]> git.bitcoin.ninja Git - youtube-dl/commitdiff
[vimeo] fix album extraction(closes #23864)
authorRemita Amine <remitamine@gmail.com>
Mon, 27 Jan 2020 22:37:29 +0000 (23:37 +0100)
committerRemita Amine <remitamine@gmail.com>
Mon, 27 Jan 2020 22:37:29 +0000 (23:37 +0100)
youtube_dl/extractor/vimeo.py

index baa46d5f3513cbde337f144c7143a9c501455ff4..f378aa283e3d61f6a2afbfa5134c73171b317bfd 100644 (file)
@@ -841,33 +841,6 @@ class VimeoChannelIE(VimeoBaseInfoExtractor):
         return self._TITLE or self._html_search_regex(
             self._TITLE_RE, webpage, 'list title', fatal=False)
 
-    def _login_list_password(self, page_url, list_id, webpage):
-        login_form = self._search_regex(
-            r'(?s)<form[^>]+?id="pw_form"(.*?)</form>',
-            webpage, 'login form', default=None)
-        if not login_form:
-            return webpage
-
-        password = self._downloader.params.get('videopassword')
-        if password is None:
-            raise ExtractorError('This album is protected by a password, use the --video-password option', expected=True)
-        fields = self._hidden_inputs(login_form)
-        token, vuid = self._extract_xsrft_and_vuid(webpage)
-        fields['token'] = token
-        fields['password'] = password
-        post = urlencode_postdata(fields)
-        password_path = self._search_regex(
-            r'action="([^"]+)"', login_form, 'password URL')
-        password_url = compat_urlparse.urljoin(page_url, password_path)
-        password_request = sanitized_Request(password_url, post)
-        password_request.add_header('Content-type', 'application/x-www-form-urlencoded')
-        self._set_vimeo_cookie('vuid', vuid)
-        self._set_vimeo_cookie('xsrft', token)
-
-        return self._download_webpage(
-            password_request, list_id,
-            'Verifying the password', 'Wrong password')
-
     def _title_and_entries(self, list_id, base_url):
         for pagenum in itertools.count(1):
             page_url = self._page_url(base_url, pagenum)
@@ -876,7 +849,6 @@ class VimeoChannelIE(VimeoBaseInfoExtractor):
                 'Downloading page %s' % pagenum)
 
             if pagenum == 1:
-                webpage = self._login_list_password(page_url, list_id, webpage)
                 yield self._extract_list_title(webpage)
 
             # Try extracting href first since not all videos are available via
@@ -923,7 +895,7 @@ class VimeoUserIE(VimeoChannelIE):
     _BASE_URL_TEMPL = 'https://vimeo.com/%s'
 
 
-class VimeoAlbumIE(VimeoChannelIE):
+class VimeoAlbumIE(VimeoBaseInfoExtractor):
     IE_NAME = 'vimeo:album'
     _VALID_URL = r'https://vimeo\.com/(?:album|showcase)/(?P<id>\d+)(?:$|[?#]|/(?!video))'
     _TITLE_RE = r'<header id="page_header">\n\s*<h1>(.*?)</h1>'
@@ -973,13 +945,39 @@ class VimeoAlbumIE(VimeoChannelIE):
     def _real_extract(self, url):
         album_id = self._match_id(url)
         webpage = self._download_webpage(url, album_id)
-        webpage = self._login_list_password(url, album_id, webpage)
-        api_config = self._extract_vimeo_config(webpage, album_id)['api']
+        viewer = self._parse_json(self._search_regex(
+            r'bootstrap_data\s*=\s*({.+?})</script>',
+            webpage, 'bootstrap data'), album_id)['viewer']
+        jwt = viewer['jwt']
+        album = self._download_json(
+            'https://api.vimeo.com/albums/' + album_id,
+            album_id, headers={'Authorization': 'jwt ' + jwt},
+            query={'fields': 'description,name,privacy'})
+        hashed_pass = None
+        if try_get(album, lambda x: x['privacy']['view']) == 'password':
+            password = self._downloader.params.get('videopassword')
+            if not password:
+                raise ExtractorError(
+                    'This album is protected by a password, use the --video-password option',
+                    expected=True)
+            self._set_vimeo_cookie('vuid', viewer['vuid'])
+            try:
+                hashed_pass = self._download_json(
+                    'https://vimeo.com/showcase/%s/auth' % album_id,
+                    album_id, 'Verifying the password', data=urlencode_postdata({
+                        'password': password,
+                        'token': viewer['xsrft'],
+                    }), headers={
+                        'X-Requested-With': 'XMLHttpRequest',
+                    })['hashed_pass']
+            except ExtractorError as e:
+                if isinstance(e.cause, compat_HTTPError) and e.cause.code == 401:
+                    raise ExtractorError('Wrong password', expected=True)
+                raise
         entries = OnDemandPagedList(functools.partial(
-            self._fetch_page, album_id, api_config['jwt'],
-            api_config.get('hashed_pass')), self._PAGE_SIZE)
-        return self.playlist_result(entries, album_id, self._html_search_regex(
-            r'<title>\s*(.+?)(?:\s+on Vimeo)?</title>', webpage, 'title', fatal=False))
+            self._fetch_page, album_id, jwt, hashed_pass), self._PAGE_SIZE)
+        return self.playlist_result(
+            entries, album_id, album.get('name'), album.get('description'))
 
 
 class VimeoGroupsIE(VimeoChannelIE):