162 lines
		
	
	
		
			5.6 KiB
		
	
	
	
		
			Python
		
	
	
	
	
	
			
		
		
	
	
			162 lines
		
	
	
		
			5.6 KiB
		
	
	
	
		
			Python
		
	
	
	
	
	
| # coding: utf-8
 | |
| from __future__ import unicode_literals
 | |
| 
 | |
| import re
 | |
| 
 | |
| from .common import InfoExtractor
 | |
| from ..compat import (
 | |
|     compat_urllib_parse_urlencode,
 | |
|     compat_urlparse,
 | |
| )
 | |
| from ..utils import (
 | |
|     clean_html,
 | |
|     parse_duration,
 | |
|     str_to_int,
 | |
|     unified_strdate,
 | |
| )
 | |
| 
 | |
| 
 | |
| class CamdemyIE(InfoExtractor):
 | |
|     _VALID_URL = r'https?://(?:www\.)?camdemy\.com/media/(?P<id>\d+)'
 | |
|     _TESTS = [{
 | |
|         # single file
 | |
|         'url': 'http://www.camdemy.com/media/5181/',
 | |
|         'md5': '5a5562b6a98b37873119102e052e311b',
 | |
|         'info_dict': {
 | |
|             'id': '5181',
 | |
|             'ext': 'mp4',
 | |
|             'title': 'Ch1-1 Introduction, Signals (02-23-2012)',
 | |
|             'thumbnail': 're:^https?://.*\.jpg$',
 | |
|             'creator': 'ss11spring',
 | |
|             'duration': 1591,
 | |
|             'upload_date': '20130114',
 | |
|             'view_count': int,
 | |
|         }
 | |
|     }, {
 | |
|         # With non-empty description
 | |
|         # webpage returns "No permission or not login"
 | |
|         'url': 'http://www.camdemy.com/media/13885',
 | |
|         'md5': '4576a3bb2581f86c61044822adbd1249',
 | |
|         'info_dict': {
 | |
|             'id': '13885',
 | |
|             'ext': 'mp4',
 | |
|             'title': 'EverCam + Camdemy QuickStart',
 | |
|             'thumbnail': 're:^https?://.*\.jpg$',
 | |
|             'description': 'md5:2a9f989c2b153a2342acee579c6e7db6',
 | |
|             'creator': 'evercam',
 | |
|             'duration': 318,
 | |
|         }
 | |
|     }, {
 | |
|         # External source (YouTube)
 | |
|         'url': 'http://www.camdemy.com/media/14842',
 | |
|         'info_dict': {
 | |
|             'id': '2vsYQzNIsJo',
 | |
|             'ext': 'mp4',
 | |
|             'title': 'Excel 2013 Tutorial - How to add Password Protection',
 | |
|             'description': 'Excel 2013 Tutorial for Beginners - How to add Password Protection',
 | |
|             'upload_date': '20130211',
 | |
|             'uploader': 'Hun Kim',
 | |
|             'uploader_id': 'hunkimtutorials',
 | |
|         },
 | |
|         'params': {
 | |
|             'skip_download': True,
 | |
|         },
 | |
|     }]
 | |
| 
 | |
|     def _real_extract(self, url):
 | |
|         video_id = self._match_id(url)
 | |
| 
 | |
|         webpage = self._download_webpage(url, video_id)
 | |
| 
 | |
|         src_from = self._html_search_regex(
 | |
|             r"class=['\"]srcFrom['\"][^>]*>Sources?(?:\s+from)?\s*:\s*<a[^>]+(?:href|title)=(['\"])(?P<url>(?:(?!\1).)+)\1",
 | |
|             webpage, 'external source', default=None, group='url')
 | |
|         if src_from:
 | |
|             return self.url_result(src_from)
 | |
| 
 | |
|         oembed_obj = self._download_json(
 | |
|             'http://www.camdemy.com/oembed/?format=json&url=' + url, video_id)
 | |
| 
 | |
|         title = oembed_obj['title']
 | |
|         thumb_url = oembed_obj['thumbnail_url']
 | |
|         video_folder = compat_urlparse.urljoin(thumb_url, 'video/')
 | |
|         file_list_doc = self._download_xml(
 | |
|             compat_urlparse.urljoin(video_folder, 'fileList.xml'),
 | |
|             video_id, 'Downloading filelist XML')
 | |
|         file_name = file_list_doc.find('./video/item/fileName').text
 | |
|         video_url = compat_urlparse.urljoin(video_folder, file_name)
 | |
| 
 | |
|         # Some URLs return "No permission or not login" in a webpage despite being
 | |
|         # freely available via oembed JSON URL (e.g. http://www.camdemy.com/media/13885)
 | |
|         upload_date = unified_strdate(self._search_regex(
 | |
|             r'>published on ([^<]+)<', webpage,
 | |
|             'upload date', default=None))
 | |
|         view_count = str_to_int(self._search_regex(
 | |
|             r'role=["\']viewCnt["\'][^>]*>([\d,.]+) views',
 | |
|             webpage, 'view count', default=None))
 | |
|         description = self._html_search_meta(
 | |
|             'description', webpage, default=None) or clean_html(
 | |
|             oembed_obj.get('description'))
 | |
| 
 | |
|         return {
 | |
|             'id': video_id,
 | |
|             'url': video_url,
 | |
|             'title': title,
 | |
|             'thumbnail': thumb_url,
 | |
|             'description': description,
 | |
|             'creator': oembed_obj.get('author_name'),
 | |
|             'duration': parse_duration(oembed_obj.get('duration')),
 | |
|             'upload_date': upload_date,
 | |
|             'view_count': view_count,
 | |
|         }
 | |
| 
 | |
| 
 | |
| class CamdemyFolderIE(InfoExtractor):
 | |
|     _VALID_URL = r'https?://www.camdemy.com/folder/(?P<id>\d+)'
 | |
|     _TESTS = [{
 | |
|         # links with trailing slash
 | |
|         'url': 'http://www.camdemy.com/folder/450',
 | |
|         'info_dict': {
 | |
|             'id': '450',
 | |
|             'title': '信號與系統 2012 & 2011 (Signals and Systems)',
 | |
|         },
 | |
|         'playlist_mincount': 145
 | |
|     }, {
 | |
|         # links without trailing slash
 | |
|         # and multi-page
 | |
|         'url': 'http://www.camdemy.com/folder/853',
 | |
|         'info_dict': {
 | |
|             'id': '853',
 | |
|             'title': '科學計算 - 使用 Matlab'
 | |
|         },
 | |
|         'playlist_mincount': 20
 | |
|     }, {
 | |
|         # with displayMode parameter. For testing the codes to add parameters
 | |
|         'url': 'http://www.camdemy.com/folder/853/?displayMode=defaultOrderByOrg',
 | |
|         'info_dict': {
 | |
|             'id': '853',
 | |
|             'title': '科學計算 - 使用 Matlab'
 | |
|         },
 | |
|         'playlist_mincount': 20
 | |
|     }]
 | |
| 
 | |
|     def _real_extract(self, url):
 | |
|         folder_id = self._match_id(url)
 | |
| 
 | |
|         # Add displayMode=list so that all links are displayed in a single page
 | |
|         parsed_url = list(compat_urlparse.urlparse(url))
 | |
|         query = dict(compat_urlparse.parse_qsl(parsed_url[4]))
 | |
|         query.update({'displayMode': 'list'})
 | |
|         parsed_url[4] = compat_urllib_parse_urlencode(query)
 | |
|         final_url = compat_urlparse.urlunparse(parsed_url)
 | |
| 
 | |
|         page = self._download_webpage(final_url, folder_id)
 | |
|         matches = re.findall(r"href='(/media/\d+/?)'", page)
 | |
| 
 | |
|         entries = [self.url_result('http://www.camdemy.com' + media_path)
 | |
|                    for media_path in matches]
 | |
| 
 | |
|         folder_title = self._html_search_meta('keywords', page)
 | |
| 
 | |
|         return self.playlist_result(entries, folder_id, folder_title)
 |