ismindex: Add an option for outputting files elsewhere than in the current directory
[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  * avconv <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  * By adding -path-prefix path/, the produced foo.ism will refer to the
29  * files foo.ismv as "path/foo.ismv" - the prefix for the generated ismc
30  * file can be set with the -ismc-prefix option similarly.
31  *
32  * To pre-split files for serving as static files by a web server without
33  * any extra server support, create the ismv file as above, and split it:
34  * ismindex -split foo.ismv
35  * This step creates a file Manifest and directories QualityLevel(...),
36  * that can be read directly by a smooth streaming player.
37  *
38  * The -output dir option can be used to request that output files
39  * (both .ism/.ismc, or Manifest/QualityLevels* when splitting)
40  * should be written to this directory instead of in the current directory.
41  * (The directory itself isn't created if it doesn't already exist.)
42  */
43
44 #include <stdio.h>
45 #include <string.h>
46
47 #include "libavformat/avformat.h"
48 #include "libavformat/os_support.h"
49 #include "libavutil/intreadwrite.h"
50 #include "libavutil/mathematics.h"
51
52 static int usage(const char *argv0, int ret)
53 {
54     fprintf(stderr, "%s [-split] [-n basename] [-path-prefix prefix] "
55                     "[-ismc-prefix prefix] [-output dir] file1 [file2] ...\n", argv0);
56     return ret;
57 }
58
59 struct MoofOffset {
60     int64_t time;
61     int64_t offset;
62     int64_t duration;
63 };
64
65 struct Track {
66     const char *name;
67     int64_t duration;
68     int bitrate;
69     int track_id;
70     int is_audio, is_video;
71     int width, height;
72     int chunks;
73     int sample_rate, channels;
74     uint8_t *codec_private;
75     int codec_private_size;
76     struct MoofOffset *offsets;
77     int timescale;
78     const char *fourcc;
79     int blocksize;
80     int tag;
81 };
82
83 struct Tracks {
84     int nb_tracks;
85     int64_t duration;
86     struct Track **tracks;
87     int video_track, audio_track;
88     int nb_video_tracks, nb_audio_tracks;
89 };
90
91 static int copy_tag(AVIOContext *in, AVIOContext *out, int32_t tag_name)
92 {
93     int32_t size, tag;
94
95     size = avio_rb32(in);
96     tag  = avio_rb32(in);
97     avio_wb32(out, size);
98     avio_wb32(out, tag);
99     if (tag != tag_name)
100         return -1;
101     size -= 8;
102     while (size > 0) {
103         char buf[1024];
104         int len = FFMIN(sizeof(buf), size);
105         if (avio_read(in, buf, len) != len)
106             break;
107         avio_write(out, buf, len);
108         size -= len;
109     }
110     return 0;
111 }
112
113 static int write_fragment(const char *filename, AVIOContext *in)
114 {
115     AVIOContext *out = NULL;
116     int ret;
117
118     if ((ret = avio_open2(&out, filename, AVIO_FLAG_WRITE, NULL, NULL)) < 0)
119         return ret;
120     copy_tag(in, out, MKBETAG('m', 'o', 'o', 'f'));
121     copy_tag(in, out, MKBETAG('m', 'd', 'a', 't'));
122
123     avio_flush(out);
124     avio_close(out);
125
126     return ret;
127 }
128
129 static int write_fragments(struct Tracks *tracks, int start_index,
130                            AVIOContext *in, const char *output_prefix)
131 {
132     char dirname[2048], filename[2048];
133     int i, j;
134
135     for (i = start_index; i < tracks->nb_tracks; i++) {
136         struct Track *track = tracks->tracks[i];
137         const char *type    = track->is_video ? "video" : "audio";
138         snprintf(dirname, sizeof(dirname), "%sQualityLevels(%d)", output_prefix, track->bitrate);
139         mkdir(dirname, 0777);
140         for (j = 0; j < track->chunks; j++) {
141             snprintf(filename, sizeof(filename), "%s/Fragments(%s=%"PRId64")",
142                      dirname, type, track->offsets[j].time);
143             avio_seek(in, track->offsets[j].offset, SEEK_SET);
144             write_fragment(filename, in);
145         }
146     }
147     return 0;
148 }
149
150 static int read_tfra(struct Tracks *tracks, int start_index, AVIOContext *f)
151 {
152     int ret = AVERROR_EOF, track_id;
153     int version, fieldlength, i, j;
154     int64_t pos   = avio_tell(f);
155     uint32_t size = avio_rb32(f);
156     struct Track *track = NULL;
157
158     if (avio_rb32(f) != MKBETAG('t', 'f', 'r', 'a'))
159         goto fail;
160     version = avio_r8(f);
161     avio_rb24(f);
162     track_id = avio_rb32(f); /* track id */
163     for (i = start_index; i < tracks->nb_tracks && !track; i++)
164         if (tracks->tracks[i]->track_id == track_id)
165             track = tracks->tracks[i];
166     if (!track) {
167         /* Ok, continue parsing the next atom */
168         ret = 0;
169         goto fail;
170     }
171     fieldlength = avio_rb32(f);
172     track->chunks  = avio_rb32(f);
173     track->offsets = av_mallocz(sizeof(*track->offsets) * track->chunks);
174     if (!track->offsets) {
175         ret = AVERROR(ENOMEM);
176         goto fail;
177     }
178     for (i = 0; i < track->chunks; i++) {
179         if (version == 1) {
180             track->offsets[i].time   = avio_rb64(f);
181             track->offsets[i].offset = avio_rb64(f);
182         } else {
183             track->offsets[i].time   = avio_rb32(f);
184             track->offsets[i].offset = avio_rb32(f);
185         }
186         for (j = 0; j < ((fieldlength >> 4) & 3) + 1; j++)
187             avio_r8(f);
188         for (j = 0; j < ((fieldlength >> 2) & 3) + 1; j++)
189             avio_r8(f);
190         for (j = 0; j < ((fieldlength >> 0) & 3) + 1; j++)
191             avio_r8(f);
192         if (i > 0)
193             track->offsets[i - 1].duration = track->offsets[i].time -
194                                              track->offsets[i - 1].time;
195     }
196     if (track->chunks > 0)
197         track->offsets[track->chunks - 1].duration = track->duration -
198                                                      track->offsets[track->chunks - 1].time;
199     ret = 0;
200
201 fail:
202     avio_seek(f, pos + size, SEEK_SET);
203     return ret;
204 }
205
206 static int read_mfra(struct Tracks *tracks, int start_index,
207                      const char *file, int split, const char *output_prefix)
208 {
209     int err = 0;
210     AVIOContext *f = NULL;
211     int32_t mfra_size;
212
213     if ((err = avio_open2(&f, file, AVIO_FLAG_READ, NULL, NULL)) < 0)
214         goto fail;
215     avio_seek(f, avio_size(f) - 4, SEEK_SET);
216     mfra_size = avio_rb32(f);
217     avio_seek(f, -mfra_size, SEEK_CUR);
218     if (avio_rb32(f) != mfra_size) {
219         err = AVERROR_INVALIDDATA;
220         goto fail;
221     }
222     if (avio_rb32(f) != MKBETAG('m', 'f', 'r', 'a')) {
223         err = AVERROR_INVALIDDATA;
224         goto fail;
225     }
226     while (!read_tfra(tracks, start_index, f)) {
227         /* Empty */
228     }
229
230     if (split)
231         write_fragments(tracks, start_index, f, output_prefix);
232
233 fail:
234     if (f)
235         avio_close(f);
236     if (err)
237         fprintf(stderr, "Unable to read the MFRA atom in %s\n", file);
238     return err;
239 }
240
241 static int get_private_data(struct Track *track, AVCodecContext *codec)
242 {
243     track->codec_private_size = codec->extradata_size;
244     track->codec_private      = av_mallocz(codec->extradata_size);
245     if (!track->codec_private)
246         return AVERROR(ENOMEM);
247     memcpy(track->codec_private, codec->extradata, codec->extradata_size);
248     return 0;
249 }
250
251 static int get_video_private_data(struct Track *track, AVCodecContext *codec)
252 {
253     AVIOContext *io = NULL;
254     uint16_t sps_size, pps_size;
255     int err = AVERROR(EINVAL);
256
257     if (codec->codec_id == AV_CODEC_ID_VC1)
258         return get_private_data(track, codec);
259
260     avio_open_dyn_buf(&io);
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                        const char *output_prefix)
282 {
283     AVFormatContext *ctx = NULL;
284     int err = 0, i, orig_tracks = tracks->nb_tracks;
285     char errbuf[50], *ptr;
286     struct Track *track;
287
288     err = avformat_open_input(&ctx, file, NULL, NULL);
289     if (err < 0) {
290         av_strerror(err, errbuf, sizeof(errbuf));
291         fprintf(stderr, "Unable to open %s: %s\n", file, errbuf);
292         return 1;
293     }
294
295     err = avformat_find_stream_info(ctx, NULL);
296     if (err < 0) {
297         av_strerror(err, errbuf, sizeof(errbuf));
298         fprintf(stderr, "Unable to identify %s: %s\n", file, errbuf);
299         goto fail;
300     }
301
302     if (ctx->nb_streams < 1) {
303         fprintf(stderr, "No streams found in %s\n", file);
304         goto fail;
305     }
306
307     for (i = 0; i < ctx->nb_streams; i++) {
308         struct Track **temp;
309         AVStream *st = ctx->streams[i];
310         track = av_mallocz(sizeof(*track));
311         if (!track) {
312             err = AVERROR(ENOMEM);
313             goto fail;
314         }
315         temp = av_realloc(tracks->tracks,
316                           sizeof(*tracks->tracks) * (tracks->nb_tracks + 1));
317         if (!temp) {
318             av_free(track);
319             err = AVERROR(ENOMEM);
320             goto fail;
321         }
322         tracks->tracks = temp;
323         tracks->tracks[tracks->nb_tracks] = track;
324
325         track->name = file;
326         if ((ptr = strrchr(file, '/')) != NULL)
327             track->name = ptr + 1;
328
329         track->bitrate   = st->codec->bit_rate;
330         track->track_id  = st->id;
331         track->timescale = st->time_base.den;
332         track->duration  = st->duration;
333         track->is_audio  = st->codec->codec_type == AVMEDIA_TYPE_AUDIO;
334         track->is_video  = st->codec->codec_type == AVMEDIA_TYPE_VIDEO;
335
336         if (!track->is_audio && !track->is_video) {
337             fprintf(stderr,
338                     "Track %d in %s is neither video nor audio, skipping\n",
339                     track->track_id, file);
340             av_freep(&tracks->tracks[tracks->nb_tracks]);
341             continue;
342         }
343
344         tracks->duration = FFMAX(tracks->duration,
345                                  av_rescale_rnd(track->duration, AV_TIME_BASE,
346                                                 track->timescale, AV_ROUND_UP));
347
348         if (track->is_audio) {
349             if (tracks->audio_track < 0)
350                 tracks->audio_track = tracks->nb_tracks;
351             tracks->nb_audio_tracks++;
352             track->channels    = st->codec->channels;
353             track->sample_rate = st->codec->sample_rate;
354             if (st->codec->codec_id == AV_CODEC_ID_AAC) {
355                 track->fourcc    = "AACL";
356                 track->tag       = 255;
357                 track->blocksize = 4;
358             } else if (st->codec->codec_id == AV_CODEC_ID_WMAPRO) {
359                 track->fourcc    = "WMAP";
360                 track->tag       = st->codec->codec_tag;
361                 track->blocksize = st->codec->block_align;
362             }
363             get_private_data(track, st->codec);
364         }
365         if (track->is_video) {
366             if (tracks->video_track < 0)
367                 tracks->video_track = tracks->nb_tracks;
368             tracks->nb_video_tracks++;
369             track->width  = st->codec->width;
370             track->height = st->codec->height;
371             if (st->codec->codec_id == AV_CODEC_ID_H264)
372                 track->fourcc = "H264";
373             else if (st->codec->codec_id == AV_CODEC_ID_VC1)
374                 track->fourcc = "WVC1";
375             get_video_private_data(track, st->codec);
376         }
377
378         tracks->nb_tracks++;
379     }
380
381     avformat_close_input(&ctx);
382
383     err = read_mfra(tracks, orig_tracks, file, split, output_prefix);
384
385 fail:
386     if (ctx)
387         avformat_close_input(&ctx);
388     return err;
389 }
390
391 static void output_server_manifest(struct Tracks *tracks, const char *basename,
392                                    const char *output_prefix,
393                                    const char *path_prefix,
394                                    const char *ismc_prefix)
395 {
396     char filename[1000];
397     FILE *out;
398     int i;
399
400     snprintf(filename, sizeof(filename), "%s%s.ism", output_prefix, basename);
401     out = fopen(filename, "w");
402     if (!out) {
403         perror(filename);
404         return;
405     }
406     fprintf(out, "<?xml version=\"1.0\" encoding=\"utf-8\"?>\n");
407     fprintf(out, "<smil xmlns=\"http://www.w3.org/2001/SMIL20/Language\">\n");
408     fprintf(out, "\t<head>\n");
409     fprintf(out, "\t\t<meta name=\"clientManifestRelativePath\" "
410                  "content=\"%s%s.ismc\" />\n", ismc_prefix, basename);
411     fprintf(out, "\t</head>\n");
412     fprintf(out, "\t<body>\n");
413     fprintf(out, "\t\t<switch>\n");
414     for (i = 0; i < tracks->nb_tracks; i++) {
415         struct Track *track = tracks->tracks[i];
416         const char *type    = track->is_video ? "video" : "audio";
417         fprintf(out, "\t\t\t<%s src=\"%s%s\" systemBitrate=\"%d\">\n",
418                 type, path_prefix, track->name, track->bitrate);
419         fprintf(out, "\t\t\t\t<param name=\"trackID\" value=\"%d\" "
420                      "valueType=\"data\" />\n", track->track_id);
421         fprintf(out, "\t\t\t</%s>\n", type);
422     }
423     fprintf(out, "\t\t</switch>\n");
424     fprintf(out, "\t</body>\n");
425     fprintf(out, "</smil>\n");
426     fclose(out);
427 }
428
429 static void print_track_chunks(FILE *out, struct Tracks *tracks, int main,
430                                const char *type)
431 {
432     int i, j;
433     struct Track *track = tracks->tracks[main];
434     for (i = 0; i < track->chunks; i++) {
435         for (j = main + 1; j < tracks->nb_tracks; j++) {
436             if (tracks->tracks[j]->is_audio == track->is_audio &&
437                 track->offsets[i].duration != tracks->tracks[j]->offsets[i].duration)
438                 fprintf(stderr, "Mismatched duration of %s chunk %d in %s and %s\n",
439                         type, i, track->name, tracks->tracks[j]->name);
440         }
441         fprintf(out, "\t\t<c n=\"%d\" d=\"%"PRId64"\" />\n",
442                 i, track->offsets[i].duration);
443     }
444 }
445
446 static void output_client_manifest(struct Tracks *tracks, const char *basename,
447                                    const char *output_prefix, int split)
448 {
449     char filename[1000];
450     FILE *out;
451     int i, j;
452
453     if (split)
454         snprintf(filename, sizeof(filename), "%sManifest", output_prefix);
455     else
456         snprintf(filename, sizeof(filename), "%s%s.ismc", output_prefix, basename);
457     out = fopen(filename, "w");
458     if (!out) {
459         perror(filename);
460         return;
461     }
462     fprintf(out, "<?xml version=\"1.0\" encoding=\"utf-8\"?>\n");
463     fprintf(out, "<SmoothStreamingMedia MajorVersion=\"2\" MinorVersion=\"0\" "
464                  "Duration=\"%"PRId64 "\">\n", tracks->duration * 10);
465     if (tracks->video_track >= 0) {
466         struct Track *track = tracks->tracks[tracks->video_track];
467         struct Track *first_track = track;
468         int index = 0;
469         fprintf(out,
470                 "\t<StreamIndex Type=\"video\" QualityLevels=\"%d\" "
471                 "Chunks=\"%d\" "
472                 "Url=\"QualityLevels({bitrate})/Fragments(video={start time})\">\n",
473                 tracks->nb_video_tracks, track->chunks);
474         for (i = 0; i < tracks->nb_tracks; i++) {
475             track = tracks->tracks[i];
476             if (!track->is_video)
477                 continue;
478             fprintf(out,
479                     "\t\t<QualityLevel Index=\"%d\" Bitrate=\"%d\" "
480                     "FourCC=\"%s\" MaxWidth=\"%d\" MaxHeight=\"%d\" "
481                     "CodecPrivateData=\"",
482                     index, track->bitrate, track->fourcc, track->width, track->height);
483             for (j = 0; j < track->codec_private_size; j++)
484                 fprintf(out, "%02X", track->codec_private[j]);
485             fprintf(out, "\" />\n");
486             index++;
487             if (track->chunks != first_track->chunks)
488                 fprintf(stderr, "Mismatched number of video chunks in %s and %s\n",
489                         track->name, first_track->name);
490         }
491         print_track_chunks(out, tracks, tracks->video_track, "video");
492         fprintf(out, "\t</StreamIndex>\n");
493     }
494     if (tracks->audio_track >= 0) {
495         struct Track *track = tracks->tracks[tracks->audio_track];
496         struct Track *first_track = track;
497         int index = 0;
498         fprintf(out,
499                 "\t<StreamIndex Type=\"audio\" QualityLevels=\"%d\" "
500                 "Chunks=\"%d\" "
501                 "Url=\"QualityLevels({bitrate})/Fragments(audio={start time})\">\n",
502                 tracks->nb_audio_tracks, track->chunks);
503         for (i = 0; i < tracks->nb_tracks; i++) {
504             track = tracks->tracks[i];
505             if (!track->is_audio)
506                 continue;
507             fprintf(out,
508                     "\t\t<QualityLevel Index=\"%d\" Bitrate=\"%d\" "
509                     "FourCC=\"%s\" SamplingRate=\"%d\" Channels=\"%d\" "
510                     "BitsPerSample=\"16\" PacketSize=\"%d\" "
511                     "AudioTag=\"%d\" CodecPrivateData=\"",
512                     index, track->bitrate, track->fourcc, track->sample_rate,
513                     track->channels, track->blocksize, track->tag);
514             for (j = 0; j < track->codec_private_size; j++)
515                 fprintf(out, "%02X", track->codec_private[j]);
516             fprintf(out, "\" />\n");
517             index++;
518             if (track->chunks != first_track->chunks)
519                 fprintf(stderr, "Mismatched number of audio chunks in %s and %s\n",
520                         track->name, first_track->name);
521         }
522         print_track_chunks(out, tracks, tracks->audio_track, "audio");
523         fprintf(out, "\t</StreamIndex>\n");
524     }
525     fprintf(out, "</SmoothStreamingMedia>\n");
526     fclose(out);
527 }
528
529 static void clean_tracks(struct Tracks *tracks)
530 {
531     int i;
532     for (i = 0; i < tracks->nb_tracks; i++) {
533         av_freep(&tracks->tracks[i]->codec_private);
534         av_freep(&tracks->tracks[i]->offsets);
535         av_freep(&tracks->tracks[i]);
536     }
537     av_freep(&tracks->tracks);
538     tracks->nb_tracks = 0;
539 }
540
541 int main(int argc, char **argv)
542 {
543     const char *basename = NULL;
544     const char *path_prefix = "", *ismc_prefix = "";
545     const char *output_prefix = "";
546     char output_prefix_buf[2048];
547     int split = 0, i;
548     struct Tracks tracks = { 0, .video_track = -1, .audio_track = -1 };
549
550     av_register_all();
551
552     for (i = 1; i < argc; i++) {
553         if (!strcmp(argv[i], "-n")) {
554             basename = argv[i + 1];
555             i++;
556         } else if (!strcmp(argv[i], "-path-prefix")) {
557             path_prefix = argv[i + 1];
558             i++;
559         } else if (!strcmp(argv[i], "-ismc-prefix")) {
560             ismc_prefix = argv[i + 1];
561             i++;
562         } else if (!strcmp(argv[i], "-output")) {
563             output_prefix = argv[i + 1];
564             i++;
565             if (output_prefix[strlen(output_prefix) - 1] != '/') {
566                 snprintf(output_prefix_buf, sizeof(output_prefix_buf),
567                          "%s/", output_prefix);
568                 output_prefix = output_prefix_buf;
569             }
570         } else if (!strcmp(argv[i], "-split")) {
571             split = 1;
572         } else if (argv[i][0] == '-') {
573             return usage(argv[0], 1);
574         } else {
575             if (handle_file(&tracks, argv[i], split, output_prefix))
576                 return 1;
577         }
578     }
579     if (!tracks.nb_tracks || (!basename && !split))
580         return usage(argv[0], 1);
581
582     if (!split)
583         output_server_manifest(&tracks, basename, output_prefix,
584                                path_prefix, ismc_prefix);
585     output_client_manifest(&tracks, basename, output_prefix, split);
586
587     clean_tracks(&tracks);
588
589     return 0;
590 }