Merge commit '30dfc1dad4285e7362ce3f596d7c5d5d9b7fb33d'
[ffmpeg.git] / tools / sidxindex.c
1 /*
2  * Copyright (c) 2014 Martin Storsjo
3  *
4  * This file is part of FFmpeg.
5  *
6  * FFmpeg 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  * FFmpeg 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 FFmpeg; 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(AVCodecContext *codec, char *str, int size)
56 {
57     switch (codec->codec_id) {
58     case AV_CODEC_ID_H264:
59         snprintf(str, size, "avc1");
60         if (codec->extradata_size >= 4 && codec->extradata[0] == 1) {
61             av_strlcatf(str, size, ".%02x%02x%02x",
62                         codec->extradata[1], codec->extradata[2], codec->extradata[3]);
63         }
64         break;
65     case AV_CODEC_ID_AAC:
66         snprintf(str, size, "mp4a.40"); // 0x40 is the mp4 object type for AAC
67         if (codec->extradata_size >= 2) {
68             int aot = codec->extradata[0] >> 3;
69             if (aot == 31)
70                 aot = ((AV_RB16(codec->extradata) >> 5) & 0x3f) + 32;
71             av_strlcatf(str, size, ".%d", aot);
72         }
73         break;
74     }
75 }
76
77 static int find_sidx(struct Tracks *tracks, int start_index,
78                      const char *file)
79 {
80     int err = 0;
81     AVIOContext *f = NULL;
82     int i;
83
84     if ((err = avio_open2(&f, file, AVIO_FLAG_READ, NULL, NULL)) < 0)
85         goto fail;
86
87     while (!f->eof_reached) {
88         int64_t pos = avio_tell(f);
89         int32_t size, tag;
90
91         size = avio_rb32(f);
92         tag  = avio_rb32(f);
93         if (size < 8)
94             break;
95         if (tag == MKBETAG('s', 'i', 'd', 'x')) {
96             for (i = start_index; i < tracks->nb_tracks; i++) {
97                 struct Track *track = tracks->tracks[i];
98                 if (!track->sidx_start) {
99                     track->sidx_start  = pos;
100                     track->sidx_length = size;
101                 } else if (pos == track->sidx_start + track->sidx_length) {
102                     track->sidx_length = pos + size - track->sidx_start;
103                 }
104             }
105         }
106         if (avio_seek(f, pos + size, SEEK_SET) != pos + size)
107             break;
108     }
109
110 fail:
111     if (f)
112         avio_close(f);
113     return err;
114 }
115
116 static int handle_file(struct Tracks *tracks, const char *file)
117 {
118     AVFormatContext *ctx = NULL;
119     int err = 0, i, orig_tracks = tracks->nb_tracks;
120     char errbuf[50], *ptr;
121     struct Track *track;
122
123     err = avformat_open_input(&ctx, file, NULL, NULL);
124     if (err < 0) {
125         av_strerror(err, errbuf, sizeof(errbuf));
126         fprintf(stderr, "Unable to open %s: %s\n", file, errbuf);
127         return 1;
128     }
129
130     err = avformat_find_stream_info(ctx, NULL);
131     if (err < 0) {
132         av_strerror(err, errbuf, sizeof(errbuf));
133         fprintf(stderr, "Unable to identify %s: %s\n", file, errbuf);
134         goto fail;
135     }
136
137     if (ctx->nb_streams < 1) {
138         fprintf(stderr, "No streams found in %s\n", file);
139         goto fail;
140     }
141     if (ctx->nb_streams > 1)
142         tracks->multiple_tracks_per_file = 1;
143
144     for (i = 0; i < ctx->nb_streams; i++) {
145         struct Track **temp;
146         AVStream *st = ctx->streams[i];
147
148         if (st->codec->bit_rate == 0) {
149             fprintf(stderr, "Skipping track %d in %s as it has zero bitrate\n",
150                     st->id, file);
151             continue;
152         }
153
154         track = av_mallocz(sizeof(*track));
155         if (!track) {
156             err = AVERROR(ENOMEM);
157             goto fail;
158         }
159         temp = av_realloc_array(tracks->tracks, tracks->nb_tracks + 1,
160                                 sizeof(*tracks->tracks));
161         if (!temp) {
162             av_free(track);
163             err = AVERROR(ENOMEM);
164             goto fail;
165         }
166         tracks->tracks = temp;
167         tracks->tracks[tracks->nb_tracks] = track;
168
169         track->name = file;
170         if ((ptr = strrchr(file, '/')))
171             track->name = ptr + 1;
172
173         track->bitrate   = st->codec->bit_rate;
174         track->track_id  = st->id;
175         track->timescale = st->time_base.den;
176         track->duration  = st->duration;
177         track->is_audio  = st->codec->codec_type == AVMEDIA_TYPE_AUDIO;
178         track->is_video  = st->codec->codec_type == AVMEDIA_TYPE_VIDEO;
179
180         if (!track->is_audio && !track->is_video) {
181             fprintf(stderr,
182                     "Track %d in %s is neither video nor audio, skipping\n",
183                     track->track_id, file);
184             av_freep(&tracks->tracks[tracks->nb_tracks]);
185             continue;
186         }
187
188         tracks->duration = FFMAX(tracks->duration,
189                                  av_rescale_rnd(track->duration, AV_TIME_BASE,
190                                                 track->timescale, AV_ROUND_UP));
191
192         if (track->is_audio) {
193             track->channels    = st->codec->channels;
194             track->sample_rate = st->codec->sample_rate;
195         }
196         if (track->is_video) {
197             track->width  = st->codec->width;
198             track->height = st->codec->height;
199         }
200         set_codec_str(st->codec, track->codec_str, sizeof(track->codec_str));
201
202         tracks->nb_tracks++;
203     }
204
205     avformat_close_input(&ctx);
206
207     err = find_sidx(tracks, orig_tracks, file);
208
209 fail:
210     if (ctx)
211         avformat_close_input(&ctx);
212     return err;
213 }
214
215 static void write_time(FILE *out, int64_t time, int decimals, enum AVRounding round)
216 {
217     int seconds = time / AV_TIME_BASE;
218     int fractions = time % AV_TIME_BASE;
219     int minutes = seconds / 60;
220     int hours = minutes / 60;
221     fractions = av_rescale_rnd(fractions, pow(10, decimals), AV_TIME_BASE, round);
222     seconds %= 60;
223     minutes %= 60;
224     fprintf(out, "PT");
225     if (hours)
226         fprintf(out, "%dH", hours);
227     if (hours || minutes)
228         fprintf(out, "%dM", minutes);
229     fprintf(out, "%d.%0*dS", seconds, decimals, fractions);
230 }
231
232 static int output_mpd(struct Tracks *tracks, const char *filename)
233 {
234     FILE *out;
235     int i, j, ret = 0;
236     struct Track **adaptation_sets_buf[2] = { NULL };
237     struct Track ***adaptation_sets;
238     int nb_tracks_buf[2] = { 0 };
239     int *nb_tracks;
240     int set, nb_sets;
241
242     if (!tracks->multiple_tracks_per_file) {
243         adaptation_sets = adaptation_sets_buf;
244         nb_tracks = nb_tracks_buf;
245         nb_sets = 2;
246         for (i = 0; i < 2; i++) {
247             adaptation_sets[i] = av_malloc_array(tracks->nb_tracks, sizeof(*adaptation_sets[i]));
248             if (!adaptation_sets[i]) {
249                 ret = AVERROR(ENOMEM);
250                 goto err;
251             }
252         }
253         for (i = 0; i < tracks->nb_tracks; i++) {
254             int set_index = -1;
255             if (tracks->tracks[i]->is_video)
256                 set_index = 0;
257             else if (tracks->tracks[i]->is_audio)
258                 set_index = 1;
259             else
260                 continue;
261             adaptation_sets[set_index][nb_tracks[set_index]++] = tracks->tracks[i];
262         }
263     } else {
264         adaptation_sets = &tracks->tracks;
265         nb_tracks = &tracks->nb_tracks;
266         nb_sets = 1;
267     }
268
269     out = fopen(filename, "w");
270     if (!out) {
271         ret = AVERROR(errno);
272         perror(filename);
273         return ret;
274     }
275     fprintf(out, "<?xml version=\"1.0\" encoding=\"utf-8\"?>\n");
276     fprintf(out, "<MPD xmlns:xsi=\"http://www.w3.org/2001/XMLSchema-instance\"\n"
277                 "\txmlns=\"urn:mpeg:dash:schema:mpd:2011\"\n"
278                 "\txmlns:xlink=\"http://www.w3.org/1999/xlink\"\n"
279                 "\txsi:schemaLocation=\"urn:mpeg:DASH:schema:MPD:2011 http://standards.iso.org/ittf/PubliclyAvailableStandards/MPEG-DASH_schema_files/DASH-MPD.xsd\"\n"
280                 "\tprofiles=\"urn:mpeg:dash:profile:isoff-on-demand:2011\"\n"
281                 "\ttype=\"static\"\n");
282     fprintf(out, "\tmediaPresentationDuration=\"");
283     write_time(out, tracks->duration, 1, AV_ROUND_DOWN);
284     fprintf(out, "\"\n");
285     fprintf(out, "\tminBufferTime=\"PT5S\">\n");
286
287     fprintf(out, "\t<Period start=\"PT0.0S\">\n");
288
289     for (set = 0; set < nb_sets; set++) {
290         if (nb_tracks[set] == 0)
291             continue;
292         fprintf(out, "\t\t<AdaptationSet segmentAlignment=\"true\">\n");
293         if (nb_sets == 1) {
294             for (i = 0; i < nb_tracks[set]; i++) {
295                 struct Track *track = adaptation_sets[set][i];
296                 if (strcmp(track->name, adaptation_sets[set][0]->name))
297                     break;
298                 fprintf(out, "\t\t\t<ContentComponent id=\"%d\" contentType=\"%s\" />\n", track->track_id, track->is_audio ? "audio" : "video");
299             }
300         }
301
302         for (i = 0; i < nb_tracks[set]; ) {
303             struct Track *first_track = adaptation_sets[set][i];
304             int width = 0, height = 0, sample_rate = 0, channels = 0, bitrate = 0;
305             fprintf(out, "\t\t\t<Representation id=\"%d\" codecs=\"", i);
306             for (j = i; j < nb_tracks[set]; j++) {
307                 struct Track *track = adaptation_sets[set][j];
308                 if (strcmp(track->name, first_track->name))
309                     break;
310                 if (track->is_audio) {
311                     sample_rate = track->sample_rate;
312                     channels = track->channels;
313                 }
314                 if (track->is_video) {
315                     width = track->width;
316                     height = track->height;
317                 }
318                 bitrate += track->bitrate;
319                 if (j > i)
320                     fprintf(out, ",");
321                 fprintf(out, "%s", track->codec_str);
322             }
323             fprintf(out, "\" mimeType=\"%s/mp4\" bandwidth=\"%d\"",
324                     width ? "video" : "audio", bitrate);
325             if (width > 0 && height > 0)
326                 fprintf(out, " width=\"%d\" height=\"%d\"", width, height);
327             if (sample_rate > 0)
328                 fprintf(out, " audioSamplingRate=\"%d\"", sample_rate);
329             fprintf(out, ">\n");
330             if (channels > 0)
331                 fprintf(out, "\t\t\t\t<AudioChannelConfiguration schemeIdUri=\"urn:mpeg:dash:23003:3:audio_channel_configuration:2011\" value=\"%d\" />\n", channels);
332             fprintf(out, "\t\t\t\t<BaseURL>%s</BaseURL>\n", first_track->name);
333             fprintf(out, "\t\t\t\t<SegmentBase indexRange=\"%"PRId64"-%"PRId64"\" />\n", first_track->sidx_start, first_track->sidx_start + first_track->sidx_length - 1);
334             fprintf(out, "\t\t\t</Representation>\n");
335             i = j;
336         }
337         fprintf(out, "\t\t</AdaptationSet>\n");
338     }
339     fprintf(out, "\t</Period>\n");
340     fprintf(out, "</MPD>\n");
341
342     fclose(out);
343 err:
344     for (i = 0; i < 2; i++)
345         av_free(adaptation_sets_buf[i]);
346     return ret;
347 }
348
349 static void clean_tracks(struct Tracks *tracks)
350 {
351     int i;
352     for (i = 0; i < tracks->nb_tracks; i++) {
353         av_freep(&tracks->tracks[i]);
354     }
355     av_freep(&tracks->tracks);
356     tracks->nb_tracks = 0;
357 }
358
359 int main(int argc, char **argv)
360 {
361     const char *out = NULL;
362     struct Tracks tracks = { 0 };
363     int i;
364
365     av_register_all();
366
367     for (i = 1; i < argc; i++) {
368         if (!strcmp(argv[i], "-out")) {
369             out = argv[i + 1];
370             i++;
371         } else if (argv[i][0] == '-') {
372             return usage(argv[0], 1);
373         } else {
374             if (handle_file(&tracks, argv[i]))
375                 return 1;
376         }
377     }
378     if (!tracks.nb_tracks || !out)
379         return usage(argv[0], 1);
380
381     output_mpd(&tracks, out);
382
383     clean_tracks(&tracks);
384
385     return 0;
386 }