|
|
|
@ -34,14 +34,20 @@ |
|
|
|
|
#include "internal.h" |
|
|
|
|
|
|
|
|
|
static const char * const var_names[] = { |
|
|
|
|
"ch", ///< the value of the current channel
|
|
|
|
|
"n", ///< number of frame
|
|
|
|
|
"nb_in_channels", |
|
|
|
|
"nb_out_channels", |
|
|
|
|
"t", ///< timestamp expressed in seconds
|
|
|
|
|
"s", ///< sample rate
|
|
|
|
|
NULL |
|
|
|
|
}; |
|
|
|
|
|
|
|
|
|
enum var_name { |
|
|
|
|
VAR_CH, |
|
|
|
|
VAR_N, |
|
|
|
|
VAR_NB_IN_CHANNELS, |
|
|
|
|
VAR_NB_OUT_CHANNELS, |
|
|
|
|
VAR_T, |
|
|
|
|
VAR_S, |
|
|
|
|
VAR_VARS_NB |
|
|
|
@ -53,7 +59,9 @@ typedef struct { |
|
|
|
|
int sample_rate; |
|
|
|
|
int64_t chlayout; |
|
|
|
|
char *chlayout_str; |
|
|
|
|
int nb_channels; |
|
|
|
|
int nb_channels; ///< number of output channels
|
|
|
|
|
int nb_in_channels; ///< number of input channels
|
|
|
|
|
int same_chlayout; ///< set output as input channel layout
|
|
|
|
|
int64_t pts; |
|
|
|
|
AVExpr **expr; |
|
|
|
|
char *exprs; |
|
|
|
@ -61,8 +69,19 @@ typedef struct { |
|
|
|
|
int64_t duration; |
|
|
|
|
uint64_t n; |
|
|
|
|
double var_values[VAR_VARS_NB]; |
|
|
|
|
double *channel_values; |
|
|
|
|
int64_t out_channel_layout; |
|
|
|
|
} EvalContext; |
|
|
|
|
|
|
|
|
|
static double val(void *priv, double ch) |
|
|
|
|
{ |
|
|
|
|
EvalContext *eval = priv; |
|
|
|
|
return eval->channel_values[FFMIN((int)ch, eval->nb_in_channels-1)]; |
|
|
|
|
} |
|
|
|
|
|
|
|
|
|
static double (* const aeval_func1[])(void *, double) = { val, NULL }; |
|
|
|
|
static const char * const aeval_func1_names[] = { "val", NULL }; |
|
|
|
|
|
|
|
|
|
#define OFFSET(x) offsetof(EvalContext, x) |
|
|
|
|
#define FLAGS AV_OPT_FLAG_AUDIO_PARAM|AV_OPT_FLAG_FILTERING_PARAM |
|
|
|
|
|
|
|
|
@ -81,17 +100,27 @@ static const AVOption aevalsrc_options[]= { |
|
|
|
|
|
|
|
|
|
AVFILTER_DEFINE_CLASS(aevalsrc); |
|
|
|
|
|
|
|
|
|
static av_cold int init(AVFilterContext *ctx) |
|
|
|
|
static int parse_channel_expressions(AVFilterContext *ctx, |
|
|
|
|
int expected_nb_channels) |
|
|
|
|
{ |
|
|
|
|
EvalContext *eval = ctx->priv; |
|
|
|
|
char *args1 = av_strdup(eval->exprs); |
|
|
|
|
char *expr, *last_expr, *buf; |
|
|
|
|
int ret; |
|
|
|
|
double (* const *func1)(void *, double) = NULL; |
|
|
|
|
const char * const *func1_names = NULL; |
|
|
|
|
int i, ret = 0; |
|
|
|
|
|
|
|
|
|
if (!args1) |
|
|
|
|
return AVERROR(ENOMEM); |
|
|
|
|
|
|
|
|
|
if (!args1) { |
|
|
|
|
if (!eval->exprs) { |
|
|
|
|
av_log(ctx, AV_LOG_ERROR, "Channels expressions list is empty\n"); |
|
|
|
|
ret = eval->exprs ? AVERROR(ENOMEM) : AVERROR(EINVAL); |
|
|
|
|
goto end; |
|
|
|
|
return AVERROR(EINVAL); |
|
|
|
|
} |
|
|
|
|
|
|
|
|
|
if (!strcmp(ctx->filter->name, "aeval")) { |
|
|
|
|
func1 = aeval_func1; |
|
|
|
|
func1_names = aeval_func1_names; |
|
|
|
|
} |
|
|
|
|
|
|
|
|
|
#define ADD_EXPRESSION(expr_) do { \ |
|
|
|
@ -102,56 +131,79 @@ static av_cold int init(AVFilterContext *ctx) |
|
|
|
|
} \
|
|
|
|
|
eval->expr[eval->nb_channels-1] = NULL; \
|
|
|
|
|
ret = av_expr_parse(&eval->expr[eval->nb_channels - 1], expr_, \
|
|
|
|
|
var_names, NULL, NULL, \
|
|
|
|
|
var_names, func1_names, func1, \
|
|
|
|
|
NULL, NULL, 0, ctx); \
|
|
|
|
|
if (ret < 0) \
|
|
|
|
|
goto end; \
|
|
|
|
|
} while (0) |
|
|
|
|
|
|
|
|
|
/* parse expressions */ |
|
|
|
|
/* reset expressions */ |
|
|
|
|
for (i = 0; i < eval->nb_channels; i++) { |
|
|
|
|
av_expr_free(eval->expr[i]); |
|
|
|
|
eval->expr[i] = NULL; |
|
|
|
|
} |
|
|
|
|
av_freep(&eval->expr); |
|
|
|
|
eval->nb_channels = 0; |
|
|
|
|
|
|
|
|
|
buf = args1; |
|
|
|
|
while (expr = av_strtok(buf, "|", &buf)) { |
|
|
|
|
ADD_EXPRESSION(expr); |
|
|
|
|
last_expr = expr; |
|
|
|
|
} |
|
|
|
|
|
|
|
|
|
if (eval->chlayout_str) { |
|
|
|
|
int i, n; |
|
|
|
|
ret = ff_parse_channel_layout(&eval->chlayout, NULL, eval->chlayout_str, ctx); |
|
|
|
|
if (ret < 0) |
|
|
|
|
goto end; |
|
|
|
|
|
|
|
|
|
n = av_get_channel_layout_nb_channels(eval->chlayout); |
|
|
|
|
if (n > eval->nb_channels) { |
|
|
|
|
for (i = eval->nb_channels; i < n; i++) |
|
|
|
|
if (expected_nb_channels > eval->nb_channels) |
|
|
|
|
for (i = eval->nb_channels; i < expected_nb_channels; i++) |
|
|
|
|
ADD_EXPRESSION(last_expr); |
|
|
|
|
} |
|
|
|
|
|
|
|
|
|
if (n != eval->nb_channels) { |
|
|
|
|
if (expected_nb_channels > 0 && eval->nb_channels != expected_nb_channels) { |
|
|
|
|
av_log(ctx, AV_LOG_ERROR, |
|
|
|
|
"Mismatch between the specified number of channels '%d' " |
|
|
|
|
"and the number of channels '%d' in the specified channel layout '%s'\n", |
|
|
|
|
eval->nb_channels, n, eval->chlayout_str); |
|
|
|
|
"Mismatch between the specified number of channel expressions '%d' " |
|
|
|
|
"and the number of expected output channels '%d' for the specified channel layout\n", |
|
|
|
|
eval->nb_channels, expected_nb_channels); |
|
|
|
|
ret = AVERROR(EINVAL); |
|
|
|
|
goto end; |
|
|
|
|
} |
|
|
|
|
|
|
|
|
|
end: |
|
|
|
|
av_free(args1); |
|
|
|
|
return ret; |
|
|
|
|
} |
|
|
|
|
|
|
|
|
|
static av_cold int init(AVFilterContext *ctx) |
|
|
|
|
{ |
|
|
|
|
EvalContext *eval = ctx->priv; |
|
|
|
|
int ret; |
|
|
|
|
|
|
|
|
|
if (eval->chlayout_str) { |
|
|
|
|
if (!strcmp(eval->chlayout_str, "same") && !strcmp(ctx->filter->name, "aeval")) { |
|
|
|
|
eval->same_chlayout = 1; |
|
|
|
|
} else { |
|
|
|
|
ret = ff_parse_channel_layout(&eval->chlayout, NULL, eval->chlayout_str, ctx); |
|
|
|
|
if (ret < 0) |
|
|
|
|
return ret; |
|
|
|
|
|
|
|
|
|
ret = parse_channel_expressions(ctx, av_get_channel_layout_nb_channels(eval->chlayout)); |
|
|
|
|
if (ret < 0) |
|
|
|
|
return ret; |
|
|
|
|
} |
|
|
|
|
} else { |
|
|
|
|
/* guess channel layout from nb expressions/channels */ |
|
|
|
|
if ((ret = parse_channel_expressions(ctx, -1)) < 0) |
|
|
|
|
return ret; |
|
|
|
|
|
|
|
|
|
eval->chlayout = av_get_default_channel_layout(eval->nb_channels); |
|
|
|
|
if (!eval->chlayout && eval->nb_channels <= 0) { |
|
|
|
|
av_log(ctx, AV_LOG_ERROR, "Invalid number of channels '%d' provided\n", |
|
|
|
|
eval->nb_channels); |
|
|
|
|
ret = AVERROR(EINVAL); |
|
|
|
|
goto end; |
|
|
|
|
return AVERROR(EINVAL); |
|
|
|
|
} |
|
|
|
|
} |
|
|
|
|
|
|
|
|
|
if (eval->sample_rate_str) |
|
|
|
|
if ((ret = ff_parse_sample_rate(&eval->sample_rate, eval->sample_rate_str, ctx))) |
|
|
|
|
goto end; |
|
|
|
|
return ret; |
|
|
|
|
eval->n = 0; |
|
|
|
|
|
|
|
|
|
end: |
|
|
|
|
av_free(args1); |
|
|
|
|
return ret; |
|
|
|
|
} |
|
|
|
|
|
|
|
|
@ -176,6 +228,8 @@ static int config_props(AVFilterLink *outlink) |
|
|
|
|
outlink->sample_rate = eval->sample_rate; |
|
|
|
|
|
|
|
|
|
eval->var_values[VAR_S] = eval->sample_rate; |
|
|
|
|
eval->var_values[VAR_NB_IN_CHANNELS] = NAN; |
|
|
|
|
eval->var_values[VAR_NB_OUT_CHANNELS] = outlink->channels; |
|
|
|
|
|
|
|
|
|
av_get_channel_layout_string(buf, sizeof(buf), 0, eval->chlayout); |
|
|
|
|
|
|
|
|
@ -232,6 +286,7 @@ static int request_frame(AVFilterLink *outlink) |
|
|
|
|
return ff_filter_frame(outlink, samplesref); |
|
|
|
|
} |
|
|
|
|
|
|
|
|
|
#if CONFIG_AEVALSRC_FILTER |
|
|
|
|
static const AVFilterPad aevalsrc_outputs[] = { |
|
|
|
|
{ |
|
|
|
|
.name = "default", |
|
|
|
@ -253,3 +308,159 @@ AVFilter ff_asrc_aevalsrc = { |
|
|
|
|
.outputs = aevalsrc_outputs, |
|
|
|
|
.priv_class = &aevalsrc_class, |
|
|
|
|
}; |
|
|
|
|
|
|
|
|
|
#endif /* CONFIG_AEVALSRC_FILTER */ |
|
|
|
|
|
|
|
|
|
#define OFFSET(x) offsetof(EvalContext, x) |
|
|
|
|
#define FLAGS AV_OPT_FLAG_AUDIO_PARAM|AV_OPT_FLAG_FILTERING_PARAM |
|
|
|
|
|
|
|
|
|
static const AVOption aeval_options[]= { |
|
|
|
|
{ "exprs", "set the '|'-separated list of channels expressions", OFFSET(exprs), AV_OPT_TYPE_STRING, {.str = NULL}, .flags = FLAGS }, |
|
|
|
|
{ "channel_layout", "set channel layout", OFFSET(chlayout_str), AV_OPT_TYPE_STRING, {.str = NULL}, 0, 0, FLAGS }, |
|
|
|
|
{ "c", "set channel layout", OFFSET(chlayout_str), AV_OPT_TYPE_STRING, {.str = NULL}, 0, 0, FLAGS }, |
|
|
|
|
{ NULL } |
|
|
|
|
}; |
|
|
|
|
|
|
|
|
|
AVFILTER_DEFINE_CLASS(aeval); |
|
|
|
|
|
|
|
|
|
static int aeval_query_formats(AVFilterContext *ctx) |
|
|
|
|
{ |
|
|
|
|
AVFilterFormats *formats = NULL; |
|
|
|
|
AVFilterChannelLayouts *layouts; |
|
|
|
|
AVFilterLink *inlink = ctx->inputs[0]; |
|
|
|
|
AVFilterLink *outlink = ctx->outputs[0]; |
|
|
|
|
EvalContext *eval = ctx->priv; |
|
|
|
|
static const enum AVSampleFormat sample_fmts[] = { |
|
|
|
|
AV_SAMPLE_FMT_DBLP, AV_SAMPLE_FMT_NONE |
|
|
|
|
}; |
|
|
|
|
|
|
|
|
|
// inlink supports any channel layout
|
|
|
|
|
layouts = ff_all_channel_counts(); |
|
|
|
|
ff_channel_layouts_ref(layouts, &inlink->out_channel_layouts); |
|
|
|
|
|
|
|
|
|
if (eval->same_chlayout) { |
|
|
|
|
layouts = ff_all_channel_counts(); |
|
|
|
|
if (!layouts) |
|
|
|
|
return AVERROR(ENOMEM); |
|
|
|
|
ff_set_common_channel_layouts(ctx, layouts); |
|
|
|
|
} else { |
|
|
|
|
// outlink supports only requested output channel layout
|
|
|
|
|
layouts = NULL; |
|
|
|
|
ff_add_channel_layout(&layouts, |
|
|
|
|
eval->out_channel_layout ? eval->out_channel_layout : |
|
|
|
|
FF_COUNT2LAYOUT(eval->nb_channels)); |
|
|
|
|
ff_channel_layouts_ref(layouts, &outlink->in_channel_layouts); |
|
|
|
|
} |
|
|
|
|
|
|
|
|
|
formats = ff_make_format_list(sample_fmts); |
|
|
|
|
if (!formats) |
|
|
|
|
return AVERROR(ENOMEM); |
|
|
|
|
ff_set_common_formats(ctx, formats); |
|
|
|
|
|
|
|
|
|
formats = ff_all_samplerates(); |
|
|
|
|
if (!formats) |
|
|
|
|
return AVERROR(ENOMEM); |
|
|
|
|
ff_set_common_samplerates(ctx, formats); |
|
|
|
|
|
|
|
|
|
return 0; |
|
|
|
|
} |
|
|
|
|
|
|
|
|
|
static int aeval_config_output(AVFilterLink *outlink) |
|
|
|
|
{ |
|
|
|
|
AVFilterContext *ctx = outlink->src; |
|
|
|
|
EvalContext *eval = ctx->priv; |
|
|
|
|
AVFilterLink *inlink = ctx->inputs[0]; |
|
|
|
|
int ret; |
|
|
|
|
|
|
|
|
|
if (eval->same_chlayout) { |
|
|
|
|
eval->chlayout = inlink->channel_layout; |
|
|
|
|
|
|
|
|
|
if ((ret = parse_channel_expressions(ctx, inlink->channels)) < 0) |
|
|
|
|
return ret; |
|
|
|
|
} |
|
|
|
|
|
|
|
|
|
eval->n = 0; |
|
|
|
|
eval->nb_in_channels = eval->var_values[VAR_NB_IN_CHANNELS] = inlink->channels; |
|
|
|
|
eval->var_values[VAR_NB_OUT_CHANNELS] = outlink->channels; |
|
|
|
|
eval->var_values[VAR_S] = inlink->sample_rate; |
|
|
|
|
eval->var_values[VAR_T] = NAN; |
|
|
|
|
|
|
|
|
|
eval->channel_values = av_realloc_f(eval->channel_values, |
|
|
|
|
inlink->channels, sizeof(*eval->channel_values)); |
|
|
|
|
if (!eval->channel_values) |
|
|
|
|
return AVERROR(ENOMEM); |
|
|
|
|
|
|
|
|
|
return 0; |
|
|
|
|
} |
|
|
|
|
|
|
|
|
|
#define TS2T(ts, tb) ((ts) == AV_NOPTS_VALUE ? NAN : (double)(ts)*av_q2d(tb)) |
|
|
|
|
|
|
|
|
|
static int filter_frame(AVFilterLink *inlink, AVFrame *in) |
|
|
|
|
{ |
|
|
|
|
EvalContext *eval = inlink->dst->priv; |
|
|
|
|
AVFilterLink *outlink = inlink->dst->outputs[0]; |
|
|
|
|
int nb_samples = in->nb_samples; |
|
|
|
|
AVFrame *out; |
|
|
|
|
double t0; |
|
|
|
|
int i, j; |
|
|
|
|
|
|
|
|
|
/* do volume scaling in-place if input buffer is writable */ |
|
|
|
|
out = ff_get_audio_buffer(outlink, nb_samples); |
|
|
|
|
if (!out) |
|
|
|
|
return AVERROR(ENOMEM); |
|
|
|
|
av_frame_copy_props(out, in); |
|
|
|
|
|
|
|
|
|
t0 = TS2T(in->pts, inlink->time_base); |
|
|
|
|
|
|
|
|
|
/* evaluate expression for each single sample and for each channel */ |
|
|
|
|
for (i = 0; i < nb_samples; i++, eval->n++) { |
|
|
|
|
eval->var_values[VAR_N] = eval->n; |
|
|
|
|
eval->var_values[VAR_T] = t0 + i * (double)1/inlink->sample_rate; |
|
|
|
|
|
|
|
|
|
for (j = 0; j < inlink->channels; j++) |
|
|
|
|
eval->channel_values[j] = *((double *) in->extended_data[j] + i); |
|
|
|
|
|
|
|
|
|
for (j = 0; j < outlink->channels; j++) { |
|
|
|
|
eval->var_values[VAR_CH] = j; |
|
|
|
|
*((double *) out->extended_data[j] + i) = |
|
|
|
|
av_expr_eval(eval->expr[j], eval->var_values, eval); |
|
|
|
|
} |
|
|
|
|
} |
|
|
|
|
|
|
|
|
|
av_frame_free(&in); |
|
|
|
|
return ff_filter_frame(outlink, out); |
|
|
|
|
} |
|
|
|
|
|
|
|
|
|
#if CONFIG_AEVAL_FILTER |
|
|
|
|
|
|
|
|
|
static const AVFilterPad aeval_inputs[] = { |
|
|
|
|
{ |
|
|
|
|
.name = "default", |
|
|
|
|
.type = AVMEDIA_TYPE_AUDIO, |
|
|
|
|
.filter_frame = filter_frame, |
|
|
|
|
}, |
|
|
|
|
{ NULL } |
|
|
|
|
}; |
|
|
|
|
|
|
|
|
|
static const AVFilterPad aeval_outputs[] = { |
|
|
|
|
{ |
|
|
|
|
.name = "default", |
|
|
|
|
.type = AVMEDIA_TYPE_AUDIO, |
|
|
|
|
.config_props = aeval_config_output, |
|
|
|
|
}, |
|
|
|
|
{ NULL } |
|
|
|
|
}; |
|
|
|
|
|
|
|
|
|
AVFilter ff_af_aeval = { |
|
|
|
|
.name = "aeval", |
|
|
|
|
.description = NULL_IF_CONFIG_SMALL("Filter audio signal according to a specified expression."), |
|
|
|
|
.query_formats = aeval_query_formats, |
|
|
|
|
.init = init, |
|
|
|
|
.uninit = uninit, |
|
|
|
|
.priv_size = sizeof(EvalContext), |
|
|
|
|
.inputs = aeval_inputs, |
|
|
|
|
.outputs = aeval_outputs, |
|
|
|
|
.priv_class = &aeval_class, |
|
|
|
|
}; |
|
|
|
|
|
|
|
|
|
#endif /* CONFIG_AEVAL_FILTER */ |
|
|
|
|