mirror of
https://github.com/yt-dlp/yt-dlp.git
synced 2024-11-17 01:25:11 +00:00
Merge branch 'naglis-izlesene'
This commit is contained in:
commit
9afa6ede21
3 changed files with 100 additions and 0 deletions
|
@ -143,6 +143,7 @@
|
||||||
IviIE,
|
IviIE,
|
||||||
IviCompilationIE
|
IviCompilationIE
|
||||||
)
|
)
|
||||||
|
from .izlesene import IzleseneIE
|
||||||
from .jadorecettepub import JadoreCettePubIE
|
from .jadorecettepub import JadoreCettePubIE
|
||||||
from .jeuxvideo import JeuxVideoIE
|
from .jeuxvideo import JeuxVideoIE
|
||||||
from .jukebox import JukeboxIE
|
from .jukebox import JukeboxIE
|
||||||
|
|
97
youtube_dl/extractor/izlesene.py
Normal file
97
youtube_dl/extractor/izlesene.py
Normal file
|
@ -0,0 +1,97 @@
|
||||||
|
# coding: utf-8
|
||||||
|
from __future__ import unicode_literals
|
||||||
|
|
||||||
|
import re
|
||||||
|
|
||||||
|
from .common import InfoExtractor
|
||||||
|
from ..utils import (
|
||||||
|
get_element_by_id,
|
||||||
|
parse_iso8601,
|
||||||
|
determine_ext,
|
||||||
|
int_or_none,
|
||||||
|
str_to_int,
|
||||||
|
)
|
||||||
|
|
||||||
|
|
||||||
|
class IzleseneIE(InfoExtractor):
|
||||||
|
_VALID_URL = r'https?://(?:(?:www|m)\.)?izlesene\.com/(?:video|embedplayer)/(?:[^/]+/)?(?P<id>[0-9]+)'
|
||||||
|
_STREAM_URL = 'http://panel.izlesene.com/api/streamurl/{id:}/{format:}'
|
||||||
|
_TEST = {
|
||||||
|
'url': 'http://www.izlesene.com/video/sevincten-cildirtan-dogum-gunu-hediyesi/7599694',
|
||||||
|
'md5': '4384f9f0ea65086734b881085ee05ac2',
|
||||||
|
'info_dict': {
|
||||||
|
'id': '7599694',
|
||||||
|
'ext': 'mp4',
|
||||||
|
'title': 'Sevinçten Çıldırtan Doğum Günü Hediyesi',
|
||||||
|
'description': 'Annesi oğluna doğum günü hediyesi olarak minecraft cd si alıyor, ve çocuk hunharca seviniyor',
|
||||||
|
'thumbnail': 're:^http://.*\.jpg',
|
||||||
|
'uploader_id': 'pelikzzle',
|
||||||
|
'timestamp': 1404298698,
|
||||||
|
'upload_date': '20140702',
|
||||||
|
'duration': 95.395,
|
||||||
|
'age_limit': 0,
|
||||||
|
}
|
||||||
|
}
|
||||||
|
|
||||||
|
def _real_extract(self, url):
|
||||||
|
mobj = re.match(self._VALID_URL, url)
|
||||||
|
video_id = mobj.group('id')
|
||||||
|
url = 'http://www.izlesene.com/video/%s' % video_id
|
||||||
|
|
||||||
|
webpage = self._download_webpage(url, video_id)
|
||||||
|
|
||||||
|
title = self._og_search_title(webpage)
|
||||||
|
description = self._og_search_description(webpage)
|
||||||
|
thumbnail = self._og_search_thumbnail(webpage)
|
||||||
|
|
||||||
|
uploader = self._html_search_regex(
|
||||||
|
r"adduserUsername\s*=\s*'([^']+)';", webpage, 'uploader', fatal=False, default='')
|
||||||
|
timestamp = parse_iso8601(self._html_search_meta(
|
||||||
|
'uploadDate', webpage, 'upload date', fatal=False))
|
||||||
|
|
||||||
|
duration = int_or_none(self._html_search_regex(
|
||||||
|
r'"videoduration"\s*:\s*"([^"]+)"', webpage, 'duration', fatal=False))
|
||||||
|
if duration:
|
||||||
|
duration /= 1000.0
|
||||||
|
|
||||||
|
view_count = str_to_int(get_element_by_id('videoViewCount', webpage))
|
||||||
|
comment_count = self._html_search_regex(
|
||||||
|
r'comment_count\s*=\s*\'([^\']+)\';', webpage, 'uploader', fatal=False)
|
||||||
|
|
||||||
|
family_friendly = self._html_search_meta(
|
||||||
|
'isFamilyFriendly', webpage, 'age limit', fatal=False)
|
||||||
|
|
||||||
|
content_url = self._html_search_meta(
|
||||||
|
'contentURL', webpage, 'content URL', fatal=False)
|
||||||
|
ext = determine_ext(content_url, 'mp4')
|
||||||
|
|
||||||
|
# Might be empty for some videos.
|
||||||
|
qualities = self._html_search_regex(
|
||||||
|
r'"quality"\s*:\s*"([^"]+)"', webpage, 'qualities', fatal=False, default='')
|
||||||
|
|
||||||
|
formats = []
|
||||||
|
for quality in qualities.split('|'):
|
||||||
|
json = self._download_json(
|
||||||
|
self._STREAM_URL.format(id=video_id, format=quality), video_id,
|
||||||
|
note='Getting video URL for "%s" quality' % quality,
|
||||||
|
errnote='Failed to get video URL for "%s" quality' % quality
|
||||||
|
)
|
||||||
|
formats.append({
|
||||||
|
'url': json.get('streamurl'),
|
||||||
|
'ext': ext,
|
||||||
|
'format_id': '%sp' % quality if quality else 'sd',
|
||||||
|
})
|
||||||
|
|
||||||
|
return {
|
||||||
|
'id': video_id,
|
||||||
|
'title': title,
|
||||||
|
'description': description,
|
||||||
|
'thumbnail': thumbnail,
|
||||||
|
'uploader_id': uploader,
|
||||||
|
'timestamp': timestamp,
|
||||||
|
'duration': duration,
|
||||||
|
'view_count': int_or_none(view_count),
|
||||||
|
'comment_count': int_or_none(comment_count),
|
||||||
|
'age_limit': 18 if family_friendly == 'False' else 0,
|
||||||
|
'formats': formats,
|
||||||
|
}
|
|
@ -852,6 +852,8 @@ def unified_strdate(date_str):
|
||||||
return upload_date
|
return upload_date
|
||||||
|
|
||||||
def determine_ext(url, default_ext=u'unknown_video'):
|
def determine_ext(url, default_ext=u'unknown_video'):
|
||||||
|
if url is None:
|
||||||
|
return default_ext
|
||||||
guess = url.partition(u'?')[0].rpartition(u'.')[2]
|
guess = url.partition(u'?')[0].rpartition(u'.')[2]
|
||||||
if re.match(r'^[A-Za-z0-9]+$', guess):
|
if re.match(r'^[A-Za-z0-9]+$', guess):
|
||||||
return guess
|
return guess
|
||||||
|
|
Loading…
Reference in a new issue