123456789101112131415161718192021222324252627282930313233343536373839404142434445464748495051525354555657585960616263646566676869707172737475767778798081828384858687888990 |
- # SPDX-License-Identifier: AGPL-3.0-or-later
- """
- BTDigg (Videos, Music, Files)
- """
- from lxml import html
- from urllib.parse import quote, urljoin
- from searx.utils import extract_text, get_torrent_size
- # about
- about = {
- "website": 'https://btdig.com',
- "wikidata_id": 'Q4836698',
- "official_api_documentation": {
- 'url': 'https://btdig.com/contacts',
- 'comment': 'on demand'
- },
- "use_official_api": False,
- "require_api_key": False,
- "results": 'HTML',
- }
- # engine dependent config
- categories = ['videos', 'music', 'files']
- paging = True
- # search-url
- url = 'https://btdig.com'
- search_url = url + '/search?q={search_term}&p={pageno}'
- # do search-request
- def request(query, params):
- params['url'] = search_url.format(search_term=quote(query),
- pageno=params['pageno'] - 1)
- return params
- # get response from search-request
- def response(resp):
- results = []
- dom = html.fromstring(resp.text)
- search_res = dom.xpath('//div[@class="one_result"]')
- # return empty array if nothing is found
- if not search_res:
- return []
- # parse results
- for result in search_res:
- link = result.xpath('.//div[@class="torrent_name"]//a')[0]
- href = urljoin(url, link.attrib.get('href'))
- title = extract_text(link)
- excerpt = result.xpath('.//div[@class="torrent_excerpt"]')[0]
- content = html.tostring(excerpt, encoding='unicode', method='text', with_tail=False)
- # it is better to emit <br/> instead of |, but html tags are verboten
- content = content.strip().replace('\n', ' | ')
- content = ' '.join(content.split())
- filesize = result.xpath('.//span[@class="torrent_size"]/text()')[0].split()[0]
- filesize_multiplier = result.xpath('.//span[@class="torrent_size"]/text()')[0].split()[1]
- files = (result.xpath('.//span[@class="torrent_files"]/text()') or ['1'])[0]
- # convert filesize to byte if possible
- filesize = get_torrent_size(filesize, filesize_multiplier)
- # convert files to int if possible
- try:
- files = int(files)
- except:
- files = None
- magnetlink = result.xpath('.//div[@class="torrent_magnet"]//a')[0].attrib['href']
- # append result
- results.append({'url': href,
- 'title': title,
- 'content': content,
- 'filesize': filesize,
- 'files': files,
- 'magnetlink': magnetlink,
- 'template': 'torrent.html'})
- # return results sorted by seeder
- return results
|