[NRK] Ambiguous subtitle duration thousands

[forthrin]

• I've verified and I assure that I'm running youtube-dl 2018.12.03
• At least skimmed through the README, most notably the FAQ and BUGS sections
• Searched the bugtracker for similar issues including closed ones
• Checked that provided video/audio/playlist URLs (if any) are alive and playable in a browser
• Bug report (encountered problems with youtube-dl)

$ youtube-dl --write-sub --sub-lang no --convert-subtitles srt --verbose https://tv.nrk.no/serie/unge-lovende/sesong/4/episode/2/avspiller
[debug] System config: []
[debug] User config: []
[debug] Custom config: []
[debug] Command-line args: [u'--write-sub', u'--sub-lang', u'no', u'--convert-subtitles', u'srt', u'--verbose', u'https://tv.nrk.no/serie/unge-lovende/sesong/4/episode/2/avspiller']
[debug] Encodings: locale UTF-8, fs utf-8, out utf8, pref UTF-8
[debug] youtube-dl version 2018.12.03
[debug] Python version 2.7.15 (CPython) - Darwin-16.7.0-x86_64-i386-64bit
[debug] exe versions: ffmpeg 4.0.2, ffprobe 4.0.2, rtmpdump 2.4
[debug] Proxy map: {}
[debug] Default format spec: bestvideo+bestaudio/best
[debug] Invoking downloader on u'http://nordond34c-f.akamaihd.net/i/no/open/ps/km/kmte20006218/kmte20006218ca/kmte20006218ca_,141,316,563,1266,2250,.mp4.csmil/index_4_av.m3u8?null=0'

Please see the subtitles at: https://undertekst.nrk.no/prod/KMTE20/00/KMTE20006218CA/TMP/KMTE20006218CA.ttml

  <p begin="00:01:59.320" dur="00:00:04.60" style="left">-Vil dere ha dette?<br />-Det tar vi etterpå.</p>
  <p begin="00:02:03.680" dur="00:00:04.940" style="left">Jeg tenkte du skulle<br />sitte der og Elise her.</p>

Notice the duration that ends with ".60". When converting to .srt, youtube-dl incorrectly interprets this as ".600" rather than the ".060" evidently intended by the author, causing several subtitles to overrun the succeeding subtitle, which in turn makes VLC annoyingly display the subtitles over each other. The problem applies for all subtitled videos on the entire site, so it's a major nuisance.

[evexoio]

04.60 is 04.600

I don't get it, why would we convert .60 to .060? isn't this a mistake from the subtitles author end?

[forthrin]

@evexoio: They probably do something like sprintf("...:%d", thousands) rather than sprintf("...:%03d", thousands).Will fixing it break any other sites?

[evexoio]

@forthrin can you open a PR please?

[forthrin]

I've found a way to get around the issue.

The site provides two subtitle formats, VTT and TTML, but only the latter has the thousands issue.

However, youtube-dl doesn't pick up the VTT file, because the site metadata points to a m3u8 file, which again points to a VTT file. This may be non-standard, but that's how they do it.

Getting the VTT file and adding --sub-format vtt/ttml on the command line, the issue disappears.

I am not proud of this crappy code. I assume it breaks with several of the project's programming conventions, but I couldn't find any other way to do it at the moment, and it works.

Maybe someone can suggest a way to clean up the code, so it can be submitted to the project?

diff --git a/youtube_dl/extractor/nrk.py b/youtube_dl/extractor/nrk.py
index 072f920a9..7a120376a 100644
--- a/youtube_dl/extractor/nrk.py
+++ b/youtube_dl/extractor/nrk.py
@@ -4,2 +4,3 @@ from __future__ import unicode_literals
 import re
+import os
 
@@ -18,2 +19,3 @@ from ..utils import (
     try_get,
+    determine_ext,
 )
@@ -78,5 +80,12 @@ class NRKBaseIE(InfoExtractor):
                     subtitle_url = asset.get('%sSubtitlesUrl' % subtitle)
+                    subtitle_url = compat_urllib_parse_unquote(subtitle_url)
                     if subtitle_url:
+                        ext = determine_ext(subtitle_url)
+                        if ext == "m3u8":
+                            m3u8 = self._download_webpage(subtitle_url, ext)
+                            head = os.path.dirname(subtitle_url)
+                            tail = re.findall(r'(?i)(.*\.vtt)', m3u8)[0]
+                            subtitle_url = '%s/%s' % (head, tail)
                         subtitles.setdefault('no', []).append({
-                            'url': compat_urllib_parse_unquote(subtitle_url)
+                            'url': subtitle_url
                         })
diff --git a/youtube_dl/utils.py b/youtube_dl/utils.py
index 08af57bae..a203a7e19 100644
--- a/youtube_dl/utils.py
+++ b/youtube_dl/utils.py
@@ -2760,2 +2760,5 @@ def parse_dfxp_time_expr(time_expr):
 
+    if re.match(r'^(\d+):(\d\d):(\d\d)[\.:]\d{1,2}$', time_expr):
+        print('NOTICE: Ambiguous timecode in subtitles: ' + time_expr)
+
     mobj = re.match(r'^(\d+):(\d\d):(\d\d(?:(?:\.|:)\d+)?)$', time_expr)