searxng/searx/engines/bing_news.py
Markus Heiser d0f465e6fa [mod] bing: add time_range support & upgrade to data_type: traits_v1
Signed-off-by: Markus Heiser <markus.heiser@darmarit.de>
2023-03-24 10:37:42 +01:00

156 lines
4.8 KiB
Python

# SPDX-License-Identifier: AGPL-3.0-or-later
# lint: pylint
"""Bing (News)
"""
from urllib.parse import (
urlencode,
urlparse,
parse_qsl,
quote,
)
from datetime import datetime
from dateutil import parser
from lxml import etree
from lxml.etree import XPath
from searx.utils import match_language, eval_xpath_getindex
from searx.enginelib.traits import EngineTraits
# about
about = {
"website": 'https://www.bing.com/news',
"wikidata_id": 'Q2878637',
"official_api_documentation": 'https://www.microsoft.com/en-us/bing/apis/bing-news-search-api',
"use_official_api": False,
"require_api_key": False,
"results": 'RSS',
}
# engine dependent config
categories = ['news']
paging = True
time_range_support = True
send_accept_language_header = True
# search-url
base_url = 'https://www.bing.com/'
search_string = 'news/search?{query}&first={offset}&format=RSS'
search_string_with_time = 'news/search?{query}&first={offset}&qft=interval%3d"{interval}"&format=RSS'
time_range_dict = {'day': '7', 'week': '8', 'month': '9'}
def url_cleanup(url_string):
"""remove click"""
parsed_url = urlparse(url_string)
if parsed_url.netloc == 'www.bing.com' and parsed_url.path == '/news/apiclick.aspx':
query = dict(parse_qsl(parsed_url.query))
url_string = query.get('url', None)
return url_string
def image_url_cleanup(url_string):
"""replace the http://*bing.com/th?id=... by https://www.bing.com/th?id=..."""
parsed_url = urlparse(url_string)
if parsed_url.netloc.endswith('bing.com') and parsed_url.path == '/th':
query = dict(parse_qsl(parsed_url.query))
url_string = "https://www.bing.com/th?id=" + quote(query.get('id'))
return url_string
def _get_url(query, language, offset, time_range):
if time_range in time_range_dict:
search_path = search_string_with_time.format(
# fmt: off
query = urlencode({
'q': query,
'setmkt': language
}),
offset = offset,
interval = time_range_dict[time_range]
# fmt: on
)
else:
# e.g. setmkt=de-de&setlang=de
search_path = search_string.format(
# fmt: off
query = urlencode({
'q': query,
'setmkt': language
}),
offset = offset
# fmt: on
)
return base_url + search_path
def request(query, params):
if params['time_range'] and params['time_range'] not in time_range_dict:
return params
offset = (params['pageno'] - 1) * 10 + 1
if params['language'] == 'all':
language = 'en-US'
else:
language = match_language(params['language'], supported_languages, language_aliases)
params['url'] = _get_url(query, language, offset, params['time_range'])
return params
def response(resp):
results = []
rss = etree.fromstring(resp.content)
namespaces = rss.nsmap
for item in rss.xpath('./channel/item'):
# url / title / content
url = url_cleanup(eval_xpath_getindex(item, './link/text()', 0, default=None))
title = eval_xpath_getindex(item, './title/text()', 0, default=url)
content = eval_xpath_getindex(item, './description/text()', 0, default='')
# publishedDate
publishedDate = eval_xpath_getindex(item, './pubDate/text()', 0, default=None)
try:
publishedDate = parser.parse(publishedDate, dayfirst=False)
except TypeError:
publishedDate = datetime.now()
except ValueError:
publishedDate = datetime.now()
# thumbnail
thumbnail = eval_xpath_getindex(item, XPath('./News:Image/text()', namespaces=namespaces), 0, default=None)
if thumbnail is not None:
thumbnail = image_url_cleanup(thumbnail)
# append result
if thumbnail is not None:
results.append(
{'url': url, 'title': title, 'publishedDate': publishedDate, 'content': content, 'img_src': thumbnail}
)
else:
results.append({'url': url, 'title': title, 'publishedDate': publishedDate, 'content': content})
return results
def fetch_traits(engine_traits: EngineTraits):
"""Fetch languages and regions from Bing-News."""
# pylint: disable=import-outside-toplevel
from searx.engines.bing import _fetch_traits
url = 'https://learn.microsoft.com/en-us/bing/search-apis/bing-news-search/reference/market-codes'
# The description of the first table says "query parameter when calling the
# Video Search API." .. thats why I use the 4. table "News Category API markets"
xpath_market_codes = '//table[4]/tbody/tr/td[3]'
# xpath_country_codes = '//table[2]/tbody/tr/td[2]'
xpath_language_codes = '//table[3]/tbody/tr/td[2]'
_fetch_traits(engine_traits, url, xpath_language_codes, xpath_market_codes)