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

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

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);
}

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

    xas_spatial_object *synth_l,
                         *synth_r,
                         *voice;

    xas_audio_stream *output,
                       *wave;

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

    size_t buffer_size = 735;

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

    if ((seq = xas_seq_new(scene, buffer_size)) == NULL) {
        goto error_seq_new;
    }

    if ((voice = xas_spatial_scene_add_vox(scene,
                                             (xas_spatial_coord){ 0.0, 0.0, -20.0 },
                                             "/usr/bin/text2wave")) == NULL) {
        goto error_spatial_scene_add_vox;
    }

    if ((synth_l = xas_spatial_scene_add_synth(scene,
                                                 (xas_spatial_coord){ -20.0, 0.0, 0.0 },
                                                 XAS_SYNTH_SINE)) == NULL) {
        goto error_spatial_scene_add_synth_l;
    }

    if ((synth_r = xas_spatial_scene_add_synth(scene,
                                                 (xas_spatial_coord){ 20.0, 0.0, 0.0 },
                                                 XAS_SYNTH_SINE)) == NULL) {
        goto error_spatial_scene_add_synth_r;
    }

    if ((output = xas_spatial_scene_stream_new(scene,
                                                 buffer_size)) == NULL) {
        goto error_spatial_scene_stream_new;
    }

    xas_vox_set_voice(voice->ctx, "voice_cmu_us_slt_cg");
    xas_vox_set_parameter_float(voice->ctx, "Duration_Stretch", 1.3);

    xas_seq_add_set_frequency(seq, synth_l, (struct timeval){ 0, 0 }, 220);
    xas_seq_add_set_frequency(seq, synth_r, (struct timeval){ 0, 0 }, 420);
    xas_seq_add_event_on(     seq, synth_l, (struct timeval){ 0, 0 });
    xas_seq_add_event_on(     seq, synth_r, (struct timeval){ 0, 0 });

    xas_seq_add_set_position(seq,
                               synth_l,
                               (struct timeval){ 15, 0 },
                               (xas_spatial_coord){ -10.0, 0.0, 0.0 });

    xas_seq_add_phrase(seq,
                         voice,
                         (struct timeval){ 10, 0 },
                         "I will eat your soul.");

    xas_seq_add_event_off(seq, synth_l, (struct timeval){ 60, 0 });

    xas_seq_play(seq, wave);

    xas_seq_destroy(seq);
    xas_spatial_scene_destroy(scene);
    xas_audio_stream_destroy(wave);

    return EX_OK;

error_spatial_scene_stream_new:
error_spatial_scene_add_synth_r:
error_spatial_scene_add_synth_l:
error_spatial_scene_add_vox:
    xas_seq_destroy(seq);

error_seq_new:
    xas_spatial_scene_destroy(scene);

error_spatial_scene_new:
    xas_audio_stream_destroy(wave);

error_riff_new_file:
    return EX_OSERR;
}