mirror of
https://github.com/searxng/searxng.git
synced 2024-11-09 06:30:10 +01:00
1113f7e616
related to #101
54 lines
1.7 KiB
Python
54 lines
1.7 KiB
Python
# SPDX-License-Identifier: AGPL-3.0-or-later
|
|
"""
|
|
1337x
|
|
"""
|
|
|
|
from urllib.parse import quote, urljoin
|
|
from lxml import html
|
|
from searx.utils import extract_text, get_torrent_size, eval_xpath, eval_xpath_list, eval_xpath_getindex
|
|
|
|
# about
|
|
about = {
|
|
"website": 'https://1337x.to/',
|
|
"wikidata_id": 'Q28134166',
|
|
"official_api_documentation": None,
|
|
"use_official_api": False,
|
|
"require_api_key": False,
|
|
"results": 'HTML',
|
|
}
|
|
|
|
url = 'https://1337x.to/'
|
|
search_url = url + 'search/{search_term}/{pageno}/'
|
|
categories = ['files']
|
|
paging = True
|
|
|
|
|
|
def request(query, params):
|
|
params['url'] = search_url.format(search_term=quote(query), pageno=params['pageno'])
|
|
|
|
return params
|
|
|
|
|
|
def response(resp):
|
|
results = []
|
|
|
|
dom = html.fromstring(resp.text)
|
|
|
|
for result in eval_xpath_list(dom, '//table[contains(@class, "table-list")]/tbody//tr'):
|
|
href = urljoin(url, eval_xpath_getindex(result, './td[contains(@class, "name")]/a[2]/@href', 0))
|
|
title = extract_text(eval_xpath(result, './td[contains(@class, "name")]/a[2]'))
|
|
seed = extract_text(eval_xpath(result, './/td[contains(@class, "seeds")]'))
|
|
leech = extract_text(eval_xpath(result, './/td[contains(@class, "leeches")]'))
|
|
filesize_info = extract_text(eval_xpath(result, './/td[contains(@class, "size")]/text()'))
|
|
filesize, filesize_multiplier = filesize_info.split()
|
|
filesize = get_torrent_size(filesize, filesize_multiplier)
|
|
|
|
results.append({'url': href,
|
|
'title': title,
|
|
'seed': seed,
|
|
'leech': leech,
|
|
'filesize': filesize,
|
|
'template': 'torrent.html'})
|
|
|
|
return results
|