mirror of
https://github.com/yt-dlp/yt-dlp.git
synced 2024-11-09 01:43:37 +00:00
124 lines
4.4 KiB
Python
124 lines
4.4 KiB
Python
from __future__ import unicode_literals
|
|
|
|
import re
|
|
|
|
from .common import InfoExtractor
|
|
from ..utils import (
|
|
ExtractorError,
|
|
parse_duration,
|
|
str_to_int,
|
|
urljoin,
|
|
)
|
|
|
|
|
|
class VpornIE(InfoExtractor):
|
|
_VALID_URL = r'https?://(?:www\.)?vporn\.com/[^/]+/(?P<display_id>[^/]+)/(?P<id>\d+)'
|
|
_TESTS = [
|
|
{
|
|
'url': 'http://www.vporn.com/masturbation/violet-on-her-th-birthday/497944/',
|
|
'md5': 'facf37c1b86546fa0208058546842c55',
|
|
'info_dict': {
|
|
'id': '497944',
|
|
'display_id': 'violet-on-her-th-birthday',
|
|
'ext': 'mp4',
|
|
'title': 'Violet on her 19th birthday',
|
|
'description': 'Violet dances in front of the camera which is sure to get you horny.',
|
|
'thumbnail': r're:^https?://.*\.jpg$',
|
|
'uploader': 'kileyGrope',
|
|
'categories': ['Masturbation', 'Teen'],
|
|
'duration': 393,
|
|
'age_limit': 18,
|
|
'view_count': int,
|
|
},
|
|
'skip': 'video removed',
|
|
},
|
|
{
|
|
'url': 'http://www.vporn.com/female/hana-shower/523564/',
|
|
'md5': 'ced35a4656198a1664cf2cda1575a25f',
|
|
'info_dict': {
|
|
'id': '523564',
|
|
'display_id': 'hana-shower',
|
|
'ext': 'mp4',
|
|
'title': 'Hana Shower',
|
|
'description': 'Hana showers at the bathroom.',
|
|
'thumbnail': r're:^https?://.*\.jpg$',
|
|
'uploader': 'Hmmmmm',
|
|
'categories': ['Big Boobs', 'Erotic', 'Teen', 'Female', '720p'],
|
|
'duration': 588,
|
|
'age_limit': 18,
|
|
'view_count': int,
|
|
}
|
|
},
|
|
]
|
|
|
|
def _real_extract(self, url):
|
|
mobj = re.match(self._VALID_URL, url)
|
|
video_id = mobj.group('id')
|
|
display_id = mobj.group('display_id')
|
|
|
|
webpage = self._download_webpage(url, display_id)
|
|
|
|
errmsg = 'This video has been deleted due to Copyright Infringement or by the account owner!'
|
|
if errmsg in webpage:
|
|
raise ExtractorError('%s said: %s' % (self.IE_NAME, errmsg), expected=True)
|
|
|
|
title = self._html_search_regex(
|
|
r'videoname\s*=\s*\'([^\']+)\'', webpage, 'title').strip()
|
|
description = self._html_search_regex(
|
|
r'class="(?:descr|description_txt)">(.*?)</div>',
|
|
webpage, 'description', fatal=False)
|
|
thumbnail = urljoin('http://www.vporn.com', self._html_search_regex(
|
|
r'flashvars\.imageUrl\s*=\s*"([^"]+)"', webpage, 'description',
|
|
default=None))
|
|
|
|
uploader = self._html_search_regex(
|
|
r'(?s)Uploaded by:.*?<a href="/user/[^"]+"[^>]*>(.+?)</a>',
|
|
webpage, 'uploader', fatal=False)
|
|
|
|
categories = re.findall(r'<a href="/cat/[^"]+"[^>]*>([^<]+)</a>', webpage)
|
|
|
|
duration = parse_duration(self._search_regex(
|
|
r'Runtime:\s*</span>\s*(\d+ min \d+ sec)',
|
|
webpage, 'duration', fatal=False))
|
|
|
|
view_count = str_to_int(self._search_regex(
|
|
r'class="views">([\d,\.]+) [Vv]iews<',
|
|
webpage, 'view count', fatal=False))
|
|
comment_count = str_to_int(self._html_search_regex(
|
|
r"'Comments \(([\d,\.]+)\)'",
|
|
webpage, 'comment count', default=None))
|
|
|
|
formats = []
|
|
|
|
for video in re.findall(r'flashvars\.videoUrl([^=]+?)\s*=\s*"(https?://[^"]+)"', webpage):
|
|
video_url = video[1]
|
|
fmt = {
|
|
'url': video_url,
|
|
'format_id': video[0],
|
|
}
|
|
m = re.search(r'_(?P<width>\d+)x(?P<height>\d+)_(?P<vbr>\d+)k\.mp4$', video_url)
|
|
if m:
|
|
fmt.update({
|
|
'width': int(m.group('width')),
|
|
'height': int(m.group('height')),
|
|
'vbr': int(m.group('vbr')),
|
|
})
|
|
formats.append(fmt)
|
|
|
|
self._sort_formats(formats)
|
|
|
|
return {
|
|
'id': video_id,
|
|
'display_id': display_id,
|
|
'title': title,
|
|
'description': description,
|
|
'thumbnail': thumbnail,
|
|
'uploader': uploader,
|
|
'categories': categories,
|
|
'duration': duration,
|
|
'view_count': view_count,
|
|
'comment_count': comment_count,
|
|
'age_limit': 18,
|
|
'formats': formats,
|
|
}
|