#include <stdio.h>
#include <stdlib.h>
#include <stdarg.h>
#include <string.h>
#include <sysexits.h>
#include <fcntl.h>
#include <math.h>

#include <xas/spatial.h>
#include <xas/synth.h>
#include <xas/vox.h>
#include <xas/audio.h>
#include <xas/riff.h>

#define SYNTH_STATUS_CLEAR  0
#define SYNTH_STATUS_ON    (1 << 0)

typedef struct _synth_sine {
    int flags;
    float phase;

    size_t frequency;
} synth_sine;

static void usage(int argc, char **argv, const char *message, ...) {
    va_list args;

    va_start(args, message);

    if (message) {
        vfprintf(stderr, message, args);
        fputc('\n', stderr);
    }

    va_end(args);

    fprintf(stderr, "usage: %s output.wav\n", argv[0]);

    exit(EX_USAGE);
}

static int16_t sine_sample(xas_synth *synth, synth_sine *sine) {
    int16_t ret;
    static float tau = 2.0f * M_PI;

    if (sine->flags & SYNTH_STATUS_ON) {
        ret = (int16_t)roundf((INT16_MAX >> 2) * sinf(sine->phase));

        sine->phase += tau / (synth->format.sample_rate / sine->frequency);

        if (sine->phase > tau) {
            sine->phase -= tau;
        }
    } else {
        ret = 0;
    }

    return ret;
}

static void sine_cleanup(xas_synth *synth, synth_sine *sine) {
    return;
}

int main(int argc, char **argv) {
    xas_spatial_scene *scene;

    xas_audio_stream *synth,
                       *voice,
                       *output,
                       *wave;

    xas_vox *vox;

    synth_sine sine_channels[2] = {
        { SYNTH_STATUS_ON, 0.0f, 2600 },
        { SYNTH_STATUS_ON, 0.0f,  420 },
    };

    xas_audio_format format = {
        .channels    = XAS_AUDIO_STEREO,
        .sample_size = XAS_AUDIO_PCM_16_BIT,
        .sample_rate = 44100
    };

    size_t buffer_size = 735,
           duration_s  = 300,
           i;

    xas_spatial_coord speakers[2] = {
        { -0.09, 0.0, 0.0 },
        {  0.09, 0.0, 0.0 }
    };

    if (argc != 2) {
        usage(argc, argv, "No output file provided");
    }

    if ((wave = xas_riff_new_file(argv[1],
                                    format,
                                    O_WRONLY | O_CREAT | O_TRUNC)) == NULL) {
        goto error_riff_new_file;
    }

    if ((synth = xas_synth_new((xas_synth_callback_sample)sine_sample,
                                 (xas_synth_callback_cleanup)sine_cleanup,
                                 format,
                                 buffer_size,
                                 &sine_channels[0])) == NULL) {
        goto error_synth_new;
    }

    if ((vox = xas_vox_new("/usr/bin/text2wave",
                             format,
                             buffer_size)) == NULL) {
        goto error_vox_new;
    }

    if ((voice = xas_vox_stream_new(vox)) == NULL) {
        goto error_vox_stream_new;
    }

    if ((scene = xas_spatial_scene_new(format,
                                         speakers[0],
                                         speakers[1])) == NULL) {
        goto error_spatial_scene_new;
    }

    if ((output = xas_spatial_scene_new_stream(scene,
                                                 buffer_size)) == NULL) {
        goto error_spatial_scene_new_stream;
    }

    if (xas_spatial_scene_add_object(scene,
                                       (xas_spatial_coord){ 5.2, 0.0, 0.0 },
                                       synth) == NULL) {
        goto error_spatial_scene_add_object;
    }

    if (xas_spatial_scene_add_object(scene,
                                       (xas_spatial_coord){ -5.2, 0.0, 0.0 },
                                       voice) == NULL) {
        goto error_spatial_scene_add_object;
    }

    xas_vox_say(vox, "I want to eat your soul.  You don't understand.  I really want to eat your soul.\n");

    xas_vox_generate(vox);

    for (i=0; i<duration_s; i++) {
        void *buf;
        ssize_t readlen;

        if ((readlen = xas_audio_stream_read(output,
                                               &buf,
                                               buffer_size)) < 0) {
            goto error_audio_stream_read;
        }

        if (xas_audio_stream_write(wave, buf, readlen) < 0) {
            goto error_audio_stream_write;
        }
    }

    xas_audio_stream_flush(wave);

    xas_audio_stream_destroy(output);
    xas_spatial_scene_destroy(scene);
    xas_audio_stream_destroy(synth);
    xas_audio_stream_destroy(wave);

    return EX_OK;

error_audio_stream_read:
error_audio_stream_write:
error_spatial_scene_add_object:
    xas_audio_stream_destroy(output);

error_spatial_scene_new_stream:
    xas_spatial_scene_destroy(scene);

error_spatial_scene_new:
    xas_audio_stream_destroy(voice);

error_vox_stream_new:
    xas_vox_destroy(vox);

error_vox_new:
    xas_audio_stream_destroy(synth);

error_synth_new:
    xas_audio_stream_destroy(wave);

error_riff_new_file:
    return EX_OSERR;
}