#include <stdlib.h>
#include <string.h>

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

static size_t DEFAULT_BUFFER_SIZE = 735;

static int record_speech_sample(xas_bank *bank,
                                size_t entry_index,
                                float speed,
                                const char *text) {
    xas_vox *vox;
    xas_audio_stream *source;

    if ((vox = xas_vox_new(bank->format,
                             DEFAULT_BUFFER_SIZE,
                             "/usr/bin/text2wave")) == NULL) {
        goto error_vox_new;
    }

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

    xas_vox_set_voice(vox, "voice_cmu_us_slt_cg");
    xas_vox_set_parameter_float(vox, "Duration_Stretch", speed);
    xas_vox_say(vox, text);
    xas_vox_generate(vox);

    if (xas_bank_record(bank, source, entry_index, bank->entry_size) < 0) {
        goto error_bank_record;
    }

    xas_audio_stream_destroy(source);
    xas_vox_destroy(vox);

    return 0;

error_bank_record:
    xas_audio_stream_destroy(source);

error_vox_stream_new:
    xas_vox_destroy(vox);

error_vox_new:
    return -1;
}

xas_drone *xas_drone_new(xas_spatial_scene *scene,
                             xas_spatial_coord position,
                             float speech_speed,
                             size_t speech_sample_count,
                             size_t speech_line_count,
                             const char **speech_lines) {
    xas_drone *drone;

    size_t i;

    if ((drone = malloc(sizeof(*drone))) == NULL) {
        goto error_malloc_drone;
    }

    if ((drone->bank = xas_bank_new(scene->format,
                                      speech_sample_count,
                                      speech_line_count)) == NULL) {
        goto error_bank_new;
    }

    for (i=0; i<speech_line_count; i++) {
        if (record_speech_sample(drone->bank,
                                 i,
                                 speech_speed,
                                 speech_lines[i]) < 0) {
            goto error_record_speech_sample;
        }
    }

    if ((drone->obj = xas_spatial_scene_add_bank_player(scene,
                                                          position,
                                                          drone->bank)) == NULL) {
        goto error_spatial_scene_add_bank_player;
    }

    return drone;

error_spatial_scene_add_bank_player:
error_record_speech_sample:
    xas_bank_destroy(drone->bank);

error_bank_new:
    free(drone);

error_malloc_drone:
    return NULL;
}

void xas_drone_destroy(xas_drone *drone) {
    xas_bank_destroy(drone->bank);

    free(drone);
}

xas_spatial_object *xas_drone_get_spatial_object(xas_drone *drone) {
    return drone->obj;
}

xas_drone_chamber *xas_drone_chamber_new(xas_spatial_scene *scene,
                                             xas_spatial_coord location,
                                             size_t drone_count) {
    xas_drone_chamber *chamber;

    size_t total = drone_count * sizeof(xas_drone *);

    if ((chamber = malloc(sizeof(*chamber))) == NULL) {
        goto error_malloc_chamber;
    }

    if ((chamber->drones = malloc(total)) == NULL) {
        goto error_malloc_chamber_drones;
    }

    if ((chamber->synth_bass = xas_spatial_scene_add_synth(scene,
                                                             (xas_spatial_coord){
                                                                location.x,
                                                                location.y,
                                                                location.z + 5.0
                                                             },
                                                             XAS_SYNTH_SQUARE)) == NULL) {
        goto error_spatial_scene_add_synth;
    }

    if ((chamber->synth_l = xas_spatial_scene_add_synth(scene,
                                                          (xas_spatial_coord){
                                                              location.x - 5.0,
                                                              location.y,
                                                              location.z
                                                          },
                                                          XAS_SYNTH_SINE)) == NULL) {
        goto error_spatial_scene_add_synth;
    }

    if ((chamber->synth_r = xas_spatial_scene_add_synth(scene,
                                                          (xas_spatial_coord){
                                                              location.x + 5.0,
                                                              location.y,
                                                              location.z
                                                          },
                                                          XAS_SYNTH_SINE)) == NULL) {
        goto error_spatial_scene_add_synth;
    }

    xas_synth_set_frequency(chamber->synth_bass->ctx,
                              XAS_DRONE_CHAMBER_BASS_FREQUENCY);

    xas_synth_start(chamber->synth_bass->ctx);
    xas_synth_start(chamber->synth_l->ctx);
    xas_synth_start(chamber->synth_r->ctx);

    chamber->drone_count = drone_count;

    memset(chamber->drones, '\0', total);

    return chamber;

error_spatial_scene_add_synth:
error_malloc_chamber_drones:
    free(chamber);

error_malloc_chamber:
    return NULL;
}

void xas_drone_chamber_destroy(xas_drone_chamber *chamber) {
    free(chamber);
}

void xas_drone_chamber_insert_drone(xas_drone_chamber *chamber,
                                      xas_drone *drone,
                                      size_t index) {
    chamber->drones[index] = drone;
}

int xas_drone_chamber_seq_intervals(xas_drone_chamber *chamber,
                                      xas_drone_chamber_interval *intervals,
                                      xas_seq *seq,
                                      size_t count,
                                      struct timeval *now) {
    size_t i;

    for (i=0; i<count; i++) {
        struct timeval cur = {
            .tv_sec  = now->tv_sec,
            .tv_usec = now->tv_usec
        };

        if (xas_seq_add_set_frequency(seq,
                                        chamber->synth_l,
                                        *now,
                                        intervals[i].freq_l) < 0) {
            goto error_seq_add;
        }

        if (xas_seq_add_set_synth_type(seq,
                                         chamber->synth_l,
                                         *now,
                                         intervals[i].type_l) < 0) {
            goto error_seq_add;
        }

        if (xas_seq_add_set_frequency(seq,
                                        chamber->synth_r,
                                        *now,
                                        intervals[i].freq_r) < 0) {
            goto error_seq_add;
        }

        if (xas_seq_add_set_synth_type(seq,
                                         chamber->synth_r,
                                         *now,
                                         intervals[i].type_r) < 0) {
            goto error_seq_add;
        }

        timeradd(&cur, &intervals[i].duration, now);
    }

    return 0;

error_seq_add:
    return -1;
}

int xas_drone_chamber_seq_speech(xas_drone_chamber *chamber,
                                   xas_drone *drone,
                                   xas_seq *seq,
                                   size_t speech_part,
                                   struct timeval *now) {
    struct timeval duration,
                   tmp;

    if (xas_seq_add_set_bank(seq,
                               drone->obj,
                               *now,
                               speech_part) < 0) {
        goto error_xas_seq_add;
    }

    if (xas_seq_add_event_on(seq,
                               drone->obj,
                               *now) < 0) {
        goto error_xas_seq_add;
    }

    xas_bank_entry_duration(drone->bank,
                              speech_part,
                              &duration);

    timeradd(now, &duration, &tmp);

    now->tv_sec  = tmp.tv_sec;
    now->tv_usec = tmp.tv_usec;

    return 0;

error_xas_seq_add:
    return -1;
}

static void max_speech_duration(xas_drone_chamber *chamber,
                                size_t speech_part,
                                struct timeval *max) {
    size_t i;

    max->tv_sec  = 0;
    max->tv_usec = 0;

    for (i=0; i<chamber->drone_count; i++) {
        xas_drone *drone = chamber->drones[i];

        struct timeval duration;

        xas_bank_entry_duration(drone->bank,
                                  speech_part,
                                  &duration);

        if (timercmp(&duration, max, >)) {
            max->tv_sec  = duration.tv_sec;
            max->tv_usec = duration.tv_usec;
        }
    }
}

int xas_drone_chamber_seq_chorus(xas_drone_chamber *chamber,
                                   xas_seq *seq,
                                   size_t speech_part,
                                   struct timeval *now) {
    struct timeval duration,
                   tmp,
                   delay = { 0, 500000 };

    size_t i;

    max_speech_duration(chamber, speech_part, &duration);

    for (i=0; i<chamber->drone_count; i++) {
        xas_drone *drone = chamber->drones[i];

        if (xas_seq_add_set_bank(seq,
                                   drone->obj,
                                   *now,
                                   speech_part) < 0) {
            goto error_seq_add;
        }

        if (xas_seq_add_event_on(seq, drone->obj, *now) < 0) {
            goto error_seq_add;
        }
    }

    timeradd(now, &duration, &tmp);
    timeradd(&tmp, &delay, now);

    return 0;

error_seq_add:
    return -1;
}