path: root/src/vox.c

#include <stdio.h>
#include <stdlib.h>
#include <string.h>
#include <unistd.h>
#include <fcntl.h>
#include <sysexits.h>
#include <sys/wait.h>
#include <errno.h>

#include <xas/riff.h>
#include <xas/vox.h>

static int vox_stop(xas_vox *vox) {
    if (vox->stdin >= 0) {
        (void)close(vox->stdin);

        vox->stdin = -1;
    }

    if (vox->stdout >= 0) {
        (void)close(vox->stdout);

        vox->stdout = -1;
    }

    if (vox->in) {
        (void)fclose(vox->in);

        vox->in = NULL;
    }

    if (vox->pid) {
        int status;

        if (waitpid(vox->pid, &status, 0) < 0) {
            goto error_waitpid;
        }

        vox->pid = -1;
    }

    vox->state = XAS_VOX_IDLE;

    return 0;

error_waitpid:
    return -1;
}

static char **args_concat(int argn1, char **args1,
                          int argn2, char **args2,
                          int *total) {
    char **ret;

    int count = argn1 + argn2,
        i, o;

    if ((ret = malloc((count + 1) * sizeof(char *))) == NULL) {
        goto error_malloc;
    }

    for (i=0, o=0; i<argn1; i++, o++) {
        ret[o] = args1[i];
    }

    for (i=0; i<argn2; i++, o++) {
        ret[o] = args2[i];
    }

    *total = count;
    ret[o] = NULL;

    return ret;

error_malloc:
    return NULL;
}

static int vox_start(xas_vox *vox) {
    int pipe_stdin[2],
        pipe_stdout[2];

    int pid;

    char sample_rate[40];

    if (vox->state == XAS_VOX_ACTIVE) {
        (void)vox_stop(vox);
    }

    if (vox->tmpfh) {
        if (fflush(vox->tmpfh) < 0) {
            goto error_fflush;
        }
    }

    if (pipe(pipe_stdin) < 0) {
        goto error_pipe_stdin;
    }

    if (pipe(pipe_stdout) < 0) {
        goto error_pipe_stdout;
    }

    if ((pid = fork()) < 0) {
        goto error_fork;
    }

    if (pid == 0) {
        char *args[] = {
            (char *)vox->text2wave_path,
            "-F",
            sample_rate,
            "-",
            "-o",
            "-",
            NULL,
            NULL,
            NULL
        };

        int argn = 6;

        int argc;
        char **argv;

        int fd;

        snprintf(sample_rate, sizeof(sample_rate)-1, "%zu", vox->format.sample_rate);

        if (vox->tmpfile[0] && vox->tmpfd > 0) {
            args[argn++] = "-eval";
            args[argn++] = vox->tmpfile;
        }

        if ((argv = args_concat(argn, args, vox->argn, vox->args, &argc)) == NULL) {
            goto error_child;
        }

        if ((fd = open("/dev/null", O_WRONLY)) < 0) {
            goto error_child;
        }

        close(pipe_stdin[1]);
        close(pipe_stdout[0]);

        if (dup2(pipe_stdin[0], 0) < 0) {
            goto error_child;
        }

        if (dup2(pipe_stdout[1], 1) < 0) {
            goto error_child;
        }

        if (dup2(fd, 2) < 0) {
            goto error_child;
        }

        if (execv(vox->text2wave_path, argv) < 0) {
            goto error_child;
        }

error_child:
        exit(EX_OSERR);
    }

    close(pipe_stdin[0]);
    close(pipe_stdout[1]);

    vox->pid    = pid;
    vox->stdin  = pipe_stdin[1];
    vox->stdout = pipe_stdout[0];
    vox->in     = fdopen(pipe_stdin[1], "w");
    vox->state  = XAS_VOX_ACTIVE;

    return 0;

error_fork:
    (void)close(pipe_stdout[1]);
    (void)close(pipe_stdout[0]);

error_pipe_stdout:
    (void)close(pipe_stdin[1]);
    (void)close(pipe_stdin[0]);

error_pipe_stdin:
error_fflush:
    return -1;
}

static void vox_cleanup(xas_vox *vox) {
    (void)vox_stop(vox);
}

static ssize_t vox_fill(xas_vox *vox,
                        int16_t *samples,
                        size_t count) {
    ssize_t readlen,
            readcount;

    if (vox->state != XAS_VOX_ACTIVE) {
        return 0;
    }

    if ((readlen = read(vox->stdout,
                        samples,
                        count * vox->format.sample_size)) < 0) {
        goto error_read;
    }

    if (readlen == 0) {
        vox_stop(vox);
    }

    readcount = readlen / vox->format.sample_size;

    if (readcount > 0 && vox->gain != 1.0f) {
        xas_audio_apply_gain(vox->format,
                               samples,
                               vox->gain,
                               0,
                               readcount);
    }

    return readcount;

error_read:
    return -1;
}

static int tmpfile_open(xas_vox *vox) {
    if (vox->tmpfile[0] && vox->tmpfd > 0) {
        return 0;
    }

    memset(vox->tmpfile, '\0', sizeof(vox->tmpfile));

    memcpy(vox->tmpfile,
           XAS_VOX_SETTINGS_TMP_PATH,
           strlen(XAS_VOX_SETTINGS_TMP_PATH));

    if ((vox->tmpfd = mkstemp(vox->tmpfile)) < 0) {
        goto error_mkstemp;
    }

    if ((vox->tmpfh = fdopen(vox->tmpfd, "w")) == NULL) {
        goto error_fdopen;
    }

    return 0;

error_fdopen:
    close(vox->tmpfd);
    unlink(vox->tmpfile);

error_mkstemp:
    return -1;
}

static void tmpfile_close(xas_vox *vox) {
    fclose(vox->tmpfh);
    (void)close(vox->tmpfd);
    unlink(vox->tmpfile);

    memset(vox->tmpfile, '\0', sizeof(vox->tmpfile));
    vox->tmpfd = -1;
    vox->tmpfh = NULL;
}

static int set_gain(xas_vox *vox, float gain) {
    xas_vox_set_gain(vox, gain);

    return 0;
}

xas_vox *xas_vox_new_args(xas_audio_format format,
                              size_t buffer_size,
                              const char *text2wave_path,
                              int argn,
                              char **args) {
    xas_vox *vox;

    if ((vox = malloc(sizeof(*vox))) == NULL) {
        goto error_malloc_vox;
    }

    vox->obj.start      = (xas_object_start_callback)xas_vox_generate;
    vox->obj.stop       = (xas_object_stop_callback)xas_vox_stop;
    vox->obj.set_gain   = (xas_object_set_gain_callback)set_gain;
    vox->obj.stream_new = (xas_object_stream_new_callback)xas_vox_stream_new;
    vox->obj.destroy    = (xas_object_destroy_callback)xas_vox_destroy;

    vox->text2wave_path = text2wave_path;

    vox->format.channels    = XAS_AUDIO_MONO;
    vox->format.sample_size = format.sample_size;
    vox->format.sample_rate = format.sample_rate;

    vox->buffer_size = buffer_size;
    vox->gain        = XAS_VOX_DEFAULT_GAIN;
    vox->state       = XAS_VOX_IDLE;
    vox->pid         = -1;
    vox->stdin       = -1;
    vox->stdout      = -1;
    vox->in          = NULL;

    vox->argn = argn;
    vox->args = args;

    memset(vox->tmpfile, '\0', sizeof(vox->tmpfile));
    vox->tmpfd = -1;
    vox->tmpfh = NULL;

    return vox;

error_malloc_vox:
    return NULL;
}

xas_vox *xas_vox_new(xas_audio_format format,
                         size_t buffer_size,
                         const char *text2wave_path) {
    return xas_vox_new_args(format,
                              buffer_size,
                              text2wave_path,
                              0,
                              NULL);
}

void xas_vox_destroy(xas_vox *vox) {
    if (vox->tmpfile[0] && vox->tmpfd > 0) {
        tmpfile_close(vox);
    }

    free(vox);
}

void xas_vox_set_gain(xas_vox *vox, float gain) {
    vox->gain = gain;
}

int xas_vox_set_voice(xas_vox *vox, const char *voice) {
    if (tmpfile_open(vox) < 0) {
        goto error_tmpfile_open;
    }

    if (fprintf(vox->tmpfh, "(%s)\n", voice) < 0) {
        goto error_fprintf;
    }

    return 0;

error_fprintf:
error_tmpfile_open:
    return -1;
}

static int set_parameter_va(xas_vox *vox,
                            const char *name,
                            const char *format,
                            ...) {
    va_list args;

    if (tmpfile_open(vox) < 0) {
        goto error_tmpfile_open;
    }

    va_start(args, format);

    if (fprintf(vox->tmpfh, "(Parameter.set '%s ", name) < 0) {
        goto error_fprintf;
    }

    if (vfprintf(vox->tmpfh, format, args) < 0) {
        goto error_fprintf;
    }

    if (fprintf(vox->tmpfh, ")\n") < 0) {
        goto error_fprintf;
    }

    va_end(args);

    return 0;

error_fprintf:
    va_end(args);

error_tmpfile_open:
    return -1;
}

int xas_vox_set_parameter(xas_vox *vox,
                            const char *name,
                            const char *value) {
    return set_parameter_va(vox, name, "%s", value);
}

int xas_vox_set_parameter_str(xas_vox *vox,
                                const char *name,
                                const char *value) {
    return set_parameter_va(vox, name, "\"%s\"", value);
}

int xas_vox_set_parameter_float(xas_vox *vox,
                                  const char *name,
                                  float value) {
    return set_parameter_va(vox, name, "%f", value);
}

int xas_vox_set_parameter_int(xas_vox *vox,
                                const char *name,
                                int value) {
    return set_parameter_va(vox, name, "%d", value);
}

int xas_vox_active(xas_vox *vox) {
    return vox->state == XAS_VOX_ACTIVE;
}

int xas_vox_generate(xas_vox *vox) {
    xas_audio_stream *output;

    (void)fflush(vox->in);
    (void)fclose(vox->in);
    (void)close(vox->stdin);

    vox->stdin = -1;
    vox->in    = NULL;

    if ((output = xas_riff_open_fd(vox->stdout)) == NULL) {
        goto error_riff_open_fd;
    }

    if (!xas_audio_format_eq(output->format, vox->format)) {
        errno = EINVAL;

        goto error_invalid_stream;
    }

    xas_audio_stream_destroy(output);

    return 0;

error_invalid_stream:
error_riff_open_fd:
    return -1;
}

int xas_vox_stop(xas_vox *vox) {
    if (vox->state == XAS_VOX_ACTIVE) {
        return vox_stop(vox);
    }

    return 0;
}

int xas_vox_vsayf(xas_vox *vox, const char *format, va_list args) {
    if (vox->state == XAS_VOX_IDLE && vox_start(vox) < 0) {
        goto error_vox_start;
    }

    return vfprintf(vox->in, format, args);

error_vox_start:
    return -1;
}

int xas_vox_sayf(xas_vox *vox, const char *format, ...) {
    int ret;
    va_list args;

    va_start(args, format);
    ret = xas_vox_vsayf(vox, format, args);
    va_end(args);

    return ret;
}

int xas_vox_say(xas_vox *vox, const char *message) {
    return xas_vox_sayf(vox, "%s\n", message);
}

xas_audio_stream *xas_vox_stream_new(xas_vox *vox) {
    return xas_audio_stream_new_source((xas_audio_fill)vox_fill,
                                         (xas_audio_cleanup)vox_cleanup,
                                         vox->format,
                                         vox->buffer_size,
                                         vox);
}