1
0
mirror of https://github.com/yt-dlp/yt-dlp.git synced 2024-11-02 09:12:40 +01:00

[nationalgeographic] Add support for National Geographic Episode Guide

This commit is contained in:
Remita Amine 2016-08-04 12:18:10 +01:00
parent 1094074c04
commit 1891ea2d76

View File

@ -1,16 +1,19 @@
from __future__ import unicode_literals from __future__ import unicode_literals
import re
from .common import InfoExtractor from .common import InfoExtractor
from .theplatform import ThePlatformIE from .theplatform import ThePlatformIE
from ..utils import ( from ..utils import (
smuggle_url, smuggle_url,
url_basename, url_basename,
update_url_query, update_url_query,
get_element_by_class,
) )
class NationalGeographicIE(InfoExtractor): class NationalGeographicVideoIE(InfoExtractor):
IE_NAME = 'natgeo' IE_NAME = 'natgeo:video'
_VALID_URL = r'https?://video\.nationalgeographic\.com/.*?' _VALID_URL = r'https?://video\.nationalgeographic\.com/.*?'
_TESTS = [ _TESTS = [
@ -62,9 +65,9 @@ def _real_extract(self, url):
} }
class NationalGeographicChannelIE(ThePlatformIE): class NationalGeographicIE(ThePlatformIE):
IE_NAME = 'natgeo:channel' IE_NAME = 'natgeo'
_VALID_URL = r'https?://channel\.nationalgeographic\.com/(?:wild/)?[^/]+/videos/(?P<id>[^/?]+)' _VALID_URL = r'https?://channel\.nationalgeographic\.com/(?:wild/)?[^/]+/(?:videos|episodes)/(?P<id>[^/?]+)'
_TESTS = [ _TESTS = [
{ {
@ -95,6 +98,10 @@ class NationalGeographicChannelIE(ThePlatformIE):
}, },
'add_ie': ['ThePlatform'], 'add_ie': ['ThePlatform'],
}, },
{
'url': 'http://channel.nationalgeographic.com/the-story-of-god-with-morgan-freeman/episodes/the-power-of-miracles/',
'only_matching': True,
}
] ]
def _real_extract(self, url): def _real_extract(self, url):
@ -122,3 +129,40 @@ def _real_extract(self, url):
{'force_smil_url': True}), {'force_smil_url': True}),
'display_id': display_id, 'display_id': display_id,
} }
class NationalGeographicEpisodeGuideIE(ThePlatformIE):
IE_NAME = 'natgeo:episodeguide'
_VALID_URL = r'https?://channel\.nationalgeographic\.com/(?:wild/)?(?P<id>[^/]+)/episode-guide'
_TESTS = [
{
'url': 'http://channel.nationalgeographic.com/the-story-of-god-with-morgan-freeman/episode-guide/',
'info_dict': {
'id': 'the-story-of-god-with-morgan-freeman-season-1',
'title': 'The Story of God with Morgan Freeman - Season 1',
},
'playlist_mincount': 6,
},
{
'url': 'http://channel.nationalgeographic.com/underworld-inc/episode-guide/?s=2',
'info_dict': {
'id': 'underworld-inc-season-2',
'title': 'Underworld, Inc. - Season 2',
},
'playlist_mincount': 7,
},
]
def _real_extract(self, url):
display_id = self._match_id(url)
webpage = self._download_webpage(url, display_id)
show = get_element_by_class('show', webpage)
selected_season = self._search_regex(
r'<div[^>]+class="select-seasons[^"]*".*?<a[^>]*>(.*?)</a>',
webpage, 'selected season')
entries = [
self.url_result(self._proto_relative_url(url), 'NationalGeographic')
for url in re.findall('(?s)<div[^>]+class="col-inner"[^>]*?>.*?<a[^>]+href="([^"]+)"', webpage)]
return self.playlist_result(
entries, '%s-%s' % (display_id, selected_season.lower().replace(' ', '-')),
'%s - %s' % (show, selected_season))