Add new extractor
This commit is contained in:
parent
12f5304556
commit
bfabd17b33
2 changed files with 75 additions and 0 deletions
74
youtube_dl/extractor/clippit.py
Normal file
74
youtube_dl/extractor/clippit.py
Normal file
|
@ -0,0 +1,74 @@
|
||||||
|
# coding: utf-8
|
||||||
|
|
||||||
|
from __future__ import unicode_literals
|
||||||
|
|
||||||
|
from .common import InfoExtractor
|
||||||
|
from ..utils import (
|
||||||
|
parse_iso8601,
|
||||||
|
qualities,
|
||||||
|
)
|
||||||
|
|
||||||
|
import re
|
||||||
|
|
||||||
|
|
||||||
|
class ClippitIE(InfoExtractor):
|
||||||
|
|
||||||
|
_VALID_URL = r'https?://(?:www\.)?clippituser\.tv/c/(?P<id>[a-z]+)'
|
||||||
|
_TEST = {
|
||||||
|
'url': 'https://www.clippituser.tv/c/evmgm',
|
||||||
|
'md5': '963ae7a59a2ec4572ab8bf2f2d2c5f09',
|
||||||
|
'info_dict': {
|
||||||
|
'id': 'evmgm',
|
||||||
|
'ext': 'mp4',
|
||||||
|
'title': 'Bye bye Brutus. #BattleBots - Clippit',
|
||||||
|
'uploader': 'lizllove',
|
||||||
|
'uploader_url': 'https://www.clippituser.tv/p/lizllove',
|
||||||
|
'timestamp': 1472183818,
|
||||||
|
'upload_date': '20160826',
|
||||||
|
'description': 'BattleBots | ABC',
|
||||||
|
'thumbnail': r're:^https?://.*\.jpg$',
|
||||||
|
}
|
||||||
|
}
|
||||||
|
|
||||||
|
def _real_extract(self, url):
|
||||||
|
video_id = self._match_id(url)
|
||||||
|
webpage = self._download_webpage(url, video_id)
|
||||||
|
|
||||||
|
title = self._html_search_regex(r'<title.*>(.+?)</title>', webpage, 'title')
|
||||||
|
|
||||||
|
FORMATS = ('sd', 'hd')
|
||||||
|
quality = qualities(FORMATS)
|
||||||
|
formats = []
|
||||||
|
for format_id in FORMATS:
|
||||||
|
url = self._html_search_regex(r'data-%s-file="(.+?)"' % format_id,
|
||||||
|
webpage, 'url', fatal=False)
|
||||||
|
if not url:
|
||||||
|
continue
|
||||||
|
match = re.search(r'/(?P<height>\d+)\.mp4', url)
|
||||||
|
formats.append({
|
||||||
|
'url': url,
|
||||||
|
'format_id': format_id,
|
||||||
|
'quality': quality(format_id),
|
||||||
|
'height': int(match.group('height')) if match else None,
|
||||||
|
})
|
||||||
|
|
||||||
|
uploader = self._html_search_regex(r'class="username".*>\s+(.+?)\n',
|
||||||
|
webpage, 'uploader', fatal=False)
|
||||||
|
uploader_url = ('https://www.clippituser.tv/p/' + uploader
|
||||||
|
if uploader else None)
|
||||||
|
|
||||||
|
timestamp = self._html_search_regex(r'datetime="(.+?)"',
|
||||||
|
webpage, 'date', fatal=False)
|
||||||
|
thumbnail = self._html_search_regex(r'data-image="(.+?)"',
|
||||||
|
webpage, 'thumbnail', fatal=False)
|
||||||
|
|
||||||
|
return {
|
||||||
|
'id': video_id,
|
||||||
|
'title': title,
|
||||||
|
'formats': formats,
|
||||||
|
'uploader': uploader,
|
||||||
|
'uploader_url': uploader_url,
|
||||||
|
'timestamp': parse_iso8601(timestamp),
|
||||||
|
'description': self._og_search_description(webpage),
|
||||||
|
'thumbnail': thumbnail,
|
||||||
|
}
|
|
@ -187,6 +187,7 @@ from .chirbit import (
|
||||||
from .cinchcast import CinchcastIE
|
from .cinchcast import CinchcastIE
|
||||||
from .cjsw import CJSWIE
|
from .cjsw import CJSWIE
|
||||||
from .cliphunter import CliphunterIE
|
from .cliphunter import CliphunterIE
|
||||||
|
from .clippit import ClippitIE
|
||||||
from .cliprs import ClipRsIE
|
from .cliprs import ClipRsIE
|
||||||
from .clipsyndicate import ClipsyndicateIE
|
from .clipsyndicate import ClipsyndicateIE
|
||||||
from .closertotruth import CloserToTruthIE
|
from .closertotruth import CloserToTruthIE
|
||||||
|
|
Loading…
Reference in a new issue