mirror of
https://github.com/yt-dlp/yt-dlp.git
synced 2024-11-03 02:03:23 +01:00
[ie/dailymotion] Support search (#8292)
Closes #6126 Authored by: drzraf, seproDev Co-authored-by: sepro <4618135+seproDev@users.noreply.github.com>
This commit is contained in:
parent
ede624d1db
commit
11ffa92a61
@ -444,6 +444,7 @@
|
||||
from .dailymotion import (
|
||||
DailymotionIE,
|
||||
DailymotionPlaylistIE,
|
||||
DailymotionSearchIE,
|
||||
DailymotionUserIE,
|
||||
)
|
||||
from .dailywire import (
|
||||
|
@ -1,6 +1,7 @@
|
||||
import functools
|
||||
import json
|
||||
import re
|
||||
import urllib.parse
|
||||
|
||||
from .common import InfoExtractor
|
||||
from ..networking.exceptions import HTTPError
|
||||
@ -44,36 +45,41 @@ def _real_initialize(self):
|
||||
self._FAMILY_FILTER = ff == 'on' if ff else age_restricted(18, self.get_param('age_limit'))
|
||||
self._set_dailymotion_cookie('ff', 'on' if self._FAMILY_FILTER else 'off')
|
||||
|
||||
def _get_token(self, xid):
|
||||
cookies = self._get_dailymotion_cookies()
|
||||
token = self._get_cookie_value(cookies, 'access_token') or self._get_cookie_value(cookies, 'client_token')
|
||||
if token:
|
||||
return token
|
||||
|
||||
data = {
|
||||
'client_id': 'f1a362d288c1b98099c7',
|
||||
'client_secret': 'eea605b96e01c796ff369935357eca920c5da4c5',
|
||||
}
|
||||
username, password = self._get_login_info()
|
||||
if username:
|
||||
data.update({
|
||||
'grant_type': 'password',
|
||||
'password': password,
|
||||
'username': username,
|
||||
})
|
||||
else:
|
||||
data['grant_type'] = 'client_credentials'
|
||||
try:
|
||||
token = self._download_json(
|
||||
'https://graphql.api.dailymotion.com/oauth/token',
|
||||
None, 'Downloading Access Token',
|
||||
data=urlencode_postdata(data))['access_token']
|
||||
except ExtractorError as e:
|
||||
if isinstance(e.cause, HTTPError) and e.cause.status == 400:
|
||||
raise ExtractorError(self._parse_json(
|
||||
e.cause.response.read().decode(), xid)['error_description'], expected=True)
|
||||
raise
|
||||
self._set_dailymotion_cookie('access_token' if username else 'client_token', token)
|
||||
return token
|
||||
|
||||
def _call_api(self, object_type, xid, object_fields, note, filter_extra=None):
|
||||
if not self._HEADERS.get('Authorization'):
|
||||
cookies = self._get_dailymotion_cookies()
|
||||
token = self._get_cookie_value(cookies, 'access_token') or self._get_cookie_value(cookies, 'client_token')
|
||||
if not token:
|
||||
data = {
|
||||
'client_id': 'f1a362d288c1b98099c7',
|
||||
'client_secret': 'eea605b96e01c796ff369935357eca920c5da4c5',
|
||||
}
|
||||
username, password = self._get_login_info()
|
||||
if username:
|
||||
data.update({
|
||||
'grant_type': 'password',
|
||||
'password': password,
|
||||
'username': username,
|
||||
})
|
||||
else:
|
||||
data['grant_type'] = 'client_credentials'
|
||||
try:
|
||||
token = self._download_json(
|
||||
'https://graphql.api.dailymotion.com/oauth/token',
|
||||
None, 'Downloading Access Token',
|
||||
data=urlencode_postdata(data))['access_token']
|
||||
except ExtractorError as e:
|
||||
if isinstance(e.cause, HTTPError) and e.cause.status == 400:
|
||||
raise ExtractorError(self._parse_json(
|
||||
e.cause.response.read().decode(), xid)['error_description'], expected=True)
|
||||
raise
|
||||
self._set_dailymotion_cookie('access_token' if username else 'client_token', token)
|
||||
self._HEADERS['Authorization'] = 'Bearer ' + token
|
||||
self._HEADERS['Authorization'] = f'Bearer {self._get_token(xid)}'
|
||||
|
||||
resp = self._download_json(
|
||||
'https://graphql.api.dailymotion.com/', xid, note, data=json.dumps({
|
||||
@ -393,9 +399,55 @@ def _extract_embed_urls(cls, url, webpage):
|
||||
yield '//dailymotion.com/playlist/%s' % p
|
||||
|
||||
|
||||
class DailymotionSearchIE(DailymotionPlaylistBaseIE):
|
||||
IE_NAME = 'dailymotion:search'
|
||||
_VALID_URL = r'https?://(?:www\.)?dailymotion\.[a-z]{2,3}/search/(?P<id>[^/?#]+)/videos'
|
||||
_PAGE_SIZE = 20
|
||||
_TESTS = [{
|
||||
'url': 'http://www.dailymotion.com/search/king of turtles/videos',
|
||||
'info_dict': {
|
||||
'id': 'king of turtles',
|
||||
'title': 'king of turtles',
|
||||
},
|
||||
'playlist_mincount': 90,
|
||||
}]
|
||||
_SEARCH_QUERY = 'query SEARCH_QUERY( $query: String! $page: Int $limit: Int ) { search { videos( query: $query first: $limit page: $page ) { edges { node { xid } } } } } '
|
||||
|
||||
def _call_search_api(self, term, page, note):
|
||||
if not self._HEADERS.get('Authorization'):
|
||||
self._HEADERS['Authorization'] = f'Bearer {self._get_token(term)}'
|
||||
resp = self._download_json(
|
||||
'https://graphql.api.dailymotion.com/', None, note, data=json.dumps({
|
||||
'operationName': 'SEARCH_QUERY',
|
||||
'query': self._SEARCH_QUERY,
|
||||
'variables': {
|
||||
'limit': 20,
|
||||
'page': page,
|
||||
'query': term,
|
||||
}
|
||||
}).encode(), headers=self._HEADERS)
|
||||
obj = traverse_obj(resp, ('data', 'search', {dict}))
|
||||
if not obj:
|
||||
raise ExtractorError(
|
||||
traverse_obj(resp, ('errors', 0, 'message', {str})) or 'Could not fetch search data')
|
||||
|
||||
return obj
|
||||
|
||||
def _fetch_page(self, term, page):
|
||||
page += 1
|
||||
response = self._call_search_api(term, page, f'Searching "{term}" page {page}')
|
||||
for xid in traverse_obj(response, ('videos', 'edges', ..., 'node', 'xid')):
|
||||
yield self.url_result(f'https://www.dailymotion.com/video/{xid}', DailymotionIE, xid)
|
||||
|
||||
def _real_extract(self, url):
|
||||
term = urllib.parse.unquote_plus(self._match_id(url))
|
||||
return self.playlist_result(
|
||||
OnDemandPagedList(functools.partial(self._fetch_page, term), self._PAGE_SIZE), term, term)
|
||||
|
||||
|
||||
class DailymotionUserIE(DailymotionPlaylistBaseIE):
|
||||
IE_NAME = 'dailymotion:user'
|
||||
_VALID_URL = r'https?://(?:www\.)?dailymotion\.[a-z]{2,3}/(?!(?:embed|swf|#|video|playlist)/)(?:(?:old/)?user/)?(?P<id>[^/]+)'
|
||||
_VALID_URL = r'https?://(?:www\.)?dailymotion\.[a-z]{2,3}/(?!(?:embed|swf|#|video|playlist|search)/)(?:(?:old/)?user/)?(?P<id>[^/?#]+)'
|
||||
_TESTS = [{
|
||||
'url': 'https://www.dailymotion.com/user/nqtv',
|
||||
'info_dict': {
|
||||
|
Loading…
Reference in New Issue
Block a user