# get first meaningful paragraph
# this should filter out disambiguation pages and notes above first paragraph
# "magic numbers" were obtained by fine tuning
def extract_first_paragraph(content, title, image):
first_paragraph = None
failed_attempts = 0
for paragraph in content.split('\n'):
starts_with_title = paragraph.lower().find(title.lower(), 0, len(title) + 35)
length = len(paragraph)
if length >= 200 or (starts_with_title >= 0 and (image or length >= 150)):
first_paragraph = paragraph
break
failed_attempts += 1
if failed_attempts > 3:
return None
return first_paragraph
# get response from search-request
def response(resp):
results = []
search_result = loads(resp.content)
# wikipedia article's unique id
# first valid id is assumed to be the requested article
for article_id in search_result['query']['pages']:
page = search_result['query']['pages'][article_id]
if int(article_id) > 0:
break
if int(article_id) < 0:
return []
title = page.get('title')
image = page.get('thumbnail')
if image:
image = image.get('source')
extract = page.get('extract')
summary = extract_first_paragraph(extract, title, image)
if not summary:
return []
# link to wikipedia article
# parenthesis are not quoted to make infobox mergeable with wikidata's
wikipedia_link = url_lang(resp.search_params['language']) \
+ 'wiki/' + quote(title.replace(' ', '_').encode('utf8')).replace('%28', '(').replace('%29', ')')
results.append({'url': wikipedia_link, 'title': title})
results.append({'infobox': title,
'id': wikipedia_link,
'content': summary,
'img_src': image,
'urls': [{'title': 'Wikipedia', 'url': wikipedia_link}]})
return results

View File

@ -37,7 +37,7 @@ def merge_two_infoboxes(infobox1, infobox2):
urls1 = infobox1.get('urls', None)
if urls1 is None:
urls1 = []
infobox1.set('urls', urls1)
infobox1['urls'] = urls1
urlSet = set()
for url in infobox1.get('urls', []):
@ -47,11 +47,17 @@ def merge_two_infoboxes(infobox1, infobox2):
if url.get('url', None) not in urlSet:
urls1.append(url)
if 'img_src' in infobox2:
img1 = infobox1.get('img_src', None)
img2 = infobox2.get('img_src')
if img1 is None:
infobox1['img_src'] = img2
if 'attributes' in infobox2:
attributes1 = infobox1.get('attributes', None)
if attributes1 is None:
attributes1 = []
infobox1.set('attributes', attributes1)
infobox1['attributes'] = attributes1
attributeSet = set()
for attribute in infobox1.get('attributes', []):
@ -68,7 +74,7 @@ def merge_two_infoboxes(infobox1, infobox2):
if result_content_len(content2) > result_content_len(content1):
infobox1['content'] = content2
else:
infobox1.set('content', content2)
infobox1['content'] = content2
def result_score(result):

View File

@ -43,10 +43,9 @@ engines:
shortcut : bs
- name : wikipedia
engine : mediawiki
engine : wikipedia
shortcut : wp
base_url : 'https://{language}.wikipedia.org/'
number_of_results : 1
- name : bing
engine : bing
@ -93,6 +92,7 @@ engines:
- name : ddg definitions
engine : duckduckgo_definitions
shortcut : ddd
disabled : True
- name : digg
engine : digg

View File

@ -1,8 +1,9 @@
<div class="panel panel-default infobox">
<div class="panel-heading">
<h4 class="panel-title infobox_part">{{ infobox.infobox }}</h4>
<bdi><h4 class="panel-title infobox_part">{{ infobox.infobox }}</h4></bdi>
</div>
<div class="panel-body">
<bdi>
{% if infobox.img_src %}<img class="img-responsive center-block infobox_part" src="{{ image_proxify(infobox.img_src) }}" alt="{{ infobox.infobox }}" />{% endif %}
{% if infobox.content %}<p class="infobox_part">{{ infobox.content }}</p>{% endif %}
@ -28,5 +29,6 @@
{% endfor %}
</div>
{% endif %}
</bdi>
</div>
</div>

View File

@ -123,7 +123,7 @@ class TestDDGDefinitionsEngine(SearxTestCase):
self.assertEqual(results[1]['url'], 'result first url')
self.assertEqual(results[2]['suggestion'], 'text')
self.assertEqual(results[3]['infobox'], 'heading')
self.assertEqual(results[3]['id'], 'http://definition.url')
self.assertEqual(results[3]['id'], 'https://definition.url')
self.assertEqual(results[3]['entity'], 'Entity')
self.assertIn('abstract', results[3]['content'])
self.assertIn('this is the definition', results[3]['content'])
@ -240,7 +240,7 @@ class TestDDGDefinitionsEngine(SearxTestCase):
self.assertEqual(type(results), list)
self.assertEqual(len(results), 1)
self.assertEqual(results[0]['infobox'], 'heading')
self.assertEqual(results[0]['id'], 'http://definition.url')
self.assertEqual(results[0]['id'], 'https://definition.url')
self.assertEqual(results[0]['entity'], 'Entity')
self.assertIn('abstract', results[0]['content'])
self.assertIn('this is the definition', results[0]['content'])

View File

@ -0,0 +1,160 @@
# -*- coding: utf-8 -*-
from collections import defaultdict
import mock
from searx.engines import wikipedia
from searx.testing import SearxTestCase
class TestWikipediaEngine(SearxTestCase):
def test_request(self):
query = 'test_query'
dicto = defaultdict(dict)
dicto['language'] = 'fr_FR'
params = wikipedia.request(query, dicto)
self.assertIn('url', params)
self.assertIn(query, params['url'])
self.assertIn('test_query', params['url'])
self.assertIn('Test_Query', params['url'])
self.assertIn('fr.wikipedia.org', params['url'])
query = 'Test_Query'
params = wikipedia.request(query, dicto)
self.assertIn('Test_Query', params['url'])
self.assertNotIn('test_query', params['url'])
dicto['language'] = 'all'
params = wikipedia.request(query, dicto)
self.assertIn('en', params['url'])
def test_response(self):
dicto = defaultdict(dict)
dicto['language'] = 'fr'
self.assertRaises(AttributeError, wikipedia.response, None)
self.assertRaises(AttributeError, wikipedia.response, [])
self.assertRaises(AttributeError, wikipedia.response, '')
self.assertRaises(AttributeError, wikipedia.response, '[]')
# page not found
json = """
{
"batchcomplete": "",
"query": {
"normalized": [],
"pages": {
"-1": {
"ns": 0,
"title": "",
"missing": ""
}
}
}
}"""
response = mock.Mock(content=json, search_params=dicto)
self.assertEqual(wikipedia.response(response), [])
# normal case
json = """
{
"batchcomplete": "",
"query": {
"normalized": [],
"pages": {
"12345": {
"pageid": 12345,
"ns": 0,
"title": "The Title",
"extract": "The Title is...",
"thumbnail": {
"source": "img_src.jpg"
},
"pageimage": "img_name.jpg"
}
}
}
}"""
response = mock.Mock(content=json, search_params=dicto)
results = wikipedia.response(response)
self.assertEqual(type(results), list)
self.assertEqual(len(results), 2)
self.assertEqual(results[0]['title'], u'The Title')
self.assertIn('fr.wikipedia.org/wiki/The_Title', results[0]['url'])
self.assertEqual(results[1]['infobox'], u'The Title')
self.assertIn('fr.wikipedia.org/wiki/The_Title', results[1]['id'])
self.assertIn('The Title is...', results[1]['content'])
self.assertEqual(results[1]['img_src'], 'img_src.jpg')
# disambiguation page
json = """
{
"batchcomplete": "",
"query": {
"normalized": [],
"pages": {
"12345": {
"pageid": 12345,
"ns": 0,
"title": "The Title",
"extract": "The Title can be:\\nThe Title 1\\nThe Title 2\\nThe Title 3\\nThe Title 4......................................................................................................................................." """ # noqa
json += """
}
}
}
}"""
response = mock.Mock(content=json, search_params=dicto)
results = wikipedia.response(response)
self.assertEqual(type(results), list)
self.assertEqual(len(results), 0)
# no image
json = """
{
"batchcomplete": "",
"query": {
"normalized": [],
"pages": {
"12345": {
"pageid": 12345,
"ns": 0,
"title": "The Title",
"extract": "The Title is......................................................................................................................................................................................." """ # noqa
json += """
}
}
}
}"""
response = mock.Mock(content=json, search_params=dicto)
results = wikipedia.response(response)
self.assertEqual(type(results), list)
self.assertEqual(len(results), 2)
self.assertIn('The Title is...', results[1]['content'])
self.assertEqual(results[1]['img_src'], None)
# title not in first paragraph
json = u"""
{
"batchcomplete": "",
"query": {
"normalized": [],
"pages": {
"12345": {
"pageid": 12345,
"ns": 0,
"title": "披頭四樂隊",
"extract": "披头士乐队....................................................................................................................................................................................................\\n披頭四樂隊...", """ # noqa
json += """
"thumbnail": {
"source": "img_src.jpg"
},
"pageimage": "img_name.jpg"
}
}
}
}"""
response = mock.Mock(content=json, search_params=dicto)
results = wikipedia.response(response)
self.assertEqual(type(results), list)
self.assertEqual(len(results), 2)
self.assertEqual(results[1]['infobox'], u'披頭四樂隊')
self.assertIn(u'披头士乐队...', results[1]['content'])