forked from minhngoc25a/youtube-dl
ceskatelevize: Closed captions support
This commit is contained in:
parent
324b2c78fa
commit
c067545c17
|
@ -3,7 +3,7 @@ from __future__ import unicode_literals
|
|||
|
||||
import re
|
||||
|
||||
from .common import InfoExtractor
|
||||
from .subtitles import SubtitlesInfoExtractor
|
||||
from ..compat import (
|
||||
compat_urllib_request,
|
||||
compat_urllib_parse,
|
||||
|
@ -12,10 +12,11 @@ from ..compat import (
|
|||
from ..utils import (
|
||||
ExtractorError,
|
||||
float_or_none,
|
||||
HEADRequest,
|
||||
)
|
||||
|
||||
|
||||
class CeskaTelevizeIE(InfoExtractor):
|
||||
class CeskaTelevizeIE(SubtitlesInfoExtractor):
|
||||
_VALID_URL = r'https?://www\.ceskatelevize\.cz/(porady|ivysilani)/(.+/)?(?P<id>[^?#]+)'
|
||||
|
||||
_TESTS = [
|
||||
|
@ -104,6 +105,15 @@ class CeskaTelevizeIE(InfoExtractor):
|
|||
duration = float_or_none(item.get('duration'))
|
||||
thumbnail = item.get('previewImageUrl')
|
||||
|
||||
# subtitles
|
||||
subtitles = self.extract_subtitles(video_id, webpage)
|
||||
|
||||
if self._downloader.params.get('listsubtitles', False):
|
||||
self._list_available_subtitles(video_id, webpage)
|
||||
return
|
||||
|
||||
subtitles = self._fix_subtitles(self.extract_subtitles(video_id, webpage))
|
||||
|
||||
return {
|
||||
'id': episode_id,
|
||||
'title': title,
|
||||
|
@ -111,4 +121,45 @@ class CeskaTelevizeIE(InfoExtractor):
|
|||
'thumbnail': thumbnail,
|
||||
'duration': duration,
|
||||
'formats': formats,
|
||||
'subtitles': subtitles,
|
||||
}
|
||||
|
||||
def _fix_subtitles(self, subtitles):
|
||||
""" Convert milisecond-based subtitles to SRT """
|
||||
if subtitles is None:
|
||||
return subtitles # subtitles not requested
|
||||
|
||||
def _msectotimecode(msec):
|
||||
""" Helper utility to convert miliseconds to timecode """
|
||||
components = []
|
||||
for divider in [1000, 60, 60, 100]:
|
||||
components.append(msec % divider)
|
||||
msec //= divider
|
||||
return "{3:02}:{2:02}:{1:02},{0:03}".format(*components)
|
||||
|
||||
def _fix_subtitle(subtitle):
|
||||
for line in subtitle.splitlines():
|
||||
m = re.match(r"^ *([0-9]+); *([0-9]+) +([0-9]+) *$", line)
|
||||
if m:
|
||||
yield m.group(1)
|
||||
start, stop = (_msectotimecode(int(t)) for t in m.groups()[1:])
|
||||
yield "{} --> {}".format(start, stop)
|
||||
else:
|
||||
yield line
|
||||
|
||||
fixed_subtitles = {}
|
||||
for k, v in subtitles.items():
|
||||
fixed_subtitles[k] = "\r\n".join(_fix_subtitle(v))
|
||||
return fixed_subtitles
|
||||
|
||||
def _get_available_subtitles(self, video_id, webpage):
|
||||
video_id = video_id.partition('-')[0]
|
||||
url = 'http://imgct.ceskatelevize.cz/cache/data/ivysilani/' \
|
||||
'subtitles/{}/{}/sub.txt'.format(video_id[:3], video_id)
|
||||
req = HEADRequest(url)
|
||||
sub = self._request_webpage(
|
||||
req, video_id,
|
||||
note="Checking subtitles",
|
||||
errnote="No subtitles found",
|
||||
fatal=False)
|
||||
return {'cs': url} if sub else {}
|
||||
|
|
Loading…
Reference in New Issue