- # encoding: utf-8
-
- from __future__ import unicode_literals
-
- from .common import InfoExtractor
- from ..compat import compat_urllib_parse
- from ..utils import (
- int_or_none,
- str_to_int,
- xpath_text,
- )
-
-
- class DaumIE(InfoExtractor):
- _VALID_URL = r'https?://(?:m\.)?tvpot\.daum\.net/v/(?P<id>[^?#&]+)'
- IE_NAME = 'daum.net'
-
- _TESTS = [{
- 'url': 'http://tvpot.daum.net/v/vab4dyeDBysyBssyukBUjBz',
- 'info_dict': {
- 'id': 'vab4dyeDBysyBssyukBUjBz',
- 'ext': 'mp4',
- 'title': '마크 헌트 vs 안토니오 실바',
- 'description': 'Mark Hunt vs Antonio Silva',
- 'upload_date': '20131217',
- 'duration': 2117,
- 'view_count': int,
- 'comment_count': int,
- },
- }, {
- 'url': 'http://tvpot.daum.net/v/07dXWRka62Y%24',
- 'only_matching': True,
- }]
-
- def _real_extract(self, url):
- video_id = self._match_id(url)
- query = compat_urllib_parse.urlencode({'vid': video_id})
- info = self._download_xml(
- 'http://tvpot.daum.net/clip/ClipInfoXml.do?' + query, video_id,
- 'Downloading video info')
- movie_data = self._download_json(
- 'http://videofarm.daum.net/controller/api/closed/v1_2/IntegratedMovieData.json?' + query,
- video_id, 'Downloading video formats info')
-
- formats = []
- for format_el in movie_data['output_list']['output_list']:
- profile = format_el['profile']
- format_query = compat_urllib_parse.urlencode({
- 'vid': video_id,
- 'profile': profile,
- })
- url_doc = self._download_xml(
- 'http://videofarm.daum.net/controller/api/open/v1_2/MovieLocation.apixml?' + format_query,
- video_id, note='Downloading video data for %s format' % profile)
- format_url = url_doc.find('result/url').text
- formats.append({
- 'url': format_url,
- 'format_id': profile,
- 'width': int_or_none(format_el.get('width')),
- 'height': int_or_none(format_el.get('height')),
- 'filesize': int_or_none(format_el.get('filesize')),
- })
- self._sort_formats(formats)
-
- return {
- 'id': video_id,
- 'title': info.find('TITLE').text,
- 'formats': formats,
- 'thumbnail': xpath_text(info, 'THUMB_URL'),
- 'description': xpath_text(info, 'CONTENTS'),
- 'duration': int_or_none(xpath_text(info, 'DURATION')),
- 'upload_date': info.find('REGDTTM').text[:8],
- 'view_count': str_to_int(xpath_text(info, 'PLAY_CNT')),
- 'comment_count': str_to_int(xpath_text(info, 'COMMENT_CNT')),
- }
-
-
- class DaumClipIE(InfoExtractor):
- _VALID_URL = r'https?://(?:m\.)?tvpot\.daum\.net/(?:clip/ClipView.do|mypot/View.do)\?.*?clipid=(?P<id>\d+)'
- IE_NAME = 'daum.net:clip'
-
- _TESTS = [{
- 'url': 'http://tvpot.daum.net/clip/ClipView.do?clipid=52554690',
- 'info_dict': {
- 'id': '52554690',
- 'ext': 'mp4',
- 'title': 'DOTA 2GETHER 시즌2 6회 - 2부',
- 'description': 'DOTA 2GETHER 시즌2 6회 - 2부',
- 'upload_date': '20130831',
- 'duration': 3868,
- 'view_count': int,
- },
- }]
-
- def _real_extract(self, url):
- video_id = self._match_id(url)
- clip_info = self._download_json(
- 'http://tvpot.daum.net/mypot/json/GetClipInfo.do?clipid=%s' % video_id,
- video_id, 'Downloading clip info')['clip_bean']
-
- return {
- '_type': 'url_transparent',
- 'id': video_id,
- 'url': 'http://tvpot.daum.net/v/%s' % clip_info['vid'],
- 'title': clip_info['title'],
- 'thumbnail': clip_info.get('thumb_url'),
- 'description': clip_info.get('contents'),
- 'duration': int_or_none(clip_info.get('duration')),
- 'upload_date': clip_info.get('up_date')[:8],
- 'view_count': int_or_none(clip_info.get('play_count')),
- 'ie_key': 'Daum',
- }
|