yading@10: /*
yading@10:  * Copyright (c) 2012 Nicolas George
yading@10:  *
yading@10:  * This file is part of FFmpeg.
yading@10:  *
yading@10:  * FFmpeg is free software; you can redistribute it and/or
yading@10:  * modify it under the terms of the GNU Lesser General Public
yading@10:  * License as published by the Free Software Foundation; either
yading@10:  * version 2.1 of the License, or (at your option) any later version.
yading@10:  *
yading@10:  * FFmpeg is distributed in the hope that it will be useful,
yading@10:  * but WITHOUT ANY WARRANTY; without even the implied warranty of
yading@10:  * MERCHANTABILITY or FITNESS FOR A PARTICULAR PURPOSE.
yading@10:  * See the GNU Lesser General Public License for more details.
yading@10:  *
yading@10:  * You should have received a copy of the GNU Lesser General Public License
yading@10:  * along with FFmpeg; if not, write to the Free Software Foundation, Inc.,
yading@10:  * 51 Franklin Street, Fifth Floor, Boston, MA 02110-1301 USA
yading@10:  */
yading@10: 
yading@10: /**
yading@10:  * @file
yading@10:  * concat audio-video filter
yading@10:  */
yading@10: 
yading@10: #include "libavutil/avassert.h"
yading@10: #include "libavutil/avstring.h"
yading@10: #include "libavutil/channel_layout.h"
yading@10: #include "libavutil/opt.h"
yading@10: #include "avfilter.h"
yading@10: #define FF_BUFQUEUE_SIZE 256
yading@10: #include "bufferqueue.h"
yading@10: #include "internal.h"
yading@10: #include "video.h"
yading@10: #include "audio.h"
yading@10: 
yading@10: #define TYPE_ALL 2
yading@10: 
yading@10: typedef struct {
yading@10:     const AVClass *class;
yading@10:     unsigned nb_streams[TYPE_ALL]; /**< number of out streams of each type */
yading@10:     unsigned nb_segments;
yading@10:     unsigned cur_idx; /**< index of the first input of current segment */
yading@10:     int64_t delta_ts; /**< timestamp to add to produce output timestamps */
yading@10:     unsigned nb_in_active; /**< number of active inputs in current segment */
yading@10:     unsigned unsafe;
yading@10:     struct concat_in {
yading@10:         int64_t pts;
yading@10:         int64_t nb_frames;
yading@10:         unsigned eof;
yading@10:         struct FFBufQueue queue;
yading@10:     } *in;
yading@10: } ConcatContext;
yading@10: 
yading@10: #define OFFSET(x) offsetof(ConcatContext, x)
yading@10: #define A AV_OPT_FLAG_AUDIO_PARAM
yading@10: #define F AV_OPT_FLAG_FILTERING_PARAM
yading@10: #define V AV_OPT_FLAG_VIDEO_PARAM
yading@10: 
yading@10: static const AVOption concat_options[] = {
yading@10:     { "n", "specify the number of segments", OFFSET(nb_segments),
yading@10:       AV_OPT_TYPE_INT, { .i64 = 2 }, 2, INT_MAX, V|A|F},
yading@10:     { "v", "specify the number of video streams",
yading@10:       OFFSET(nb_streams[AVMEDIA_TYPE_VIDEO]),
yading@10:       AV_OPT_TYPE_INT, { .i64 = 1 }, 0, INT_MAX, V|F },
yading@10:     { "a", "specify the number of audio streams",
yading@10:       OFFSET(nb_streams[AVMEDIA_TYPE_AUDIO]),
yading@10:       AV_OPT_TYPE_INT, { .i64 = 0 }, 0, INT_MAX, A|F},
yading@10:     { "unsafe", "enable unsafe mode",
yading@10:       OFFSET(unsafe),
yading@10:       AV_OPT_TYPE_INT, { .i64 = 0 }, 0, INT_MAX, A|A|F},
yading@10:     { 0 }
yading@10: };
yading@10: 
yading@10: AVFILTER_DEFINE_CLASS(concat);
yading@10: 
yading@10: static int query_formats(AVFilterContext *ctx)
yading@10: {
yading@10:     ConcatContext *cat = ctx->priv;
yading@10:     unsigned type, nb_str, idx0 = 0, idx, str, seg;
yading@10:     AVFilterFormats *formats, *rates = NULL;
yading@10:     AVFilterChannelLayouts *layouts = NULL;
yading@10: 
yading@10:     for (type = 0; type < TYPE_ALL; type++) {
yading@10:         nb_str = cat->nb_streams[type];
yading@10:         for (str = 0; str < nb_str; str++) {
yading@10:             idx = idx0;
yading@10: 
yading@10:             /* Set the output formats */
yading@10:             formats = ff_all_formats(type);
yading@10:             if (!formats)
yading@10:                 return AVERROR(ENOMEM);
yading@10:             ff_formats_ref(formats, &ctx->outputs[idx]->in_formats);
yading@10:             if (type == AVMEDIA_TYPE_AUDIO) {
yading@10:                 rates = ff_all_samplerates();
yading@10:                 if (!rates)
yading@10:                     return AVERROR(ENOMEM);
yading@10:                 ff_formats_ref(rates, &ctx->outputs[idx]->in_samplerates);
yading@10:                 layouts = ff_all_channel_layouts();
yading@10:                 if (!layouts)
yading@10:                     return AVERROR(ENOMEM);
yading@10:                 ff_channel_layouts_ref(layouts, &ctx->outputs[idx]->in_channel_layouts);
yading@10:             }
yading@10: 
yading@10:             /* Set the same formats for each corresponding input */
yading@10:             for (seg = 0; seg < cat->nb_segments; seg++) {
yading@10:                 ff_formats_ref(formats, &ctx->inputs[idx]->out_formats);
yading@10:                 if (type == AVMEDIA_TYPE_AUDIO) {
yading@10:                     ff_formats_ref(rates, &ctx->inputs[idx]->out_samplerates);
yading@10:                     ff_channel_layouts_ref(layouts, &ctx->inputs[idx]->out_channel_layouts);
yading@10:                 }
yading@10:                 idx += ctx->nb_outputs;
yading@10:             }
yading@10: 
yading@10:             idx0++;
yading@10:         }
yading@10:     }
yading@10:     return 0;
yading@10: }
yading@10: 
yading@10: static int config_output(AVFilterLink *outlink)
yading@10: {
yading@10:     AVFilterContext *ctx = outlink->src;
yading@10:     ConcatContext *cat   = ctx->priv;
yading@10:     unsigned out_no = FF_OUTLINK_IDX(outlink);
yading@10:     unsigned in_no  = out_no, seg;
yading@10:     AVFilterLink *inlink = ctx->inputs[in_no];
yading@10: 
yading@10:     /* enhancement: find a common one */
yading@10:     outlink->time_base           = AV_TIME_BASE_Q;
yading@10:     outlink->w                   = inlink->w;
yading@10:     outlink->h                   = inlink->h;
yading@10:     outlink->sample_aspect_ratio = inlink->sample_aspect_ratio;
yading@10:     outlink->format              = inlink->format;
yading@10:     for (seg = 1; seg < cat->nb_segments; seg++) {
yading@10:         inlink = ctx->inputs[in_no += ctx->nb_outputs];
yading@10:         /* possible enhancement: unsafe mode, do not check */
yading@10:         if (outlink->w                       != inlink->w                       ||
yading@10:             outlink->h                       != inlink->h                       ||
yading@10:             outlink->sample_aspect_ratio.num != inlink->sample_aspect_ratio.num ||
yading@10:             outlink->sample_aspect_ratio.den != inlink->sample_aspect_ratio.den) {
yading@10:             av_log(ctx, AV_LOG_ERROR, "Input link %s parameters "
yading@10:                    "(size %dx%d, SAR %d:%d) do not match the corresponding "
yading@10:                    "output link %s parameters (%dx%d, SAR %d:%d)\n",
yading@10:                    ctx->input_pads[in_no].name, inlink->w, inlink->h,
yading@10:                    inlink->sample_aspect_ratio.num,
yading@10:                    inlink->sample_aspect_ratio.den,
yading@10:                    ctx->input_pads[out_no].name, outlink->w, outlink->h,
yading@10:                    outlink->sample_aspect_ratio.num,
yading@10:                    outlink->sample_aspect_ratio.den);
yading@10:             if (!cat->unsafe)
yading@10:                 return AVERROR(EINVAL);
yading@10:         }
yading@10:     }
yading@10: 
yading@10:     return 0;
yading@10: }
yading@10: 
yading@10: static int push_frame(AVFilterContext *ctx, unsigned in_no, AVFrame *buf)
yading@10: {
yading@10:     ConcatContext *cat = ctx->priv;
yading@10:     unsigned out_no = in_no % ctx->nb_outputs;
yading@10:     AVFilterLink * inlink = ctx-> inputs[ in_no];
yading@10:     AVFilterLink *outlink = ctx->outputs[out_no];
yading@10:     struct concat_in *in = &cat->in[in_no];
yading@10: 
yading@10:     buf->pts = av_rescale_q(buf->pts, inlink->time_base, outlink->time_base);
yading@10:     in->pts = buf->pts;
yading@10:     in->nb_frames++;
yading@10:     /* add duration to input PTS */
yading@10:     if (inlink->sample_rate)
yading@10:         /* use number of audio samples */
yading@10:         in->pts += av_rescale_q(buf->nb_samples,
yading@10:                                 (AVRational){ 1, inlink->sample_rate },
yading@10:                                 outlink->time_base);
yading@10:     else if (in->nb_frames >= 2)
yading@10:         /* use mean duration */
yading@10:         in->pts = av_rescale(in->pts, in->nb_frames, in->nb_frames - 1);
yading@10: 
yading@10:     buf->pts += cat->delta_ts;
yading@10:     return ff_filter_frame(outlink, buf);
yading@10: }
yading@10: 
yading@10: static int process_frame(AVFilterLink *inlink, AVFrame *buf)
yading@10: {
yading@10:     AVFilterContext *ctx  = inlink->dst;
yading@10:     ConcatContext *cat    = ctx->priv;
yading@10:     unsigned in_no = FF_INLINK_IDX(inlink);
yading@10: 
yading@10:     if (in_no < cat->cur_idx) {
yading@10:         av_log(ctx, AV_LOG_ERROR, "Frame after EOF on input %s\n",
yading@10:                ctx->input_pads[in_no].name);
yading@10:         av_frame_free(&buf);
yading@10:     } else if (in_no >= cat->cur_idx + ctx->nb_outputs) {
yading@10:         ff_bufqueue_add(ctx, &cat->in[in_no].queue, buf);
yading@10:     } else {
yading@10:         return push_frame(ctx, in_no, buf);
yading@10:     }
yading@10:     return 0;
yading@10: }
yading@10: 
yading@10: static AVFrame *get_video_buffer(AVFilterLink *inlink, int w, int h)
yading@10: {
yading@10:     AVFilterContext *ctx = inlink->dst;
yading@10:     unsigned in_no = FF_INLINK_IDX(inlink);
yading@10:     AVFilterLink *outlink = ctx->outputs[in_no % ctx->nb_outputs];
yading@10: 
yading@10:     return ff_get_video_buffer(outlink, w, h);
yading@10: }
yading@10: 
yading@10: static AVFrame *get_audio_buffer(AVFilterLink *inlink, int nb_samples)
yading@10: {
yading@10:     AVFilterContext *ctx = inlink->dst;
yading@10:     unsigned in_no = FF_INLINK_IDX(inlink);
yading@10:     AVFilterLink *outlink = ctx->outputs[in_no % ctx->nb_outputs];
yading@10: 
yading@10:     return ff_get_audio_buffer(outlink, nb_samples);
yading@10: }
yading@10: 
yading@10: static int filter_frame(AVFilterLink *inlink, AVFrame *buf)
yading@10: {
yading@10:     return process_frame(inlink, buf);
yading@10: }
yading@10: 
yading@10: static void close_input(AVFilterContext *ctx, unsigned in_no)
yading@10: {
yading@10:     ConcatContext *cat = ctx->priv;
yading@10: 
yading@10:     cat->in[in_no].eof = 1;
yading@10:     cat->nb_in_active--;
yading@10:     av_log(ctx, AV_LOG_VERBOSE, "EOF on %s, %d streams left in segment.\n",
yading@10:            ctx->input_pads[in_no].name, cat->nb_in_active);
yading@10: }
yading@10: 
yading@10: static void find_next_delta_ts(AVFilterContext *ctx, int64_t *seg_delta)
yading@10: {
yading@10:     ConcatContext *cat = ctx->priv;
yading@10:     unsigned i = cat->cur_idx;
yading@10:     unsigned imax = i + ctx->nb_outputs;
yading@10:     int64_t pts;
yading@10: 
yading@10:     pts = cat->in[i++].pts;
yading@10:     for (; i < imax; i++)
yading@10:         pts = FFMAX(pts, cat->in[i].pts);
yading@10:     cat->delta_ts += pts;
yading@10:     *seg_delta = pts;
yading@10: }
yading@10: 
yading@10: static int send_silence(AVFilterContext *ctx, unsigned in_no, unsigned out_no,
yading@10:                         int64_t seg_delta)
yading@10: {
yading@10:     ConcatContext *cat = ctx->priv;
yading@10:     AVFilterLink *outlink = ctx->outputs[out_no];
yading@10:     int64_t base_pts = cat->in[in_no].pts + cat->delta_ts - seg_delta;
yading@10:     int64_t nb_samples, sent = 0;
yading@10:     int frame_nb_samples, ret;
yading@10:     AVRational rate_tb = { 1, ctx->inputs[in_no]->sample_rate };
yading@10:     AVFrame *buf;
yading@10:     int nb_channels = av_get_channel_layout_nb_channels(outlink->channel_layout);
yading@10: 
yading@10:     if (!rate_tb.den)
yading@10:         return AVERROR_BUG;
yading@10:     nb_samples = av_rescale_q(seg_delta - cat->in[in_no].pts,
yading@10:                               outlink->time_base, rate_tb);
yading@10:     frame_nb_samples = FFMAX(9600, rate_tb.den / 5); /* arbitrary */
yading@10:     while (nb_samples) {
yading@10:         frame_nb_samples = FFMIN(frame_nb_samples, nb_samples);
yading@10:         buf = ff_get_audio_buffer(outlink, frame_nb_samples);
yading@10:         if (!buf)
yading@10:             return AVERROR(ENOMEM);
yading@10:         av_samples_set_silence(buf->extended_data, 0, frame_nb_samples,
yading@10:                                nb_channels, outlink->format);
yading@10:         buf->pts = base_pts + av_rescale_q(sent, rate_tb, outlink->time_base);
yading@10:         ret = ff_filter_frame(outlink, buf);
yading@10:         if (ret < 0)
yading@10:             return ret;
yading@10:         sent       += frame_nb_samples;
yading@10:         nb_samples -= frame_nb_samples;
yading@10:     }
yading@10:     return 0;
yading@10: }
yading@10: 
yading@10: static int flush_segment(AVFilterContext *ctx)
yading@10: {
yading@10:     int ret;
yading@10:     ConcatContext *cat = ctx->priv;
yading@10:     unsigned str, str_max;
yading@10:     int64_t seg_delta;
yading@10: 
yading@10:     find_next_delta_ts(ctx, &seg_delta);
yading@10:     cat->cur_idx += ctx->nb_outputs;
yading@10:     cat->nb_in_active = ctx->nb_outputs;
yading@10:     av_log(ctx, AV_LOG_VERBOSE, "Segment finished at pts=%"PRId64"\n",
yading@10:            cat->delta_ts);
yading@10: 
yading@10:     if (cat->cur_idx < ctx->nb_inputs) {
yading@10:         /* pad audio streams with silence */
yading@10:         str = cat->nb_streams[AVMEDIA_TYPE_VIDEO];
yading@10:         str_max = str + cat->nb_streams[AVMEDIA_TYPE_AUDIO];
yading@10:         for (; str < str_max; str++) {
yading@10:             ret = send_silence(ctx, cat->cur_idx - ctx->nb_outputs + str, str,
yading@10:                                seg_delta);
yading@10:             if (ret < 0)
yading@10:                 return ret;
yading@10:         }
yading@10:         /* flush queued buffers */
yading@10:         /* possible enhancement: flush in PTS order */
yading@10:         str_max = cat->cur_idx + ctx->nb_outputs;
yading@10:         for (str = cat->cur_idx; str < str_max; str++) {
yading@10:             while (cat->in[str].queue.available) {
yading@10:                 ret = push_frame(ctx, str, ff_bufqueue_get(&cat->in[str].queue));
yading@10:                 if (ret < 0)
yading@10:                     return ret;
yading@10:             }
yading@10:         }
yading@10:     }
yading@10:     return 0;
yading@10: }
yading@10: 
yading@10: static int request_frame(AVFilterLink *outlink)
yading@10: {
yading@10:     AVFilterContext *ctx = outlink->src;
yading@10:     ConcatContext *cat   = ctx->priv;
yading@10:     unsigned out_no = FF_OUTLINK_IDX(outlink);
yading@10:     unsigned in_no  = out_no + cat->cur_idx;
yading@10:     unsigned str, str_max;
yading@10:     int ret;
yading@10: 
yading@10:     while (1) {
yading@10:         if (in_no >= ctx->nb_inputs)
yading@10:             return AVERROR_EOF;
yading@10:         if (!cat->in[in_no].eof) {
yading@10:             ret = ff_request_frame(ctx->inputs[in_no]);
yading@10:             if (ret != AVERROR_EOF)
yading@10:                 return ret;
yading@10:             close_input(ctx, in_no);
yading@10:         }
yading@10:         /* cycle on all inputs to finish the segment */
yading@10:         /* possible enhancement: request in PTS order */
yading@10:         str_max = cat->cur_idx + ctx->nb_outputs - 1;
yading@10:         for (str = cat->cur_idx; cat->nb_in_active;
yading@10:              str = str == str_max ? cat->cur_idx : str + 1) {
yading@10:             if (cat->in[str].eof)
yading@10:                 continue;
yading@10:             ret = ff_request_frame(ctx->inputs[str]);
yading@10:             if (ret == AVERROR_EOF)
yading@10:                 close_input(ctx, str);
yading@10:             else if (ret < 0)
yading@10:                 return ret;
yading@10:         }
yading@10:         ret = flush_segment(ctx);
yading@10:         if (ret < 0)
yading@10:             return ret;
yading@10:         in_no += ctx->nb_outputs;
yading@10:     }
yading@10: }
yading@10: 
yading@10: static av_cold int init(AVFilterContext *ctx)
yading@10: {
yading@10:     ConcatContext *cat = ctx->priv;
yading@10:     unsigned seg, type, str;
yading@10: 
yading@10:     /* create input pads */
yading@10:     for (seg = 0; seg < cat->nb_segments; seg++) {
yading@10:         for (type = 0; type < TYPE_ALL; type++) {
yading@10:             for (str = 0; str < cat->nb_streams[type]; str++) {
yading@10:                 AVFilterPad pad = {
yading@10:                     .type             = type,
yading@10:                     .get_video_buffer = get_video_buffer,
yading@10:                     .get_audio_buffer = get_audio_buffer,
yading@10:                     .filter_frame     = filter_frame,
yading@10:                 };
yading@10:                 pad.name = av_asprintf("in%d:%c%d", seg, "va"[type], str);
yading@10:                 ff_insert_inpad(ctx, ctx->nb_inputs, &pad);
yading@10:             }
yading@10:         }
yading@10:     }
yading@10:     /* create output pads */
yading@10:     for (type = 0; type < TYPE_ALL; type++) {
yading@10:         for (str = 0; str < cat->nb_streams[type]; str++) {
yading@10:             AVFilterPad pad = {
yading@10:                 .type          = type,
yading@10:                 .config_props  = config_output,
yading@10:                 .request_frame = request_frame,
yading@10:             };
yading@10:             pad.name = av_asprintf("out:%c%d", "va"[type], str);
yading@10:             ff_insert_outpad(ctx, ctx->nb_outputs, &pad);
yading@10:         }
yading@10:     }
yading@10: 
yading@10:     cat->in = av_calloc(ctx->nb_inputs, sizeof(*cat->in));
yading@10:     if (!cat->in)
yading@10:         return AVERROR(ENOMEM);
yading@10:     cat->nb_in_active = ctx->nb_outputs;
yading@10:     return 0;
yading@10: }
yading@10: 
yading@10: static av_cold void uninit(AVFilterContext *ctx)
yading@10: {
yading@10:     ConcatContext *cat = ctx->priv;
yading@10:     unsigned i;
yading@10: 
yading@10:     for (i = 0; i < ctx->nb_inputs; i++) {
yading@10:         av_freep(&ctx->input_pads[i].name);
yading@10:         ff_bufqueue_discard_all(&cat->in[i].queue);
yading@10:     }
yading@10:     for (i = 0; i < ctx->nb_outputs; i++)
yading@10:         av_freep(&ctx->output_pads[i].name);
yading@10:     av_free(cat->in);
yading@10: }
yading@10: 
yading@10: AVFilter avfilter_avf_concat = {
yading@10:     .name          = "concat",
yading@10:     .description   = NULL_IF_CONFIG_SMALL("Concatenate audio and video streams."),
yading@10:     .init          = init,
yading@10:     .uninit        = uninit,
yading@10:     .query_formats = query_formats,
yading@10:     .priv_size     = sizeof(ConcatContext),
yading@10:     .inputs        = NULL,
yading@10:     .outputs       = NULL,
yading@10:     .priv_class    = &concat_class,
yading@10:     .flags         = AVFILTER_FLAG_DYNAMIC_INPUTS | AVFILTER_FLAG_DYNAMIC_OUTPUTS,
yading@10: };