2 * This file is part of Libav.
4 * Libav 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.
9 * Libav 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.
14 * You should have received a copy of the GNU Lesser General Public
15 * License along with Libav; if not, write to the Free Software
16 * Foundation, Inc., 51 Franklin Street, Fifth Floor, Boston, MA 02110-1301 USA
23 * Join multiple audio inputs as different channels in
27 #include "libavutil/avassert.h"
28 #include "libavutil/channel_layout.h"
29 #include "libavutil/common.h"
30 #include "libavutil/opt.h"
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
44 typedef struct JoinContext {
49 char *channel_layout_str;
50 uint64_t channel_layout;
56 * Temporary storage for input frames, until we get one on each input.
58 AVFrame **input_frames;
61 * Temporary storage for buffer references, for assembling the output frame.
63 AVBufferRef **buffers;
66 #define OFFSET(x) offsetof(JoinContext, x)
67 #define A AV_OPT_FLAG_AUDIO_PARAM
68 static const AVOption join_options[] = {
69 { "inputs", "Number of input streams.", OFFSET(inputs), AV_OPT_TYPE_INT, { .i64 = 2 }, 1, INT_MAX, A },
70 { "channel_layout", "Channel layout of the "
71 "output stream.", OFFSET(channel_layout_str), AV_OPT_TYPE_STRING, {.str = "stereo"}, 0, 0, A },
72 { "map", "A comma-separated list of channels maps in the format "
73 "'input_stream.input_channel-output_channel.",
74 OFFSET(map), AV_OPT_TYPE_STRING, .flags = A },
78 static const AVClass join_class = {
79 .class_name = "join filter",
80 .item_name = av_default_item_name,
81 .option = join_options,
82 .version = LIBAVUTIL_VERSION_INT,
85 static int filter_frame(AVFilterLink *link, AVFrame *frame)
87 AVFilterContext *ctx = link->dst;
88 JoinContext *s = ctx->priv;
91 for (i = 0; i < ctx->nb_inputs; i++)
92 if (link == ctx->inputs[i])
94 av_assert0(i < ctx->nb_inputs);
95 av_assert0(!s->input_frames[i]);
96 s->input_frames[i] = frame;
101 static int parse_maps(AVFilterContext *ctx)
103 JoinContext *s = ctx->priv;
104 char separator = '|';
107 while (cur && *cur) {
108 char *sep, *next, *p;
109 uint64_t in_channel = 0, out_channel = 0;
110 int input_idx, out_ch_idx, in_ch_idx;
112 next = strchr(cur, separator);
116 /* split the map into input and output parts */
117 if (!(sep = strchr(cur, '-'))) {
118 av_log(ctx, AV_LOG_ERROR, "Missing separator '-' in channel "
120 return AVERROR(EINVAL);
124 #define PARSE_CHANNEL(str, var, inout) \
125 if (!(var = av_get_channel_layout(str))) { \
126 av_log(ctx, AV_LOG_ERROR, "Invalid " inout " channel: %s.\n", str);\
127 return AVERROR(EINVAL); \
129 if (av_get_channel_layout_nb_channels(var) != 1) { \
130 av_log(ctx, AV_LOG_ERROR, "Channel map describes more than one " \
131 inout " channel.\n"); \
132 return AVERROR(EINVAL); \
135 /* parse output channel */
136 PARSE_CHANNEL(sep, out_channel, "output");
137 if (!(out_channel & s->channel_layout)) {
138 av_log(ctx, AV_LOG_ERROR, "Output channel '%s' is not present in "
139 "requested channel layout.\n", sep);
140 return AVERROR(EINVAL);
143 out_ch_idx = av_get_channel_layout_channel_index(s->channel_layout,
145 if (s->channels[out_ch_idx].input >= 0) {
146 av_log(ctx, AV_LOG_ERROR, "Multiple maps for output channel "
148 return AVERROR(EINVAL);
151 /* parse input channel */
152 input_idx = strtol(cur, &cur, 0);
153 if (input_idx < 0 || input_idx >= s->inputs) {
154 av_log(ctx, AV_LOG_ERROR, "Invalid input stream index: %d.\n",
156 return AVERROR(EINVAL);
162 in_ch_idx = strtol(cur, &p, 0);
164 /* channel specifier is not a number,
165 * try to parse as channel name */
166 PARSE_CHANNEL(cur, in_channel, "input");
169 s->channels[out_ch_idx].input = input_idx;
171 s->channels[out_ch_idx].in_channel = in_channel;
173 s->channels[out_ch_idx].in_channel_idx = in_ch_idx;
180 static av_cold int join_init(AVFilterContext *ctx)
182 JoinContext *s = ctx->priv;
185 if (!(s->channel_layout = av_get_channel_layout(s->channel_layout_str))) {
186 av_log(ctx, AV_LOG_ERROR, "Error parsing channel layout '%s'.\n",
187 s->channel_layout_str);
188 ret = AVERROR(EINVAL);
192 s->nb_channels = av_get_channel_layout_nb_channels(s->channel_layout);
193 s->channels = av_mallocz(sizeof(*s->channels) * s->nb_channels);
194 s->buffers = av_mallocz(sizeof(*s->buffers) * s->nb_channels);
195 s->input_frames = av_mallocz(sizeof(*s->input_frames) * s->inputs);
196 if (!s->channels || !s->buffers|| !s->input_frames) {
197 ret = AVERROR(ENOMEM);
201 for (i = 0; i < s->nb_channels; i++) {
202 s->channels[i].out_channel = av_channel_layout_extract_channel(s->channel_layout, i);
203 s->channels[i].input = -1;
206 if ((ret = parse_maps(ctx)) < 0)
209 for (i = 0; i < s->inputs; i++) {
211 AVFilterPad pad = { 0 };
213 snprintf(name, sizeof(name), "input%d", i);
214 pad.type = AVMEDIA_TYPE_AUDIO;
215 pad.name = av_strdup(name);
216 pad.filter_frame = filter_frame;
220 ff_insert_inpad(ctx, i, &pad);
228 static av_cold void join_uninit(AVFilterContext *ctx)
230 JoinContext *s = ctx->priv;
233 for (i = 0; i < ctx->nb_inputs; i++) {
234 av_freep(&ctx->input_pads[i].name);
235 av_frame_free(&s->input_frames[i]);
238 av_freep(&s->channels);
239 av_freep(&s->buffers);
240 av_freep(&s->input_frames);
243 static int join_query_formats(AVFilterContext *ctx)
245 JoinContext *s = ctx->priv;
246 AVFilterChannelLayouts *layouts = NULL;
249 ff_add_channel_layout(&layouts, s->channel_layout);
250 ff_channel_layouts_ref(layouts, &ctx->outputs[0]->in_channel_layouts);
252 for (i = 0; i < ctx->nb_inputs; i++)
253 ff_channel_layouts_ref(ff_all_channel_layouts(),
254 &ctx->inputs[i]->out_channel_layouts);
256 ff_set_common_formats (ctx, ff_planar_sample_fmts());
257 ff_set_common_samplerates(ctx, ff_all_samplerates());
262 static void guess_map_matching(AVFilterContext *ctx, ChannelMap *ch,
267 for (i = 0; i < ctx->nb_inputs; i++) {
268 AVFilterLink *link = ctx->inputs[i];
270 if (ch->out_channel & link->channel_layout &&
271 !(ch->out_channel & inputs[i])) {
273 ch->in_channel = ch->out_channel;
274 inputs[i] |= ch->out_channel;
280 static void guess_map_any(AVFilterContext *ctx, ChannelMap *ch,
285 for (i = 0; i < ctx->nb_inputs; i++) {
286 AVFilterLink *link = ctx->inputs[i];
288 if ((inputs[i] & link->channel_layout) != link->channel_layout) {
289 uint64_t unused = link->channel_layout & ~inputs[i];
292 ch->in_channel = av_channel_layout_extract_channel(unused, 0);
293 inputs[i] |= ch->in_channel;
299 static int join_config_output(AVFilterLink *outlink)
301 AVFilterContext *ctx = outlink->src;
302 JoinContext *s = ctx->priv;
303 uint64_t *inputs; // nth element tracks which channels are used from nth input
306 /* initialize inputs to user-specified mappings */
307 if (!(inputs = av_mallocz(sizeof(*inputs) * ctx->nb_inputs)))
308 return AVERROR(ENOMEM);
309 for (i = 0; i < s->nb_channels; i++) {
310 ChannelMap *ch = &s->channels[i];
311 AVFilterLink *inlink;
316 inlink = ctx->inputs[ch->input];
319 ch->in_channel = av_channel_layout_extract_channel(inlink->channel_layout,
322 if (!(ch->in_channel & inlink->channel_layout)) {
323 av_log(ctx, AV_LOG_ERROR, "Requested channel %s is not present in "
324 "input stream #%d.\n", av_get_channel_name(ch->in_channel),
326 ret = AVERROR(EINVAL);
330 inputs[ch->input] |= ch->in_channel;
333 /* guess channel maps when not explicitly defined */
334 /* first try unused matching channels */
335 for (i = 0; i < s->nb_channels; i++) {
336 ChannelMap *ch = &s->channels[i];
339 guess_map_matching(ctx, ch, inputs);
342 /* if the above failed, try to find _any_ unused input channel */
343 for (i = 0; i < s->nb_channels; i++) {
344 ChannelMap *ch = &s->channels[i];
347 guess_map_any(ctx, ch, inputs);
350 av_log(ctx, AV_LOG_ERROR, "Could not find input channel for "
351 "output channel '%s'.\n",
352 av_get_channel_name(ch->out_channel));
356 ch->in_channel_idx = av_get_channel_layout_channel_index(ctx->inputs[ch->input]->channel_layout,
361 av_log(ctx, AV_LOG_VERBOSE, "mappings: ");
362 for (i = 0; i < s->nb_channels; i++) {
363 ChannelMap *ch = &s->channels[i];
364 av_log(ctx, AV_LOG_VERBOSE, "%d.%s => %s ", ch->input,
365 av_get_channel_name(ch->in_channel),
366 av_get_channel_name(ch->out_channel));
368 av_log(ctx, AV_LOG_VERBOSE, "\n");
370 for (i = 0; i < ctx->nb_inputs; i++) {
372 av_log(ctx, AV_LOG_WARNING, "No channels are used from input "
381 static int join_request_frame(AVFilterLink *outlink)
383 AVFilterContext *ctx = outlink->src;
384 JoinContext *s = ctx->priv;
386 int linesize = INT_MAX;
391 /* get a frame on each input */
392 for (i = 0; i < ctx->nb_inputs; i++) {
393 AVFilterLink *inlink = ctx->inputs[i];
395 if (!s->input_frames[i] &&
396 (ret = ff_request_frame(inlink)) < 0)
399 /* request the same number of samples on all inputs */
401 nb_samples = s->input_frames[0]->nb_samples;
403 for (j = 1; !i && j < ctx->nb_inputs; j++)
404 ctx->inputs[j]->request_samples = nb_samples;
408 /* setup the output frame */
409 frame = av_frame_alloc();
411 return AVERROR(ENOMEM);
412 if (s->nb_channels > FF_ARRAY_ELEMS(frame->data)) {
413 frame->extended_data = av_mallocz(s->nb_channels *
414 sizeof(*frame->extended_data));
415 if (!frame->extended_data) {
416 ret = AVERROR(ENOMEM);
421 /* copy the data pointers */
422 for (i = 0; i < s->nb_channels; i++) {
423 ChannelMap *ch = &s->channels[i];
424 AVFrame *cur = s->input_frames[ch->input];
427 frame->extended_data[i] = cur->extended_data[ch->in_channel_idx];
428 linesize = FFMIN(linesize, cur->linesize[0]);
430 /* add the buffer where this plan is stored to the list if it's
431 * not already there */
432 buf = av_frame_get_plane_buffer(cur, ch->in_channel_idx);
434 ret = AVERROR(EINVAL);
437 for (j = 0; j < nb_buffers; j++)
438 if (s->buffers[j]->buffer == buf->buffer)
441 s->buffers[nb_buffers++] = buf;
444 /* create references to the buffers we copied to output */
445 if (nb_buffers > FF_ARRAY_ELEMS(frame->buf)) {
446 frame->nb_extended_buf = nb_buffers - FF_ARRAY_ELEMS(frame->buf);
447 frame->extended_buf = av_mallocz(sizeof(*frame->extended_buf) *
448 frame->nb_extended_buf);
449 if (!frame->extended_buf) {
450 frame->nb_extended_buf = 0;
451 ret = AVERROR(ENOMEM);
455 for (i = 0; i < FFMIN(FF_ARRAY_ELEMS(frame->buf), nb_buffers); i++) {
456 frame->buf[i] = av_buffer_ref(s->buffers[i]);
457 if (!frame->buf[i]) {
458 ret = AVERROR(ENOMEM);
462 for (i = 0; i < frame->nb_extended_buf; i++) {
463 frame->extended_buf[i] = av_buffer_ref(s->buffers[i +
464 FF_ARRAY_ELEMS(frame->buf)]);
465 if (!frame->extended_buf[i]) {
466 ret = AVERROR(ENOMEM);
471 frame->nb_samples = nb_samples;
472 frame->channel_layout = outlink->channel_layout;
473 frame->sample_rate = outlink->sample_rate;
474 frame->format = outlink->format;
475 frame->pts = s->input_frames[0]->pts;
476 frame->linesize[0] = linesize;
477 if (frame->data != frame->extended_data) {
478 memcpy(frame->data, frame->extended_data, sizeof(*frame->data) *
479 FFMIN(FF_ARRAY_ELEMS(frame->data), s->nb_channels));
482 ret = ff_filter_frame(outlink, frame);
484 for (i = 0; i < ctx->nb_inputs; i++)
485 av_frame_free(&s->input_frames[i]);
490 av_frame_free(&frame);
494 static const AVFilterPad avfilter_af_join_outputs[] = {
497 .type = AVMEDIA_TYPE_AUDIO,
498 .config_props = join_config_output,
499 .request_frame = join_request_frame,
504 AVFilter ff_af_join = {
506 .description = NULL_IF_CONFIG_SMALL("Join multiple audio streams into "
507 "multi-channel output"),
508 .priv_size = sizeof(JoinContext),
509 .priv_class = &join_class,
512 .uninit = join_uninit,
513 .query_formats = join_query_formats,
516 .outputs = avfilter_af_join_outputs,
518 .flags = AVFILTER_FLAG_DYNAMIC_INPUTS,