1 /*
2 * Copyright (c) 2011 Stefano Sabatini
3 *
4 * This file is part of FFmpeg.
5 *
6 * FFmpeg is free software; you can redistribute it and/or
7 * modify it under the terms of the GNU Lesser General Public
8 * License as published by the Free Software Foundation; either
9 * version 2.1 of the License, or (at your option) any later version.
10 *
11 * FFmpeg is distributed in the hope that it will be useful,
12 * but WITHOUT ANY WARRANTY; without even the implied warranty of
13 * MERCHANTABILITY or FITNESS FOR A PARTICULAR PURPOSE. See the GNU
14 * Lesser General Public License for more details.
15 *
16 * You should have received a copy of the GNU Lesser General Public
17 * License along with FFmpeg; if not, write to the Free Software
18 * Foundation, Inc., 51 Franklin Street, Fifth Floor, Boston, MA 02110-1301 USA
19 */
20
21 /**
22 * @file
23 * eval audio source
24 */
25
26 #include "libavutil/avassert.h"
27 #include "libavutil/avstring.h"
28 #include "libavutil/channel_layout.h"
29 #include "libavutil/eval.h"
30 #include "libavutil/opt.h"
31 #include "libavutil/parseutils.h"
32 #include "avfilter.h"
33 #include "audio.h"
34 #include "internal.h"
35
36 static const char * const var_names[] = {
37 "ch", ///< the value of the current channel
38 "n", ///< number of frame
39 "nb_in_channels",
40 "nb_out_channels",
41 "t", ///< timestamp expressed in seconds
42 "s", ///< sample rate
43 NULL
44 };
45
46 enum var_name {
47 VAR_CH,
48 VAR_N,
49 VAR_NB_IN_CHANNELS,
50 VAR_NB_OUT_CHANNELS,
51 VAR_T,
52 VAR_S,
53 VAR_VARS_NB
54 };
55
56 typedef struct EvalContext {
57 const AVClass *class;
58 char *sample_rate_str;
59 int sample_rate;
60 int64_t chlayout;
61 char *chlayout_str;
62 int nb_channels; ///< number of output channels
63 int nb_in_channels; ///< number of input channels
64 int same_chlayout; ///< set output as input channel layout
65 int64_t pts;
66 AVExpr **expr;
67 char *exprs;
68 int nb_samples; ///< number of samples per requested frame
69 int64_t duration;
70 uint64_t n;
71 double var_values[VAR_VARS_NB];
72 double *channel_values;
73 int64_t out_channel_layout;
74 } EvalContext;
75
val(void * priv,double ch)76 static double val(void *priv, double ch)
77 {
78 EvalContext *eval = priv;
79 return eval->channel_values[FFMIN((int)ch, eval->nb_in_channels-1)];
80 }
81
82 static double (* const aeval_func1[])(void *, double) = { val, NULL };
83 static const char * const aeval_func1_names[] = { "val", NULL };
84
85 #define OFFSET(x) offsetof(EvalContext, x)
86 #define FLAGS AV_OPT_FLAG_AUDIO_PARAM|AV_OPT_FLAG_FILTERING_PARAM
87
88 static const AVOption aevalsrc_options[]= {
89 { "exprs", "set the '|'-separated list of channels expressions", OFFSET(exprs), AV_OPT_TYPE_STRING, {.str = NULL}, .flags = FLAGS },
90 { "nb_samples", "set the number of samples per requested frame", OFFSET(nb_samples), AV_OPT_TYPE_INT, {.i64 = 1024}, 0, INT_MAX, FLAGS },
91 { "n", "set the number of samples per requested frame", OFFSET(nb_samples), AV_OPT_TYPE_INT, {.i64 = 1024}, 0, INT_MAX, FLAGS },
92 { "sample_rate", "set the sample rate", OFFSET(sample_rate_str), AV_OPT_TYPE_STRING, {.str = "44100"}, 0, 0, FLAGS },
93 { "s", "set the sample rate", OFFSET(sample_rate_str), AV_OPT_TYPE_STRING, {.str = "44100"}, 0, 0, FLAGS },
94 { "duration", "set audio duration", OFFSET(duration), AV_OPT_TYPE_DURATION, {.i64 = -1}, -1, INT64_MAX, FLAGS },
95 { "d", "set audio duration", OFFSET(duration), AV_OPT_TYPE_DURATION, {.i64 = -1}, -1, INT64_MAX, FLAGS },
96 { "channel_layout", "set channel layout", OFFSET(chlayout_str), AV_OPT_TYPE_STRING, {.str = NULL}, 0, 0, FLAGS },
97 { "c", "set channel layout", OFFSET(chlayout_str), AV_OPT_TYPE_STRING, {.str = NULL}, 0, 0, FLAGS },
98 { NULL }
99 };
100
101 AVFILTER_DEFINE_CLASS(aevalsrc);
102
parse_channel_expressions(AVFilterContext * ctx,int expected_nb_channels)103 static int parse_channel_expressions(AVFilterContext *ctx,
104 int expected_nb_channels)
105 {
106 EvalContext *eval = ctx->priv;
107 char *args1 = av_strdup(eval->exprs);
108 char *expr, *last_expr = NULL, *buf;
109 double (* const *func1)(void *, double) = NULL;
110 const char * const *func1_names = NULL;
111 int i, ret = 0;
112
113 if (!args1)
114 return AVERROR(ENOMEM);
115
116 if (!eval->exprs) {
117 av_log(ctx, AV_LOG_ERROR, "Channels expressions list is empty\n");
118 return AVERROR(EINVAL);
119 }
120
121 if (!strcmp(ctx->filter->name, "aeval")) {
122 func1 = aeval_func1;
123 func1_names = aeval_func1_names;
124 }
125
126 #define ADD_EXPRESSION(expr_) do { \
127 if (!av_dynarray2_add((void **)&eval->expr, &eval->nb_channels, \
128 sizeof(*eval->expr), NULL)) { \
129 ret = AVERROR(ENOMEM); \
130 goto end; \
131 } \
132 eval->expr[eval->nb_channels-1] = NULL; \
133 ret = av_expr_parse(&eval->expr[eval->nb_channels - 1], expr_, \
134 var_names, func1_names, func1, \
135 NULL, NULL, 0, ctx); \
136 if (ret < 0) \
137 goto end; \
138 } while (0)
139
140 /* reset expressions */
141 for (i = 0; i < eval->nb_channels; i++) {
142 av_expr_free(eval->expr[i]);
143 eval->expr[i] = NULL;
144 }
145 av_freep(&eval->expr);
146 eval->nb_channels = 0;
147
148 buf = args1;
149 while (expr = av_strtok(buf, "|", &buf)) {
150 ADD_EXPRESSION(expr);
151 last_expr = expr;
152 }
153
154 if (expected_nb_channels > eval->nb_channels)
155 for (i = eval->nb_channels; i < expected_nb_channels; i++)
156 ADD_EXPRESSION(last_expr);
157
158 if (expected_nb_channels > 0 && eval->nb_channels != expected_nb_channels) {
159 av_log(ctx, AV_LOG_ERROR,
160 "Mismatch between the specified number of channel expressions '%d' "
161 "and the number of expected output channels '%d' for the specified channel layout\n",
162 eval->nb_channels, expected_nb_channels);
163 ret = AVERROR(EINVAL);
164 goto end;
165 }
166
167 end:
168 av_free(args1);
169 return ret;
170 }
171
init(AVFilterContext * ctx)172 static av_cold int init(AVFilterContext *ctx)
173 {
174 EvalContext *eval = ctx->priv;
175 int ret = 0;
176
177 if (eval->chlayout_str) {
178 if (!strcmp(eval->chlayout_str, "same") && !strcmp(ctx->filter->name, "aeval")) {
179 eval->same_chlayout = 1;
180 } else {
181 ret = ff_parse_channel_layout(&eval->chlayout, NULL, eval->chlayout_str, ctx);
182 if (ret < 0)
183 return ret;
184
185 ret = parse_channel_expressions(ctx, av_get_channel_layout_nb_channels(eval->chlayout));
186 if (ret < 0)
187 return ret;
188 }
189 } else {
190 /* guess channel layout from nb expressions/channels */
191 if ((ret = parse_channel_expressions(ctx, -1)) < 0)
192 return ret;
193
194 eval->chlayout = av_get_default_channel_layout(eval->nb_channels);
195 if (!eval->chlayout && eval->nb_channels <= 0) {
196 av_log(ctx, AV_LOG_ERROR, "Invalid number of channels '%d' provided\n",
197 eval->nb_channels);
198 return AVERROR(EINVAL);
199 }
200 }
201
202 if (eval->sample_rate_str)
203 if ((ret = ff_parse_sample_rate(&eval->sample_rate, eval->sample_rate_str, ctx)))
204 return ret;
205 eval->n = 0;
206
207 return ret;
208 }
209
uninit(AVFilterContext * ctx)210 static av_cold void uninit(AVFilterContext *ctx)
211 {
212 EvalContext *eval = ctx->priv;
213 int i;
214
215 for (i = 0; i < eval->nb_channels; i++) {
216 av_expr_free(eval->expr[i]);
217 eval->expr[i] = NULL;
218 }
219 av_freep(&eval->expr);
220 av_freep(&eval->channel_values);
221 }
222
config_props(AVFilterLink * outlink)223 static int config_props(AVFilterLink *outlink)
224 {
225 EvalContext *eval = outlink->src->priv;
226 char buf[128];
227
228 outlink->time_base = (AVRational){1, eval->sample_rate};
229 outlink->sample_rate = eval->sample_rate;
230
231 eval->var_values[VAR_S] = eval->sample_rate;
232 eval->var_values[VAR_NB_IN_CHANNELS] = NAN;
233 eval->var_values[VAR_NB_OUT_CHANNELS] = outlink->channels;
234
235 av_get_channel_layout_string(buf, sizeof(buf), 0, eval->chlayout);
236
237 av_log(outlink->src, AV_LOG_VERBOSE,
238 "sample_rate:%d chlayout:%s duration:%"PRId64"\n",
239 eval->sample_rate, buf, eval->duration);
240
241 return 0;
242 }
243
query_formats(AVFilterContext * ctx)244 static int query_formats(AVFilterContext *ctx)
245 {
246 EvalContext *eval = ctx->priv;
247 static const enum AVSampleFormat sample_fmts[] = { AV_SAMPLE_FMT_DBLP, AV_SAMPLE_FMT_NONE };
248 int64_t chlayouts[] = { eval->chlayout ? eval->chlayout : FF_COUNT2LAYOUT(eval->nb_channels) , -1 };
249 int sample_rates[] = { eval->sample_rate, -1 };
250 AVFilterFormats *formats;
251 AVFilterChannelLayouts *layouts;
252 int ret;
253
254 formats = ff_make_format_list(sample_fmts);
255 if (!formats)
256 return AVERROR(ENOMEM);
257 ret = ff_set_common_formats (ctx, formats);
258 if (ret < 0)
259 return ret;
260
261 layouts = ff_make_format64_list(chlayouts);
262 if (!layouts)
263 return AVERROR(ENOMEM);
264 ret = ff_set_common_channel_layouts(ctx, layouts);
265 if (ret < 0)
266 return ret;
267
268 formats = ff_make_format_list(sample_rates);
269 if (!formats)
270 return AVERROR(ENOMEM);
271 return ff_set_common_samplerates(ctx, formats);
272 }
273
request_frame(AVFilterLink * outlink)274 static int request_frame(AVFilterLink *outlink)
275 {
276 EvalContext *eval = outlink->src->priv;
277 AVFrame *samplesref;
278 int i, j;
279 int64_t t = av_rescale(eval->n, AV_TIME_BASE, eval->sample_rate);
280 int nb_samples;
281
282 if (eval->duration >= 0 && t >= eval->duration)
283 return AVERROR_EOF;
284
285 if (eval->duration >= 0) {
286 nb_samples = FFMIN(eval->nb_samples, av_rescale(eval->duration, eval->sample_rate, AV_TIME_BASE) - eval->pts);
287 if (!nb_samples)
288 return AVERROR_EOF;
289 } else {
290 nb_samples = eval->nb_samples;
291 }
292 samplesref = ff_get_audio_buffer(outlink, nb_samples);
293 if (!samplesref)
294 return AVERROR(ENOMEM);
295
296 /* evaluate expression for each single sample and for each channel */
297 for (i = 0; i < nb_samples; i++, eval->n++) {
298 eval->var_values[VAR_N] = eval->n;
299 eval->var_values[VAR_T] = eval->var_values[VAR_N] * (double)1/eval->sample_rate;
300
301 for (j = 0; j < eval->nb_channels; j++) {
302 *((double *) samplesref->extended_data[j] + i) =
303 av_expr_eval(eval->expr[j], eval->var_values, NULL);
304 }
305 }
306
307 samplesref->pts = eval->pts;
308 samplesref->sample_rate = eval->sample_rate;
309 eval->pts += nb_samples;
310
311 return ff_filter_frame(outlink, samplesref);
312 }
313
314 #if CONFIG_AEVALSRC_FILTER
315 static const AVFilterPad aevalsrc_outputs[] = {
316 {
317 .name = "default",
318 .type = AVMEDIA_TYPE_AUDIO,
319 .config_props = config_props,
320 .request_frame = request_frame,
321 },
322 { NULL }
323 };
324
325 AVFilter ff_asrc_aevalsrc = {
326 .name = "aevalsrc",
327 .description = NULL_IF_CONFIG_SMALL("Generate an audio signal generated by an expression."),
328 .query_formats = query_formats,
329 .init = init,
330 .uninit = uninit,
331 .priv_size = sizeof(EvalContext),
332 .inputs = NULL,
333 .outputs = aevalsrc_outputs,
334 .priv_class = &aevalsrc_class,
335 };
336
337 #endif /* CONFIG_AEVALSRC_FILTER */
338
339 #define OFFSET(x) offsetof(EvalContext, x)
340 #define FLAGS AV_OPT_FLAG_AUDIO_PARAM|AV_OPT_FLAG_FILTERING_PARAM
341
342 static const AVOption aeval_options[]= {
343 { "exprs", "set the '|'-separated list of channels expressions", OFFSET(exprs), AV_OPT_TYPE_STRING, {.str = NULL}, .flags = FLAGS },
344 { "channel_layout", "set channel layout", OFFSET(chlayout_str), AV_OPT_TYPE_STRING, {.str = NULL}, 0, 0, FLAGS },
345 { "c", "set channel layout", OFFSET(chlayout_str), AV_OPT_TYPE_STRING, {.str = NULL}, 0, 0, FLAGS },
346 { NULL }
347 };
348
349 AVFILTER_DEFINE_CLASS(aeval);
350
aeval_query_formats(AVFilterContext * ctx)351 static int aeval_query_formats(AVFilterContext *ctx)
352 {
353 AVFilterFormats *formats = NULL;
354 AVFilterChannelLayouts *layouts;
355 AVFilterLink *inlink = ctx->inputs[0];
356 AVFilterLink *outlink = ctx->outputs[0];
357 EvalContext *eval = ctx->priv;
358 static const enum AVSampleFormat sample_fmts[] = {
359 AV_SAMPLE_FMT_DBLP, AV_SAMPLE_FMT_NONE
360 };
361 int ret;
362
363 // inlink supports any channel layout
364 layouts = ff_all_channel_counts();
365 if ((ret = ff_channel_layouts_ref(layouts, &inlink->outcfg.channel_layouts)) < 0)
366 return ret;
367
368 if (eval->same_chlayout) {
369 layouts = ff_all_channel_counts();
370 if ((ret = ff_set_common_channel_layouts(ctx, layouts)) < 0)
371 return ret;
372 } else {
373 // outlink supports only requested output channel layout
374 layouts = NULL;
375 if ((ret = ff_add_channel_layout(&layouts,
376 eval->out_channel_layout ? eval->out_channel_layout :
377 FF_COUNT2LAYOUT(eval->nb_channels))) < 0)
378 return ret;
379 if ((ret = ff_channel_layouts_ref(layouts, &outlink->incfg.channel_layouts)) < 0)
380 return ret;
381 }
382
383 formats = ff_make_format_list(sample_fmts);
384 if ((ret = ff_set_common_formats(ctx, formats)) < 0)
385 return ret;
386
387 formats = ff_all_samplerates();
388 return ff_set_common_samplerates(ctx, formats);
389 }
390
aeval_config_output(AVFilterLink * outlink)391 static int aeval_config_output(AVFilterLink *outlink)
392 {
393 AVFilterContext *ctx = outlink->src;
394 EvalContext *eval = ctx->priv;
395 AVFilterLink *inlink = ctx->inputs[0];
396 int ret;
397
398 if (eval->same_chlayout) {
399 eval->chlayout = inlink->channel_layout;
400
401 if ((ret = parse_channel_expressions(ctx, inlink->channels)) < 0)
402 return ret;
403 }
404
405 eval->n = 0;
406 eval->nb_in_channels = eval->var_values[VAR_NB_IN_CHANNELS] = inlink->channels;
407 eval->var_values[VAR_NB_OUT_CHANNELS] = outlink->channels;
408 eval->var_values[VAR_S] = inlink->sample_rate;
409 eval->var_values[VAR_T] = NAN;
410
411 eval->channel_values = av_realloc_f(eval->channel_values,
412 inlink->channels, sizeof(*eval->channel_values));
413 if (!eval->channel_values)
414 return AVERROR(ENOMEM);
415
416 return 0;
417 }
418
filter_frame(AVFilterLink * inlink,AVFrame * in)419 static int filter_frame(AVFilterLink *inlink, AVFrame *in)
420 {
421 EvalContext *eval = inlink->dst->priv;
422 AVFilterLink *outlink = inlink->dst->outputs[0];
423 int nb_samples = in->nb_samples;
424 AVFrame *out;
425 double t0;
426 int i, j;
427
428 out = ff_get_audio_buffer(outlink, nb_samples);
429 if (!out) {
430 av_frame_free(&in);
431 return AVERROR(ENOMEM);
432 }
433 av_frame_copy_props(out, in);
434
435 t0 = TS2T(in->pts, inlink->time_base);
436
437 /* evaluate expression for each single sample and for each channel */
438 for (i = 0; i < nb_samples; i++, eval->n++) {
439 eval->var_values[VAR_N] = eval->n;
440 eval->var_values[VAR_T] = t0 + i * (double)1/inlink->sample_rate;
441
442 for (j = 0; j < inlink->channels; j++)
443 eval->channel_values[j] = *((double *) in->extended_data[j] + i);
444
445 for (j = 0; j < outlink->channels; j++) {
446 eval->var_values[VAR_CH] = j;
447 *((double *) out->extended_data[j] + i) =
448 av_expr_eval(eval->expr[j], eval->var_values, eval);
449 }
450 }
451
452 av_frame_free(&in);
453 return ff_filter_frame(outlink, out);
454 }
455
456 #if CONFIG_AEVAL_FILTER
457
458 static const AVFilterPad aeval_inputs[] = {
459 {
460 .name = "default",
461 .type = AVMEDIA_TYPE_AUDIO,
462 .filter_frame = filter_frame,
463 },
464 { NULL }
465 };
466
467 static const AVFilterPad aeval_outputs[] = {
468 {
469 .name = "default",
470 .type = AVMEDIA_TYPE_AUDIO,
471 .config_props = aeval_config_output,
472 },
473 { NULL }
474 };
475
476 AVFilter ff_af_aeval = {
477 .name = "aeval",
478 .description = NULL_IF_CONFIG_SMALL("Filter audio signal according to a specified expression."),
479 .query_formats = aeval_query_formats,
480 .init = init,
481 .uninit = uninit,
482 .priv_size = sizeof(EvalContext),
483 .inputs = aeval_inputs,
484 .outputs = aeval_outputs,
485 .priv_class = &aeval_class,
486 .flags = AVFILTER_FLAG_SUPPORT_TIMELINE_GENERIC,
487 };
488
489 #endif /* CONFIG_AEVAL_FILTER */
490