doc/general: s/Libav/FFmpeg
[ffmpeg.git] / tools / ismindex.c
1 /*
2  * Copyright (c) 2012 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 /*
22  * To create a simple file for smooth streaming:
23  * ffmpeg <normal input/transcoding options> -movflags frag_keyframe foo.ismv
24  * ismindex -n foo foo.ismv
25  * This step creates foo.ism and foo.ismc that is required by IIS for
26  * serving it.
27  *
28  * To pre-split files for serving as static files by a web server without
29  * any extra server support, create the ismv file as above, and split it:
30  * ismindex -split foo.ismv
31  * This step creates a file Manifest and directories QualityLevel(...),
32  * that can be read directly by a smooth streaming player.
33  */
34
35 #include <stdio.h>
36 #include <string.h>
37 #include <sys/stat.h>
38 #ifdef _WIN32
39 #include <direct.h>
40 #define mkdir(a, b) _mkdir(a)
41 #endif
42
43 #include "cmdutils.h"
44
45 #include "libavformat/avformat.h"
46 #include "libavutil/intreadwrite.h"
47 #include "libavutil/mathematics.h"
48
49 static int usage(const char *argv0, int ret)
50 {
51     fprintf(stderr, "%s [-split] [-n basename] file1 [file2] ...\n", argv0);
52     return ret;
53 }
54
55 struct MoofOffset {
56     int64_t time;
57     int64_t offset;
58     int duration;
59 };
60
61 struct Track {
62     const char *name;
63     int64_t duration;
64     int bitrate;
65     int track_id;
66     int is_audio, is_video;
67     int width, height;
68     int chunks;
69     int sample_rate, channels;
70     uint8_t *codec_private;
71     int codec_private_size;
72     struct MoofOffset *offsets;
73     int timescale;
74     const char *fourcc;
75     int blocksize;
76     int tag;
77 };
78
79 struct Tracks {
80     int nb_tracks;
81     int64_t duration;
82     struct Track **tracks;
83     int video_track, audio_track;
84     int nb_video_tracks, nb_audio_tracks;
85 };
86
87 static int copy_tag(AVIOContext *in, AVIOContext *out, int32_t tag_name)
88 {
89     int32_t size, tag;
90
91     size = avio_rb32(in);
92     tag  = avio_rb32(in);
93     avio_wb32(out, size);
94     avio_wb32(out, tag);
95     if (tag != tag_name)
96         return -1;
97     size -= 8;
98     while (size > 0) {
99         char buf[1024];
100         int len = FFMIN(sizeof(buf), size);
101         if (avio_read(in, buf, len) != len)
102             break;
103         avio_write(out, buf, len);
104         size -= len;
105     }
106     return 0;
107 }
108
109 static int write_fragment(const char *filename, AVIOContext *in)
110 {
111     AVIOContext *out = NULL;
112     int ret;
113
114     if ((ret = avio_open2(&out, filename, AVIO_FLAG_WRITE, NULL, NULL)) < 0)
115         return ret;
116     copy_tag(in, out, MKBETAG('m', 'o', 'o', 'f'));
117     copy_tag(in, out, MKBETAG('m', 'd', 'a', 't'));
118
119     avio_flush(out);
120     avio_close(out);
121
122     return ret;
123 }
124
125 static int write_fragments(struct Tracks *tracks, int start_index,
126                            AVIOContext *in)
127 {
128     char dirname[100], filename[500];
129     int i, j;
130
131     for (i = start_index; i < tracks->nb_tracks; i++) {
132         struct Track *track = tracks->tracks[i];
133         const char *type    = track->is_video ? "video" : "audio";
134         snprintf(dirname, sizeof(dirname), "QualityLevels(%d)", track->bitrate);
135         if (mkdir(dirname, 0777) == -1)
136             return AVERROR(errno);
137         for (j = 0; j < track->chunks; j++) {
138             snprintf(filename, sizeof(filename), "%s/Fragments(%s=%"PRId64")",
139                      dirname, type, track->offsets[j].time);
140             avio_seek(in, track->offsets[j].offset, SEEK_SET);
141             write_fragment(filename, in);
142         }
143     }
144     return 0;
145 }
146
147 static int read_tfra(struct Tracks *tracks, int start_index, AVIOContext *f)
148 {
149     int ret = AVERROR_EOF, track_id;
150     int version, fieldlength, i, j;
151     int64_t pos   = avio_tell(f);
152     uint32_t size = avio_rb32(f);
153     struct Track *track = NULL;
154
155     if (avio_rb32(f) != MKBETAG('t', 'f', 'r', 'a'))
156         goto fail;
157     version = avio_r8(f);
158     avio_rb24(f);
159     track_id = avio_rb32(f); /* track id */
160     for (i = start_index; i < tracks->nb_tracks && !track; i++)
161         if (tracks->tracks[i]->track_id == track_id)
162             track = tracks->tracks[i];
163     if (!track) {
164         /* Ok, continue parsing the next atom */
165         ret = 0;
166         goto fail;
167     }
168     fieldlength = avio_rb32(f);
169     track->chunks  = avio_rb32(f);
170     track->offsets = av_mallocz(sizeof(*track->offsets) * track->chunks);
171     if (!track->offsets) {
172         ret = AVERROR(ENOMEM);
173         goto fail;
174     }
175     for (i = 0; i < track->chunks; i++) {
176         if (version == 1) {
177             track->offsets[i].time   = avio_rb64(f);
178             track->offsets[i].offset = avio_rb64(f);
179         } else {
180             track->offsets[i].time   = avio_rb32(f);
181             track->offsets[i].offset = avio_rb32(f);
182         }
183         for (j = 0; j < ((fieldlength >> 4) & 3) + 1; j++)
184             avio_r8(f);
185         for (j = 0; j < ((fieldlength >> 2) & 3) + 1; j++)
186             avio_r8(f);
187         for (j = 0; j < ((fieldlength >> 0) & 3) + 1; j++)
188             avio_r8(f);
189         if (i > 0)
190             track->offsets[i - 1].duration = track->offsets[i].time -
191                                              track->offsets[i - 1].time;
192     }
193     if (track->chunks > 0)
194         track->offsets[track->chunks - 1].duration = track->duration -
195                                                      track->offsets[track->chunks - 1].time;
196     ret = 0;
197
198 fail:
199     avio_seek(f, pos + size, SEEK_SET);
200     return ret;
201 }
202
203 static int read_mfra(struct Tracks *tracks, int start_index,
204                      const char *file, int split)
205 {
206     int err = 0;
207     AVIOContext *f = NULL;
208     int32_t mfra_size;
209
210     if ((err = avio_open2(&f, file, AVIO_FLAG_READ, NULL, NULL)) < 0)
211         goto fail;
212     avio_seek(f, avio_size(f) - 4, SEEK_SET);
213     mfra_size = avio_rb32(f);
214     avio_seek(f, -mfra_size, SEEK_CUR);
215     if (avio_rb32(f) != mfra_size) {
216         err = AVERROR_INVALIDDATA;
217         goto fail;
218     }
219     if (avio_rb32(f) != MKBETAG('m', 'f', 'r', 'a')) {
220         err = AVERROR_INVALIDDATA;
221         goto fail;
222     }
223     while (!read_tfra(tracks, start_index, f)) {
224         /* Empty */
225     }
226
227     if (split)
228         err = write_fragments(tracks, start_index, f);
229
230 fail:
231     if (f)
232         avio_close(f);
233     if (err)
234         fprintf(stderr, "Unable to read the MFRA atom in %s\n", file);
235     return err;
236 }
237
238 static int get_private_data(struct Track *track, AVCodecContext *codec)
239 {
240     track->codec_private_size = codec->extradata_size;
241     track->codec_private      = av_mallocz(codec->extradata_size);
242     if (!track->codec_private)
243         return AVERROR(ENOMEM);
244     memcpy(track->codec_private, codec->extradata, codec->extradata_size);
245     return 0;
246 }
247
248 static int get_video_private_data(struct Track *track, AVCodecContext *codec)
249 {
250     AVIOContext *io = NULL;
251     uint16_t sps_size, pps_size;
252     int err = AVERROR(EINVAL);
253
254     if (codec->codec_id == AV_CODEC_ID_VC1)
255         return get_private_data(track, codec);
256
257     if (avio_open_dyn_buf(&io) < 0)  {
258         err = AVERROR(ENOMEM);
259         goto fail;
260     }
261     if (codec->extradata_size < 11 || codec->extradata[0] != 1)
262         goto fail;
263     sps_size = AV_RB16(&codec->extradata[6]);
264     if (11 + sps_size > codec->extradata_size)
265         goto fail;
266     avio_wb32(io, 0x00000001);
267     avio_write(io, &codec->extradata[8], sps_size);
268     pps_size = AV_RB16(&codec->extradata[9 + sps_size]);
269     if (11 + sps_size + pps_size > codec->extradata_size)
270         goto fail;
271     avio_wb32(io, 0x00000001);
272     avio_write(io, &codec->extradata[11 + sps_size], pps_size);
273     err = 0;
274
275 fail:
276     track->codec_private_size = avio_close_dyn_buf(io, &track->codec_private);
277     return err;
278 }
279
280 static int handle_file(struct Tracks *tracks, const char *file, int split)
281 {
282     AVFormatContext *ctx = NULL;
283     int err = 0, i, orig_tracks = tracks->nb_tracks;
284     char errbuf[50], *ptr;
285     struct Track *track;
286
287     err = avformat_open_input(&ctx, file, NULL, NULL);
288     if (err < 0) {
289         av_strerror(err, errbuf, sizeof(errbuf));
290         fprintf(stderr, "Unable to open %s: %s\n", file, errbuf);
291         return 1;
292     }
293
294     err = avformat_find_stream_info(ctx, NULL);
295     if (err < 0) {
296         av_strerror(err, errbuf, sizeof(errbuf));
297         fprintf(stderr, "Unable to identify %s: %s\n", file, errbuf);
298         goto fail;
299     }
300
301     if (ctx->nb_streams < 1) {
302         fprintf(stderr, "No streams found in %s\n", file);
303         goto fail;
304     }
305     if (!tracks->duration)
306         tracks->duration = ctx->duration;
307
308     for (i = 0; i < ctx->nb_streams; i++) {
309         struct Track **temp;
310         AVStream *st = ctx->streams[i];
311         track = av_mallocz(sizeof(*track));
312         if (!track) {
313             err = AVERROR(ENOMEM);
314             goto fail;
315         }
316         temp = av_realloc(tracks->tracks,
317                           sizeof(*tracks->tracks) * (tracks->nb_tracks + 1));
318         if (!temp) {
319             av_free(track);
320             err = AVERROR(ENOMEM);
321             goto fail;
322         }
323         tracks->tracks = temp;
324         tracks->tracks[tracks->nb_tracks] = track;
325
326         track->name = file;
327         if ((ptr = strrchr(file, '/')) != NULL)
328             track->name = ptr + 1;
329
330         track->bitrate   = st->codec->bit_rate;
331         track->track_id  = st->id;
332         track->timescale = st->time_base.den;
333         track->duration  = av_rescale_rnd(ctx->duration, track->timescale,
334                                           AV_TIME_BASE, AV_ROUND_UP);
335         track->is_audio  = st->codec->codec_type == AVMEDIA_TYPE_AUDIO;
336         track->is_video  = st->codec->codec_type == AVMEDIA_TYPE_VIDEO;
337
338         if (!track->is_audio && !track->is_video) {
339             fprintf(stderr,
340                     "Track %d in %s is neither video nor audio, skipping\n",
341                     track->track_id, file);
342             av_freep(&tracks->tracks[tracks->nb_tracks]);
343             continue;
344         }
345
346         if (track->is_audio) {
347             if (tracks->audio_track < 0)
348                 tracks->audio_track = tracks->nb_tracks;
349             tracks->nb_audio_tracks++;
350             track->channels    = st->codec->channels;
351             track->sample_rate = st->codec->sample_rate;
352             if (st->codec->codec_id == AV_CODEC_ID_AAC) {
353                 track->fourcc    = "AACL";
354                 track->tag       = 255;
355                 track->blocksize = 4;
356             } else if (st->codec->codec_id == AV_CODEC_ID_WMAPRO) {
357                 track->fourcc    = "WMAP";
358                 track->tag       = st->codec->codec_tag;
359                 track->blocksize = st->codec->block_align;
360             }
361             get_private_data(track, st->codec);
362         }
363         if (track->is_video) {
364             if (tracks->video_track < 0)
365                 tracks->video_track = tracks->nb_tracks;
366             tracks->nb_video_tracks++;
367             track->width  = st->codec->width;
368             track->height = st->codec->height;
369             if (st->codec->codec_id == AV_CODEC_ID_H264)
370                 track->fourcc = "H264";
371             else if (st->codec->codec_id == AV_CODEC_ID_VC1)
372                 track->fourcc = "WVC1";
373             get_video_private_data(track, st->codec);
374         }
375
376         tracks->nb_tracks++;
377     }
378
379     avformat_close_input(&ctx);
380
381     err = read_mfra(tracks, orig_tracks, file, split);
382
383 fail:
384     if (ctx)
385         avformat_close_input(&ctx);
386     return err;
387 }
388
389 static void output_server_manifest(struct Tracks *tracks,
390                                    const char *basename)
391 {
392     char filename[1000];
393     FILE *out;
394     int i;
395
396     snprintf(filename, sizeof(filename), "%s.ism", basename);
397     out = fopen(filename, "w");
398     if (!out) {
399         perror(filename);
400         return;
401     }
402     fprintf(out, "<?xml version=\"1.0\" encoding=\"utf-8\"?>\n");
403     fprintf(out, "<smil xmlns=\"http://www.w3.org/2001/SMIL20/Language\">\n");
404     fprintf(out, "\t<head>\n");
405     fprintf(out, "\t\t<meta name=\"clientManifestRelativePath\" "
406                  "content=\"%s.ismc\" />\n", basename);
407     fprintf(out, "\t</head>\n");
408     fprintf(out, "\t<body>\n");
409     fprintf(out, "\t\t<switch>\n");
410     for (i = 0; i < tracks->nb_tracks; i++) {
411         struct Track *track = tracks->tracks[i];
412         const char *type    = track->is_video ? "video" : "audio";
413         fprintf(out, "\t\t\t<%s src=\"%s\" systemBitrate=\"%d\">\n",
414                 type, track->name, track->bitrate);
415         fprintf(out, "\t\t\t\t<param name=\"trackID\" value=\"%d\" "
416                      "valueType=\"data\" />\n", track->track_id);
417         fprintf(out, "\t\t\t</%s>\n", type);
418     }
419     fprintf(out, "\t\t</switch>\n");
420     fprintf(out, "\t</body>\n");
421     fprintf(out, "</smil>\n");
422     fclose(out);
423 }
424
425 static void print_track_chunks(FILE *out, struct Tracks *tracks, int main,
426                                const char *type)
427 {
428     int i, j;
429     struct Track *track = tracks->tracks[main];
430     for (i = 0; i < track->chunks; i++) {
431         for (j = main + 1; j < tracks->nb_tracks; j++) {
432             if (tracks->tracks[j]->is_audio == track->is_audio &&
433                 track->offsets[i].duration != tracks->tracks[j]->offsets[i].duration)
434                 fprintf(stderr, "Mismatched duration of %s chunk %d in %s and %s\n",
435                         type, i, track->name, tracks->tracks[j]->name);
436         }
437         fprintf(out, "\t\t<c n=\"%d\" d=\"%d\" />\n",
438                 i, track->offsets[i].duration);
439     }
440 }
441
442 static void output_client_manifest(struct Tracks *tracks,
443                                    const char *basename, int split)
444 {
445     char filename[1000];
446     FILE *out;
447     int i, j;
448
449     if (split)
450         snprintf(filename, sizeof(filename), "Manifest");
451     else
452         snprintf(filename, sizeof(filename), "%s.ismc", basename);
453     out = fopen(filename, "w");
454     if (!out) {
455         perror(filename);
456         return;
457     }
458     fprintf(out, "<?xml version=\"1.0\" encoding=\"utf-8\"?>\n");
459     fprintf(out, "<SmoothStreamingMedia MajorVersion=\"2\" MinorVersion=\"0\" "
460                  "Duration=\"%"PRId64 "\">\n", tracks->duration * 10);
461     if (tracks->video_track >= 0) {
462         struct Track *track = tracks->tracks[tracks->video_track];
463         struct Track *first_track = track;
464         int index = 0;
465         fprintf(out,
466                 "\t<StreamIndex Type=\"video\" QualityLevels=\"%d\" "
467                 "Chunks=\"%d\" "
468                 "Url=\"QualityLevels({bitrate})/Fragments(video={start time})\">\n",
469                 tracks->nb_video_tracks, track->chunks);
470         for (i = 0; i < tracks->nb_tracks; i++) {
471             track = tracks->tracks[i];
472             if (!track->is_video)
473                 continue;
474             fprintf(out,
475                     "\t\t<QualityLevel Index=\"%d\" Bitrate=\"%d\" "
476                     "FourCC=\"%s\" MaxWidth=\"%d\" MaxHeight=\"%d\" "
477                     "CodecPrivateData=\"",
478                     index, track->bitrate, track->fourcc, track->width, track->height);
479             for (j = 0; j < track->codec_private_size; j++)
480                 fprintf(out, "%02X", track->codec_private[j]);
481             fprintf(out, "\" />\n");
482             index++;
483             if (track->chunks != first_track->chunks)
484                 fprintf(stderr, "Mismatched number of video chunks in %s and %s\n",
485                         track->name, first_track->name);
486         }
487         print_track_chunks(out, tracks, tracks->video_track, "video");
488         fprintf(out, "\t</StreamIndex>\n");
489     }
490     if (tracks->audio_track >= 0) {
491         struct Track *track = tracks->tracks[tracks->audio_track];
492         struct Track *first_track = track;
493         int index = 0;
494         fprintf(out,
495                 "\t<StreamIndex Type=\"audio\" QualityLevels=\"%d\" "
496                 "Chunks=\"%d\" "
497                 "Url=\"QualityLevels({bitrate})/Fragments(audio={start time})\">\n",
498                 tracks->nb_audio_tracks, track->chunks);
499         for (i = 0; i < tracks->nb_tracks; i++) {
500             track = tracks->tracks[i];
501             if (!track->is_audio)
502                 continue;
503             fprintf(out,
504                     "\t\t<QualityLevel Index=\"%d\" Bitrate=\"%d\" "
505                     "FourCC=\"%s\" SamplingRate=\"%d\" Channels=\"%d\" "
506                     "BitsPerSample=\"16\" PacketSize=\"%d\" "
507                     "AudioTag=\"%d\" CodecPrivateData=\"",
508                     index, track->bitrate, track->fourcc, track->sample_rate,
509                     track->channels, track->blocksize, track->tag);
510             for (j = 0; j < track->codec_private_size; j++)
511                 fprintf(out, "%02X", track->codec_private[j]);
512             fprintf(out, "\" />\n");
513             index++;
514             if (track->chunks != first_track->chunks)
515                 fprintf(stderr, "Mismatched number of audio chunks in %s and %s\n",
516                         track->name, first_track->name);
517         }
518         print_track_chunks(out, tracks, tracks->audio_track, "audio");
519         fprintf(out, "\t</StreamIndex>\n");
520     }
521     fprintf(out, "</SmoothStreamingMedia>\n");
522     fclose(out);
523 }
524
525 static void clean_tracks(struct Tracks *tracks)
526 {
527     int i;
528     for (i = 0; i < tracks->nb_tracks; i++) {
529         av_freep(&tracks->tracks[i]->codec_private);
530         av_freep(&tracks->tracks[i]->offsets);
531         av_freep(&tracks->tracks[i]);
532     }
533     av_freep(&tracks->tracks);
534     tracks->nb_tracks = 0;
535 }
536
537 int main(int argc, char **argv)
538 {
539     const char *basename = NULL;
540     int split = 0, i;
541     struct Tracks tracks = { 0, .video_track = -1, .audio_track = -1 };
542
543     av_register_all();
544
545     for (i = 1; i < argc; i++) {
546         if (!strcmp(argv[i], "-n")) {
547             basename = argv[i + 1];
548             i++;
549         } else if (!strcmp(argv[i], "-split")) {
550             split = 1;
551         } else if (argv[i][0] == '-') {
552             return usage(argv[0], 1);
553         } else {
554             if (handle_file(&tracks, argv[i], split))
555                 return 1;
556         }
557     }
558     if (!tracks.nb_tracks || (!basename && !split))
559         return usage(argv[0], 1);
560
561     if (!split)
562         output_server_manifest(&tracks, basename);
563     output_client_manifest(&tracks, basename, split);
564
565     clean_tracks(&tracks);
566
567     return 0;
568 }