mirror of
https://github.com/searxng/searxng.git
synced 2024-11-23 19:31:00 +00:00
68 lines
1.7 KiB
Python
68 lines
1.7 KiB
Python
"""
|
|
Deezer (Music)
|
|
|
|
@website https://deezer.com
|
|
@provide-api yes (http://developers.deezer.com/api/)
|
|
|
|
@using-api yes
|
|
@results JSON
|
|
@stable yes
|
|
@parse url, title, content, embedded
|
|
"""
|
|
|
|
from json import loads
|
|
from urllib import urlencode
|
|
|
|
# engine dependent config
|
|
categories = ['music']
|
|
paging = True
|
|
|
|
# search-url
|
|
url = 'https://api.deezer.com/'
|
|
search_url = url + 'search?{query}&index={offset}'
|
|
|
|
embedded_url = '<iframe scrolling="no" frameborder="0" allowTransparency="true" ' +\
|
|
'data-src="https://www.deezer.com/plugins/player?type=tracks&id={audioid}" ' +\
|
|
'width="540" height="80"></iframe>'
|
|
|
|
|
|
# do search-request
|
|
def request(query, params):
|
|
offset = (params['pageno'] - 1) * 25
|
|
|
|
params['url'] = search_url.format(query=urlencode({'q': query}),
|
|
offset=offset)
|
|
|
|
return params
|
|
|
|
|
|
# get response from search-request
|
|
def response(resp):
|
|
results = []
|
|
|
|
search_res = loads(resp.text)
|
|
|
|
# parse results
|
|
for result in search_res.get('data', []):
|
|
if result['type'] == 'track':
|
|
title = result['title']
|
|
url = result['link']
|
|
|
|
if url.startswith('http://'):
|
|
url = 'https' + url[4:]
|
|
|
|
content = u'{} - {} - {}'.format(
|
|
result['artist']['name'],
|
|
result['album']['title'],
|
|
result['title'])
|
|
|
|
embedded = embedded_url.format(audioid=result['id'])
|
|
|
|
# append result
|
|
results.append({'url': url,
|
|
'title': title,
|
|
'embedded': embedded,
|
|
'content': content})
|
|
|
|
# return results
|
|
return results
|