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/avstring.h"
29 #include "libavutil/channel_layout.h"
30 #include "libavutil/common.h"
31 #include "libavutil/opt.h"
32
33 #include "audio.h"
34 #include "avfilter.h"
35 #include "formats.h"
36 #include "filters.h"
37 #include "internal.h"
38
39 typedef struct ChannelMap {
40 int input; ///< input stream index
41 int in_channel_idx; ///< index of in_channel in the input stream data
42 uint64_t in_channel; ///< layout describing the input channel
43 uint64_t out_channel; ///< layout describing the output channel
44 } ChannelMap;
45
46 typedef struct JoinContext {
47 const AVClass *class;
48
49 int inputs;
50 char *map;
51 char *channel_layout_str;
52 uint64_t channel_layout;
53
54 int nb_channels;
55 ChannelMap *channels;
56
57 /**
58 * Temporary storage for input frames, until we get one on each input.
59 */
60 AVFrame **input_frames;
61
62 /**
63 * Temporary storage for buffer references, for assembling the output frame.
64 */
65 AVBufferRef **buffers;
66 } JoinContext;
67
68 #define OFFSET(x) offsetof(JoinContext, x)
69 #define A AV_OPT_FLAG_AUDIO_PARAM
70 #define F AV_OPT_FLAG_FILTERING_PARAM
71 static const AVOption join_options[] = {
72 { "inputs", "Number of input streams.", OFFSET(inputs), AV_OPT_TYPE_INT, { .i64 = 2 }, 1, INT_MAX, A|F },
73 { "channel_layout", "Channel layout of the "
74 "output stream.", OFFSET(channel_layout_str), AV_OPT_TYPE_STRING, {.str = "stereo"}, 0, 0, A|F },
75 { "map", "A comma-separated list of channels maps in the format "
76 "'input_stream.input_channel-output_channel.",
77 OFFSET(map), AV_OPT_TYPE_STRING, .flags = A|F },
78 { NULL }
79 };
80
81 AVFILTER_DEFINE_CLASS(join);
82
parse_maps(AVFilterContext * ctx)83 static int parse_maps(AVFilterContext *ctx)
84 {
85 JoinContext *s = ctx->priv;
86 char separator = '|';
87 char *cur = s->map;
88
89 while (cur && *cur) {
90 char *sep, *next, *p;
91 uint64_t in_channel = 0, out_channel = 0;
92 int input_idx, out_ch_idx, in_ch_idx;
93
94 next = strchr(cur, separator);
95 if (next)
96 *next++ = 0;
97
98 /* split the map into input and output parts */
99 if (!(sep = strchr(cur, '-'))) {
100 av_log(ctx, AV_LOG_ERROR, "Missing separator '-' in channel "
101 "map '%s'\n", cur);
102 return AVERROR(EINVAL);
103 }
104 *sep++ = 0;
105
106 #define PARSE_CHANNEL(str, var, inout) \
107 if (!(var = av_get_channel_layout(str))) { \
108 av_log(ctx, AV_LOG_ERROR, "Invalid " inout " channel: %s.\n", str);\
109 return AVERROR(EINVAL); \
110 } \
111 if (av_get_channel_layout_nb_channels(var) != 1) { \
112 av_log(ctx, AV_LOG_ERROR, "Channel map describes more than one " \
113 inout " channel.\n"); \
114 return AVERROR(EINVAL); \
115 }
116
117 /* parse output channel */
118 PARSE_CHANNEL(sep, out_channel, "output");
119 if (!(out_channel & s->channel_layout)) {
120 av_log(ctx, AV_LOG_ERROR, "Output channel '%s' is not present in "
121 "requested channel layout.\n", sep);
122 return AVERROR(EINVAL);
123 }
124
125 out_ch_idx = av_get_channel_layout_channel_index(s->channel_layout,
126 out_channel);
127 if (s->channels[out_ch_idx].input >= 0) {
128 av_log(ctx, AV_LOG_ERROR, "Multiple maps for output channel "
129 "'%s'.\n", sep);
130 return AVERROR(EINVAL);
131 }
132
133 /* parse input channel */
134 input_idx = strtol(cur, &cur, 0);
135 if (input_idx < 0 || input_idx >= s->inputs) {
136 av_log(ctx, AV_LOG_ERROR, "Invalid input stream index: %d.\n",
137 input_idx);
138 return AVERROR(EINVAL);
139 }
140
141 if (*cur)
142 cur++;
143
144 in_ch_idx = strtol(cur, &p, 0);
145 if (p == cur) {
146 /* channel specifier is not a number,
147 * try to parse as channel name */
148 PARSE_CHANNEL(cur, in_channel, "input");
149 }
150
151 s->channels[out_ch_idx].input = input_idx;
152 if (in_channel)
153 s->channels[out_ch_idx].in_channel = in_channel;
154 else
155 s->channels[out_ch_idx].in_channel_idx = in_ch_idx;
156
157 cur = next;
158 }
159 return 0;
160 }
161
join_init(AVFilterContext * ctx)162 static av_cold int join_init(AVFilterContext *ctx)
163 {
164 JoinContext *s = ctx->priv;
165 int ret, i;
166
167 if (!(s->channel_layout = av_get_channel_layout(s->channel_layout_str))) {
168 av_log(ctx, AV_LOG_ERROR, "Error parsing channel layout '%s'.\n",
169 s->channel_layout_str);
170 return AVERROR(EINVAL);
171 }
172
173 s->nb_channels = av_get_channel_layout_nb_channels(s->channel_layout);
174 s->channels = av_mallocz_array(s->nb_channels, sizeof(*s->channels));
175 s->buffers = av_mallocz_array(s->nb_channels, sizeof(*s->buffers));
176 s->input_frames = av_mallocz_array(s->inputs, sizeof(*s->input_frames));
177 if (!s->channels || !s->buffers|| !s->input_frames)
178 return AVERROR(ENOMEM);
179
180 for (i = 0; i < s->nb_channels; i++) {
181 s->channels[i].out_channel = av_channel_layout_extract_channel(s->channel_layout, i);
182 s->channels[i].input = -1;
183 }
184
185 if ((ret = parse_maps(ctx)) < 0)
186 return ret;
187
188 for (i = 0; i < s->inputs; i++) {
189 AVFilterPad pad = { 0 };
190
191 pad.type = AVMEDIA_TYPE_AUDIO;
192 pad.name = av_asprintf("input%d", i);
193 if (!pad.name)
194 return AVERROR(ENOMEM);
195
196 if ((ret = ff_insert_inpad(ctx, i, &pad)) < 0) {
197 av_freep(&pad.name);
198 return ret;
199 }
200 }
201
202 return 0;
203 }
204
join_uninit(AVFilterContext * ctx)205 static av_cold void join_uninit(AVFilterContext *ctx)
206 {
207 JoinContext *s = ctx->priv;
208 int i;
209
210 for (i = 0; i < s->inputs && s->input_frames; i++) {
211 av_frame_free(&s->input_frames[i]);
212 }
213
214 for (i = 0; i < ctx->nb_inputs; i++) {
215 av_freep(&ctx->input_pads[i].name);
216 }
217
218 av_freep(&s->channels);
219 av_freep(&s->buffers);
220 av_freep(&s->input_frames);
221 }
222
join_query_formats(AVFilterContext * ctx)223 static int join_query_formats(AVFilterContext *ctx)
224 {
225 JoinContext *s = ctx->priv;
226 AVFilterChannelLayouts *layouts = NULL;
227 int i, ret;
228
229 if ((ret = ff_add_channel_layout(&layouts, s->channel_layout)) < 0 ||
230 (ret = ff_channel_layouts_ref(layouts, &ctx->outputs[0]->in_channel_layouts)) < 0)
231 return ret;
232
233 for (i = 0; i < ctx->nb_inputs; i++) {
234 layouts = ff_all_channel_layouts();
235 if ((ret = ff_channel_layouts_ref(layouts, &ctx->inputs[i]->out_channel_layouts)) < 0)
236 return ret;
237 }
238
239 if ((ret = ff_set_common_formats(ctx, ff_planar_sample_fmts())) < 0 ||
240 (ret = ff_set_common_samplerates(ctx, ff_all_samplerates())) < 0)
241 return ret;
242
243 return 0;
244 }
245
guess_map_matching(AVFilterContext * ctx,ChannelMap * ch,uint64_t * inputs)246 static void guess_map_matching(AVFilterContext *ctx, ChannelMap *ch,
247 uint64_t *inputs)
248 {
249 int i;
250
251 for (i = 0; i < ctx->nb_inputs; i++) {
252 AVFilterLink *link = ctx->inputs[i];
253
254 if (ch->out_channel & link->channel_layout &&
255 !(ch->out_channel & inputs[i])) {
256 ch->input = i;
257 ch->in_channel = ch->out_channel;
258 inputs[i] |= ch->out_channel;
259 return;
260 }
261 }
262 }
263
guess_map_any(AVFilterContext * ctx,ChannelMap * ch,uint64_t * inputs)264 static void guess_map_any(AVFilterContext *ctx, ChannelMap *ch,
265 uint64_t *inputs)
266 {
267 int i;
268
269 for (i = 0; i < ctx->nb_inputs; i++) {
270 AVFilterLink *link = ctx->inputs[i];
271
272 if ((inputs[i] & link->channel_layout) != link->channel_layout) {
273 uint64_t unused = link->channel_layout & ~inputs[i];
274
275 ch->input = i;
276 ch->in_channel = av_channel_layout_extract_channel(unused, 0);
277 inputs[i] |= ch->in_channel;
278 return;
279 }
280 }
281 }
282
join_config_output(AVFilterLink * outlink)283 static int join_config_output(AVFilterLink *outlink)
284 {
285 AVFilterContext *ctx = outlink->src;
286 JoinContext *s = ctx->priv;
287 uint64_t *inputs; // nth element tracks which channels are used from nth input
288 int i, ret = 0;
289
290 /* initialize inputs to user-specified mappings */
291 if (!(inputs = av_mallocz_array(ctx->nb_inputs, sizeof(*inputs))))
292 return AVERROR(ENOMEM);
293 for (i = 0; i < s->nb_channels; i++) {
294 ChannelMap *ch = &s->channels[i];
295 AVFilterLink *inlink;
296
297 if (ch->input < 0)
298 continue;
299
300 inlink = ctx->inputs[ch->input];
301
302 if (!ch->in_channel)
303 ch->in_channel = av_channel_layout_extract_channel(inlink->channel_layout,
304 ch->in_channel_idx);
305
306 if (!(ch->in_channel & inlink->channel_layout)) {
307 av_log(ctx, AV_LOG_ERROR, "Requested channel %s is not present in "
308 "input stream #%d.\n", av_get_channel_name(ch->in_channel),
309 ch->input);
310 ret = AVERROR(EINVAL);
311 goto fail;
312 }
313
314 inputs[ch->input] |= ch->in_channel;
315 }
316
317 /* guess channel maps when not explicitly defined */
318 /* first try unused matching channels */
319 for (i = 0; i < s->nb_channels; i++) {
320 ChannelMap *ch = &s->channels[i];
321
322 if (ch->input < 0)
323 guess_map_matching(ctx, ch, inputs);
324 }
325
326 /* if the above failed, try to find _any_ unused input channel */
327 for (i = 0; i < s->nb_channels; i++) {
328 ChannelMap *ch = &s->channels[i];
329
330 if (ch->input < 0)
331 guess_map_any(ctx, ch, inputs);
332
333 if (ch->input < 0) {
334 av_log(ctx, AV_LOG_ERROR, "Could not find input channel for "
335 "output channel '%s'.\n",
336 av_get_channel_name(ch->out_channel));
337 goto fail;
338 }
339
340 ch->in_channel_idx = av_get_channel_layout_channel_index(ctx->inputs[ch->input]->channel_layout,
341 ch->in_channel);
342 }
343
344 /* print mappings */
345 av_log(ctx, AV_LOG_VERBOSE, "mappings: ");
346 for (i = 0; i < s->nb_channels; i++) {
347 ChannelMap *ch = &s->channels[i];
348 av_log(ctx, AV_LOG_VERBOSE, "%d.%s => %s ", ch->input,
349 av_get_channel_name(ch->in_channel),
350 av_get_channel_name(ch->out_channel));
351 }
352 av_log(ctx, AV_LOG_VERBOSE, "\n");
353
354 for (i = 0; i < ctx->nb_inputs; i++) {
355 if (!inputs[i])
356 av_log(ctx, AV_LOG_WARNING, "No channels are used from input "
357 "stream %d.\n", i);
358 }
359
360 fail:
361 av_freep(&inputs);
362 return ret;
363 }
364
try_push_frame(AVFilterContext * ctx)365 static int try_push_frame(AVFilterContext *ctx)
366 {
367 AVFilterLink *outlink = ctx->outputs[0];
368 JoinContext *s = ctx->priv;
369 AVFrame *frame;
370 int linesize = INT_MAX;
371 int nb_samples = INT_MAX;
372 int nb_buffers = 0;
373 int i, j, ret;
374
375 for (i = 0; i < ctx->nb_inputs; i++) {
376 if (!s->input_frames[i])
377 return 0;
378 nb_samples = FFMIN(nb_samples, s->input_frames[i]->nb_samples);
379 }
380 if (!nb_samples)
381 return 0;
382
383 /* setup the output frame */
384 frame = av_frame_alloc();
385 if (!frame)
386 return AVERROR(ENOMEM);
387 if (s->nb_channels > FF_ARRAY_ELEMS(frame->data)) {
388 frame->extended_data = av_mallocz_array(s->nb_channels,
389 sizeof(*frame->extended_data));
390 if (!frame->extended_data) {
391 ret = AVERROR(ENOMEM);
392 goto fail;
393 }
394 }
395
396 /* copy the data pointers */
397 for (i = 0; i < s->nb_channels; i++) {
398 ChannelMap *ch = &s->channels[i];
399 AVFrame *cur = s->input_frames[ch->input];
400 AVBufferRef *buf;
401
402 frame->extended_data[i] = cur->extended_data[ch->in_channel_idx];
403 linesize = FFMIN(linesize, cur->linesize[0]);
404
405 /* add the buffer where this plan is stored to the list if it's
406 * not already there */
407 buf = av_frame_get_plane_buffer(cur, ch->in_channel_idx);
408 if (!buf) {
409 ret = AVERROR(EINVAL);
410 goto fail;
411 }
412 for (j = 0; j < nb_buffers; j++)
413 if (s->buffers[j]->buffer == buf->buffer)
414 break;
415 if (j == i)
416 s->buffers[nb_buffers++] = buf;
417 }
418
419 /* create references to the buffers we copied to output */
420 if (nb_buffers > FF_ARRAY_ELEMS(frame->buf)) {
421 frame->nb_extended_buf = nb_buffers - FF_ARRAY_ELEMS(frame->buf);
422 frame->extended_buf = av_mallocz_array(frame->nb_extended_buf,
423 sizeof(*frame->extended_buf));
424 if (!frame->extended_buf) {
425 frame->nb_extended_buf = 0;
426 ret = AVERROR(ENOMEM);
427 goto fail;
428 }
429 }
430 for (i = 0; i < FFMIN(FF_ARRAY_ELEMS(frame->buf), nb_buffers); i++) {
431 frame->buf[i] = av_buffer_ref(s->buffers[i]);
432 if (!frame->buf[i]) {
433 ret = AVERROR(ENOMEM);
434 goto fail;
435 }
436 }
437 for (i = 0; i < frame->nb_extended_buf; i++) {
438 frame->extended_buf[i] = av_buffer_ref(s->buffers[i +
439 FF_ARRAY_ELEMS(frame->buf)]);
440 if (!frame->extended_buf[i]) {
441 ret = AVERROR(ENOMEM);
442 goto fail;
443 }
444 }
445
446 frame->nb_samples = nb_samples;
447 frame->channel_layout = outlink->channel_layout;
448 frame->channels = outlink->channels;
449 frame->sample_rate = outlink->sample_rate;
450 frame->format = outlink->format;
451 frame->pts = s->input_frames[0]->pts;
452 frame->linesize[0] = linesize;
453 if (frame->data != frame->extended_data) {
454 memcpy(frame->data, frame->extended_data, sizeof(*frame->data) *
455 FFMIN(FF_ARRAY_ELEMS(frame->data), s->nb_channels));
456 }
457
458 ret = ff_filter_frame(outlink, frame);
459
460 for (i = 0; i < ctx->nb_inputs; i++)
461 av_frame_free(&s->input_frames[i]);
462
463 return ret;
464
465 fail:
466 av_frame_free(&frame);
467 return ret;
468 }
469
activate(AVFilterContext * ctx)470 static int activate(AVFilterContext *ctx)
471 {
472 JoinContext *s = ctx->priv;
473 int i, ret, status;
474 int nb_samples = 0;
475 int64_t pts;
476
477 FF_FILTER_FORWARD_STATUS_BACK_ALL(ctx->outputs[0], ctx);
478
479 if (!s->input_frames[0]) {
480 ret = ff_inlink_consume_frame(ctx->inputs[0], &s->input_frames[0]);
481 if (ret < 0) {
482 return ret;
483 } else if (ff_inlink_acknowledge_status(ctx->inputs[0], &status, &pts)) {
484 ff_outlink_set_status(ctx->outputs[0], status, pts);
485 return 0;
486 } else {
487 if (ff_outlink_frame_wanted(ctx->outputs[0]) && !s->input_frames[0]) {
488 ff_inlink_request_frame(ctx->inputs[0]);
489 return 0;
490 }
491 }
492 if (!s->input_frames[0]) {
493 return 0;
494 }
495 }
496
497 nb_samples = s->input_frames[0]->nb_samples;
498
499 for (i = 1; i < ctx->nb_inputs && nb_samples > 0; i++) {
500 if (s->input_frames[i])
501 continue;
502
503 if (ff_inlink_check_available_samples(ctx->inputs[i], nb_samples) > 0) {
504 ret = ff_inlink_consume_samples(ctx->inputs[i], nb_samples, nb_samples, &s->input_frames[i]);
505 if (ret < 0) {
506 return ret;
507 } else if (ff_inlink_acknowledge_status(ctx->inputs[i], &status, &pts)) {
508 ff_outlink_set_status(ctx->outputs[0], status, pts);
509 return 0;
510 }
511 } else {
512 if (ff_outlink_frame_wanted(ctx->outputs[0])) {
513 ff_inlink_request_frame(ctx->inputs[i]);
514 return 0;
515 }
516 }
517 }
518
519 return try_push_frame(ctx);
520 }
521
522 static const AVFilterPad avfilter_af_join_outputs[] = {
523 {
524 .name = "default",
525 .type = AVMEDIA_TYPE_AUDIO,
526 .config_props = join_config_output,
527 },
528 { NULL }
529 };
530
531 AVFilter ff_af_join = {
532 .name = "join",
533 .description = NULL_IF_CONFIG_SMALL("Join multiple audio streams into "
534 "multi-channel output."),
535 .priv_size = sizeof(JoinContext),
536 .priv_class = &join_class,
537 .init = join_init,
538 .uninit = join_uninit,
539 .activate = activate,
540 .query_formats = join_query_formats,
541 .inputs = NULL,
542 .outputs = avfilter_af_join_outputs,
543 .flags = AVFILTER_FLAG_DYNAMIC_INPUTS,
544 };
545