[webvtt] Merge daisy-chained duplicate cues (#638)

Fixes: https://github.com/yt-dlp/yt-dlp/issues/631#issuecomment-893338552

Previous deduplication algorithm only removed duplicate cues with
identical text, styles and timestamps.  This change also merges
cues that come in ‘daisy chains’, where sequences of cues with
identical text and styles appear in which the ending timestamp of
one equals the starting timestamp of the next.

This deduplication algorithm has the somewhat unfortunate side effect
that NOTE blocks between cues, if found, will be emitted in a different
order relative to their original cues.  This may be unwanted if perfect
fidelity is desired, but then so is daisy-chain deduplication itself.
NOTE blocks ought to be ignored by WebVTT players in any case.

Authored by: fstirlitz
This commit is contained in:
Felix S 2021-08-09 20:22:30 +00:00 committed by GitHub
parent ad3dc496bb
commit 25a3f4f5d6
No known key found for this signature in database
GPG Key ID: 4AEE18F83AFDEB23
3 changed files with 61 additions and 19 deletions

View File

@ -329,7 +329,7 @@ def _prepare_external_frag_download(self, ctx):
'fragment_index': 0, 'fragment_index': 0,
}) })
def download_and_append_fragments(self, ctx, fragments, info_dict, pack_func=None): def download_and_append_fragments(self, ctx, fragments, info_dict, *, pack_func=None, finish_func=None):
fragment_retries = self.params.get('fragment_retries', 0) fragment_retries = self.params.get('fragment_retries', 0)
is_fatal = (lambda idx: idx == 0) if self.params.get('skip_unavailable_fragments', True) else (lambda _: True) is_fatal = (lambda idx: idx == 0) if self.params.get('skip_unavailable_fragments', True) else (lambda _: True)
if not pack_func: if not pack_func:
@ -424,5 +424,8 @@ def _download_fragment(fragment):
if not result: if not result:
return False return False
if finish_func is not None:
ctx['dest_stream'].write(finish_func())
ctx['dest_stream'].flush()
self._finish_frag_download(ctx, info_dict) self._finish_frag_download(ctx, info_dict)
return True return True

View File

@ -260,29 +260,35 @@ def pack_fragment(frag_content, frag_index):
block.end += adjust block.end += adjust
dedup_window = extra_state.setdefault('webvtt_dedup_window', []) dedup_window = extra_state.setdefault('webvtt_dedup_window', [])
cue = block.as_json
# skip the cue if an identical one appears ready = []
# in the window of potential duplicates
# and prune the window of unviable candidates
i = 0 i = 0
skip = True is_new = True
while i < len(dedup_window): while i < len(dedup_window):
window_cue = dedup_window[i] wcue = dedup_window[i]
if window_cue == cue: wblock = webvtt.CueBlock.from_json(wcue)
break i += 1
if window_cue['end'] >= cue['start']: if wblock.hinges(block):
i += 1 wcue['end'] = block.end
is_new = False
continue continue
if wblock == block:
is_new = False
continue
if wblock.end > block.start:
continue
ready.append(wblock)
i -= 1
del dedup_window[i] del dedup_window[i]
else:
skip = False
if skip: if is_new:
continue dedup_window.append(block.as_json)
for block in ready:
block.write_into(output)
# add the cue to the window # we only emit cues once they fall out of the duplicate window
dedup_window.append(cue) continue
elif isinstance(block, webvtt.Magic): elif isinstance(block, webvtt.Magic):
# take care of MPEG PES timestamp overflow # take care of MPEG PES timestamp overflow
if block.mpegts is None: if block.mpegts is None:
@ -317,6 +323,19 @@ def pack_fragment(frag_content, frag_index):
block.write_into(output) block.write_into(output)
return output.getvalue().encode('utf-8') return output.getvalue().encode('utf-8')
def fin_fragments():
dedup_window = extra_state.get('webvtt_dedup_window')
if not dedup_window:
return b''
output = io.StringIO()
for cue in dedup_window:
webvtt.CueBlock.from_json(cue).write_into(output)
return output.getvalue().encode('utf-8')
self.download_and_append_fragments(
ctx, fragments, info_dict, pack_func=pack_fragment, finish_func=fin_fragments)
else: else:
pack_fragment = None return self.download_and_append_fragments(ctx, fragments, info_dict)
return self.download_and_append_fragments(ctx, fragments, info_dict, pack_fragment)

View File

@ -331,6 +331,26 @@ def as_json(self):
'settings': self.settings, 'settings': self.settings,
} }
def __eq__(self, other):
return self.as_json == other.as_json
@classmethod
def from_json(cls, json):
return cls(
id=json['id'],
start=json['start'],
end=json['end'],
text=json['text'],
settings=json['settings']
)
def hinges(self, other):
if self.text != other.text:
return False
if self.settings != other.settings:
return False
return self.start <= self.end == other.start <= other.end
def parse_fragment(frag_content): def parse_fragment(frag_content):
""" """