]>
jfr.im git - yt-dlp.git/blob - yt_dlp/webvtt.py
2 from __future__
import unicode_literals
, print_function
, division
5 A partial parser for WebVTT segments. Interprets enough of the WebVTT stream
6 to be able to assemble a single stand-alone subtitle file, suitably adjusting
7 timestamps on the way, while everything else is passed through unmodified.
9 Regular expressions based on the W3C WebVTT specification
10 <https://www.w3.org/TR/webvtt1/>. The X-TIMESTAMP-MAP extension is described
11 in RFC 8216 §3.5 <https://tools.ietf.org/html/rfc8216#section-3.5>.
16 from .utils
import int_or_none
, timetuple_from_msec
24 class _MatchParser(object):
26 An object that maintains the current parsing position and allows
27 conveniently advancing it as syntax elements are successfully parsed.
30 def __init__(self
, string
):
35 if isinstance(r
, compat_Pattern
):
36 return r
.match(self
._data
, self
._pos
)
37 if isinstance(r
, str):
38 if self
._data
.startswith(r
, self
._pos
):
43 def advance(self
, by
):
46 elif isinstance(by
, compat_Match
):
47 amt
= len(by
.group(0))
48 elif isinstance(by
, str):
50 elif isinstance(by
, int):
58 return self
.advance(self
.match(r
))
61 return _MatchChildParser(self
)
64 class _MatchChildParser(_MatchParser
):
66 A child parser state, which advances through the same data as
67 its parent, but has an independent position. This is useful when
68 advancing through syntax elements we might later want to backtrack
72 def __init__(self
, parent
):
73 super(_MatchChildParser
, self
).__init
__(parent
._data
)
74 self
.__parent
= parent
75 self
._pos
= parent
._pos
79 Advance the parent state to the current position of this child state.
81 self
.__parent
._pos
= self
._pos
85 class ParseError(Exception):
86 def __init__(self
, parser
):
87 super(ParseError
, self
).__init
__("Parse error at position %u (near %r)" % (
88 parser
._pos
, parser
._data
[parser
._pos
:parser
._pos
+ 20]
92 # While the specification <https://www.w3.org/TR/webvtt1/#webvtt-timestamp>
93 # prescribes that hours must be *2 or more* digits, timestamps with a single
94 # digit for the hour part has been seen in the wild.
95 # See https://github.com/yt-dlp/yt-dlp/issues/921
96 _REGEX_TS
= re
.compile(r
'''(?x)
102 _REGEX_EOF
= re
.compile(r
'\Z')
103 _REGEX_NL
= re
.compile(r
'(?:\r\n|[\r\n])')
104 _REGEX_BLANK
= re
.compile(r
'(?:\r\n|[\r\n])+')
109 Convert a parsed WebVTT timestamp (a re.Match obtained from _REGEX_TS)
110 into an MPEG PES timestamp: a tick counter at 90 kHz resolution.
113 h
, min, s
, ms
= ts
.groups()
115 int(h
or 0) * 3600000 + # noqa: W504,E221,E222
116 int(min) * 60000 + # noqa: W504,E221,E222
117 int(s
) * 1000 + # noqa: W504,E221,E222
118 int(ms
) # noqa: W504,E221,E222
124 Convert an MPEG PES timestamp into a WebVTT timestamp.
125 This will lose sub-millisecond precision.
127 return '%02u:%02u:%02u.%03u' % timetuple_from_msec(int((ts
+ 45) // 90))
132 An abstract WebVTT block.
135 def __init__(self
, **kwargs
):
136 for key
, val
in kwargs
.items():
137 setattr(self
, key
, val
)
140 def parse(cls
, parser
):
141 m
= parser
.match(cls
._REGEX
)
145 return cls(raw
=m
.group(0))
147 def write_into(self
, stream
):
148 stream
.write(self
.raw
)
151 class HeaderBlock(Block
):
153 A WebVTT block that may only appear in the header part of the file,
154 i.e. before any cue blocks.
160 class Magic(HeaderBlock
):
161 _REGEX
= re
.compile(r
'\ufeff?WEBVTT([ \t][^\r\n]*)?(?:\r\n|[\r\n])')
163 # XXX: The X-TIMESTAMP-MAP extension is described in RFC 8216 §3.5
164 # <https://tools.ietf.org/html/rfc8216#section-3.5>, but the RFC
165 # doesn’t specify the exact grammar nor where in the WebVTT
166 # syntax it should be placed; the below has been devised based
167 # on usage in the wild
169 # And strictly speaking, the presence of this extension violates
170 # the W3C WebVTT spec. Oh well.
172 _REGEX_TSMAP
= re
.compile(r
'X-TIMESTAMP-MAP=')
173 _REGEX_TSMAP_LOCAL
= re
.compile(r
'LOCAL:')
174 _REGEX_TSMAP_MPEGTS
= re
.compile(r
'MPEGTS:([0-9]+)')
175 _REGEX_TSMAP_SEP
= re
.compile(r
'[ \t]*,[ \t]*')
178 def __parse_tsmap(cls
, parser
):
179 parser
= parser
.child()
182 m
= parser
.consume(cls
._REGEX
_TSMAP
_LOCAL
)
184 m
= parser
.consume(_REGEX_TS
)
186 raise ParseError(parser
)
189 raise ParseError(parser
)
191 m
= parser
.consume(cls
._REGEX
_TSMAP
_MPEGTS
)
193 mpegts
= int_or_none(m
.group(1))
195 raise ParseError(parser
)
197 raise ParseError(parser
)
198 if parser
.consume(cls
._REGEX
_TSMAP
_SEP
):
200 if parser
.consume(_REGEX_NL
):
202 raise ParseError(parser
)
208 def parse(cls
, parser
):
209 parser
= parser
.child()
211 m
= parser
.consume(cls
._REGEX
)
213 raise ParseError(parser
)
216 local
, mpegts
= None, None
217 if parser
.consume(cls
._REGEX
_TSMAP
):
218 local
, mpegts
= cls
.__parse
_tsmap
(parser
)
219 if not parser
.consume(_REGEX_NL
):
220 raise ParseError(parser
)
222 return cls(extra
=extra
, mpegts
=mpegts
, local
=local
)
224 def write_into(self
, stream
):
225 stream
.write('WEBVTT')
226 if self
.extra
is not None:
227 stream
.write(self
.extra
)
229 if self
.local
or self
.mpegts
:
230 stream
.write('X-TIMESTAMP-MAP=LOCAL:')
231 stream
.write(_format_ts(self
.local
if self
.local
is not None else 0))
232 stream
.write(',MPEGTS:')
233 stream
.write(str(self
.mpegts
if self
.mpegts
is not None else 0))
238 class StyleBlock(HeaderBlock
):
239 _REGEX
= re
.compile(r
'''(?x)
240 STYLE[\ \t]*(?:\r\n|[\r\n])
241 ((?:(?!-->)[^\r\n])+(?:\r\n|[\r\n]))*
246 class RegionBlock(HeaderBlock
):
247 _REGEX
= re
.compile(r
'''(?x)
249 ((?:(?!-->)[^\r\n])+(?:\r\n|[\r\n]))*
254 class CommentBlock(Block
):
255 _REGEX
= re
.compile(r
'''(?x)
256 NOTE(?:\r\n|[\ \t\r\n])
257 ((?:(?!-->)[^\r\n])+(?:\r\n|[\r\n]))*
262 class CueBlock(Block
):
264 A cue block. The payload is not interpreted.
267 _REGEX_ID
= re
.compile(r
'((?:(?!-->)[^\r\n])+)(?:\r\n|[\r\n])')
268 _REGEX_ARROW
= re
.compile(r
'[ \t]+-->[ \t]+')
269 _REGEX_SETTINGS
= re
.compile(r
'[ \t]+((?:(?!-->)[^\r\n])+)')
270 _REGEX_PAYLOAD
= re
.compile(r
'[^\r\n]+(?:\r\n|[\r\n])?')
273 def parse(cls
, parser
):
274 parser
= parser
.child()
277 m
= parser
.consume(cls
._REGEX
_ID
)
281 m0
= parser
.consume(_REGEX_TS
)
284 if not parser
.consume(cls
._REGEX
_ARROW
):
286 m1
= parser
.consume(_REGEX_TS
)
289 m2
= parser
.consume(cls
._REGEX
_SETTINGS
)
290 if not parser
.consume(_REGEX_NL
):
293 start
= _parse_ts(m0
)
295 settings
= m2
.group(1) if m2
is not None else None
299 m
= parser
.consume(cls
._REGEX
_PAYLOAD
)
302 text
.write(m
.group(0))
307 start
=start
, end
=end
, settings
=settings
,
311 def write_into(self
, stream
):
312 if self
.id is not None:
313 stream
.write(self
.id)
315 stream
.write(_format_ts(self
.start
))
316 stream
.write(' --> ')
317 stream
.write(_format_ts(self
.end
))
318 if self
.settings
is not None:
320 stream
.write(self
.settings
)
322 stream
.write(self
.text
)
332 'settings': self
.settings
,
335 def __eq__(self
, other
):
336 return self
.as_json
== other
.as_json
339 def from_json(cls
, json
):
345 settings
=json
['settings']
348 def hinges(self
, other
):
349 if self
.text
!= other
.text
:
351 if self
.settings
!= other
.settings
:
353 return self
.start
<= self
.end
== other
.start
<= other
.end
356 def parse_fragment(frag_content
):
358 A generator that yields (partially) parsed WebVTT blocks when given
359 a bytes object containing the raw contents of a WebVTT file.
362 parser
= _MatchParser(frag_content
.decode('utf-8'))
364 yield Magic
.parse(parser
)
366 while not parser
.match(_REGEX_EOF
):
367 if parser
.consume(_REGEX_BLANK
):
370 block
= RegionBlock
.parse(parser
)
374 block
= StyleBlock
.parse(parser
)
378 block
= CommentBlock
.parse(parser
)
380 yield block
# XXX: or skip
385 while not parser
.match(_REGEX_EOF
):
386 if parser
.consume(_REGEX_BLANK
):
389 block
= CommentBlock
.parse(parser
)
391 yield block
# XXX: or skip
393 block
= CueBlock
.parse(parser
)
398 raise ParseError(parser
)