aboutsummaryrefslogtreecommitdiffstats
path: root/hypervideo_dl/extractor/stripchat.py
diff options
context:
space:
mode:
Diffstat (limited to 'hypervideo_dl/extractor/stripchat.py')
-rw-r--r--hypervideo_dl/extractor/stripchat.py66
1 files changed, 66 insertions, 0 deletions
diff --git a/hypervideo_dl/extractor/stripchat.py b/hypervideo_dl/extractor/stripchat.py
new file mode 100644
index 0000000..0d4a0ce
--- /dev/null
+++ b/hypervideo_dl/extractor/stripchat.py
@@ -0,0 +1,66 @@
+# coding: utf-8
+from __future__ import unicode_literals
+
+from .common import InfoExtractor
+from ..compat import (
+ compat_str,
+)
+from ..utils import (
+ ExtractorError,
+ lowercase_escape,
+ try_get,
+)
+
+
+class StripchatIE(InfoExtractor):
+ _VALID_URL = r'https?://stripchat\.com/(?P<id>[0-9A-Za-z-_]+)'
+ _TESTS = [{
+ 'url': 'https://stripchat.com/feel_me',
+ 'info_dict': {
+ 'id': 'feel_me',
+ 'ext': 'mp4',
+ 'title': 're:^feel_me [0-9]{4}-[0-9]{2}-[0-9]{2} [0-9]{2}:[0-9]{2}$',
+ 'description': str,
+ 'is_live': True,
+ 'age_limit': 18,
+ },
+ 'skip': 'Room is offline',
+ }]
+
+ def _real_extract(self, url):
+ video_id = self._match_id(url)
+ webpage = self._download_webpage(
+ 'https://stripchat.com/%s/' % video_id, video_id,
+ headers=self.geo_verification_headers())
+
+ data = self._parse_json(
+ self._search_regex(
+ r'<script\b[^>]*>\s*window\.__PRELOADED_STATE__\s*=(?P<value>.*?)<\/script>',
+ webpage, 'data', default='{}', group='value'),
+ video_id, transform_source=lowercase_escape, fatal=False)
+ if not data:
+ raise ExtractorError('Unable to find configuration for stream.')
+
+ if try_get(data, lambda x: x['viewCam']['show'], dict):
+ raise ExtractorError('Model is in private show', expected=True)
+ elif not try_get(data, lambda x: x['viewCam']['model']['isLive'], bool):
+ raise ExtractorError('Model is offline', expected=True)
+
+ server = try_get(data, lambda x: x['viewCam']['viewServers']['flashphoner-hls'], compat_str)
+ host = try_get(data, lambda x: x['config']['data']['hlsStreamHost'], compat_str)
+ model_id = try_get(data, lambda x: x['viewCam']['model']['id'], int)
+
+ formats = self._extract_m3u8_formats(
+ 'https://b-%s.%s/hls/%d/%d.m3u8' % (server, host, model_id, model_id),
+ video_id, ext='mp4', m3u8_id='hls', fatal=False, live=True)
+ self._sort_formats(formats)
+
+ return {
+ 'id': video_id,
+ 'title': video_id,
+ 'description': self._og_search_description(webpage),
+ 'is_live': True,
+ 'formats': formats,
+ # Stripchat declares the RTA meta-tag, but in an non-standard format so _rta_search() can't be used
+ 'age_limit': 18,
+ }