1
0
Fork 0
mirror of https://github.com/ytdl-org/youtube-dl.git synced 2024-11-16 22:06:11 +00:00

[tapely] Add new extractor (closes #3861)

This commit is contained in:
Naglis Jonaitis 2014-10-01 17:26:09 +03:00
parent d3c9af84fc
commit e80f40e5ca
2 changed files with 98 additions and 0 deletions

View file

@ -355,6 +355,7 @@ from .swrmediathek import SWRMediathekIE
from .syfy import SyfyIE from .syfy import SyfyIE
from .sztvhu import SztvHuIE from .sztvhu import SztvHuIE
from .tagesschau import TagesschauIE from .tagesschau import TagesschauIE
from .tapely import TapelyIE
from .teachertube import ( from .teachertube import (
TeacherTubeIE, TeacherTubeIE,
TeacherTubeUserIE, TeacherTubeUserIE,

View file

@ -0,0 +1,97 @@
# coding: utf-8
from __future__ import unicode_literals
import re
from .common import InfoExtractor
from ..utils import (
ExtractorError,
clean_html,
compat_urllib_request,
float_or_none,
parse_iso8601,
)
class TapelyIE(InfoExtractor):
_VALID_URL = r'https?://(?:www\.)?tape\.ly/(?P<id>[A-Za-z0-9\-_]+)(?:/(?P<songnr>\d+))?'
_API_URL = 'http://tape.ly/showtape?id={0:}'
_S3_SONG_URL = 'http://mytape.s3.amazonaws.com/{0:}'
_TESTS = [
{
'url': 'http://tape.ly/my-grief-as-told-by-water',
'info_dict': {
'id': 23952,
'title': 'my grief as told by water',
'thumbnail': 're:^https?://.*\.png$',
'uploader_id': 16484,
'timestamp': 1411848286,
'description': 'For Robin and Ponkers, whom the tides of life have taken out to sea.',
},
'playlist_count': 13,
},
{
'url': 'http://tape.ly/my-grief-as-told-by-water/1',
'md5': '79031f459fdec6530663b854cbc5715c',
'info_dict': {
'id': 258464,
'title': 'Dreaming Awake (My Brightest Diamond)',
'ext': 'm4a',
},
},
]
def _real_extract(self, url):
mobj = re.match(self._VALID_URL, url)
display_id = mobj.group('id')
playlist_url = self._API_URL.format(display_id)
request = compat_urllib_request.Request(playlist_url)
request.add_header('X-Requested-With', 'XMLHttpRequest')
request.add_header('Accept', 'application/json')
playlist = self._download_json(request, display_id)
tape = playlist['tape']
entries = []
for s in tape['songs']:
song = s['song']
entry = {
'id': song['id'],
'duration': float_or_none(song.get('songduration'), 1000),
'title': song['title'],
}
if song['source'] == 'S3':
entry.update({
'url': self._S3_SONG_URL.format(song['filename']),
})
entries.append(entry)
elif song['source'] == 'YT':
_, _, yt_id = song['filename'].split('/')
entry.update(self.url_result(yt_id, 'Youtube', video_id=yt_id))
entries.append(entry)
else:
self.report_warning('Unknown song source: %s' % song['source'])
if mobj.group('songnr'):
songnr = int(mobj.group('songnr')) - 1
try:
return entries[songnr]
except IndexError:
raise ExtractorError(
'No song with index: %s' % mobj.group('songnr'),
expected=True)
return {
'_type': 'playlist',
'id': tape['id'],
'display_id': display_id,
'title': tape['name'],
'entries': entries,
'thumbnail': tape.get('image_url'),
'description': clean_html(tape.get('subtext')),
'like_count': tape.get('likescount'),
'uploader_id': tape.get('user_id'),
'timestamp': parse_iso8601(tape.get('published_at')),
}