ref: d3f0b58897fb47913449b7489c2cef7ab9d0f510
dir: /src/sox.c/
/*
* Sox - The Swiss Army Knife of Audio Manipulation.
*
* This is the main function for the command line sox program.
*
* Copyright 1991 Lance Norskog And Sundry Contributors
* Copyright 1998-2006 Chris Bagnall and SoX contributors
*
* This library is free software; you can redistribute it and/or
* modify it under the terms of the GNU General Public License as
* published by the Free Software Foundation; either version 2 of the
* License, or (at your option) any later version.
*
* This library is distributed in the hope that it will be useful,
* but WITHOUT ANY WARRANTY; without even the implied warranty of
* MERCHANTABILITY or FITNESS FOR A PARTICULAR PURPOSE. See the GNU
* General Public License for more details.
*
* You should have received a copy of the GNU General Public License
* along with this library. If not, write to the Free Software
* Foundation, Fifth Floor, 51 Franklin Street, Boston, MA 02111-1301,
* USA.
*
*
* Original copyright notice:
*
* July 5, 1991
* Copyright 1991 Lance Norskog And Sundry Contributors
* This source code is freely redistributable and may be used for
* any purpose. This copyright notice must be maintained.
* Lance Norskog And Sundry Contributors are not responsible for
* the consequences of using this software.
*
*/
#include "st_i.h"
#include <math.h>
#include <stdio.h>
#include <string.h>
#include <stdlib.h> /* for malloc() */
#include <signal.h>
#ifdef HAVE_ERRNO_H
#include <errno.h>
#endif
#ifdef HAVE_UNISTD_H
#include <unistd.h> /* for unlink() */
#endif
#ifdef HAVE_GETOPT_LONG
#include <getopt.h>
#else
#include "getopt.h"
#endif
#include <sys/types.h> /* for fstat() */
#include <sys/stat.h> /* for fstat() */
#ifdef _MSC_VER
/*
* __STDC__ is defined, so these symbols aren't created.
*/
#define S_IFMT _S_IFMT
#define S_IFREG _S_IFREG
#define fstat _fstat
#define strdup _strdup
#endif
/*
* SOX main program.
*
* Rewrite for new nicer option syntax. July 13, 1991.
* Rewrite for separate effects library. Sep. 15, 1991.
* Incorporate Jimen Ching's fixes for real library operation: Aug 3, 1994.
* Rewrite for multiple effects: Aug 24, 1994.
*/
static int soxmix = 0; /* non-zero if running as soxmix */
static int clipped = 0; /* Volume change clipping errors */
static int writing = 1; /* are we writing to a file? assume yes. */
static int soxpreview = 0; /* preview mode */
static st_globalinfo_t globalinfo;
static int user_abort = 0;
static int quiet = 0;
static int status = 0;
static unsigned long input_samples = 0;
static unsigned long read_samples = 0;
static unsigned long output_samples = 0;
static st_sample_t ibufl[ST_BUFSIZ/2]; /* Left/right interleave buffers */
static st_sample_t ibufr[ST_BUFSIZ/2];
static st_sample_t obufl[ST_BUFSIZ/2];
static st_sample_t obufr[ST_BUFSIZ/2];
typedef struct file_options
{
char *filename;
char *filetype;
st_signalinfo_t info;
double volume;
char uservolume;
char *comment;
} file_options_t;
/* local forward declarations */
static void doopts(file_options_t *fo, int, char **);
static void usage(char *) NORET;
static void usage_effect(char *) NORET;
static void process(void);
static void print_input_status(int input);
static void update_status(void);
static void statistics(void);
static st_sample_t volumechange(st_sample_t *buf, st_ssize_t ct, double vol);
static void parse_effects(int argc, char **argv);
static void check_effects(void);
static int start_effects(void);
static void reserve_effect_buf(void);
static int flow_effect_out(void);
static int flow_effect(int);
static int drain_effect_out(void);
static int drain_effect(int);
static void release_effect_buf(void);
static void stop_effects(void);
void cleanup(void);
static void sigint(int s);
#define MAX_INPUT_FILES 32
#define MAX_FILES MAX_INPUT_FILES + 1
/* Array's tracking input and output files */
static file_options_t *file_opts[MAX_FILES];
static ft_t file_desc[MAX_FILES];
static int file_count = 0;
static int input_count = 0;
/* We parse effects into a temporary effects table and then place into
* the real effects table. This makes it easier to reorder some effects
* as needed. For instance, we can run a resampling effect before
* converting a mono file to stereo. This allows the resample to work
* on half the data.
*
* Real effects table only needs to be 2 entries bigger then the user
* specified table. This is because at most we will need to add
* a resample effect and a channel averaging effect.
*/
#define MAX_EFF 16
#define MAX_USER_EFF 14
/*
* efftab[0] is a dummy entry used only as an input buffer for
* reading input data into.
*
* If one was to support effects for quad-channel files, there would
* need to be an effect table for each channel to handle effects
* that don't set ST_EFF_MCHAN.
*/
static struct st_effect efftab[MAX_EFF]; /* left/mono channel effects */
static struct st_effect efftabR[MAX_EFF];/* right channel effects */
static int neffects; /* # of effects to run on data */
static int input_eff; /* last input effect with data */
static int input_eff_eof; /* has input_eff reached EOF? */
static struct st_effect user_efftab[MAX_USER_EFF];
static int nuser_effects;
int main(int argc, char **argv)
{
file_options_t *fo;
int i;
myname = argv[0];
i = strlen(myname);
if (i >= sizeof("soxmix") - 1)
soxmix = strcmp(myname + i - (sizeof("soxmix") - 1), "soxmix") == 0;
/* Loop over arguments and filenames, stop when an effect name is
* found.
*/
while (optind < argc && st_checkeffect(argv[optind]) != ST_SUCCESS)
{
if (file_count >= MAX_FILES)
{
st_fail("to many filenames. max of %d input files and 1 output file\n", MAX_INPUT_FILES);
exit(1);
}
/*
* Its possible to not specify the output filename by using
* "-e" option. This allows effects to be ran on data but
* no output file to be written.
*/
if (strcmp(argv[optind], "-e") == 0)
{
/* -e option found. Only thing valid after an -e
* option are effects.
*/
optind++;
if (optind >= argc ||
st_checkeffect(argv[optind]) == ST_SUCCESS)
{
writing = 0;
}
else
{
usage("Can only specify \"-e\" for output filenames");
}
}
else
{
fo = (file_options_t *)calloc(sizeof(file_options_t), 1);
fo->info.size = -1;
fo->info.encoding = -1;
fo->info.channels = -1;
fo->info.compression = HUGE_VAL;
fo->volume = 1.0;
file_opts[file_count++] = fo;
doopts(fo, argc, argv);
if (optind < argc)
{
fo->filename = strdup(argv[optind]);
optind++;
}
else
{
usage("missing filename");
}
}
} /* while (commandline options) */
if (writing)
input_count = file_count - 1;
else
input_count = file_count;
/* Make sure we got at least the required # of input filenames */
if (input_count < soxmix ? 2 : 1)
usage("Not enough input or output filenames specified");
for (i = 0; i < input_count; i++)
{
if (soxmix) {
/* When mixing audio, default to input side volume
* adjustments that will make sure no clipping will
* occur. Users most likely won't be happy with
* this and will want to override it.
*/
if (!file_opts[i]->uservolume)
file_opts[i]->volume = 1.0 / input_count;
}
file_desc[i] = st_open_read(file_opts[i]->filename,
&file_opts[i]->info,
file_opts[i]->filetype);
if (!file_desc[i])
{
/* st_open_read() will call st_warn for most errors.
* Rely on that printing something.
*/
exit(2);
}
if (file_opts[i]->info.compression != HUGE_VAL)
{
st_fail("-C only allowed for output file");
cleanup();
exit(1);
}
}
/* Loop through the reset of the arguments looking for effects */
parse_effects(argc, argv);
signal(SIGINT, sigint);
signal(SIGTERM, sigint);
process();
statistics();
for (i = 0; i < file_count; i++)
free(file_desc[i]);
if (status)
{
if (user_abort)
fprintf(stderr, "\n\nAborted.\n");
else
fprintf(stderr, "\n\nDone.\n");
}
return(0);
}
static char *getoptstr = "+r:v:t:c:C:phsuUAaig1b2w34lfdxVSqo";
static struct option long_options[] =
{
{"version", 0, NULL, 0},
{"help", 0, NULL, 'h'},
{"help-effect", 1, NULL, 0},
{NULL, 0, NULL, 0}
};
static void doopts(file_options_t *fo, int argc, char **argv)
{
int c, i;
int option_index;
char *str;
while ((c = getopt_long(argc, argv, getoptstr,
long_options, &option_index)) != -1) {
switch(c) {
case 0:
if (strncmp("help-effect", long_options[option_index].name,
11) == 0)
usage_effect(optarg);
else if (strncmp("version", long_options[option_index].name,
7) == 0)
{
printf("%s: ", myname);
printf("v%s\n", st_version());
exit(0);
}
/* no return from above */
break;
case 'o':
globalinfo.octave_plot_effect = true;
break;
case 'p':
soxpreview++;
break;
case 'h':
usage((char *)0);
/* no return from above */
break;
case 't':
fo->filetype = optarg;
if (fo->filetype[0] == '.')
fo->filetype++;
break;
case 'r':
str = optarg;
if ((!sscanf(optarg, "%u", &fo->info.rate)) ||
(fo->info.rate <= 0))
{
st_fail("-r must be given a positive integer");
cleanup();
exit(1);
}
break;
case 'v':
str = optarg;
if (!sscanf(str, "%lf", &fo->volume))
{
st_fail("Volume value '%s' is not a number",
optarg);
cleanup();
exit(1);
}
fo->uservolume = 1;
if (fo->volume < 0.0)
st_report("Volume adjustment is negative. This will result in a phase change\n");
break;
case 'c':
str = optarg;
if (!sscanf(str, "%d", &i))
{
st_fail("-c must be given a number");
cleanup();
exit(1);
}
/* Since we use -1 as a special internal value,
* we must do some extra logic so user doesn't
* get confused when we translate -1 to mean
* something valid.
*/
if (i < 1)
{
st_fail("-c must be given a positive number");
cleanup();
exit(1);
}
fo->info.channels = i;
break;
case 'C':
str = optarg;
if (!sscanf(str, "%lf", &fo->info.compression))
{
st_fail("-C must be given a number");
cleanup();
exit(1);
}
break;
case '1': case 'b':
fo->info.size = ST_SIZE_BYTE;
break;
case '2': case 'w':
fo->info.size = ST_SIZE_WORD;
break;
case '3':
fo->info.size = ST_SIZE_24BIT;
break;
case '4': case 'l':
fo->info.size = ST_SIZE_DWORD;
break;
case 'd':
fo->info.size = ST_SIZE_DDWORD;
break;
case 's':
fo->info.encoding = ST_ENCODING_SIGN2;
break;
case 'u':
fo->info.encoding = ST_ENCODING_UNSIGNED;
break;
case 'U':
fo->info.encoding = ST_ENCODING_ULAW;
if (fo->info.size == -1)
fo->info.size = ST_SIZE_BYTE;
break;
case 'A':
fo->info.encoding = ST_ENCODING_ALAW;
if (fo->info.size == -1)
fo->info.size = ST_SIZE_BYTE;
break;
case 'f':
fo->info.encoding = ST_ENCODING_FLOAT;
break;
case 'a':
fo->info.encoding = ST_ENCODING_ADPCM;
break;
case 'i':
fo->info.encoding = ST_ENCODING_IMA_ADPCM;
break;
case 'g':
fo->info.encoding = ST_ENCODING_GSM;
break;
case 'x':
fo->info.swap = 1;
break;
case 'V':
verbose = 1;
break;
case 'S':
status = 1;
quiet = 0;
break;
case 'q':
status = 0;
quiet = 1;
break;
case '?':
usage((char *)0);
/* no return from above */
break;
}
}
}
static int compare_input(ft_t ft1, ft_t ft2)
{
if (ft1->info.rate != ft2->info.rate)
return ST_EOF;
if (ft1->info.size != ft2->info.size)
return ST_EOF;
if (ft1->info.encoding != ft2->info.encoding)
return ST_EOF;
if (ft1->info.channels != ft2->info.channels)
return ST_EOF;
return ST_SUCCESS;
}
void optimize_trim(void)
{
/* Speed hack. If the "trim" effect is the first effect then
* peak inside its "effect descriptor" and see what the
* start location is. This has to be done after its start()
* is called to have the correct location.
* Also, only do this when only working with one input file.
* This is because the logic to do it for multiple files is
* complex and problably never used.
*/
if (input_count == 1 && neffects > 1 &&
strcmp(efftab[1].name, "trim") == 0)
{
if ((file_desc[0]->h->flags & ST_FILE_SEEK) && file_desc[0]->seekable)
{
if (st_seek(file_desc[0], st_trim_get_start(&efftab[1]),
ST_SEEK_SET) != ST_EOF)
{
/* Assuming a failed seek stayed where it was. If the
* seek worked then reset the start location of
* trim so that it thinks user didn't request a skip.
*/
st_trim_clear_start(&efftab[1]);
}
}
}
}
/*
* Process input file -> effect table -> output file
* one buffer at a time
*/
static void process(void) {
int e, f, flowstatus = ST_SUCCESS;
int current_input;
st_size_t s;
st_ssize_t ilen[MAX_INPUT_FILES];
st_sample_t *ibuf[MAX_INPUT_FILES];
for (f = 0; f < input_count; f++)
{
st_report("Input file %s: using sample rate %lu\n\tsize %s, encoding %s, %d %s",
file_desc[f]->filename, file_desc[f]->info.rate,
st_sizes_str[(unsigned char)file_desc[f]->info.size],
st_encodings_str[(unsigned char)file_desc[f]->info.encoding],
file_desc[f]->info.channels,
(file_desc[f]->info.channels > 1) ? "channels" : "channel");
if (file_desc[f]->comment)
st_report("Input file %s: comment \"%s\"\n",
file_desc[f]->filename, file_desc[f]->comment);
}
for (f = 1; f < input_count; f++)
{
if (compare_input(file_desc[0], file_desc[f]) != ST_SUCCESS)
{
st_fail("Input files must have the same rate, channels, data size, and encoding");
exit(1);
}
}
if (writing)
{
st_loopinfo_t loops[ST_MAX_NLOOPS];
double factor;
int i;
if (file_opts[file_count-1]->info.rate == 0)
file_opts[file_count-1]->info.rate = file_desc[0]->info.rate;
if (file_opts[file_count-1]->info.size == -1)
file_opts[file_count-1]->info.size = file_desc[0]->info.size;
if (file_opts[file_count-1]->info.encoding == -1)
file_opts[file_count-1]->info.encoding =
file_desc[0]->info.encoding;
if (file_opts[file_count-1]->info.channels == -1)
file_opts[file_count-1]->info.channels =
file_desc[0]->info.channels;
/*
* copy loop info, resizing appropriately
* it's in samples, so # channels don't matter
* FIXME: This doesn't work for multi-file processing or
* effects that change file length.
*/
factor = (double) file_opts[file_count-1]->info.rate / (double)
file_desc[0]->info.rate;
for(i = 0; i < ST_MAX_NLOOPS; i++) {
loops[i].start = file_desc[0]->loops[i].start * factor;
loops[i].length = file_desc[0]->loops[i].length * factor;
loops[i].count = file_desc[0]->loops[i].count;
loops[i].type = file_desc[0]->loops[i].type;
}
file_desc[file_count-1] =
st_open_write_instr(file_opts[file_count-1]->filename,
&file_opts[file_count-1]->info,
file_opts[file_count-1]->filetype,
file_desc[0]->comment,
&file_desc[0]->instr,
loops);
if (!file_desc[file_count-1])
{
/* st_open_write() will call st_warn for most errors.
* Rely on that printing something.
*/
cleanup();
exit(2);
}
/* When writing to an audio device, auto turn on the
* status display to match behavior of ogg123 status.
* That is unless user requested us not to display]
* anything.
*/
if (strcmp(file_desc[file_count-1]->filetype, "alsa") == 0 ||
strcmp(file_desc[file_count-1]->filetype, "ossdsp") == 0 ||
strcmp(file_desc[file_count-1]->filetype, "sunau") == 0)
{
if (!quiet)
status = 1;
}
st_report("Output file %s: using sample rate %lu\n\tsize %s, encoding %s, %d %s",
file_desc[file_count-1]->filename,
file_desc[file_count-1]->info.rate,
st_sizes_str[(unsigned char)file_desc[file_count-1]->info.size],
st_encodings_str[(unsigned char)file_desc[file_count-1]->info.encoding],
file_desc[file_count-1]->info.channels,
(file_desc[file_count-1]->info.channels > 1) ? "channels" : "channel");
if (file_desc[file_count-1]->comment)
st_report("Output file: comment \"%s\"\n",
file_desc[file_count-1]->comment);
}
/* build efftab */
check_effects();
/* Start all effects */
flowstatus = start_effects();
/* Reserve an output buffer for all effects */
reserve_effect_buf();
/* Try to save some time if first effect is "trim" by seeking */
optimize_trim();
if (!soxmix) {
for (f = 0; f < input_count; f++)
{
/* Treat overall length the same as longest input file. */
if (file_desc[f]->length > input_samples)
input_samples = file_desc[f]->length;
ibuf[f] = (st_sample_t *)malloc(ST_BUFSIZ * sizeof(st_sample_t));
if (!ibuf[f])
{
st_fail("could not allocate memory");
cleanup();
exit(1);
}
if (status)
print_input_status(f);
}
} else {
current_input = 0;
input_samples = file_desc[current_input]->length;
if (status)
print_input_status(current_input);
}
/*
* Just like errno, we must set st_errno to known values before
* calling I/O operations.
*/
for (f = 0; f < file_count; f++)
file_desc[f]->st_errno = 0;
input_eff = 0;
input_eff_eof = 0;
/* mark chain as empty */
for(e = 1; e < neffects; e++)
efftab[e].odone = efftab[e].olen = 0;
/* If start functions set flowstatus to ST_EOF, skip both flow and
drain; we have to have this "if" because after flow flowstatus is
supposed to be ST_EOF, so we can't test that in order to know
whether to drain. */
if (flowstatus == 0) {
/* Run input data through effects and get more until olen == 0
* (or ST_EOF).
*/
do {
if (!soxmix) {
ilen[0] = st_read(file_desc[current_input], efftab[0].obuf,
(st_ssize_t)ST_BUFSIZ);
if (ilen[0] > ST_BUFSIZ)
{
st_warn("WARNING: Corrupt value of %d! Assuming 0 bytes read.\n", ilen);
ilen[0] = 0;
}
if (ilen[0] == ST_EOF)
{
efftab[0].olen = 0;
if (file_desc[current_input]->st_errno)
{
fprintf(stderr, file_desc[current_input]->st_errstr);
}
}
else
efftab[0].olen = ilen[0];
read_samples += efftab[0].olen;
/* Some file handlers claim 0 bytes instead of returning
* ST_EOF. In either case, attempt to go to the next
* input file.
*/
if (ilen[0] == ST_EOF || efftab[0].olen == 0)
{
if (current_input < input_count-1)
{
current_input++;
input_samples = file_desc[current_input]->length;
read_samples = 0;
if (status)
print_input_status(current_input);
continue;
}
}
/* Adjust input side volume based on value specified
* by user for this file.
*/
if (file_opts[current_input]->volume != 1.0)
clipped += volumechange(efftab[0].obuf,
efftab[0].olen,
file_opts[current_input]->volume);
} else /* soxmix */ {
for (f = 0; f < input_count; f++)
{
ilen[f] = st_read(file_desc[f], ibuf[f], (st_ssize_t)ST_BUFSIZ);
if (ilen[f] == ST_EOF)
{
ilen[f] = 0;
if (file_desc[f]->st_errno)
{
fprintf(stderr, file_desc[f]->st_errstr);
}
}
/* Only count read samples for first file in mix */
if (f == 0)
read_samples += efftab[0].olen;
/* Adjust input side volume based on value specified
* by user for this file.
*/
if (file_opts[f]->volume != 1.0)
clipped += volumechange(ibuf[f],
ilen[f],
file_opts[f]->volume);
}
/* FIXME: Should report if the size of the reads are not
* the same.
*/
efftab[0].olen = 0;
for (f = 0; f < input_count; f++)
if ((st_size_t)ilen[f] > efftab[0].olen)
efftab[0].olen = ilen[f];
for (s = 0; s < efftab[0].olen; s++)
{
/* Mix data together by summing samples together.
* It is assumed that input side volume adjustments
* will take care of any possible overflow.
* By default, SoX sets the volume adjustment
* to 1/input_count but the user can override this.
* They probably will and some clipping will probably
* occur because of this.
*/
for (f = 0; f < input_count; f++)
{
if (f == 0)
efftab[0].obuf[s] =
(s<(st_size_t)ilen[f]) ? ibuf[f][s] : 0;
else
if (s < (st_size_t)ilen[f])
{
double sample;
sample = efftab[0].obuf[s] + ibuf[f][s];
ST_SAMPLE_CLIP_COUNT(sample, clipped);
efftab[0].obuf[s] = sample;
}
}
}
}
efftab[0].odone = 0;
/* If not writing and no effects are occuring then not much
* reason to continue reading. This allows this case. Mainly
* useful to print out info about input file header and quiet.
*/
if (!writing && neffects == 1)
efftab[0].olen = 0;
if (efftab[0].olen == 0)
break;
flowstatus = flow_effect_out();
if (status)
update_status();
/* Quit reading/writing on user aborts. This will close
* done the files nicely as if an EOF was reached on read.
*/
if (user_abort)
break;
/* If writing and there's an error, don't try to write more. */
if (writing && file_desc[file_count-1]->st_errno)
break;
} while (flowstatus == 0);
/* This will drain the effects */
/* Don't write if output is indicating errors. */
if (writing && file_desc[file_count-1]->st_errno == 0)
drain_effect_out();
}
if (soxmix)
/* Free input buffers now that they are not used */
for (f = 0; f < input_count; f++)
{
free(ibuf[f]);
}
/* Free output buffers now that they won't be used */
release_effect_buf();
/* Very Important:
* Effect stop is called BEFORE files close.
* Effect may write out more data after.
*/
stop_effects();
for (f = 0; f < input_count; f++)
{
/* If problems closing input file, just warn user since
* we are exiting anyways.
*/
if (st_close(file_desc[f]) == ST_EOF)
st_warn("%s", file_desc[f]->st_errstr);
}
if (writing)
{
/* problem closing output file, just warn user since we
* are exiting anyways.
*/
if (st_close(file_desc[file_count-1]) == ST_EOF)
st_warn("%s", file_desc[file_count-1]->st_errstr);
}
}
static void parse_effects(int argc, char **argv)
{
int argc_effect;
int effect_rc;
nuser_effects = 0;
while (optind < argc)
{
if (nuser_effects >= MAX_USER_EFF)
{
st_fail("Too many effects specified.\n");
cleanup();
exit(2);
}
argc_effect = st_geteffect_opt(&user_efftab[nuser_effects],
argc - optind, &argv[optind]);
if (argc_effect == ST_EOF)
{
st_fail("Effect '%s' is not known!", argv[optind]);
cleanup();
exit(2);
}
/* Skip past effect name */
optind++;
effect_rc = (*user_efftab[nuser_effects].h->getopts)
(&user_efftab[nuser_effects],
argc_effect,
&argv[optind]);
if (effect_rc == ST_EOF)
{
cleanup();
exit(2);
}
/* Skip past the effect arguments */
optind += argc_effect;
nuser_effects++;
}
}
/*
* If no effect given, decide what it should be.
* Smart ruleset for multiple effects in sequence.
* Puts user-specified effect in right place.
*/
static void check_effects(void)
{
int i;
int needchan = 0, needrate = 0, haschan = 0, hasrate = 0;
int effects_mask = 0;
int status;
if (writing)
{
needrate = (file_desc[0]->info.rate != file_desc[file_count-1]->info.rate);
needchan = (file_desc[0]->info.channels != file_desc[file_count-1]->info.channels);
}
for (i = 0; i < nuser_effects; i++)
{
user_efftab[i].globalinfo = globalinfo;
if (user_efftab[i].h->flags & ST_EFF_CHAN)
{
haschan++;
}
if (user_efftab[i].h->flags & ST_EFF_RATE)
{
hasrate++;
}
}
if (haschan > 1)
{
st_fail("Can not specify multiple effects that modify channel #");
cleanup();
exit(2);
}
if (hasrate > 1)
st_report("Can not specify multiple effects that change sample rate");
/* If not writing output then do not worry about adding
* channel and rate effects. This is just to speed things
* up.
*/
if (!writing)
{
needchan = 0;
needrate = 0;
}
/* --------- add the effects ------------------------ */
/* efftab[0] is always the input stream and always exists */
neffects = 1;
/* If reducing channels then its faster to run all effects
* after the avg effect.
*/
if (needchan && !(haschan) &&
(file_desc[0]->info.channels > file_desc[file_count-1]->info.channels))
{
/* Find effect and update initial pointers */
st_geteffect(&efftab[neffects], "avg");
/* give default opts for added effects */
status = (* efftab[neffects].h->getopts)(&efftab[neffects],(int)0,
(char **)0);
if (status == ST_EOF)
{
cleanup();
exit(2);
}
/* Copy format info to effect table */
effects_mask = st_updateeffect(&efftab[neffects],
&file_desc[0]->info,
&file_desc[file_count-1]->info,
effects_mask);
neffects++;
}
/* If reducing the number of samples, its faster to run all effects
* after the resample effect.
*/
if (needrate && !(hasrate) &&
(file_desc[0]->info.rate > file_desc[file_count-1]->info.rate))
{
if (soxpreview)
st_geteffect(&efftab[neffects], "rate");
else
st_geteffect(&efftab[neffects], "resample");
/* set up & give default opts for added effects */
status = (* efftab[neffects].h->getopts)(&efftab[neffects],(int)0,
(char **)0);
if (status == ST_EOF)
{
cleanup();
exit(2);
}
/* Copy format info to effect table */
effects_mask = st_updateeffect(&efftab[neffects],
&file_desc[0]->info,
&file_desc[file_count-1]->info,
effects_mask);
/* Rate can't handle multiple channels so be sure and
* account for that.
*/
if (efftab[neffects].ininfo.channels > 1)
{
memcpy(&efftabR[neffects], &efftab[neffects],
sizeof(struct st_effect));
}
neffects++;
}
/* Copy over user specified effects into real efftab */
for(i = 0; i < nuser_effects; i++)
{
memcpy(&efftab[neffects], &user_efftab[i],
sizeof(struct st_effect));
/* Copy format info to effect table */
effects_mask = st_updateeffect(&efftab[neffects],
&file_desc[0]->info,
&file_desc[file_count-1]->info,
effects_mask);
/* If this effect can't handle multiple channels then
* account for this.
*/
if ((efftab[neffects].ininfo.channels > 1) &&
!(efftab[neffects].h->flags & ST_EFF_MCHAN))
{
memcpy(&efftabR[neffects], &efftab[neffects],
sizeof(struct st_effect));
}
neffects++;
}
/* If rate effect hasn't been added by now then add it here.
* Check adding rate before avg because its faster to run
* rate on less channels then more.
*/
if (needrate && !(effects_mask & ST_EFF_RATE))
{
if (soxpreview)
st_geteffect(&efftab[neffects], "rate");
else
st_geteffect(&efftab[neffects], "resample");
/* set up & give default opts for added effects */
status = (* efftab[neffects].h->getopts)(&efftab[neffects],(int)0,
(char **)0);
if (status == ST_EOF)
{
cleanup();
exit(2);
}
/* Copy format info to effect table */
effects_mask = st_updateeffect(&efftab[neffects],
&file_desc[0]->info,
&file_desc[file_count-1]->info,
effects_mask);
/* Rate can't handle multiple channels so be sure and
* account for that.
*/
if (efftab[neffects].ininfo.channels > 1)
{
memcpy(&efftabR[neffects], &efftab[neffects],
sizeof(struct st_effect));
}
neffects++;
}
/* If code up until now still hasn't added avg effect then
* do it now.
*/
if (needchan && !(effects_mask & ST_EFF_CHAN))
{
st_geteffect(&efftab[neffects], "avg");
/* set up & give default opts for added effects */
status = (* efftab[neffects].h->getopts)(&efftab[neffects],(int)0,
(char **)0);
if (status == ST_EOF)
{
cleanup();
exit(2);
}
/* Copy format info to effect table */
effects_mask = st_updateeffect(&efftab[neffects],
&file_desc[0]->info,
&file_desc[file_count-1]->info,
effects_mask);
neffects++;
}
}
static int start_effects(void)
{
int e, ret = ST_SUCCESS;
for(e = 1; e < neffects; e++) {
if ((ret = (*efftab[e].h->start)(&efftab[e])) == ST_EOF)
break;
if (efftabR[e].name)
if ((ret = (*efftabR[e].h->start)(&efftabR[e])) == ST_EOF)
break;
}
return ret;
}
static void reserve_effect_buf(void)
{
int e;
for(e = 0; e < neffects; e++)
{
efftab[e].obuf = (st_sample_t *)malloc(ST_BUFSIZ *
sizeof(st_sample_t));
if (efftab[e].obuf == NULL)
{
st_fail("could not allocate memory");
cleanup();
exit(2);
}
if (efftabR[e].name)
{
efftabR[e].obuf = (st_sample_t *)malloc(ST_BUFSIZ *
sizeof(st_sample_t));
if (efftabR[e].obuf == NULL)
{
st_fail("could not allocate memory");
cleanup();
exit(2);
}
}
}
}
static int flow_effect_out(void)
{
int e, havedata, flowstatus = 0;
int len, total;
do {
/* run entire chain BACKWARDS: pull, don't push.*/
/* this is because buffering system isn't a nice queueing system */
for(e = neffects - 1; e >= input_eff; e--)
{
/* Do not call flow effect on input if its reported
* EOF already as thats a waste of time and may
* do bad things.
*/
if (e == input_eff && input_eff_eof)
continue;
/* flow_effect returns ST_EOF when it will not process
* any more samples. This is used to bail out early.
* Since we are "pulling" data, it is OK that we are not
* calling any more previous effects since their output
* would not be looked at anyways.
*/
flowstatus = flow_effect(e);
if (flowstatus == ST_EOF)
{
input_eff = e+1;
/* Assume next effect hasn't reach EOF yet */
input_eff_eof = 0;
}
/* If this buffer contains more input data then break out
* of this loop now. This will allow us to loop back around
* and reprocess the rest of this input buffer.
* I suppose this could be an issue with some effects
* if they crash when given small input buffers.
* But I was more concerned that we would need to do
* some type of garbage collection otherwise. By this I
* mean that if we went ahead and processed an effect
* lower in the chain, it might only have like 2 bytes
* left at the end of this buffer to place its data in.
* Software is more likely to refuse to handle that.
*/
if (efftab[e].odone < efftab[e].olen)
{
#if 0
fprintf(stderr, "Breaking out of loop to flush buffer\n");
#endif
break;
}
}
/* If outputing and output data was generated then write it */
if (writing && (efftab[neffects-1].olen>efftab[neffects-1].odone))
{
/* Change the volume of this output data if needed. */
if (writing && file_opts[file_count-1]->volume != 1.0)
clipped += volumechange(efftab[neffects-1].obuf,
efftab[neffects-1].olen,
file_opts[file_count-1]->volume);
total = 0;
do
{
/* Do not do any more writing during user aborts as
* we may be stuck in an infinite writing loop.
*/
if (user_abort)
return ST_EOF;
len = st_write(file_desc[file_count-1],
&efftab[neffects-1].obuf[total],
(st_ssize_t)efftab[neffects-1].olen-total);
if (len < 0 || file_desc[file_count-1]->file.eof)
{
st_warn("Error writing: %s",
file_desc[file_count-1]->st_errstr);
return ST_EOF;
}
total += len;
} while (total < efftab[neffects-1].olen);
output_samples += (total / file_desc[file_count-1]->info.channels);
efftab[neffects-1].odone = efftab[neffects-1].olen = 0;
}
else
{
/* Make it look like everything was consumed */
output_samples += (efftab[neffects-1].olen /
file_desc[file_count-1]->info.channels);
efftab[neffects-1].odone = efftab[neffects-1].olen = 0;
}
/* if stuff still in pipeline, set up to flow effects again */
/* When all effects have reported ST_EOF then this check will
* show no more data.
*/
havedata = 0;
for(e = neffects - 1; e >= input_eff; e--)
{
/* If odone and olen are the same then this buffer
* can be reused.
*/
if (efftab[e].odone == efftab[e].olen)
efftab[e].odone = efftab[e].olen = 0;
if (efftab[e].odone < efftab[e].olen)
{
/* Only mark that we have more data if a full
* frame that can be written.
* FIXME: If this error case happens for the
* input buffer then the data will be lost and
* will cause stereo channels to be inversed.
*/
if ((efftab[e].olen - efftab[e].odone) >=
file_desc[file_count-1]->info.channels)
havedata = 1;
else
st_warn("Received buffer with incomplete amount of samples.");
/* Don't break out because other things are being
* done in loop.
*/
}
}
if (!havedata && input_eff > 0)
{
/* When EOF has been detected, skip to the next input
* before looking for more data.
*/
if (input_eff_eof)
{
input_eff++;
input_eff_eof = 0;
}
/* If the input file is not returning data then
* we must prime the pump using the drain effect.
* After its primed, the loop will suck the data
* threw. Once an input_eff stop reporting samples,
* we will continue to the next until all are drained.
*/
while (input_eff < neffects)
{
int rc;
rc = drain_effect(input_eff);
if (efftab[input_eff].olen == 0)
{
input_eff++;
/* Assume next effect hasn't reached EOF yet. */
input_eff_eof = 0;
}
else
{
havedata = 1;
input_eff_eof = (rc == ST_EOF) ? 1 : 0;
break;
}
}
}
} while (havedata);
/* If input_eff isn't pointing at fake first entry then there
* is no need to read any more data from disk. Return this
* fact to caller.
*/
if (input_eff > 0)
{
#if 0
fprintf(stderr, "Effect return ST_EOF\n");
#endif
return ST_EOF;
}
return ST_SUCCESS;
}
static int flow_effect(int e)
{
st_ssize_t i, done, idone, odone, idonel, odonel, idoner, odoner;
st_sample_t *ibuf, *obuf;
int effstatus, effstatusl, effstatusr;
/* Do not attempt to do any more effect processing during
* user aborts as we may be stuck in an infinit flow loop.
*/
if (user_abort)
return ST_EOF;
/* I have no input data ? */
if (efftab[e-1].odone == efftab[e-1].olen)
{
#if 0
fprintf(stderr, "%s no data to pull to me!\n", efftab[e].name);
#endif
return 0;
}
if (! efftabR[e].name) {
/* No stereo data, or effect can handle stereo data so
* run effect over entire buffer.
*/
idone = efftab[e-1].olen - efftab[e-1].odone;
odone = ST_BUFSIZ - efftab[e].olen;
#if 0
fprintf(stderr, "pre %s idone=%d, odone=%d\n", efftab[e].name, idone, odone);
fprintf(stderr, "pre %s odone1=%d, olen1=%d odone=%d olen=%d\n", efftab[e].name, efftab[e-1].odone, efftab[e-1].olen, efftab[e].odone, efftab[e].olen);
#endif
effstatus = (* efftab[e].h->flow)(&efftab[e],
&efftab[e-1].obuf[efftab[e-1].odone],
&efftab[e].obuf[efftab[e].olen],
(st_size_t *)&idone,
(st_size_t *)&odone);
efftab[e-1].odone += idone;
/* Leave efftab[e].odone were it was since we didn't consume data */
/*efftab[e].odone = 0;*/
efftab[e].olen += odone;
#if 0
fprintf(stderr, "post %s idone=%d, odone=%d\n", efftab[e].name, idone, odone);
fprintf(stderr, "post %s odone1=%d, olen1=%d odone=%d olen=%d\n", efftab[e].name, efftab[e-1].odone, efftab[e-1].olen, efftab[e].odone, efftab[e].olen);
#endif
done = idone + odone;
}
else
{
/* Put stereo data in two seperate buffers and run effect
* on each of them.
*/
idone = efftab[e-1].olen - efftab[e-1].odone;
odone = ST_BUFSIZ - efftab[e].olen;
ibuf = &efftab[e-1].obuf[efftab[e-1].odone];
for(i = 0; i < idone; i += 2) {
ibufl[i/2] = *ibuf++;
ibufr[i/2] = *ibuf++;
}
/* left */
idonel = (idone + 1)/2; /* odd-length logic */
odonel = odone/2;
#if 0
fprintf(stderr, "pre %s idone=%d, odone=%d\n", efftab[e].name, idone, odone);
fprintf(stderr, "pre %s odone1=%d, olen1=%d odone=%d olen=%d\n", efftab[e].name, efftab[e-1].odone, efftab[e-1].olen, efftab[e].odone, efftab[e].olen);
#endif
effstatusl = (* efftab[e].h->flow)(&efftab[e],
ibufl, obufl, (st_size_t *)&idonel,
(st_size_t *)&odonel);
/* right */
idoner = idone/2; /* odd-length logic */
odoner = odone/2;
effstatusr = (* efftabR[e].h->flow)(&efftabR[e],
ibufr, obufr, (st_size_t *)&idoner,
(st_size_t *)&odoner);
obuf = &efftab[e].obuf[efftab[e].olen];
/* This loop implies left and right effect will always output
* the same amount of data.
*/
for(i = 0; i < odoner; i++) {
*obuf++ = obufl[i];
*obuf++ = obufr[i];
}
efftab[e-1].odone += idonel + idoner;
/* Don't clear since nothng has been consumed yet */
/*efftab[e].odone = 0;*/
efftab[e].olen += odonel + odoner;
#if 0
fprintf(stderr, "post %s idone=%d, odone=%d\n", efftab[e].name, idone, odone);
fprintf(stderr, "post %s odone1=%d, olen1=%d odone=%d olen=%d\n", efftab[e].name, efftab[e-1].odone, efftab[e-1].olen, efftab[e].odone, efftab[e].olen);
#endif
done = idonel + idoner + odonel + odoner;
if (effstatusl)
effstatus = effstatusl;
else
effstatus = effstatusr;
}
if (effstatus == ST_EOF)
{
return ST_EOF;
}
if (done == 0)
{
st_fail("Effect took & gave no samples!");
cleanup();
exit(2);
}
return ST_SUCCESS;
}
static int drain_effect_out(void)
{
/* Skip past input effect since we know thats not needed */
if (input_eff == 0)
{
input_eff = 1;
/* Assuming next effect hasn't reached EOF yet. */
input_eff_eof = 0;
}
/* Try to prime the pump with some data */
while (input_eff < neffects)
{
int rc;
rc = drain_effect(input_eff);
if (efftab[input_eff].olen == 0)
{
input_eff++;
/* Assuming next effect hasn't reached EOF yet. */
input_eff_eof = 0;
}
else
{
input_eff_eof = (rc == ST_EOF) ? 1 : 0;
break;
}
}
/* Just do standard flow routines after the priming. */
return flow_effect_out();
}
static int drain_effect(int e)
{
st_ssize_t i, olen, olenl, olenr;
st_sample_t *obuf;
int rc;
if (! efftabR[e].name) {
efftab[e].olen = ST_BUFSIZ;
rc = (* efftab[e].h->drain)(&efftab[e],efftab[e].obuf,
&efftab[e].olen);
efftab[e].odone = 0;
}
else {
int rc_l, rc_r;
olen = ST_BUFSIZ;
/* left */
olenl = olen/2;
rc_l = (* efftab[e].h->drain)(&efftab[e], obufl,
(st_size_t *)&olenl);
/* right */
olenr = olen/2;
rc_r = (* efftab[e].h->drain)(&efftabR[e], obufr,
(st_size_t *)&olenr);
if (rc_l == ST_EOF || rc_r == ST_EOF)
rc = ST_EOF;
else
rc = ST_SUCCESS;
obuf = efftab[e].obuf;
/* This loop implies left and right effect will always output
* the same amount of data.
*/
for(i = 0; i < olenr; i++) {
*obuf++ = obufl[i];
*obuf++ = obufr[i];
}
efftab[e].olen = olenl + olenr;
efftab[e].odone = 0;
}
return rc;
}
static void release_effect_buf(void)
{
int e;
for(e = 0; e < neffects; e++)
{
free(efftab[e].obuf);
if (efftabR[e].obuf)
free(efftabR[e].obuf);
}
}
static void stop_effects(void)
{
int e;
for (e = 1; e < neffects; e++) {
(*efftab[e].h->stop)(&efftab[e]);
if (efftabR[e].name)
(* efftabR[e].h->stop)(&efftabR[e]);
}
}
static void print_input_status(int input)
{
fprintf(stderr, "\nInput Filename : %s\n", file_desc[input]->filename);
fprintf(stderr, "Sample Size : %s\n",
st_size_bits_str[file_desc[input]->info.size]);
fprintf(stderr, "Sample Encoding: %s\n",
st_encodings_str[file_desc[input]->info.encoding]);
fprintf(stderr, "Channels : %d\n", file_desc[input]->info.channels);
fprintf(stderr, "Sample Rate : %d\n", file_desc[input]->info.rate);
if (file_desc[input]->comment && *file_desc[input]->comment)
fprintf(stderr, "Comments :\n%s\n", file_desc[input]->comment);
fprintf(stderr, "\n");
}
static void update_status(void)
{
int read_min, left_min, in_min;
double read_sec, left_sec, in_sec;
double read_time, left_time, in_time;
float completed;
double out_size;
char unit;
/* Currently, for both sox and soxmix, all input files must have
* the same sample rate. So we can always just use the rate
* of the first input file to compute time.
*/
read_time = (double)read_samples / (double)file_desc[0]->info.rate /
(double)file_desc[0]->info.channels;
read_min = read_time / 60;
read_sec = (double)read_time - 60.0f * (double)read_min;
out_size = output_samples / 1000000000.0;
if (out_size >= 1.0)
unit = 'G';
else
{
out_size = output_samples / 1000000.0;
if (out_size >= 1.0)
unit = 'M';
else
{
out_size = output_samples / 1000.0;
if (out_size >= 1.0)
unit = 'K';
else
unit = ' ';
}
}
if (input_samples)
{
in_time = (double)input_samples / (double)file_desc[0]->info.rate /
(double)file_desc[0]->info.channels;
left_time = in_time - read_time;
if (left_time < 0)
left_time = 0;
completed = ((double)read_samples / (double)input_samples) * 100;
if (completed < 0)
completed = 0;
}
else
{
in_time = 0;
left_time = 0;
completed = 0;
}
left_min = left_time / 60;
left_sec = (double)left_time - 60.0f * (double)left_min;
in_min = in_time / 60;
in_sec = (double)in_time - 60.0f * (double)in_min;
fprintf(stderr, "\rTime: %02i:%05.2f [%02i:%05.2f] of %02i:%05.2f (% 5.1f%%) Output Buffer:% 7.2f%c", read_min, read_sec, left_min, left_sec, in_min, in_sec, completed, out_size, unit);
}
static void statistics(void)
{
if (clipped > 0)
st_report("Volume change clipped %d samples", clipped);
}
static st_sample_t volumechange(st_sample_t *buf, st_ssize_t ct,
double vol)
{
double y;
st_sample_t *p,*top;
st_ssize_t clips=0;
p = buf;
top = buf+ct;
while (p < top) {
y = vol * *p;
ST_SAMPLE_CLIP_COUNT(y, clips);
*p++ = y;
}
return clips;
}
static void usage(char *opt)
{
int i;
const st_format_t *f;
const st_effect_t *e;
static char *usagestr;
if (soxmix)
usagestr = "[ gopts ] [ fopts ] ifile1 [fopts] ifile2 [ fopts ] ofile [ effect [ effopts ] ]";
else
usagestr = "[ gopts ] [ fopts ] ifile [ fopts ] ofile [ effect [ effopts ] ]";
printf("%s: ", myname);
printf("Version %s\n\n", st_version());
if (opt)
fprintf(stderr, "Failed: %s\n\n", opt);
printf("Usage: %s\n\n", usagestr);
printf(
"Global options (gopts):\n"
"\n"
"Global options can be specified anywhere on the command\n"
"\n"
"-h print version number and usage information\n"
"--help same as -h\n"
"--help-efffect=name\n"
" print usage of specified effect. use 'all' to print all\n"
"-p run in preview mode and run fast\n"
"-q run in quite mode. Inverse of -S option\n"
"-S print status while processing audio data.\n"
"--version print version number of SoX and exit\n"
"-V verbose mode. print a description during processing phase\n"
"\n"
"Format options (fopts):\n"
"\n"
"Format options only need to be supplied on input files that are\n"
"headerless otherwise they are obtained from the audio data's header.\n"
"Output files will default to the same format options as the input\n"
"file unless overriden on the command line.\n"
"\n"
"-c channels number of channels in audio data\n"
"-C compression compression factor for variably compressing output formats\n"
"-e skip processing of this filename. useful only\n"
" on output filename to prevent writing data.\n"
"-r rate sample rate of audio\n"
"-t filetype file type of audio\n"
"-v volume volume adjustment factor (floating point)\n"
"-x invert auto-detected endianess of data\n"
"-s/-u/-U/-A/ sample encoding. signed/unsigned/u-law/A-law\n"
" -a/-i/-g/-f ADPCM/IMA_ADPCM/GSM/floating point\n"
"-1/-2/-3/-4 sample size in bytes\n"
"-b/-w/-l/-d sample size. byte(8-bits)/word(16-bits)/\n"
" long(32-bits)/double long(64-bits)\n"
"\n");
printf("Supported file formats: ");
for (i = 0; st_format_fns[i]; i++) {
f = st_format_fns[i]();
if (f && f->names)
{
/* only print the first name */
printf("%s ", f->names[0]);
}
}
printf("\n\nSupported effects: ");
for (i = 0; st_effect_fns[i]; i++) {
e = st_effect_fns[i]();
if (e && e->name)
{
printf("%s ", e->name);
}
}
printf( "\n\neffopts: depends on effect\n\n");
exit(1);
}
static void usage_effect(char *effect)
{
int i;
const st_effect_t *e;
printf("%s: ", myname);
printf("v%s\n\n", st_version());
printf("Effect usage:\n\n");
for (i = 0; st_effect_fns[i]; i++)
{
e = st_effect_fns[i]();
if (e && e->name &&
(!strcmp("all", effect) || !strcmp(e->name, effect)))
{
char *p = strstr(e->usage, "Usage: ");
printf("%s\n\n", p ? p + 7 : e->usage);
}
}
if (!effect)
printf("see --help-effect=effect for effopts ('all' for effopts of all effects)\n\n");
exit(1);
}
void cleanup(void)
{
int i;
struct stat st;
char *fn;
/* Close the input file and outputfile before exiting*/
for (i = 0; i < input_count; i++)
{
if (file_desc[i])
{
st_close(file_desc[i]);
free(file_desc[i]);
}
}
if (writing && file_desc[file_count-1])
{
fstat(fileno(file_desc[file_count-1]->fp), &st);
fn = strdup(file_desc[file_count-1]->filename);
st_close(file_desc[file_count-1]);
/* remove the output file because we failed, if it's ours. */
/* Don't if its not a regular file. */
if ((st.st_mode & S_IFMT) == S_IFREG)
unlink(fn);
free(fn);
if (file_desc[file_count-1])
free(file_desc[file_count-1]);
}
}
static void sigint(int s)
{
if (s == SIGINT || s == SIGTERM)
user_abort = 1;
}