aboutsummaryrefslogtreecommitdiff
diff options
context:
space:
mode:
-rw-r--r--youtube_dl/FileDownloader.py1
-rw-r--r--youtube_dl/InfoExtractors.py78
2 files changed, 40 insertions, 39 deletions
diff --git a/youtube_dl/FileDownloader.py b/youtube_dl/FileDownloader.py
index 1fdd2071f..a7997c4f2 100644
--- a/youtube_dl/FileDownloader.py
+++ b/youtube_dl/FileDownloader.py
@@ -327,6 +327,7 @@ class FileDownloader(object):
template_dict = dict(info_dict)
template_dict['epoch'] = unicode(int(time.time()))
template_dict['autonumber'] = unicode('%05d' % self._num_downloads)
+ template_dict = dict((key, u'NA' if val is None else val) for key, val in template_dict.items())
filename = self.params['outtmpl'] % template_dict
return filename
except (ValueError, KeyError), err:
diff --git a/youtube_dl/InfoExtractors.py b/youtube_dl/InfoExtractors.py
index 7c6e1becc..b730b5773 100644
--- a/youtube_dl/InfoExtractors.py
+++ b/youtube_dl/InfoExtractors.py
@@ -380,7 +380,7 @@ class YoutubeIE(InfoExtractor):
video_thumbnail = urllib.unquote_plus(video_info['thumbnail_url'][0])
# upload date
- upload_date = u'NA'
+ upload_date = None
mobj = re.search(r'id="eow-date.*?>(.*?)</span>', video_webpage, re.DOTALL)
if mobj is not None:
upload_date = ' '.join(re.sub(r'[/,-]', r' ', mobj.group(1)).split())
@@ -631,7 +631,7 @@ class MetacafeIE(InfoExtractor):
'id': video_id.decode('utf-8'),
'url': video_url.decode('utf-8'),
'uploader': video_uploader.decode('utf-8'),
- 'upload_date': u'NA',
+ 'upload_date': None,
'title': video_title,
'ext': video_extension.decode('utf-8'),
}]
@@ -707,7 +707,7 @@ class DailymotionIE(InfoExtractor):
return
video_title = unescapeHTML(mobj.group('title').decode('utf-8'))
- video_uploader = u'NA'
+ video_uploader = None
mobj = re.search(r'(?im)<span class="owner[^\"]+?">[^<]+?<a [^>]+?>([^<]+?)</a>', webpage)
if mobj is None:
# lookin for official user
@@ -719,7 +719,7 @@ class DailymotionIE(InfoExtractor):
else:
video_uploader = mobj.group(1)
- video_upload_date = u'NA'
+ video_upload_date = None
mobj = re.search(r'<div class="[^"]*uploaded_cont[^"]*" title="[^"]*">([0-9]{2})-([0-9]{2})-([0-9]{4})</div>', webpage)
if mobj is not None:
video_upload_date = mobj.group(3) + mobj.group(2) + mobj.group(1)
@@ -820,8 +820,8 @@ class GoogleIE(InfoExtractor):
return [{
'id': video_id.decode('utf-8'),
'url': video_url.decode('utf-8'),
- 'uploader': u'NA',
- 'upload_date': u'NA',
+ 'uploader': None,
+ 'upload_date': None,
'title': video_title,
'ext': video_extension.decode('utf-8'),
}]
@@ -886,7 +886,7 @@ class PhotobucketIE(InfoExtractor):
'id': video_id.decode('utf-8'),
'url': video_url.decode('utf-8'),
'uploader': video_uploader,
- 'upload_date': u'NA',
+ 'upload_date': None,
'title': video_title,
'ext': video_extension.decode('utf-8'),
}]
@@ -1026,7 +1026,7 @@ class YahooIE(InfoExtractor):
'id': video_id.decode('utf-8'),
'url': video_url,
'uploader': video_uploader,
- 'upload_date': u'NA',
+ 'upload_date': None,
'title': video_title,
'ext': video_extension.decode('utf-8'),
'thumbnail': video_thumbnail.decode('utf-8'),
@@ -1098,7 +1098,7 @@ class VimeoIE(InfoExtractor):
else: video_description = ''
# Extract upload date
- video_upload_date = u'NA'
+ video_upload_date = None
mobj = re.search(r'<span id="clip-date" style="display:none">[^:]*: (.*?)( \([^\(]*\))?</span>', webpage)
if mobj is not None:
video_upload_date = mobj.group(1)
@@ -1286,7 +1286,7 @@ class GenericIE(InfoExtractor):
'id': video_id.decode('utf-8'),
'url': video_url.decode('utf-8'),
'uploader': video_uploader,
- 'upload_date': u'NA',
+ 'upload_date': None,
'title': video_title,
'ext': video_extension.decode('utf-8'),
}]
@@ -1889,8 +1889,8 @@ class DepositFilesIE(InfoExtractor):
return [{
'id': file_id.decode('utf-8'),
'url': file_url.decode('utf-8'),
- 'uploader': u'NA',
- 'upload_date': u'NA',
+ 'uploader': None,
+ 'upload_date': None,
'title': file_title,
'ext': file_extension.decode('utf-8'),
}]
@@ -2044,7 +2044,7 @@ class FacebookIE(InfoExtractor):
video_thumbnail = video_info['thumbnail']
# upload date
- upload_date = u'NA'
+ upload_date = None
if 'upload_date' in video_info:
upload_time = video_info['upload_date']
timetuple = email.utils.parsedate_tz(upload_time)
@@ -2142,8 +2142,8 @@ class BlipTVIE(InfoExtractor):
info = {
'id': title,
'url': url,
- 'uploader': u'NA',
- 'upload_date': u'NA',
+ 'uploader': None,
+ 'upload_date': None,
'title': title,
'ext': ext,
'urlhandle': urlh
@@ -2244,8 +2244,8 @@ class MyVideoIE(InfoExtractor):
return [{
'id': video_id,
'url': video_url,
- 'uploader': u'NA',
- 'upload_date': u'NA',
+ 'uploader': None,
+ 'upload_date': None,
'title': video_title,
'ext': u'flv',
}]
@@ -2501,7 +2501,7 @@ class EscapistIE(InfoExtractor):
'id': videoId,
'url': videoUrl,
'uploader': showName,
- 'upload_date': u'NA',
+ 'upload_date': None,
'title': showName,
'ext': 'flv',
'thumbnail': imgUrl,
@@ -2550,8 +2550,8 @@ class CollegeHumorIE(InfoExtractor):
info = {
'id': video_id,
'internal_id': internal_video_id,
- 'uploader': u'NA',
- 'upload_date': u'NA',
+ 'uploader': None,
+ 'upload_date': None,
}
self.report_extraction(video_id)
@@ -2636,8 +2636,8 @@ class XVideosIE(InfoExtractor):
info = {
'id': video_id,
'url': video_url,
- 'uploader': u'NA',
- 'upload_date': u'NA',
+ 'uploader': None,
+ 'upload_date': None,
'title': video_title,
'ext': 'flv',
'thumbnail': video_thumbnail,
@@ -2717,7 +2717,7 @@ class SoundcloudIE(InfoExtractor):
description = mobj.group(1)
# upload date
- upload_date = u'NA'
+ upload_date = None
mobj = re.search("pretty-date'>on ([\w]+ [\d]+, [\d]+ \d+:\d+)</abbr></h2>", webpage)
if mobj:
try:
@@ -2798,8 +2798,8 @@ class InfoQIE(InfoExtractor):
info = {
'id': video_id,
'url': video_url,
- 'uploader': u'NA',
- 'upload_date': u'NA',
+ 'uploader': None,
+ 'upload_date': None,
'title': video_title,
'ext': extension, # Extension is always(?) mp4, but seems to be flv
'thumbnail': None,
@@ -2912,7 +2912,7 @@ class MixcloudIE(InfoExtractor):
'id': file_id.decode('utf-8'),
'url': file_url.decode('utf-8'),
'uploader': uploader.decode('utf-8'),
- 'upload_date': u'NA',
+ 'upload_date': None,
'title': json_data['name'],
'ext': file_url.split('.')[-1].decode('utf-8'),
'format': (format_param is None and u'NA' or format_param.decode('utf-8')),
@@ -2946,8 +2946,8 @@ class StanfordOpenClassroomIE(InfoExtractor):
video = mobj.group('video')
info = {
'id': course + '_' + video,
- 'uploader': u'NA',
- 'upload_date': u'NA',
+ 'uploader': None,
+ 'upload_date': None,
}
self.report_extraction(info['id'])
@@ -2972,8 +2972,8 @@ class StanfordOpenClassroomIE(InfoExtractor):
info = {
'id': course,
'type': 'playlist',
- 'uploader': u'NA',
- 'upload_date': u'NA',
+ 'uploader': None,
+ 'upload_date': None,
}
self.report_download_webpage(info['id'])
@@ -3010,8 +3010,8 @@ class StanfordOpenClassroomIE(InfoExtractor):
info = {
'id': 'Stanford OpenClassroom',
'type': 'playlist',
- 'uploader': u'NA',
- 'upload_date': u'NA',
+ 'uploader': None,
+ 'upload_date': None,
}
self.report_download_webpage(info['id'])
@@ -3120,7 +3120,7 @@ class MTVIE(InfoExtractor):
'id': video_id,
'url': video_url,
'uploader': performer,
- 'upload_date': u'NA',
+ 'upload_date': None,
'title': video_title,
'ext': ext,
'format': format,
@@ -3241,8 +3241,8 @@ class YoukuIE(InfoExtractor):
info = {
'id': '%s_part%02d' % (video_id, index),
'url': download_url,
- 'uploader': u'NA',
- 'upload_date': u'NA',
+ 'uploader': None,
+ 'upload_date': None,
'title': video_title,
'ext': ext,
}
@@ -3305,8 +3305,8 @@ class XNXXIE(InfoExtractor):
return [{
'id': video_id,
'url': video_url,
- 'uploader': u'NA',
- 'upload_date': u'NA',
+ 'uploader': None,
+ 'upload_date': None,
'title': video_title,
'ext': 'flv',
'thumbnail': video_thumbnail,
@@ -3365,7 +3365,7 @@ class GooglePlusIE(InfoExtractor):
return
# Extract update date
- upload_date = u'NA'
+ upload_date = None
pattern = 'title="Timestamp">(.*?)</a>'
mobj = re.search(pattern, webpage)
if mobj:
@@ -3376,7 +3376,7 @@ class GooglePlusIE(InfoExtractor):
self.report_date(upload_date)
# Extract uploader
- uploader = u'NA'
+ uploader = None
pattern = r'rel\="author".*?>(.*?)</a>'
mobj = re.search(pattern, webpage)
if mobj: