1
1
mirror of https://github.com/ytdl-org/youtube-dl synced 2024-11-24 00:36:52 +01:00

[orf] Move all ORF extractors in one place

This commit is contained in:
Sergey M․ 2014-08-10 01:21:16 +07:00
parent f5273890ee
commit eb3680123a
4 changed files with 80 additions and 89 deletions

View File

@ -96,7 +96,6 @@ from .fktv import (
FKTVPosteckeIE,
)
from .flickr import FlickrIE
from .fm4 import FM4IE
from .fourtube import FourTubeIE
from .franceculture import FranceCultureIE
from .franceinter import FranceInterIE
@ -226,9 +225,12 @@ from .nrk import (
from .ntv import NTVIE
from .nytimes import NYTimesIE
from .nuvid import NuvidIE
from .oe1 import OE1IE
from .ooyala import OoyalaIE
from .orf import ORFIE
from .orf import (
ORFTVthekIE,
ORFOE1IE,
ORFFM4IE,
)
from .parliamentliveuk import ParliamentLiveUKIE
from .pbs import PBSIE
from .photobucket import PhotobucketIE

View File

@ -1,45 +0,0 @@
# coding: utf-8
from __future__ import unicode_literals
import re
from .common import InfoExtractor
# audios on fm4.orf.at are only available for 7 days, so we can't
# add tests.
class FM4IE(InfoExtractor):
IE_DESC = 'fm4.orf.at'
_VALID_URL = r'http://fm4\.orf\.at/7tage/?#(?P<date>[0-9]+)/(?P<show>\w+)'
def _real_extract(self, url):
mobj = re.match(self._VALID_URL, url)
show_date = mobj.group('date')
show_id = mobj.group('show')
data = self._download_json(
'http://audioapi.orf.at/fm4/json/2.0/broadcasts/%s/4%s' % (show_date, show_id),
show_id
)
def extract_entry_dict(info, title, subtitle):
return {
'id': info['loopStreamId'].replace('.mp3', ''),
'url': 'http://loopstream01.apa.at/?channel=fm4&id=%s' % info['loopStreamId'],
'title': title,
'description': subtitle,
'duration': (info['end'] - info['start']) / 1000,
'timestamp': info['start'] / 1000,
'ext': 'mp3'
}
entries = [extract_entry_dict(t, data['title'], data['subtitle']) for t in data['streams']]
return {
'_type': 'playlist',
'id': show_id,
'title': data['title'],
'description': data['subtitle'],
'entries': entries
}

View File

@ -1,40 +0,0 @@
# coding: utf-8
from __future__ import unicode_literals
import calendar
import datetime
import re
from .common import InfoExtractor
# audios on oe1.orf.at are only available for 7 days, so we can't
# add tests.
class OE1IE(InfoExtractor):
IE_DESC = 'oe1.orf.at'
_VALID_URL = r'http://oe1\.orf\.at/programm/(?P<id>[0-9]+)'
def _real_extract(self, url):
mobj = re.match(self._VALID_URL, url)
show_id = mobj.group('id')
data = self._download_json(
'http://oe1.orf.at/programm/%s/konsole' % show_id,
show_id
)
timestamp = datetime.datetime.strptime('%s %s' % (
data['item']['day_label'],
data['item']['time']
), '%d.%m.%Y %H:%M')
unix_timestamp = calendar.timegm(timestamp.utctimetuple())
return {
'id': show_id,
'title': data['item']['title'],
'url': data['item']['url_stream'],
'ext': 'mp3',
'description': data['item'].get('info'),
'timestamp': unix_timestamp
}

View File

@ -3,6 +3,8 @@ from __future__ import unicode_literals
import json
import re
import calendar
import datetime
from .common import InfoExtractor
from ..utils import (
@ -12,7 +14,9 @@ from ..utils import (
)
class ORFIE(InfoExtractor):
class ORFTVthekIE(InfoExtractor):
IE_NAME = 'orf:tvthek'
IE_DESC = 'ORF TVthek'
_VALID_URL = r'https?://tvthek\.orf\.at/(?:programs/.+?/episodes|topics/.+?|program/[^/]+)/(?P<id>\d+)'
_TEST = {
@ -105,3 +109,73 @@ class ORFIE(InfoExtractor):
'entries': entries,
'id': playlist_id,
}
# Audios on ORF radio are only available for 7 days, so we can't add tests.
class ORFOE1IE(InfoExtractor):
IE_NAME = 'orf:oe1'
IE_DESC = 'Radio Österreich 1'
_VALID_URL = r'http://oe1\.orf\.at/programm/(?P<id>[0-9]+)'
def _real_extract(self, url):
mobj = re.match(self._VALID_URL, url)
show_id = mobj.group('id')
data = self._download_json(
'http://oe1.orf.at/programm/%s/konsole' % show_id,
show_id
)
timestamp = datetime.datetime.strptime('%s %s' % (
data['item']['day_label'],
data['item']['time']
), '%d.%m.%Y %H:%M')
unix_timestamp = calendar.timegm(timestamp.utctimetuple())
return {
'id': show_id,
'title': data['item']['title'],
'url': data['item']['url_stream'],
'ext': 'mp3',
'description': data['item'].get('info'),
'timestamp': unix_timestamp
}
class ORFFM4IE(InfoExtractor):
IE_DESC = 'orf:fm4'
IE_DESC = 'radio FM4'
_VALID_URL = r'http://fm4\.orf\.at/7tage/?#(?P<date>[0-9]+)/(?P<show>\w+)'
def _real_extract(self, url):
mobj = re.match(self._VALID_URL, url)
show_date = mobj.group('date')
show_id = mobj.group('show')
data = self._download_json(
'http://audioapi.orf.at/fm4/json/2.0/broadcasts/%s/4%s' % (show_date, show_id),
show_id
)
def extract_entry_dict(info, title, subtitle):
return {
'id': info['loopStreamId'].replace('.mp3', ''),
'url': 'http://loopstream01.apa.at/?channel=fm4&id=%s' % info['loopStreamId'],
'title': title,
'description': subtitle,
'duration': (info['end'] - info['start']) / 1000,
'timestamp': info['start'] / 1000,
'ext': 'mp3'
}
entries = [extract_entry_dict(t, data['title'], data['subtitle']) for t in data['streams']]
return {
'_type': 'playlist',
'id': show_id,
'title': data['title'],
'description': data['subtitle'],
'entries': entries
}