tools/sidxindex.c

   1 /*
   2  * Copyright (c) 2014 Martin Storsjo
   3  *
   4  * This file is part of Libav.
   5  *
   6  * Libav is free software; you can redistribute it and/or
   7  * modify it under the terms of the GNU Lesser General Public
   8  * License as published by the Free Software Foundation; either
   9  * version 2.1 of the License, or (at your option) any later version.
  10  *
  11  * Libav is distributed in the hope that it will be useful,
  12  * but WITHOUT ANY WARRANTY; without even the implied warranty of
  13  * MERCHANTABILITY or FITNESS FOR A PARTICULAR PURPOSE.  See the GNU
  14  * Lesser General Public License for more details.
  15  *
  16  * You should have received a copy of the GNU Lesser General Public
  17  * License along with Libav; if not, write to the Free Software
  18  * Foundation, Inc., 51 Franklin Street, Fifth Floor, Boston, MA 02110-1301 USA
  19  */
  20
  21 #include <stdio.h>
  22 #include <string.h>
  23
  24 #include "libavformat/avformat.h"
  25 #include "libavutil/avstring.h"
  26 #include "libavutil/intreadwrite.h"
  27 #include "libavutil/mathematics.h"
  28
  29 static int usage(const char *argv0, int ret)
  30 {
  31     fprintf(stderr, "%s -out foo.mpd file1\n", argv0);
  32     return ret;
  33 }
  34
  35 struct Track {
  36     const char *name;
  37     int64_t duration;
  38     int bitrate;
  39     int track_id;
  40     int is_audio, is_video;
  41     int width, height;
  42     int sample_rate, channels;
  43     int timescale;
  44     char codec_str[30];
  45     int64_t sidx_start, sidx_length;
  46 };
  47
  48 struct Tracks {
  49     int nb_tracks;
  50     int64_t duration;
  51     struct Track **tracks;
  52     int multiple_tracks_per_file;
  53 };
  54
  55 static void set_codec_str(AVCodecParameters *codecpar, char *str, int size)
  56 {
  57     switch (codecpar->codec_id) {
  58     case AV_CODEC_ID_H264:
  59         snprintf(str, size, "avc1");
  60         if (codecpar->extradata_size >= 4 && codecpar->extradata[0] == 1) {
  61             av_strlcatf(str, size, ".%02x%02x%02x",
  62                         codecpar->extradata[1], codecpar->extradata[2],
  63                         codecpar->extradata[3]);
  64         }
  65         break;
  66     case AV_CODEC_ID_AAC:
  67         snprintf(str, size, "mp4a.40"); // 0x40 is the mp4 object type for AAC
  68         if (codecpar->extradata_size >= 2) {
  69             int aot = codecpar->extradata[0] >> 3;
  70             if (aot == 31)
  71                 aot = ((AV_RB16(codecpar->extradata) >> 5) & 0x3f) + 32;
  72             av_strlcatf(str, size, ".%d", aot);
  73         }
  74         break;
  75     }
  76 }
  77
  78 static int find_sidx(struct Tracks *tracks, int start_index,
  79                      const char *file)
  80 {
  81     int err = 0;
  82     AVIOContext *f = NULL;
  83     int i;
  84
  85     if ((err = avio_open2(&f, file, AVIO_FLAG_READ, NULL, NULL)) < 0)
  86         goto fail;
  87
  88     while (!f->eof_reached) {
  89         int64_t pos = avio_tell(f);
  90         int32_t size, tag;
  91
  92         size = avio_rb32(f);
  93         tag  = avio_rb32(f);
  94         if (size < 8)
  95             break;
  96         if (tag == MKBETAG('s', 'i', 'd', 'x')) {
  97             for (i = start_index; i < tracks->nb_tracks; i++) {
  98                 struct Track *track = tracks->tracks[i];
  99                 if (!track->sidx_start) {
 100                     track->sidx_start  = pos;
 101                     track->sidx_length = size;
 102                 } else if (pos == track->sidx_start + track->sidx_length) {
 103                     track->sidx_length = pos + size - track->sidx_start;
 104                 }
 105             }
 106         }
 107         if (avio_seek(f, pos + size, SEEK_SET) != pos + size)
 108             break;
 109     }
 110
 111 fail:
 112     if (f)
 113         avio_close(f);
 114     return err;
 115 }
 116
 117 static int handle_file(struct Tracks *tracks, const char *file)
 118 {
 119     AVFormatContext *ctx = NULL;
 120     int err = 0, i, orig_tracks = tracks->nb_tracks;
 121     char errbuf[50], *ptr;
 122     struct Track *track;
 123
 124     err = avformat_open_input(&ctx, file, NULL, NULL);
 125     if (err < 0) {
 126         av_strerror(err, errbuf, sizeof(errbuf));
 127         fprintf(stderr, "Unable to open %s: %s\n", file, errbuf);
 128         return 1;
 129     }
 130
 131     err = avformat_find_stream_info(ctx, NULL);
 132     if (err < 0) {
 133         av_strerror(err, errbuf, sizeof(errbuf));
 134         fprintf(stderr, "Unable to identify %s: %s\n", file, errbuf);
 135         goto fail;
 136     }
 137
 138     if (ctx->nb_streams < 1) {
 139         fprintf(stderr, "No streams found in %s\n", file);
 140         goto fail;
 141     }
 142     if (ctx->nb_streams > 1)
 143         tracks->multiple_tracks_per_file = 1;
 144
 145     for (i = 0; i < ctx->nb_streams; i++) {
 146         struct Track **temp;
 147         AVStream *st = ctx->streams[i];
 148
 149         if (st->codecpar->bit_rate == 0) {
 150             fprintf(stderr, "Skipping track %d in %s as it has zero bitrate\n",
 151                     st->id, file);
 152             continue;
 153         }
 154
 155         track = av_mallocz(sizeof(*track));
 156         if (!track) {
 157             err = AVERROR(ENOMEM);
 158             goto fail;
 159         }
 160         temp = av_realloc(tracks->tracks,
 161                           sizeof(*tracks->tracks) * (tracks->nb_tracks + 1));
 162         if (!temp) {
 163             av_free(track);
 164             err = AVERROR(ENOMEM);
 165             goto fail;
 166         }
 167         tracks->tracks = temp;
 168         tracks->tracks[tracks->nb_tracks] = track;
 169
 170         track->name = file;
 171         if ((ptr = strrchr(file, '/')))
 172             track->name = ptr + 1;
 173
 174         track->bitrate   = st->codecpar->bit_rate;
 175         track->track_id  = st->id;
 176         track->timescale = st->time_base.den;
 177         track->duration  = st->duration;
 178         track->is_audio  = st->codecpar->codec_type == AVMEDIA_TYPE_AUDIO;
 179         track->is_video  = st->codecpar->codec_type == AVMEDIA_TYPE_VIDEO;
 180
 181         if (!track->is_audio && !track->is_video) {
 182             fprintf(stderr,
 183                     "Track %d in %s is neither video nor audio, skipping\n",
 184                     track->track_id, file);
 185             av_freep(&tracks->tracks[tracks->nb_tracks]);
 186             continue;
 187         }
 188
 189         tracks->duration = FFMAX(tracks->duration,
 190                                  av_rescale_rnd(track->duration, AV_TIME_BASE,
 191                                                 track->timescale, AV_ROUND_UP));
 192
 193         if (track->is_audio) {
 194             track->channels    = st->codecpar->channels;
 195             track->sample_rate = st->codecpar->sample_rate;
 196         }
 197         if (track->is_video) {
 198             track->width  = st->codecpar->width;
 199             track->height = st->codecpar->height;
 200         }
 201         set_codec_str(st->codecpar, track->codec_str, sizeof(track->codec_str));
 202
 203         tracks->nb_tracks++;
 204     }
 205
 206     avformat_close_input(&ctx);
 207
 208     err = find_sidx(tracks, orig_tracks, file);
 209
 210 fail:
 211     if (ctx)
 212         avformat_close_input(&ctx);
 213     return err;
 214 }
 215
 216 static void write_time(FILE *out, int64_t time, int decimals, enum AVRounding round)
 217 {
 218     int seconds = time / AV_TIME_BASE;
 219     int fractions = time % AV_TIME_BASE;
 220     int minutes = seconds / 60;
 221     int hours = minutes / 60;
 222     fractions = av_rescale_rnd(fractions, pow(10, decimals), AV_TIME_BASE, round);
 223     seconds %= 60;
 224     minutes %= 60;
 225     fprintf(out, "PT");
 226     if (hours)
 227         fprintf(out, "%dH", hours);
 228     if (hours || minutes)
 229         fprintf(out, "%dM", minutes);
 230     fprintf(out, "%d.%0*dS", seconds, decimals, fractions);
 231 }
 232
 233 static int output_mpd(struct Tracks *tracks, const char *filename)
 234 {
 235     FILE *out;
 236     int i, j, ret = 0;
 237     struct Track **adaptation_sets_buf[2] = { NULL };
 238     struct Track ***adaptation_sets;
 239     int nb_tracks_buf[2] = { 0 };
 240     int *nb_tracks;
 241     int set, nb_sets;
 242
 243     if (!tracks->multiple_tracks_per_file) {
 244         adaptation_sets = adaptation_sets_buf;
 245         nb_tracks = nb_tracks_buf;
 246         nb_sets = 2;
 247         for (i = 0; i < 2; i++) {
 248             adaptation_sets[i] = av_malloc(sizeof(*adaptation_sets[i]) * tracks->nb_tracks);
 249             if (!adaptation_sets[i]) {
 250                 ret = AVERROR(ENOMEM);
 251                 goto err;
 252             }
 253         }
 254         for (i = 0; i < tracks->nb_tracks; i++) {
 255             int set_index = -1;
 256             if (tracks->tracks[i]->is_video)
 257                 set_index = 0;
 258             else if (tracks->tracks[i]->is_audio)
 259                 set_index = 1;
 260             else
 261                 continue;
 262             adaptation_sets[set_index][nb_tracks[set_index]++] = tracks->tracks[i];
 263         }
 264     } else {
 265         adaptation_sets = &tracks->tracks;
 266         nb_tracks = &tracks->nb_tracks;
 267         nb_sets = 1;
 268     }
 269
 270     out = fopen(filename, "w");
 271     if (!out) {
 272         ret = AVERROR(errno);
 273         perror(filename);
 274         return ret;
 275     }
 276     fprintf(out, "<?xml version=\"1.0\" encoding=\"utf-8\"?>\n");
 277     fprintf(out, "<MPD xmlns:xsi=\"http://www.w3.org/2001/XMLSchema-instance\"\n"
 278                 "\txmlns=\"urn:mpeg:dash:schema:mpd:2011\"\n"
 279                 "\txmlns:xlink=\"http://www.w3.org/1999/xlink\"\n"
 280                 "\txsi:schemaLocation=\"urn:mpeg:DASH:schema:MPD:2011 http://standards.iso.org/ittf/PubliclyAvailableStandards/MPEG-DASH_schema_files/DASH-MPD.xsd\"\n"
 281                 "\tprofiles=\"urn:mpeg:dash:profile:isoff-on-demand:2011\"\n"
 282                 "\ttype=\"static\"\n");
 283     fprintf(out, "\tmediaPresentationDuration=\"");
 284     write_time(out, tracks->duration, 1, AV_ROUND_DOWN);
 285     fprintf(out, "\"\n");
 286     fprintf(out, "\tminBufferTime=\"PT5S\">\n");
 287
 288     fprintf(out, "\t<Period start=\"PT0.0S\">\n");
 289
 290     for (set = 0; set < nb_sets; set++) {
 291         if (nb_tracks[set] == 0)
 292             continue;
 293         fprintf(out, "\t\t<AdaptationSet segmentAlignment=\"true\">\n");
 294         if (nb_sets == 1) {
 295             for (i = 0; i < nb_tracks[set]; i++) {
 296                 struct Track *track = adaptation_sets[set][i];
 297                 if (strcmp(track->name, adaptation_sets[set][0]->name))
 298                     break;
 299                 fprintf(out, "\t\t\t<ContentComponent id=\"%d\" contentType=\"%s\" />\n", track->track_id, track->is_audio ? "audio" : "video");
 300             }
 301         }
 302
 303         for (i = 0; i < nb_tracks[set]; ) {
 304             struct Track *first_track = adaptation_sets[set][i];
 305             int width = 0, height = 0, sample_rate = 0, channels = 0, bitrate = 0;
 306             fprintf(out, "\t\t\t<Representation id=\"%d\" codecs=\"", i);
 307             for (j = i; j < nb_tracks[set]; j++) {
 308                 struct Track *track = adaptation_sets[set][j];
 309                 if (strcmp(track->name, first_track->name))
 310                     break;
 311                 if (track->is_audio) {
 312                     sample_rate = track->sample_rate;
 313                     channels = track->channels;
 314                 }
 315                 if (track->is_video) {
 316                     width = track->width;
 317                     height = track->height;
 318                 }
 319                 bitrate += track->bitrate;
 320                 if (j > i)
 321                     fprintf(out, ",");
 322                 fprintf(out, "%s", track->codec_str);
 323             }
 324             fprintf(out, "\" mimeType=\"%s/mp4\" bandwidth=\"%d\"",
 325                     width ? "video" : "audio", bitrate);
 326             if (width > 0 && height > 0)
 327                 fprintf(out, " width=\"%d\" height=\"%d\"", width, height);
 328             if (sample_rate > 0)
 329                 fprintf(out, " audioSamplingRate=\"%d\"", sample_rate);
 330             fprintf(out, ">\n");
 331             if (channels > 0)
 332                 fprintf(out, "\t\t\t\t<AudioChannelConfiguration schemeIdUri=\"urn:mpeg:dash:23003:3:audio_channel_configuration:2011\" value=\"%d\" />\n", channels);
 333             fprintf(out, "\t\t\t\t<BaseURL>%s</BaseURL>\n", first_track->name);
 334             fprintf(out, "\t\t\t\t<SegmentBase indexRange=\"%"PRId64"-%"PRId64"\" />\n", first_track->sidx_start, first_track->sidx_start + first_track->sidx_length - 1);
 335             fprintf(out, "\t\t\t</Representation>\n");
 336             i = j;
 337         }
 338         fprintf(out, "\t\t</AdaptationSet>\n");
 339     }
 340     fprintf(out, "\t</Period>\n");
 341     fprintf(out, "</MPD>\n");
 342
 343     fclose(out);
 344 err:
 345     for (i = 0; i < 2; i++)
 346         av_free(adaptation_sets_buf[i]);
 347     return ret;
 348 }
 349
 350 static void clean_tracks(struct Tracks *tracks)
 351 {
 352     int i;
 353     for (i = 0; i < tracks->nb_tracks; i++) {
 354         av_freep(&tracks->tracks[i]);
 355     }
 356     av_freep(&tracks->tracks);
 357     tracks->nb_tracks = 0;
 358 }
 359
 360 int main(int argc, char **argv)
 361 {
 362     const char *out = NULL;
 363     struct Tracks tracks = { 0 };
 364     int i;
 365
 366     av_register_all();
 367
 368     for (i = 1; i < argc; i++) {
 369         if (!strcmp(argv[i], "-out")) {
 370             out = argv[i + 1];
 371             i++;
 372         } else if (argv[i][0] == '-') {
 373             return usage(argv[0], 1);
 374         } else {
 375             if (handle_file(&tracks, argv[i]))
 376                 return 1;
 377         }
 378     }
 379     if (!tracks.nb_tracks || !out)
 380         return usage(argv[0], 1);
 381
 382     output_mpd(&tracks, out);
 383
 384     clean_tracks(&tracks);
 385
 386     return 0;
 387 }