oggparsevorbis: export replaygain tags from Vorbis comments
[libav.git] / libavformat / oggparsevorbis.c
1 /*
2 * Copyright (C) 2005 Michael Ahlberg, Måns Rullgård
3 *
4 * Permission is hereby granted, free of charge, to any person
5 * obtaining a copy of this software and associated documentation
6 * files (the "Software"), to deal in the Software without
7 * restriction, including without limitation the rights to use, copy,
8 * modify, merge, publish, distribute, sublicense, and/or sell copies
9 * of the Software, and to permit persons to whom the Software is
10 * furnished to do so, subject to the following conditions:
11 *
12 * The above copyright notice and this permission notice shall be
13 * included in all copies or substantial portions of the Software.
14 *
15 * THE SOFTWARE IS PROVIDED "AS IS", WITHOUT WARRANTY OF ANY KIND,
16 * EXPRESS OR IMPLIED, INCLUDING BUT NOT LIMITED TO THE WARRANTIES OF
17 * MERCHANTABILITY, FITNESS FOR A PARTICULAR PURPOSE AND
18 * NONINFRINGEMENT. IN NO EVENT SHALL THE AUTHORS OR COPYRIGHT
19 * HOLDERS BE LIABLE FOR ANY CLAIM, DAMAGES OR OTHER LIABILITY,
20 * WHETHER IN AN ACTION OF CONTRACT, TORT OR OTHERWISE, ARISING FROM,
21 * OUT OF OR IN CONNECTION WITH THE SOFTWARE OR THE USE OR OTHER
22 * DEALINGS IN THE SOFTWARE.
23 */
24
25 #include <stdlib.h>
26
27 #include "libavutil/avstring.h"
28 #include "libavutil/base64.h"
29 #include "libavutil/bswap.h"
30 #include "libavutil/dict.h"
31 #include "libavcodec/bytestream.h"
32 #include "libavcodec/get_bits.h"
33 #include "libavcodec/vorbis_parser.h"
34 #include "avformat.h"
35 #include "flac_picture.h"
36 #include "internal.h"
37 #include "oggdec.h"
38 #include "vorbiscomment.h"
39 #include "replaygain.h"
40
41 static int ogm_chapter(AVFormatContext *as, uint8_t *key, uint8_t *val)
42 {
43 int i, cnum, h, m, s, ms, keylen = strlen(key);
44 AVChapter *chapter = NULL;
45
46 if (keylen < 9 || sscanf(key, "CHAPTER%03d", &cnum) != 1)
47 return 0;
48
49 if (keylen <= 10) {
50 if (sscanf(val, "%02d:%02d:%02d.%03d", &h, &m, &s, &ms) < 4)
51 return 0;
52
53 avpriv_new_chapter(as, cnum, (AVRational) { 1, 1000 },
54 ms + 1000 * (s + 60 * (m + 60 * h)),
55 AV_NOPTS_VALUE, NULL);
56 av_free(val);
57 } else if (!strcmp(key + keylen - 4, "NAME")) {
58 for (i = 0; i < as->nb_chapters; i++)
59 if (as->chapters[i]->id == cnum) {
60 chapter = as->chapters[i];
61 break;
62 }
63 if (!chapter)
64 return 0;
65
66 av_dict_set(&chapter->metadata, "title", val, AV_DICT_DONT_STRDUP_VAL);
67 } else
68 return 0;
69
70 av_free(key);
71 return 1;
72 }
73
74 int ff_vorbis_comment(AVFormatContext *as, AVDictionary **m,
75 const uint8_t *buf, int size)
76 {
77 const uint8_t *p = buf;
78 const uint8_t *end = buf + size;
79 unsigned n, j;
80 int s;
81
82 /* must have vendor_length and user_comment_list_length */
83 if (size < 8)
84 return AVERROR_INVALIDDATA;
85
86 s = bytestream_get_le32(&p);
87
88 if (end - p - 4 < s || s < 0)
89 return AVERROR_INVALIDDATA;
90
91 p += s;
92
93 n = bytestream_get_le32(&p);
94
95 while (end - p >= 4 && n > 0) {
96 const char *t, *v;
97 int tl, vl;
98
99 s = bytestream_get_le32(&p);
100
101 if (end - p < s || s < 0)
102 break;
103
104 t = p;
105 p += s;
106 n--;
107
108 v = memchr(t, '=', s);
109 if (!v)
110 continue;
111
112 tl = v - t;
113 vl = s - tl - 1;
114 v++;
115
116 if (tl && vl) {
117 char *tt, *ct;
118
119 tt = av_malloc(tl + 1);
120 ct = av_malloc(vl + 1);
121 if (!tt || !ct) {
122 av_freep(&tt);
123 av_freep(&ct);
124 return AVERROR(ENOMEM);
125 }
126
127 for (j = 0; j < tl; j++)
128 tt[j] = av_toupper(t[j]);
129 tt[tl] = 0;
130
131 memcpy(ct, v, vl);
132 ct[vl] = 0;
133
134 /* The format in which the pictures are stored is the FLAC format.
135 * Xiph says: "The binary FLAC picture structure is base64 encoded
136 * and placed within a VorbisComment with the tag name
137 * 'METADATA_BLOCK_PICTURE'. This is the preferred and
138 * recommended way of embedding cover art within VorbisComments."
139 */
140 if (!strcmp(tt, "METADATA_BLOCK_PICTURE")) {
141 int ret;
142 char *pict = av_malloc(vl);
143
144 if (!pict) {
145 av_freep(&tt);
146 av_freep(&ct);
147 return AVERROR(ENOMEM);
148 }
149 if ((ret = av_base64_decode(pict, ct, vl)) > 0)
150 ret = ff_flac_parse_picture(as, pict, ret);
151 av_freep(&tt);
152 av_freep(&ct);
153 av_freep(&pict);
154 if (ret < 0) {
155 av_log(as, AV_LOG_WARNING, "Failed to parse cover art block.\n");
156 continue;
157 }
158 } else if (!ogm_chapter(as, tt, ct))
159 av_dict_set(m, tt, ct,
160 AV_DICT_DONT_STRDUP_KEY |
161 AV_DICT_DONT_STRDUP_VAL);
162 }
163 }
164
165 if (p != end)
166 av_log(as, AV_LOG_INFO,
167 "%ti bytes of comment header remain\n", end - p);
168 if (n > 0)
169 av_log(as, AV_LOG_INFO,
170 "truncated comment header, %i comments not found\n", n);
171
172 ff_metadata_conv(m, NULL, ff_vorbiscomment_metadata_conv);
173
174 return 0;
175 }
176
177 /*
178 * Parse the vorbis header
179 *
180 * Vorbis Identification header from Vorbis_I_spec.html#vorbis-spec-codec
181 * [vorbis_version] = read 32 bits as unsigned integer | Not used
182 * [audio_channels] = read 8 bit integer as unsigned | Used
183 * [audio_sample_rate] = read 32 bits as unsigned integer | Used
184 * [bitrate_maximum] = read 32 bits as signed integer | Not used yet
185 * [bitrate_nominal] = read 32 bits as signed integer | Not used yet
186 * [bitrate_minimum] = read 32 bits as signed integer | Used as bitrate
187 * [blocksize_0] = read 4 bits as unsigned integer | Not Used
188 * [blocksize_1] = read 4 bits as unsigned integer | Not Used
189 * [framing_flag] = read one bit | Not Used
190 */
191
192 struct oggvorbis_private {
193 unsigned int len[3];
194 unsigned char *packet[3];
195 VorbisParseContext vp;
196 int64_t final_pts;
197 int final_duration;
198 };
199
200 static int fixup_vorbis_headers(AVFormatContext *as,
201 struct oggvorbis_private *priv,
202 uint8_t **buf)
203 {
204 int i, offset, len, err;
205 unsigned char *ptr;
206
207 len = priv->len[0] + priv->len[1] + priv->len[2];
208 ptr = *buf = av_mallocz(len + len / 255 + 64);
209 if (!ptr)
210 return AVERROR(ENOMEM);
211
212 ptr[0] = 2;
213 offset = 1;
214 offset += av_xiphlacing(&ptr[offset], priv->len[0]);
215 offset += av_xiphlacing(&ptr[offset], priv->len[1]);
216 for (i = 0; i < 3; i++) {
217 memcpy(&ptr[offset], priv->packet[i], priv->len[i]);
218 offset += priv->len[i];
219 av_freep(&priv->packet[i]);
220 }
221 if ((err = av_reallocp(buf, offset + FF_INPUT_BUFFER_PADDING_SIZE)) < 0)
222 return err;
223 return offset;
224 }
225
226 static void vorbis_cleanup(AVFormatContext *s, int idx)
227 {
228 struct ogg *ogg = s->priv_data;
229 struct ogg_stream *os = ogg->streams + idx;
230 struct oggvorbis_private *priv = os->private;
231 int i;
232 if (os->private)
233 for (i = 0; i < 3; i++)
234 av_freep(&priv->packet[i]);
235 }
236
237 static int vorbis_header(AVFormatContext *s, int idx)
238 {
239 struct ogg *ogg = s->priv_data;
240 AVStream *st = s->streams[idx];
241 struct ogg_stream *os = ogg->streams + idx;
242 struct oggvorbis_private *priv;
243 int pkt_type = os->buf[os->pstart];
244
245 if (!os->private) {
246 os->private = av_mallocz(sizeof(struct oggvorbis_private));
247 if (!os->private)
248 return AVERROR(ENOMEM);
249 }
250
251 if (!(pkt_type & 1))
252 return 0;
253
254 if (os->psize < 1 || pkt_type > 5)
255 return AVERROR_INVALIDDATA;
256
257 priv = os->private;
258
259 if (priv->packet[pkt_type >> 1])
260 return AVERROR_INVALIDDATA;
261 if (pkt_type > 1 && !priv->packet[0] || pkt_type > 3 && !priv->packet[1])
262 return AVERROR_INVALIDDATA;
263
264 priv->len[pkt_type >> 1] = os->psize;
265 priv->packet[pkt_type >> 1] = av_mallocz(os->psize);
266 if (!priv->packet[pkt_type >> 1])
267 return AVERROR(ENOMEM);
268 memcpy(priv->packet[pkt_type >> 1], os->buf + os->pstart, os->psize);
269 if (os->buf[os->pstart] == 1) {
270 const uint8_t *p = os->buf + os->pstart + 7; /* skip "\001vorbis" tag */
271 unsigned blocksize, bs0, bs1;
272 int srate;
273
274 if (os->psize != 30)
275 return AVERROR_INVALIDDATA;
276
277 if (bytestream_get_le32(&p) != 0) /* vorbis_version */
278 return AVERROR_INVALIDDATA;
279
280 st->codec->channels = bytestream_get_byte(&p);
281 srate = bytestream_get_le32(&p);
282 p += 4; // skip maximum bitrate
283 st->codec->bit_rate = bytestream_get_le32(&p); // nominal bitrate
284 p += 4; // skip minimum bitrate
285
286 blocksize = bytestream_get_byte(&p);
287 bs0 = blocksize & 15;
288 bs1 = blocksize >> 4;
289
290 if (bs0 > bs1)
291 return AVERROR_INVALIDDATA;
292 if (bs0 < 6 || bs1 > 13)
293 return AVERROR_INVALIDDATA;
294
295 if (bytestream_get_byte(&p) != 1) /* framing_flag */
296 return AVERROR_INVALIDDATA;
297
298 st->codec->codec_type = AVMEDIA_TYPE_AUDIO;
299 st->codec->codec_id = AV_CODEC_ID_VORBIS;
300
301 if (srate > 0) {
302 st->codec->sample_rate = srate;
303 avpriv_set_pts_info(st, 64, 1, srate);
304 }
305 } else if (os->buf[os->pstart] == 3) {
306 if (os->psize > 8 &&
307 ff_vorbis_comment(s, &st->metadata, os->buf + os->pstart + 7,
308 os->psize - 8) >= 0) {
309 unsigned new_len;
310
311 int ret = ff_replaygain_export(st, st->metadata);
312 if (ret < 0)
313 return ret;
314
315 // drop all metadata we parsed and which is not required by libvorbis
316 new_len = 7 + 4 + AV_RL32(priv->packet[1] + 7) + 4 + 1;
317 if (new_len >= 16 && new_len < os->psize) {
318 AV_WL32(priv->packet[1] + new_len - 5, 0);
319 priv->packet[1][new_len - 1] = 1;
320 priv->len[1] = new_len;
321 }
322 }
323 } else {
324 int ret = fixup_vorbis_headers(s, priv, &st->codec->extradata);
325 if (ret < 0) {
326 st->codec->extradata_size = 0;
327 return ret;
328 }
329 st->codec->extradata_size = ret;
330 if ((ret = avpriv_vorbis_parse_extradata(st->codec, &priv->vp))) {
331 av_freep(&st->codec->extradata);
332 st->codec->extradata_size = 0;
333 return ret;
334 }
335 }
336
337 return 1;
338 }
339
340 static int vorbis_packet(AVFormatContext *s, int idx)
341 {
342 struct ogg *ogg = s->priv_data;
343 struct ogg_stream *os = ogg->streams + idx;
344 struct oggvorbis_private *priv = os->private;
345 int duration;
346
347 /* first packet handling
348 * here we parse the duration of each packet in the first page and compare
349 * the total duration to the page granule to find the encoder delay and
350 * set the first timestamp */
351 if (!os->lastpts) {
352 int seg;
353 uint8_t *last_pkt = os->buf + os->pstart;
354 uint8_t *next_pkt = last_pkt;
355 int first_duration = 0;
356
357 avpriv_vorbis_parse_reset(&priv->vp);
358 duration = 0;
359 for (seg = 0; seg < os->nsegs; seg++) {
360 if (os->segments[seg] < 255) {
361 int d = avpriv_vorbis_parse_frame(&priv->vp, last_pkt, 1);
362 if (d < 0) {
363 duration = os->granule;
364 break;
365 }
366 if (!duration)
367 first_duration = d;
368 duration += d;
369 last_pkt = next_pkt + os->segments[seg];
370 }
371 next_pkt += os->segments[seg];
372 }
373 os->lastpts =
374 os->lastdts = os->granule - duration;
375 s->streams[idx]->start_time = os->lastpts + first_duration;
376 if (s->streams[idx]->duration)
377 s->streams[idx]->duration -= s->streams[idx]->start_time;
378 s->streams[idx]->cur_dts = AV_NOPTS_VALUE;
379 priv->final_pts = AV_NOPTS_VALUE;
380 avpriv_vorbis_parse_reset(&priv->vp);
381 }
382
383 /* parse packet duration */
384 if (os->psize > 0) {
385 duration = avpriv_vorbis_parse_frame(&priv->vp, os->buf + os->pstart, 1);
386 if (duration <= 0) {
387 os->pflags |= AV_PKT_FLAG_CORRUPT;
388 return 0;
389 }
390 os->pduration = duration;
391 }
392
393 /* final packet handling
394 * here we save the pts of the first packet in the final page, sum up all
395 * packet durations in the final page except for the last one, and compare
396 * to the page granule to find the duration of the final packet */
397 if (os->flags & OGG_FLAG_EOS) {
398 if (os->lastpts != AV_NOPTS_VALUE) {
399 priv->final_pts = os->lastpts;
400 priv->final_duration = 0;
401 }
402 if (os->segp == os->nsegs)
403 os->pduration = os->granule - priv->final_pts - priv->final_duration;
404 priv->final_duration += os->pduration;
405 }
406
407 return 0;
408 }
409
410 const struct ogg_codec ff_vorbis_codec = {
411 .magic = "\001vorbis",
412 .magicsize = 7,
413 .header = vorbis_header,
414 .packet = vorbis_packet,
415 .cleanup = vorbis_cleanup,
416 .nb_header = 3,
417 };