replaygain: allow exporting already decoded replaygain values
[libav.git] / libavformat / mp3dec.c
CommitLineData
115329f1 1/*
d08928bb 2 * MP3 demuxer
406792e7 3 * Copyright (c) 2003 Fabrice Bellard
6a58e151 4 *
2912e87a 5 * This file is part of Libav.
b78e7197 6 *
2912e87a 7 * Libav is free software; you can redistribute it and/or
6a58e151
FB
8 * modify it under the terms of the GNU Lesser General Public
9 * License as published by the Free Software Foundation; either
b78e7197 10 * version 2.1 of the License, or (at your option) any later version.
6a58e151 11 *
2912e87a 12 * Libav is distributed in the hope that it will be useful,
6a58e151
FB
13 * but WITHOUT ANY WARRANTY; without even the implied warranty of
14 * MERCHANTABILITY or FITNESS FOR A PARTICULAR PURPOSE. See the GNU
15 * Lesser General Public License for more details.
16 *
17 * You should have received a copy of the GNU Lesser General Public
2912e87a 18 * License along with Libav; if not, write to the Free Software
5509bffa 19 * Foundation, Inc., 51 Franklin Street, Fifth Floor, Boston, MA 02110-1301 USA
6a58e151 20 */
245976da
DB
21
22#include "libavutil/avstring.h"
ff06d76a 23#include "libavutil/intreadwrite.h"
d2d67e42 24#include "libavutil/dict.h"
0ebcdf5c 25#include "libavutil/mathematics.h"
6a58e151 26#include "avformat.h"
c3f9ebf7 27#include "internal.h"
2ea512a6 28#include "id3v2.h"
75411182 29#include "id3v1.h"
15c5a8d2
AK
30#include "replaygain.h"
31
de653ba3
DB
32#include "libavcodec/mpegaudiodecheader.h"
33
2d1a1a7f
AK
34#define XING_FLAG_FRAMES 0x01
35#define XING_FLAG_SIZE 0x02
50d1f443
AK
36#define XING_FLAG_TOC 0x04
37
38#define XING_TOC_COUNT 100
2d1a1a7f 39
505642f1
MN
40typedef struct MP3DecContext {
41 int xing_toc;
42} MP3DecContext;
43
6a58e151 44/* mp3 read */
dc4ed3b1
MR
45
46static int mp3_read_probe(AVProbeData *p)
47{
ea7599ba 48 int max_frames, first_frames = 0;
498c544a 49 int fsize, frames, sample_rate;
abade142 50 uint32_t header;
ac3ef4a4 51 uint8_t *buf, *buf0, *buf2, *end;
abade142 52 AVCodecContext avctx;
dc4ed3b1 53
ac3ef4a4 54 buf0 = p->buf;
5692ed3a
MN
55 end = p->buf + p->buf_size - sizeof(uint32_t);
56 while(buf0 < end && !*buf0)
57 buf0++;
dc4ed3b1 58
abade142 59 max_frames = 0;
ac3ef4a4 60 buf = buf0;
dc4ed3b1 61
5ef5501b 62 for(; buf < end; buf= buf2+1) {
abade142 63 buf2 = buf;
dc4ed3b1 64
d0702de9 65 for(frames = 0; buf2 < end; frames++) {
80fb8234 66 header = AV_RB32(buf2);
82ab61f9 67 fsize = avpriv_mpa_decode_header(&avctx, header, &sample_rate, &sample_rate, &sample_rate, &sample_rate);
abade142
68 if(fsize < 0)
69 break;
70 buf2 += fsize;
71 }
72 max_frames = FFMAX(max_frames, frames);
ac3ef4a4 73 if(buf == buf0)
22668218 74 first_frames= frames;
abade142 75 }
4b4ac5c7
RD
76 // keep this in sync with ac3 probe, both need to avoid
77 // issues with MPEG-files!
e0f8be64 78 if (first_frames >= 4) return AVPROBE_SCORE_EXTENSION + 1;
64bde805
RB
79
80 if (max_frames) {
81 int pes = 0, i;
82 unsigned int code = -1;
83
84#define VIDEO_ID 0x000001e0
85#define AUDIO_ID 0x000001c0
86 /* do a search for mpegps headers to be able to properly bias
87 * towards mpegps if we detect this stream as both. */
88 for (i = 0; i<p->buf_size; i++) {
89 code = (code << 8) + p->buf[i];
90 if ((code & 0xffffff00) == 0x100) {
91 if ((code & 0x1f0) == VIDEO_ID) pes++;
92 else if((code & 0x1e0) == AUDIO_ID) pes++;
93 }
94 }
95
96 if (pes)
97 max_frames = (max_frames + pes - 1) / pes;
98 }
e0f8be64
DB
99 if (max_frames > 500) return AVPROBE_SCORE_EXTENSION;
100 else if (max_frames >= 4) return AVPROBE_SCORE_EXTENSION / 2;
64bde805
RB
101 else if (max_frames >= 1) return 1;
102 else return 0;
f9f2ab82 103//mpegps_mp3_unrecognized_format.mpg has max_frames=3
dc4ed3b1
MR
104}
105
50d1f443
AK
106static void read_xing_toc(AVFormatContext *s, int64_t filesize, int64_t duration)
107{
108 int i;
505642f1 109 MP3DecContext *mp3 = s->priv_data;
50d1f443
AK
110
111 if (!filesize &&
112 !(filesize = avio_size(s->pb))) {
113 av_log(s, AV_LOG_WARNING, "Cannot determine file size, skipping TOC table.\n");
114 return;
115 }
116
117 for (i = 0; i < XING_TOC_COUNT; i++) {
118 uint8_t b = avio_r8(s->pb);
119
120 av_add_index_entry(s->streams[0],
121 av_rescale(b, filesize, 256),
122 av_rescale(i, duration, XING_TOC_COUNT),
123 0, 0, AVINDEX_KEYFRAME);
124 }
505642f1 125 mp3->xing_toc = 1;
50d1f443
AK
126}
127
ca851a3a 128/**
8862fb41 129 * Try to find Xing/Info/VBRI tags and compute duration from info therein
ca851a3a 130 */
853a9753 131static int mp3_parse_vbr_tags(AVFormatContext *s, AVStream *st, int64_t base)
ca851a3a 132{
db233e83 133 uint32_t v, spf;
2a758efe 134 unsigned frames = 0; /* Total number of frames in file */
49d7ef28 135 unsigned size = 0; /* Total number of bytes in the stream */
bc5c918e 136 const int64_t xing_offtbl[2][2] = {{32, 17}, {17,9}};
cf92cec7 137 MPADecodeHeader c;
853a9753 138 int vbrtag_size = 0;
eae0879d 139 int is_cbr;
ca851a3a 140
b7effd4e 141 v = avio_rb32(s->pb);
5b5bd7c9 142 if(ff_mpa_check_header(v) < 0)
853a9753 143 return -1;
5b5bd7c9 144
82ab61f9 145 if (avpriv_mpegaudio_decode_header(&c, v) == 0)
853a9753 146 vbrtag_size = c.frame_size;
db233e83 147 if(c.layer != 3)
853a9753 148 return -1;
ca851a3a 149
50d1f443
AK
150 spf = c.lsf ? 576 : 1152; /* Samples per frame, layer 3 */
151
db233e83 152 /* Check for Xing / Info tag */
e42500cb 153 avio_skip(s->pb, xing_offtbl[c.lsf == 1][c.nb_channels == 1]);
b7effd4e 154 v = avio_rb32(s->pb);
eae0879d
AK
155 is_cbr = v == MKBETAG('I', 'n', 'f', 'o');
156 if (v == MKBETAG('X', 'i', 'n', 'g') || is_cbr) {
b7effd4e 157 v = avio_rb32(s->pb);
2d1a1a7f 158 if(v & XING_FLAG_FRAMES)
b7effd4e 159 frames = avio_rb32(s->pb);
2d1a1a7f 160 if(v & XING_FLAG_SIZE)
b7effd4e 161 size = avio_rb32(s->pb);
50d1f443
AK
162 if (v & XING_FLAG_TOC && frames)
163 read_xing_toc(s, size, av_rescale_q(frames, (AVRational){spf, c.sample_rate},
164 st->time_base));
ca851a3a 165 }
db233e83 166
8862fb41 167 /* Check for VBRI tag (always 32 bytes after end of mpegaudio header) */
6b4aa5da 168 avio_seek(s->pb, base + 4 + 32, SEEK_SET);
b7effd4e 169 v = avio_rb32(s->pb);
8862fb41
170 if(v == MKBETAG('V', 'B', 'R', 'I')) {
171 /* Check tag version */
b7effd4e 172 if(avio_rb16(s->pb) == 1) {
49d7ef28 173 /* skip delay and quality */
45a8a02a 174 avio_skip(s->pb, 4);
b7effd4e 175 size = avio_rb32(s->pb);
c05e2be9 176 frames = avio_rb32(s->pb);
8862fb41
177 }
178 }
179
2a758efe 180 if(!frames && !size)
853a9753
YS
181 return -1;
182
183 /* Skip the vbr tag frame */
6b4aa5da 184 avio_seek(s->pb, base + vbrtag_size, SEEK_SET);
db233e83 185
2a758efe 186 if(frames)
49d7ef28
AK
187 st->duration = av_rescale_q(frames, (AVRational){spf, c.sample_rate},
188 st->time_base);
eae0879d 189 if (size && frames && !is_cbr)
49d7ef28
AK
190 st->codec->bit_rate = av_rescale(size, 8 * c.sample_rate, frames * (int64_t)spf);
191
853a9753 192 return 0;
ca851a3a
193}
194
6e9651d1 195static int mp3_read_header(AVFormatContext *s)
6a58e151
FB
196{
197 AVStream *st;
bc5c918e 198 int64_t off;
15c5a8d2 199 int ret;
6a58e151 200
3b3bbdd3 201 st = avformat_new_stream(s, NULL);
6a58e151 202 if (!st)
769e10f0 203 return AVERROR(ENOMEM);
6a58e151 204
72415b2a 205 st->codec->codec_type = AVMEDIA_TYPE_AUDIO;
36ef5369 206 st->codec->codec_id = AV_CODEC_ID_MP3;
57004ff1 207 st->need_parsing = AVSTREAM_PARSE_FULL;
f97931c1 208 st->start_time = 0;
115329f1 209
9c5a9e6b 210 // lcm of all mp3 sample rates
c3f9ebf7 211 avpriv_set_pts_info(st, 64, 1, 14112000);
9c5a9e6b 212
a2704c97 213 off = avio_tell(s->pb);
80242f9b 214
d2d67e42 215 if (!av_dict_get(s->metadata, "", NULL, AV_DICT_IGNORE_SUFFIX))
7fd5aeb3 216 ff_id3v1_read(s);
6a58e151 217
853a9753 218 if (mp3_parse_vbr_tags(s, st, off) < 0)
6b4aa5da 219 avio_seek(s->pb, off, SEEK_SET);
ca851a3a 220
15c5a8d2
AK
221 ret = ff_replaygain_export(st, s->metadata);
222 if (ret < 0)
223 return ret;
224
6a58e151
FB
225 /* the parameters will be extracted from the compressed bitstream */
226 return 0;
227}
228
229#define MP3_PACKET_SIZE 1024
230
231static int mp3_read_packet(AVFormatContext *s, AVPacket *pkt)
232{
67b1156f 233 int ret;
115329f1 234
67b1156f 235 ret = av_get_packet(s->pb, pkt, MP3_PACKET_SIZE);
f73e3938
AK
236 if (ret < 0)
237 return ret;
6a58e151
FB
238
239 pkt->stream_index = 0;
045b80e5
CB
240
241 if (ret > ID3v1_TAG_SIZE &&
242 memcmp(&pkt->data[ret - ID3v1_TAG_SIZE], "TAG", 3) == 0)
243 ret -= ID3v1_TAG_SIZE;
244
6a58e151
FB
245 /* note: we need to modify the packet size here to handle the last
246 packet */
247 pkt->size = ret;
248 return ret;
249}
250
50d1f443
AK
251static int mp3_seek(AVFormatContext *s, int stream_index, int64_t timestamp,
252 int flags)
253{
505642f1 254 MP3DecContext *mp3 = s->priv_data;
50d1f443
AK
255 AVIndexEntry *ie;
256 AVStream *st = s->streams[0];
257 int64_t ret = av_index_search_timestamp(st, timestamp, flags);
258 uint32_t header = 0;
259
505642f1
MN
260 if (!mp3->xing_toc)
261 return AVERROR(ENOSYS);
262
50d1f443
AK
263 if (ret < 0)
264 return ret;
265
266 ie = &st->index_entries[ret];
267 ret = avio_seek(s->pb, ie->pos, SEEK_SET);
268 if (ret < 0)
269 return ret;
270
271 while (!s->pb->eof_reached) {
272 header = (header << 8) + avio_r8(s->pb);
273 if (ff_mpa_check_header(header) >= 0) {
274 ff_update_cur_dts(s, st, ie->timestamp);
275 ret = avio_seek(s->pb, -4, SEEK_CUR);
276 return (ret >= 0) ? 0 : ret;
277 }
278 }
279
280 return AVERROR_EOF;
281}
282
c6610a21 283AVInputFormat ff_mp3_demuxer = {
dfc2c4d9 284 .name = "mp3",
0177b7d2 285 .long_name = NULL_IF_CONFIG_SMALL("MP2/3 (MPEG audio layer 2/3)"),
dfc2c4d9
AK
286 .read_probe = mp3_read_probe,
287 .read_header = mp3_read_header,
288 .read_packet = mp3_read_packet,
50d1f443 289 .read_seek = mp3_seek,
505642f1 290 .priv_data_size = sizeof(MP3DecContext),
20234a4b 291 .flags = AVFMT_GENERIC_INDEX,
c0779a67 292 .extensions = "mp2,mp3,m2a,mpa", /* XXX: use probe */
de653ba3 293};