1
0
mirror of https://github.com/searxng/searxng.git synced 2024-11-22 20:17:45 +01:00
searxng/searx/engines/archlinux.py
Markus Heiser e8706fb738 [fix] engine & network issues / documentation and type annotations
This patch fixes some quirks and issues related to the engines and the network.
Each engine has its own network and this network was broken for the following
engines[1]:

- archlinux
- bing
- dailymotion
- duckduckgo
- google
- peertube
- startpage
- wikipedia

Since the files have been touched anyway, the type annotaions of the engine
modules has also been completed so that error messages from the type checker are
no longer reported.

Related and (partial) fixed issue:

- [1] https://github.com/searxng/searxng/issues/762#issuecomment-1605323861
- [2] https://github.com/searxng/searxng/issues/2513
- [3] https://github.com/searxng/searxng/issues/2515

Signed-off-by: Markus Heiser <markus.heiser@darmarit.de>
2023-06-25 13:58:26 +02:00

153 lines
4.5 KiB
Python

# SPDX-License-Identifier: AGPL-3.0-or-later
# lint: pylint
"""
Arch Linux Wiki
~~~~~~~~~~~~~~~
This implementation does not use a official API: Mediawiki provides API, but
Arch Wiki blocks access to it.
"""
from typing import TYPE_CHECKING
from urllib.parse import urlencode, urljoin, urlparse
import lxml
import babel
from searx.utils import extract_text, eval_xpath_list, eval_xpath_getindex
from searx.enginelib.traits import EngineTraits
from searx.locales import language_tag
if TYPE_CHECKING:
import logging
logger: logging.Logger
traits: EngineTraits
about = {
"website": 'https://wiki.archlinux.org/',
"wikidata_id": 'Q101445877',
"official_api_documentation": None,
"use_official_api": False,
"require_api_key": False,
"results": 'HTML',
}
# engine dependent config
categories = ['it', 'software wikis']
paging = True
main_wiki = 'wiki.archlinux.org'
def request(query, params):
sxng_lang = params['searxng_locale'].split('-')[0]
netloc: str = traits.custom['wiki_netloc'].get(sxng_lang, main_wiki) # type: ignore
title: str = traits.custom['title'].get(sxng_lang, 'Special:Search') # type: ignore
base_url = 'https://' + netloc + '/index.php?'
offset = (params['pageno'] - 1) * 20
if netloc == main_wiki:
eng_lang: str = traits.get_language(sxng_lang, 'English') # type: ignore
query += ' (' + eng_lang + ')'
elif netloc == 'wiki.archlinuxcn.org':
base_url = 'https://' + netloc + '/wzh/index.php?'
args = {
'search': query,
'title': title,
'limit': 20,
'offset': offset,
'profile': 'default',
}
params['url'] = base_url + urlencode(args)
return params
def response(resp):
results = []
dom = lxml.html.fromstring(resp.text) # type: ignore
# get the base URL for the language in which request was made
sxng_lang = resp.search_params['searxng_locale'].split('-')[0]
netloc: str = traits.custom['wiki_netloc'].get(sxng_lang, main_wiki) # type: ignore
base_url = 'https://' + netloc + '/index.php?'
for result in eval_xpath_list(dom, '//ul[@class="mw-search-results"]/li'):
link = eval_xpath_getindex(result, './/div[@class="mw-search-result-heading"]/a', 0)
content = extract_text(result.xpath('.//div[@class="searchresult"]'))
results.append(
{
'url': urljoin(base_url, link.get('href')), # type: ignore
'title': extract_text(link),
'content': content,
}
)
return results
def fetch_traits(engine_traits: EngineTraits):
"""Fetch languages from Archlinix-Wiki. The location of the Wiki address of a
language is mapped in a :py:obj:`custom field
<searx.enginelib.traits.EngineTraits.custom>` (``wiki_netloc``). Depending
on the location, the ``title`` argument in the request is translated.
.. code:: python
"custom": {
"wiki_netloc": {
"de": "wiki.archlinux.de",
# ...
"zh": "wiki.archlinuxcn.org"
}
"title": {
"de": "Spezial:Suche",
# ...
"zh": "Special:\u641c\u7d22"
},
},
"""
# pylint: disable=import-outside-toplevel
from searx.network import get # see https://github.com/searxng/searxng/issues/762
engine_traits.custom['wiki_netloc'] = {}
engine_traits.custom['title'] = {}
title_map = {
'de': 'Spezial:Suche',
'fa': 'ویژه:جستجو',
'ja': '特別:検索',
'zh': 'Special:搜索',
}
resp = get('https://wiki.archlinux.org/')
if not resp.ok: # type: ignore
print("ERROR: response from wiki.archlinix.org is not OK.")
dom = lxml.html.fromstring(resp.text) # type: ignore
for a in eval_xpath_list(dom, "//a[@class='interlanguage-link-target']"):
sxng_tag = language_tag(babel.Locale.parse(a.get('lang'), sep='-'))
# zh_Hans --> zh
sxng_tag = sxng_tag.split('_')[0]
netloc = urlparse(a.get('href')).netloc
if netloc != 'wiki.archlinux.org':
title = title_map.get(sxng_tag)
if not title:
print("ERROR: title tag from %s (%s) is unknown" % (netloc, sxng_tag))
continue
engine_traits.custom['wiki_netloc'][sxng_tag] = netloc
engine_traits.custom['title'][sxng_tag] = title # type: ignore
eng_tag = extract_text(eval_xpath_list(a, ".//span"))
engine_traits.languages[sxng_tag] = eng_tag # type: ignore
engine_traits.languages['en'] = 'English'