Imported Debian version 2.5.3~trusty1
[deb_ffmpeg.git] / ffmpeg / libavfilter / af_join.c
1 /*
2 * This file is part of FFmpeg.
3 *
4 * FFmpeg is free software; you can redistribute it and/or
5 * modify it under the terms of the GNU Lesser General Public
6 * License as published by the Free Software Foundation; either
7 * version 2.1 of the License, or (at your option) any later version.
8 *
9 * FFmpeg is distributed in the hope that it will be useful,
10 * but WITHOUT ANY WARRANTY; without even the implied warranty of
11 * MERCHANTABILITY or FITNESS FOR A PARTICULAR PURPOSE. See the GNU
12 * Lesser General Public License for more details.
13 *
14 * You should have received a copy of the GNU Lesser General Public
15 * License along with FFmpeg; if not, write to the Free Software
16 * Foundation, Inc., 51 Franklin Street, Fifth Floor, Boston, MA 02110-1301 USA
17 */
18
19 /**
20 * @file
21 * Audio join filter
22 *
23 * Join multiple audio inputs as different channels in
24 * a single output
25 */
26
27 #include "libavutil/avassert.h"
28 #include "libavutil/channel_layout.h"
29 #include "libavutil/common.h"
30 #include "libavutil/opt.h"
31
32 #include "audio.h"
33 #include "avfilter.h"
34 #include "formats.h"
35 #include "internal.h"
36
37 typedef struct ChannelMap {
38 int input; ///< input stream index
39 int in_channel_idx; ///< index of in_channel in the input stream data
40 uint64_t in_channel; ///< layout describing the input channel
41 uint64_t out_channel; ///< layout describing the output channel
42 } ChannelMap;
43
44 typedef struct JoinContext {
45 const AVClass *class;
46
47 int inputs;
48 char *map;
49 char *channel_layout_str;
50 uint64_t channel_layout;
51
52 int nb_channels;
53 ChannelMap *channels;
54
55 /**
56 * Temporary storage for input frames, until we get one on each input.
57 */
58 AVFrame **input_frames;
59
60 /**
61 * Temporary storage for buffer references, for assembling the output frame.
62 */
63 AVBufferRef **buffers;
64 } JoinContext;
65
66 #define OFFSET(x) offsetof(JoinContext, x)
67 #define A AV_OPT_FLAG_AUDIO_PARAM
68 #define F AV_OPT_FLAG_FILTERING_PARAM
69 static const AVOption join_options[] = {
70 { "inputs", "Number of input streams.", OFFSET(inputs), AV_OPT_TYPE_INT, { .i64 = 2 }, 1, INT_MAX, A|F },
71 { "channel_layout", "Channel layout of the "
72 "output stream.", OFFSET(channel_layout_str), AV_OPT_TYPE_STRING, {.str = "stereo"}, 0, 0, A|F },
73 { "map", "A comma-separated list of channels maps in the format "
74 "'input_stream.input_channel-output_channel.",
75 OFFSET(map), AV_OPT_TYPE_STRING, .flags = A|F },
76 { NULL }
77 };
78
79 AVFILTER_DEFINE_CLASS(join);
80
81 static int filter_frame(AVFilterLink *link, AVFrame *frame)
82 {
83 AVFilterContext *ctx = link->dst;
84 JoinContext *s = ctx->priv;
85 int i;
86
87 for (i = 0; i < ctx->nb_inputs; i++)
88 if (link == ctx->inputs[i])
89 break;
90 av_assert0(i < ctx->nb_inputs);
91 av_assert0(!s->input_frames[i]);
92 s->input_frames[i] = frame;
93
94 return 0;
95 }
96
97 static int parse_maps(AVFilterContext *ctx)
98 {
99 JoinContext *s = ctx->priv;
100 char separator = '|';
101 char *cur = s->map;
102
103 #if FF_API_OLD_FILTER_OPTS
104 if (cur && strchr(cur, ',')) {
105 av_log(ctx, AV_LOG_WARNING, "This syntax is deprecated, use '|' to "
106 "separate the mappings.\n");
107 separator = ',';
108 }
109 #endif
110
111 while (cur && *cur) {
112 char *sep, *next, *p;
113 uint64_t in_channel = 0, out_channel = 0;
114 int input_idx, out_ch_idx, in_ch_idx;
115
116 next = strchr(cur, separator);
117 if (next)
118 *next++ = 0;
119
120 /* split the map into input and output parts */
121 if (!(sep = strchr(cur, '-'))) {
122 av_log(ctx, AV_LOG_ERROR, "Missing separator '-' in channel "
123 "map '%s'\n", cur);
124 return AVERROR(EINVAL);
125 }
126 *sep++ = 0;
127
128 #define PARSE_CHANNEL(str, var, inout) \
129 if (!(var = av_get_channel_layout(str))) { \
130 av_log(ctx, AV_LOG_ERROR, "Invalid " inout " channel: %s.\n", str);\
131 return AVERROR(EINVAL); \
132 } \
133 if (av_get_channel_layout_nb_channels(var) != 1) { \
134 av_log(ctx, AV_LOG_ERROR, "Channel map describes more than one " \
135 inout " channel.\n"); \
136 return AVERROR(EINVAL); \
137 }
138
139 /* parse output channel */
140 PARSE_CHANNEL(sep, out_channel, "output");
141 if (!(out_channel & s->channel_layout)) {
142 av_log(ctx, AV_LOG_ERROR, "Output channel '%s' is not present in "
143 "requested channel layout.\n", sep);
144 return AVERROR(EINVAL);
145 }
146
147 out_ch_idx = av_get_channel_layout_channel_index(s->channel_layout,
148 out_channel);
149 if (s->channels[out_ch_idx].input >= 0) {
150 av_log(ctx, AV_LOG_ERROR, "Multiple maps for output channel "
151 "'%s'.\n", sep);
152 return AVERROR(EINVAL);
153 }
154
155 /* parse input channel */
156 input_idx = strtol(cur, &cur, 0);
157 if (input_idx < 0 || input_idx >= s->inputs) {
158 av_log(ctx, AV_LOG_ERROR, "Invalid input stream index: %d.\n",
159 input_idx);
160 return AVERROR(EINVAL);
161 }
162
163 if (*cur)
164 cur++;
165
166 in_ch_idx = strtol(cur, &p, 0);
167 if (p == cur) {
168 /* channel specifier is not a number,
169 * try to parse as channel name */
170 PARSE_CHANNEL(cur, in_channel, "input");
171 }
172
173 s->channels[out_ch_idx].input = input_idx;
174 if (in_channel)
175 s->channels[out_ch_idx].in_channel = in_channel;
176 else
177 s->channels[out_ch_idx].in_channel_idx = in_ch_idx;
178
179 cur = next;
180 }
181 return 0;
182 }
183
184 static av_cold int join_init(AVFilterContext *ctx)
185 {
186 JoinContext *s = ctx->priv;
187 int ret, i;
188
189 if (!(s->channel_layout = av_get_channel_layout(s->channel_layout_str))) {
190 av_log(ctx, AV_LOG_ERROR, "Error parsing channel layout '%s'.\n",
191 s->channel_layout_str);
192 return AVERROR(EINVAL);
193 }
194
195 s->nb_channels = av_get_channel_layout_nb_channels(s->channel_layout);
196 s->channels = av_mallocz(sizeof(*s->channels) * s->nb_channels);
197 s->buffers = av_mallocz(sizeof(*s->buffers) * s->nb_channels);
198 s->input_frames = av_mallocz(sizeof(*s->input_frames) * s->inputs);
199 if (!s->channels || !s->buffers|| !s->input_frames)
200 return AVERROR(ENOMEM);
201
202 for (i = 0; i < s->nb_channels; i++) {
203 s->channels[i].out_channel = av_channel_layout_extract_channel(s->channel_layout, i);
204 s->channels[i].input = -1;
205 }
206
207 if ((ret = parse_maps(ctx)) < 0)
208 return ret;
209
210 for (i = 0; i < s->inputs; i++) {
211 char name[32];
212 AVFilterPad pad = { 0 };
213
214 snprintf(name, sizeof(name), "input%d", i);
215 pad.type = AVMEDIA_TYPE_AUDIO;
216 pad.name = av_strdup(name);
217 if (!pad.name)
218 return AVERROR(ENOMEM);
219 pad.filter_frame = filter_frame;
220
221 pad.needs_fifo = 1;
222
223 ff_insert_inpad(ctx, i, &pad);
224 }
225
226 return 0;
227 }
228
229 static av_cold void join_uninit(AVFilterContext *ctx)
230 {
231 JoinContext *s = ctx->priv;
232 int i;
233
234 for (i = 0; i < ctx->nb_inputs; i++) {
235 av_freep(&ctx->input_pads[i].name);
236 av_frame_free(&s->input_frames[i]);
237 }
238
239 av_freep(&s->channels);
240 av_freep(&s->buffers);
241 av_freep(&s->input_frames);
242 }
243
244 static int join_query_formats(AVFilterContext *ctx)
245 {
246 JoinContext *s = ctx->priv;
247 AVFilterChannelLayouts *layouts = NULL;
248 int i;
249
250 ff_add_channel_layout(&layouts, s->channel_layout);
251 ff_channel_layouts_ref(layouts, &ctx->outputs[0]->in_channel_layouts);
252
253 for (i = 0; i < ctx->nb_inputs; i++) {
254 layouts = ff_all_channel_layouts();
255 if (!layouts)
256 return AVERROR(ENOMEM);
257 ff_channel_layouts_ref(layouts, &ctx->inputs[i]->out_channel_layouts);
258 }
259
260 ff_set_common_formats (ctx, ff_planar_sample_fmts());
261 ff_set_common_samplerates(ctx, ff_all_samplerates());
262
263 return 0;
264 }
265
266 static void guess_map_matching(AVFilterContext *ctx, ChannelMap *ch,
267 uint64_t *inputs)
268 {
269 int i;
270
271 for (i = 0; i < ctx->nb_inputs; i++) {
272 AVFilterLink *link = ctx->inputs[i];
273
274 if (ch->out_channel & link->channel_layout &&
275 !(ch->out_channel & inputs[i])) {
276 ch->input = i;
277 ch->in_channel = ch->out_channel;
278 inputs[i] |= ch->out_channel;
279 return;
280 }
281 }
282 }
283
284 static void guess_map_any(AVFilterContext *ctx, ChannelMap *ch,
285 uint64_t *inputs)
286 {
287 int i;
288
289 for (i = 0; i < ctx->nb_inputs; i++) {
290 AVFilterLink *link = ctx->inputs[i];
291
292 if ((inputs[i] & link->channel_layout) != link->channel_layout) {
293 uint64_t unused = link->channel_layout & ~inputs[i];
294
295 ch->input = i;
296 ch->in_channel = av_channel_layout_extract_channel(unused, 0);
297 inputs[i] |= ch->in_channel;
298 return;
299 }
300 }
301 }
302
303 static int join_config_output(AVFilterLink *outlink)
304 {
305 AVFilterContext *ctx = outlink->src;
306 JoinContext *s = ctx->priv;
307 uint64_t *inputs; // nth element tracks which channels are used from nth input
308 int i, ret = 0;
309
310 /* initialize inputs to user-specified mappings */
311 if (!(inputs = av_mallocz(sizeof(*inputs) * ctx->nb_inputs)))
312 return AVERROR(ENOMEM);
313 for (i = 0; i < s->nb_channels; i++) {
314 ChannelMap *ch = &s->channels[i];
315 AVFilterLink *inlink;
316
317 if (ch->input < 0)
318 continue;
319
320 inlink = ctx->inputs[ch->input];
321
322 if (!ch->in_channel)
323 ch->in_channel = av_channel_layout_extract_channel(inlink->channel_layout,
324 ch->in_channel_idx);
325
326 if (!(ch->in_channel & inlink->channel_layout)) {
327 av_log(ctx, AV_LOG_ERROR, "Requested channel %s is not present in "
328 "input stream #%d.\n", av_get_channel_name(ch->in_channel),
329 ch->input);
330 ret = AVERROR(EINVAL);
331 goto fail;
332 }
333
334 inputs[ch->input] |= ch->in_channel;
335 }
336
337 /* guess channel maps when not explicitly defined */
338 /* first try unused matching channels */
339 for (i = 0; i < s->nb_channels; i++) {
340 ChannelMap *ch = &s->channels[i];
341
342 if (ch->input < 0)
343 guess_map_matching(ctx, ch, inputs);
344 }
345
346 /* if the above failed, try to find _any_ unused input channel */
347 for (i = 0; i < s->nb_channels; i++) {
348 ChannelMap *ch = &s->channels[i];
349
350 if (ch->input < 0)
351 guess_map_any(ctx, ch, inputs);
352
353 if (ch->input < 0) {
354 av_log(ctx, AV_LOG_ERROR, "Could not find input channel for "
355 "output channel '%s'.\n",
356 av_get_channel_name(ch->out_channel));
357 goto fail;
358 }
359
360 ch->in_channel_idx = av_get_channel_layout_channel_index(ctx->inputs[ch->input]->channel_layout,
361 ch->in_channel);
362 }
363
364 /* print mappings */
365 av_log(ctx, AV_LOG_VERBOSE, "mappings: ");
366 for (i = 0; i < s->nb_channels; i++) {
367 ChannelMap *ch = &s->channels[i];
368 av_log(ctx, AV_LOG_VERBOSE, "%d.%s => %s ", ch->input,
369 av_get_channel_name(ch->in_channel),
370 av_get_channel_name(ch->out_channel));
371 }
372 av_log(ctx, AV_LOG_VERBOSE, "\n");
373
374 for (i = 0; i < ctx->nb_inputs; i++) {
375 if (!inputs[i])
376 av_log(ctx, AV_LOG_WARNING, "No channels are used from input "
377 "stream %d.\n", i);
378 }
379
380 fail:
381 av_freep(&inputs);
382 return ret;
383 }
384
385 static int join_request_frame(AVFilterLink *outlink)
386 {
387 AVFilterContext *ctx = outlink->src;
388 JoinContext *s = ctx->priv;
389 AVFrame *frame;
390 int linesize = INT_MAX;
391 int nb_samples = 0;
392 int nb_buffers = 0;
393 int i, j, ret;
394
395 /* get a frame on each input */
396 for (i = 0; i < ctx->nb_inputs; i++) {
397 AVFilterLink *inlink = ctx->inputs[i];
398
399 if (!s->input_frames[i] &&
400 (ret = ff_request_frame(inlink)) < 0)
401 return ret;
402
403 /* request the same number of samples on all inputs */
404 if (i == 0) {
405 nb_samples = s->input_frames[0]->nb_samples;
406
407 for (j = 1; !i && j < ctx->nb_inputs; j++)
408 ctx->inputs[j]->request_samples = nb_samples;
409 }
410 }
411
412 /* setup the output frame */
413 frame = av_frame_alloc();
414 if (!frame)
415 return AVERROR(ENOMEM);
416 if (s->nb_channels > FF_ARRAY_ELEMS(frame->data)) {
417 frame->extended_data = av_mallocz(s->nb_channels *
418 sizeof(*frame->extended_data));
419 if (!frame->extended_data) {
420 ret = AVERROR(ENOMEM);
421 goto fail;
422 }
423 }
424
425 /* copy the data pointers */
426 for (i = 0; i < s->nb_channels; i++) {
427 ChannelMap *ch = &s->channels[i];
428 AVFrame *cur = s->input_frames[ch->input];
429 AVBufferRef *buf;
430
431 frame->extended_data[i] = cur->extended_data[ch->in_channel_idx];
432 linesize = FFMIN(linesize, cur->linesize[0]);
433
434 /* add the buffer where this plan is stored to the list if it's
435 * not already there */
436 buf = av_frame_get_plane_buffer(cur, ch->in_channel_idx);
437 if (!buf) {
438 ret = AVERROR(EINVAL);
439 goto fail;
440 }
441 for (j = 0; j < nb_buffers; j++)
442 if (s->buffers[j]->buffer == buf->buffer)
443 break;
444 if (j == i)
445 s->buffers[nb_buffers++] = buf;
446 }
447
448 /* create references to the buffers we copied to output */
449 if (nb_buffers > FF_ARRAY_ELEMS(frame->buf)) {
450 frame->nb_extended_buf = nb_buffers - FF_ARRAY_ELEMS(frame->buf);
451 frame->extended_buf = av_mallocz(sizeof(*frame->extended_buf) *
452 frame->nb_extended_buf);
453 if (!frame->extended_buf) {
454 frame->nb_extended_buf = 0;
455 ret = AVERROR(ENOMEM);
456 goto fail;
457 }
458 }
459 for (i = 0; i < FFMIN(FF_ARRAY_ELEMS(frame->buf), nb_buffers); i++) {
460 frame->buf[i] = av_buffer_ref(s->buffers[i]);
461 if (!frame->buf[i]) {
462 ret = AVERROR(ENOMEM);
463 goto fail;
464 }
465 }
466 for (i = 0; i < frame->nb_extended_buf; i++) {
467 frame->extended_buf[i] = av_buffer_ref(s->buffers[i +
468 FF_ARRAY_ELEMS(frame->buf)]);
469 if (!frame->extended_buf[i]) {
470 ret = AVERROR(ENOMEM);
471 goto fail;
472 }
473 }
474
475 frame->nb_samples = nb_samples;
476 frame->channel_layout = outlink->channel_layout;
477 av_frame_set_channels(frame, outlink->channels);
478 frame->sample_rate = outlink->sample_rate;
479 frame->format = outlink->format;
480 frame->pts = s->input_frames[0]->pts;
481 frame->linesize[0] = linesize;
482 if (frame->data != frame->extended_data) {
483 memcpy(frame->data, frame->extended_data, sizeof(*frame->data) *
484 FFMIN(FF_ARRAY_ELEMS(frame->data), s->nb_channels));
485 }
486
487 ret = ff_filter_frame(outlink, frame);
488
489 for (i = 0; i < ctx->nb_inputs; i++)
490 av_frame_free(&s->input_frames[i]);
491
492 return ret;
493
494 fail:
495 av_frame_free(&frame);
496 return ret;
497 }
498
499 static const AVFilterPad avfilter_af_join_outputs[] = {
500 {
501 .name = "default",
502 .type = AVMEDIA_TYPE_AUDIO,
503 .config_props = join_config_output,
504 .request_frame = join_request_frame,
505 },
506 { NULL }
507 };
508
509 AVFilter ff_af_join = {
510 .name = "join",
511 .description = NULL_IF_CONFIG_SMALL("Join multiple audio streams into "
512 "multi-channel output."),
513 .priv_size = sizeof(JoinContext),
514 .priv_class = &join_class,
515 .init = join_init,
516 .uninit = join_uninit,
517 .query_formats = join_query_formats,
518 .inputs = NULL,
519 .outputs = avfilter_af_join_outputs,
520 .flags = AVFILTER_FLAG_DYNAMIC_INPUTS,
521 };