X-Git-Url: https://git.rapsys.eu/youtubedl/blobdiff_plain/33cd347759d6d999325ebf3c69b7ed5692c343b2..252064fa082f84e06d1251185428b44c75c59109:/youtube_dl/extractor/wat.py?ds=inline diff --git a/youtube_dl/extractor/wat.py b/youtube_dl/extractor/wat.py index 7d228ed..a584e08 100644 --- a/youtube_dl/extractor/wat.py +++ b/youtube_dl/extractor/wat.py @@ -1,38 +1,37 @@ # coding: utf-8 +from __future__ import unicode_literals -import json import re from .common import InfoExtractor - from ..utils import ( - compat_urllib_parse, unified_strdate, ) class WatIE(InfoExtractor): - _VALID_URL=r'http://www.wat.tv/.*-(?P.*?)_.*?.html' + _VALID_URL = r'http://www\.wat\.tv/.*-(?P.*?)_.*?\.html' IE_NAME = 'wat.tv' _TEST = { - u'url': u'http://www.wat.tv/video/world-war-philadelphia-vost-6bv55_2fjr7_.html', - u'file': u'10631273.mp4', - u'md5': u'd8b2231e1e333acd12aad94b80937e19', - u'info_dict': { - u'title': u'World War Z - Philadelphia VOST', - u'description': u'La menace est partout. Que se passe-t-il à Philadelphia ?\r\nWORLD WAR Z, avec Brad Pitt, au cinéma le 3 juillet.\r\nhttp://www.worldwarz.fr', + 'url': 'http://www.wat.tv/video/world-war-philadelphia-vost-6bv55_2fjr7_.html', + 'info_dict': { + 'id': '10631273', + 'ext': 'mp4', + 'title': 'World War Z - Philadelphia VOST', + 'description': 'La menace est partout. Que se passe-t-il à Philadelphia ?\r\nWORLD WAR Z, avec Brad Pitt, au cinéma le 3 juillet.\r\nhttp://www.worldwarz.fr', + }, + 'params': { + # Sometimes wat serves the whole file with the --test option + 'skip_download': True, }, - u'skip': u'Sometimes wat serves the whole file with the --test option', } - + def download_video_info(self, real_id): # 'contentv4' is used in the website, but it also returns the related # videos, we don't need them - info = self._download_webpage('http://www.wat.tv/interface/contentv3/' + real_id, real_id, 'Downloading video info') - info = json.loads(info) + info = self._download_json('http://www.wat.tv/interface/contentv3/' + real_id, real_id) return info['media'] - def _real_extract(self, url): def real_id_for_chapter(chapter): return chapter['tc_start'].split('-')[0] @@ -57,17 +56,17 @@ class WatIE(InfoExtractor): entries = [self.url_result(chapter_url) for chapter_url in chapter_urls] return self.playlist_result(entries, real_id, video_info['title']) + upload_date = None + if 'date_diffusion' in first_chapter: + upload_date = unified_strdate(first_chapter['date_diffusion']) # Otherwise we can continue and extract just one part, we have to use # the short id for getting the video url - info = {'id': real_id, - 'url': 'http://wat.tv/get/android5/%s.mp4' % real_id, - 'ext': 'mp4', - 'title': first_chapter['title'], - 'thumbnail': first_chapter['preview'], - 'description': first_chapter['description'], - 'view_count': video_info['views'], - } - if 'date_diffusion' in first_chapter: - info['upload_date'] = unified_strdate(first_chapter['date_diffusion']) - - return info + return { + 'id': real_id, + 'url': 'http://wat.tv/get/android5/%s.mp4' % real_id, + 'title': first_chapter['title'], + 'thumbnail': first_chapter['preview'], + 'description': first_chapter['description'], + 'view_count': video_info['views'], + 'upload_date': upload_date, + }