summaryrefslogtreecommitdiffstats
path: root/yt_dlp/webvtt.py
blob: 9f1a5086b8b9b929bc0b5758ef19958f8ee5c82a (plain)
1
2
3
4
5
6
7
8
9
10
11
12
13
14
15
16
17
18
19
20
21
22
23
24
25
26
27
28
29
30
31
32
33
34
35
36
37
38
39
40
41
42
43
44
45
46
47
48
49
50
51
52
53
54
55
56
57
58
59
60
61
62
63
64
65
66
67
68
69
70
71
72
73
74
75
76
77
78
79
80
81
82
83
84
85
86
87
88
89
90
91
92
93
94
95
96
97
98
99
100
101
102
103
104
105
106
107
108
109
110
111
112
113
114
115
116
117
118
119
120
121
122
123
124
125
126
127
128
129
130
131
132
133
134
135
136
137
138
139
140
141
142
143
144
145
146
147
148
149
150
151
152
153
154
155
156
157
158
159
160
161
162
163
164
165
166
167
168
169
170
171
172
173
174
175
176
177
178
179
180
181
182
183
184
185
186
187
188
189
190
191
192
193
194
195
196
197
198
199
200
201
202
203
204
205
206
207
208
209
210
211
212
213
214
215
216
217
218
219
220
221
222
223
224
225
226
227
228
229
230
231
232
233
234
235
236
237
238
239
240
241
242
243
244
245
246
247
248
249
250
251
252
253
254
255
256
257
258
259
260
261
262
263
264
265
266
267
268
269
270
271
272
273
274
275
276
277
278
279
280
281
282
283
284
285
286
287
288
289
290
291
292
293
294
295
296
297
298
299
300
301
302
303
304
305
306
307
308
309
310
311
312
313
314
315
316
317
318
319
320
321
322
323
324
325
326
327
328
329
330
331
332
333
334
335
336
337
338
339
340
341
342
343
344
345
346
347
348
349
350
351
352
353
354
355
356
357
358
359
360
361
362
363
364
365
366
367
368
369
370
371
372
373
374
375
376
377
378
379
380
381
382
383
384
385
386
387
388
389
390
391
392
393
394
395
396
397
398
"""
A partial parser for WebVTT segments. Interprets enough of the WebVTT stream
to be able to assemble a single stand-alone subtitle file, suitably adjusting
timestamps on the way, while everything else is passed through unmodified.

Regular expressions based on the W3C WebVTT specification
<https://www.w3.org/TR/webvtt1/>. The X-TIMESTAMP-MAP extension is described
in RFC 8216 §3.5 <https://tools.ietf.org/html/rfc8216#section-3.5>.
"""

import io
import re

from .utils import int_or_none, timetuple_from_msec


class _MatchParser:
    """
    An object that maintains the current parsing position and allows
    conveniently advancing it as syntax elements are successfully parsed.
    """

    def __init__(self, string):
        self._data = string
        self._pos = 0

    def match(self, r):
        if isinstance(r, re.Pattern):
            return r.match(self._data, self._pos)
        if isinstance(r, str):
            if self._data.startswith(r, self._pos):
                return len(r)
            return None
        raise ValueError(r)

    def advance(self, by):
        if by is None:
            amt = 0
        elif isinstance(by, re.Match):
            amt = len(by.group(0))
        elif isinstance(by, str):
            amt = len(by)
        elif isinstance(by, int):
            amt = by
        else:
            raise ValueError(by)
        self._pos += amt
        return by

    def consume(self, r):
        return self.advance(self.match(r))

    def child(self):
        return _MatchChildParser(self)


class _MatchChildParser(_MatchParser):
    """
    A child parser state, which advances through the same data as
    its parent, but has an independent position. This is useful when
    advancing through syntax elements we might later want to backtrack
    from.
    """

    def __init__(self, parent):
        super().__init__(parent._data)
        self.__parent = parent
        self._pos = parent._pos

    def commit(self):
        """
        Advance the parent state to the current position of this child state.
        """
        self.__parent._pos = self._pos
        return self.__parent


class ParseError(Exception):
    def __init__(self, parser):
        data = parser._data[parser._pos:parser._pos + 100]
        super().__init__(f'Parse error at position {parser._pos} (near {data!r})')


# While the specification <https://www.w3.org/TR/webvtt1/#webvtt-timestamp>
# prescribes that hours must be *2 or more* digits, timestamps with a single
# digit for the hour part has been seen in the wild.
# See https://github.com/yt-dlp/yt-dlp/issues/921
_REGEX_TS = re.compile(r'''(?x)
    (?:([0-9]{1,}):)?
    ([0-9]{2}):
    ([0-9]{2})\.
    ([0-9]{3})?
''')
_REGEX_EOF = re.compile(r'\Z')
_REGEX_NL = re.compile(r'(?:\r\n|[\r\n]|$)')
_REGEX_BLANK = re.compile(r'(?:\r\n|[\r\n])+')
_REGEX_OPTIONAL_WHITESPACE = re.compile(r'[ \t]*')


def _parse_ts(ts):
    """
    Convert a parsed WebVTT timestamp (a re.Match obtained from _REGEX_TS)
    into an MPEG PES timestamp: a tick counter at 90 kHz resolution.
    """
    return 90 * sum(
        int(part or 0) * mult for part, mult in zip(ts.groups(), (3600_000, 60_000, 1000, 1)))


def _format_ts(ts):
    """
    Convert an MPEG PES timestamp into a WebVTT timestamp.
    This will lose sub-millisecond precision.
    """
    return '%02u:%02u:%02u.%03u' % timetuple_from_msec(int((ts + 45) // 90))


class Block:
    """
    An abstract WebVTT block.
    """

    def __init__(self, **kwargs):
        for key, val in kwargs.items():
            setattr(self, key, val)

    @classmethod
    def parse(cls, parser):
        m = parser.match(cls._REGEX)
        if not m:
            return None
        parser.advance(m)
        return cls(raw=m.group(0))

    def write_into(self, stream):
        stream.write(self.raw)


class HeaderBlock(Block):
    """
    A WebVTT block that may only appear in the header part of the file,
    i.e. before any cue blocks.
    """
    pass


class Magic(HeaderBlock):
    _REGEX = re.compile(r'\ufeff?WEBVTT([ \t][^\r\n]*)?(?:\r\n|[\r\n])')

    # XXX: The X-TIMESTAMP-MAP extension is described in RFC 8216 §3.5
    # <https://tools.ietf.org/html/rfc8216#section-3.5>, but the RFC
    # doesn't specify the exact grammar nor where in the WebVTT
    # syntax it should be placed; the below has been devised based
    # on usage in the wild
    #
    # And strictly speaking, the presence of this extension violates
    # the W3C WebVTT spec. Oh well.

    _REGEX_TSMAP = re.compile(r'X-TIMESTAMP-MAP=')
    _REGEX_TSMAP_LOCAL = re.compile(r'LOCAL:')
    _REGEX_TSMAP_MPEGTS = re.compile(r'MPEGTS:([0-9]+)')
    _REGEX_TSMAP_SEP = re.compile(r'[ \t]*,[ \t]*')

    # This was removed from the spec in the 2017 revision;
    # the last spec draft to describe this syntax element is
    # <https://www.w3.org/TR/2015/WD-webvtt1-20151208/#webvtt-metadata-header>.
    # Nevertheless, YouTube keeps serving those
    _REGEX_META = re.compile(r'(?:(?!-->)[^\r\n])+:(?:(?!-->)[^\r\n])+(?:\r\n|[\r\n])')

    @classmethod
    def __parse_tsmap(cls, parser):
        parser = parser.child()

        while True:
            m = parser.consume(cls._REGEX_TSMAP_LOCAL)
            if m:
                m = parser.consume(_REGEX_TS)
                if m is None:
                    raise ParseError(parser)
                local = _parse_ts(m)
                if local is None:
                    raise ParseError(parser)
            else:
                m = parser.consume(cls._REGEX_TSMAP_MPEGTS)
                if m:
                    mpegts = int_or_none(m.group(1))
                    if mpegts is None:
                        raise ParseError(parser)
                else:
                    raise ParseError(parser)
            if parser.consume(cls._REGEX_TSMAP_SEP):
                continue
            if parser.consume(_REGEX_NL):
                break
            raise ParseError(parser)

        parser.commit()
        return local, mpegts

    @classmethod
    def parse(cls, parser):
        parser = parser.child()

        m = parser.consume(cls._REGEX)
        if not m:
            raise ParseError(parser)

        extra = m.group(1)
        local, mpegts, meta = None, None, ''
        while not parser.consume(_REGEX_NL):
            if parser.consume(cls._REGEX_TSMAP):
                local, mpegts = cls.__parse_tsmap(parser)
                continue
            m = parser.consume(cls._REGEX_META)
            if m:
                meta += m.group(0)
                continue
            raise ParseError(parser)
        parser.commit()
        return cls(extra=extra, mpegts=mpegts, local=local, meta=meta)

    def write_into(self, stream):
        stream.write('WEBVTT')
        if self.extra is not None:
            stream.write(self.extra)
        stream.write('\n')
        if self.local or self.mpegts:
            stream.write('X-TIMESTAMP-MAP=LOCAL:')
            stream.write(_format_ts(self.local if self.local is not None else 0))
            stream.write(',MPEGTS:')
            stream.write(str(self.mpegts if self.mpegts is not None else 0))
            stream.write('\n')
        if self.meta:
            stream.write(self.meta)
        stream.write('\n')


class StyleBlock(HeaderBlock):
    _REGEX = re.compile(r'''(?x)
        STYLE[\ \t]*(?:\r\n|[\r\n])
        ((?:(?!-->)[^\r\n])+(?:\r\n|[\r\n]))*
        (?:\r\n|[\r\n])
    ''')


class RegionBlock(HeaderBlock):
    _REGEX = re.compile(r'''(?x)
        REGION[\ \t]*
        ((?:(?!-->)[^\r\n])+(?:\r\n|[\r\n]))*
        (?:\r\n|[\r\n])
    ''')


class CommentBlock(Block):
    _REGEX = re.compile(r'''(?x)
        NOTE(?:\r\n|[\ \t\r\n])
        ((?:(?!-->)[^\r\n])+(?:\r\n|[\r\n]))*
        (?:\r\n|[\r\n])
    ''')


class CueBlock(Block):
    """
    A cue block. The payload is not interpreted.
    """

    _REGEX_ID = re.compile(r'((?:(?!-->)[^\r\n])+)(?:\r\n|[\r\n])')
    _REGEX_ARROW = re.compile(r'[ \t]+-->[ \t]+')
    _REGEX_SETTINGS = re.compile(r'[ \t]+((?:(?!-->)[^\r\n])+)')
    _REGEX_PAYLOAD = re.compile(r'[^\r\n]+(?:\r\n|[\r\n])?')

    @classmethod
    def parse(cls, parser):
        parser = parser.child()

        id_ = None
        m = parser.consume(cls._REGEX_ID)
        if m:
            id_ = m.group(1)

        m0 = parser.consume(_REGEX_TS)
        if not m0:
            return None
        if not parser.consume(cls._REGEX_ARROW):
            return None
        m1 = parser.consume(_REGEX_TS)
        if not m1:
            return None
        m2 = parser.consume(cls._REGEX_SETTINGS)
        parser.consume(_REGEX_OPTIONAL_WHITESPACE)
        if not parser.consume(_REGEX_NL):
            return None

        start = _parse_ts(m0)
        end = _parse_ts(m1)
        settings = m2.group(1) if m2 is not None else None

        text = io.StringIO()
        while True:
            m = parser.consume(cls._REGEX_PAYLOAD)
            if not m:
                break
            text.write(m.group(0))

        parser.commit()
        return cls(
            id=id_,
            start=start, end=end, settings=settings,
            text=text.getvalue(),
        )

    def write_into(self, stream):
        if self.id is not None:
            stream.write(self.id)
            stream.write('\n')
        stream.write(_format_ts(self.start))
        stream.write(' --> ')
        stream.write(_format_ts(self.end))
        if self.settings is not None:
            stream.write(' ')
            stream.write(self.settings)
        stream.write('\n')
        stream.write(self.text)
        stream.write('\n')

    @property
    def as_json(self):
        return {
            'id': self.id,
            'start': self.start,
            'end': self.end,
            'text': self.text,
            '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):
    """
    A generator that yields (partially) parsed WebVTT blocks when given
    a bytes object containing the raw contents of a WebVTT file.
    """

    parser = _MatchParser(frag_content.decode())

    yield Magic.parse(parser)

    while not parser.match(_REGEX_EOF):
        if parser.consume(_REGEX_BLANK):
            continue

        block = RegionBlock.parse(parser)
        if block:
            yield block
            continue
        block = StyleBlock.parse(parser)
        if block:
            yield block
            continue
        block = CommentBlock.parse(parser)
        if block:
            yield block  # XXX: or skip
            continue

        break

    while not parser.match(_REGEX_EOF):
        if parser.consume(_REGEX_BLANK):
            continue

        block = CommentBlock.parse(parser)
        if block:
            yield block  # XXX: or skip
            continue
        block = CueBlock.parse(parser)
        if block:
            yield block
            continue

        raise ParseError(parser)