|
@@ -0,0 +1,83 @@
|
|
|
+# coding: utf-8
|
|
|
+from __future__ import unicode_literals
|
|
|
+
|
|
|
+import re
|
|
|
+
|
|
|
+from .common import InfoExtractor
|
|
|
+from ..compat import compat_urlparse
|
|
|
+from ..utils import (
|
|
|
+ get_element_by_class,
|
|
|
+ urlencode_postdata,
|
|
|
+)
|
|
|
+
|
|
|
+
|
|
|
+class NJPWWorldIE(InfoExtractor):
|
|
|
+ _VALID_URL = r'https?://njpwworld\.com/p/(?P<id>[a-z0-9_]+)'
|
|
|
+ IE_DESC = '新日本プロレスワールド'
|
|
|
+ _NETRC_MACHINE = 'njpwworld'
|
|
|
+
|
|
|
+ _TEST = {
|
|
|
+ 'url': 'http://njpwworld.com/p/s_series_00155_1_9/',
|
|
|
+ 'info_dict': {
|
|
|
+ 'id': 's_series_00155_1_9',
|
|
|
+ 'ext': 'mp4',
|
|
|
+ 'title': '第9試合 ランディ・サベージ vs リック・スタイナー',
|
|
|
+ 'tags': list,
|
|
|
+ },
|
|
|
+ 'params': {
|
|
|
+ 'skip_download': True, # AES-encrypted m3u8
|
|
|
+ },
|
|
|
+ 'skip': 'Requires login',
|
|
|
+ }
|
|
|
+
|
|
|
+ def _real_initialize(self):
|
|
|
+ self._login()
|
|
|
+
|
|
|
+ def _login(self):
|
|
|
+ username, password = self._get_login_info()
|
|
|
+ # No authentication to be performed
|
|
|
+ if not username:
|
|
|
+ return True
|
|
|
+
|
|
|
+ webpage, urlh = self._download_webpage_handle(
|
|
|
+ 'https://njpwworld.com/auth/login', None,
|
|
|
+ note='Logging in', errnote='Unable to login',
|
|
|
+ data=urlencode_postdata({'login_id': username, 'pw': password}))
|
|
|
+ # /auth/login will return 302 for successful logins
|
|
|
+ if urlh.geturl() == 'https://njpwworld.com/auth/login':
|
|
|
+ self.report_warning('unable to login')
|
|
|
+ return False
|
|
|
+
|
|
|
+ return True
|
|
|
+
|
|
|
+ def _real_extract(self, url):
|
|
|
+ video_id = self._match_id(url)
|
|
|
+
|
|
|
+ webpage = self._download_webpage(url, video_id)
|
|
|
+
|
|
|
+ formats = []
|
|
|
+ for player_url, kind in re.findall(r'<a[^>]+href="(/player[^"]+)".+?<img[^>]+src="[^"]+qf_btn_([^".]+)', webpage):
|
|
|
+ player_url = compat_urlparse.urljoin(url, player_url)
|
|
|
+
|
|
|
+ player_page = self._download_webpage(
|
|
|
+ player_url, video_id, note='Downloading player page')
|
|
|
+
|
|
|
+ entries = self._parse_html5_media_entries(
|
|
|
+ player_url, player_page, video_id, m3u8_id='hls-%s' % kind,
|
|
|
+ m3u8_entry_protocol='m3u8_native',
|
|
|
+ preference=2 if 'hq' in kind else 1)
|
|
|
+ formats.extend(entries[0]['formats'])
|
|
|
+
|
|
|
+ self._sort_formats(formats)
|
|
|
+
|
|
|
+ post_content = get_element_by_class('post-content', webpage)
|
|
|
+ tags = re.findall(
|
|
|
+ r'<li[^>]+class="tag-[^"]+"><a[^>]*>([^<]+)</a></li>', post_content
|
|
|
+ ) if post_content else None
|
|
|
+
|
|
|
+ return {
|
|
|
+ 'id': video_id,
|
|
|
+ 'title': self._og_search_title(webpage),
|
|
|
+ 'formats': formats,
|
|
|
+ 'tags': tags,
|
|
|
+ }
|