xtc/src/midi/instrument.c

#include "xtc/xtc_midi.h"
#include "../mixer/mix_internal.h"
#include <math.h>
#include <stdlib.h>
#include <string.h>

#define TAU 6.28318530717958647692

typedef struct {
    uint64_t start, end, age;
    unsigned int pitch, velocity, channel;
    size_t track;
} span;

typedef struct {
    uint64_t time;
    int change;
} endpoint;

typedef struct {
    unsigned int active, channel, pitch, velocity;
    uint64_t age;
    double phase;
} voice;

struct xtc_midi_plan {
    xtc_mix_snapshot mix;
    span notes[XTC_PROJECT_MAX_NOTES];
    size_t count;
    voice voices[XTC_MIDI_VOICES];
    double step[128];
    uint64_t cutoff[16], next_position, age;
};

xtc_result xtc_midi_decode(const unsigned char *bytes, size_t size, unsigned int offset, xtc_midi_event *out)
{
    xtc_midi_event e;
    unsigned int command;
    if (!bytes || !size || !out)
        return XTC_ERR_INVALID_ARGUMENT;
    command = bytes[0] & 240U;
    if (command != 128 && command != 144 && command != 176)
        return XTC_ERR_UNSUPPORTED;
    if (size != 3 || bytes[1] > 127 || bytes[2] > 127)
        return XTC_ERR_INVALID_ARGUMENT;
    if (command == 176 && bytes[1] != 120 && bytes[1] != 123)
        return XTC_ERR_UNSUPPORTED;
    e.offset = offset;
    e.channel = bytes[0] & 15U;
    e.pitch = bytes[1];
    e.velocity = bytes[2];
    e.type = command == 176 ? XTC_MIDI_ALL_OFF : command == 128 || !e.velocity ? XTC_MIDI_NOTE_OFF : XTC_MIDI_NOTE_ON;
    *out = e;
    return XTC_OK;
}

static int compare_endpoint(const void *left, const void *right)
{
    const endpoint *a = left, *b = right;
    if (a->time != b->time)
        return a->time < b->time ? -1 : 1;
    return a->change - b->change; /* End before start at the same sample. */
}

void xtc_midi_plan_reset(xtc_midi_plan *p)
{
    if (!p)
        return;
    memset(p->voices, 0, sizeof(p->voices));
    memset(p->cutoff, 0, sizeof(p->cutoff));
    p->next_position = UINT64_MAX;
    p->age = 0;
}

xtc_result xtc_midi_plan_create(const xtc_project *project, unsigned int rate, xtc_midi_plan **out)
{
    xtc_midi_plan *p;
    endpoint *points;
    size_t i;
    int active = 0;
    if (!out)
        return XTC_ERR_INVALID_ARGUMENT;
    *out = NULL;
    if (!project || !rate || rate > 768000)
        return XTC_ERR_INVALID_ARGUMENT;
    p = calloc(1, sizeof(*p));
    points = calloc(XTC_PROJECT_MAX_NOTES * 2U, sizeof(*points));
    if (!p || !points) {
        free(p);
        free(points);
        return XTC_ERR_OUT_OF_MEMORY;
    }
    {
        xtc_result r = xtc_mix_prepare(project, rate, &p->mix);
        if (r != XTC_OK) {
            free(points);
            free(p);
            return r;
        }
    }
    for (i = 0; i < 128; ++i) {
        double frequency = 440.0 * pow(2.0, ((double)i - 69.0) / 12.0);
        p->step[i] = frequency < rate * 0.5 ? frequency / rate : 0;
    }
    for (i = 0; i < project->note_count; ++i) {
        const xtc_note_info *n = project->notes + i;
        xtc_clip_info c;
        uint64_t begin, end;
        span *s = p->notes + p->count;
        xtc_result r = xtc_project_find_clip(project, n->clip_id, &c);
        if (r != XTC_OK) {
            free(points);
            free(p);
            return XTC_ERR_INVALID_STATE;
        }
        begin = n->start_tick > c.offset_tick ? n->start_tick : c.offset_tick;
        end = n->start_tick + n->length_tick < c.offset_tick + c.length_tick ? n->start_tick + n->length_tick
                                                                             : c.offset_tick + c.length_tick;
        if (end <= begin)
            continue;
        r = xtc_project_tick_to_sample(project, c.start_tick + begin - c.offset_tick, rate, &s->start);
        if (r == XTC_OK)
            r = xtc_project_tick_to_sample(project, c.start_tick + end - c.offset_tick, rate, &s->end);
        if (r == XTC_OK)
            r = xtc_project_tick_to_sample(project, begin - n->start_tick, rate, &s->age);
        if (r != XTC_OK) {
            free(points);
            free(p);
            return r;
        }
        if (s->end <= s->start)
            continue;
        s->track = xtc_mix_track(&p->mix, c.track_id);
        s->pitch = n->pitch;
        s->velocity = n->velocity;
        s->channel = n->channel;
        points[p->count * 2].time = s->start;
        points[p->count * 2].change = 1;
        points[p->count * 2 + 1].time = s->end;
        points[p->count * 2 + 1].change = -1;
        ++p->count;
    }
    qsort(points, p->count * 2, sizeof(*points), compare_endpoint);
    for (i = 0; i < p->count * 2; ++i) {
        active += points[i].change;
        if (active > (int)XTC_MIDI_VOICES) {
            free(points);
            free(p);
            return XTC_ERR_CAPACITY;
        }
    }
    free(points);
    xtc_midi_plan_reset(p);
    *out = p;
    return XTC_OK;
}

void xtc_midi_plan_destroy(xtc_midi_plan *p)
{
    free(p);
}

static void channel_off(xtc_midi_plan *p, unsigned int channel, uint64_t position)
{
    size_t i;
    for (i = 0; i < XTC_MIDI_VOICES; ++i)
        if (p->voices[i].channel == channel)
            p->voices[i].active = 0;
    p->cutoff[channel] = position == UINT64_MAX ? position : position + 1;
}

void xtc_midi_plan_panic(xtc_midi_plan *p, uint64_t position)
{
    unsigned int c;
    if (!p)
        return;
    for (c = 0; c < 16; ++c)
        channel_off(p, c, position);
    p->next_position = position;
}

static void event(xtc_midi_plan *p, const xtc_midi_event *e, uint64_t position)
{
    size_t i, slot = SIZE_MAX;
    if (e->type == XTC_MIDI_ALL_OFF) {
        channel_off(p, e->channel, position);
        return;
    }
    for (i = 0; i < XTC_MIDI_VOICES; ++i) {
        if (p->voices[i].active && p->voices[i].channel == e->channel && p->voices[i].pitch == e->pitch) {
            slot = i;
            break;
        }
    }
    if (e->type == XTC_MIDI_NOTE_OFF) {
        if (slot != SIZE_MAX)
            p->voices[slot].active = 0;
        return;
    }
    if (slot == SIZE_MAX) {
        for (i = 0; i < XTC_MIDI_VOICES; ++i)
            if (!p->voices[i].active) {
                slot = i;
                break;
            }
    }
    if (slot == SIZE_MAX) {
        slot = 0;
        for (i = 1; i < XTC_MIDI_VOICES; ++i)
            if (p->voices[i].age < p->voices[slot].age)
                slot = i;
    }
    p->voices[slot].active = 1;
    p->voices[slot].channel = e->channel;
    p->voices[slot].pitch = e->pitch;
    p->voices[slot].velocity = e->velocity;
    p->voices[slot].phase = 0;
    p->voices[slot].age = p->age++;
}

static void segment(xtc_midi_plan *p, uint64_t position, float *out, float *right, size_t frames)
{
    size_t i, j;
    for (i = 0; i < p->count; ++i) {
        const span *s = p->notes + i;
        uint64_t begin = position > s->start ? position : s->start;
        uint64_t end = position + frames < s->end ? position + frames : s->end;
        double step = p->step[s->pitch];
        if (step == 0.0 || s->start < p->cutoff[s->channel] || end <= begin)
            continue;
        for (j = (size_t)(begin - position); j < (size_t)(end - position); ++j) {
            double phase = fmod((double)(position + j - s->start + s->age) * step, 1.0);
            float sample = (float)(sin(TAU * phase) * 0.1 * s->velocity / 127.0);
            out[j] += sample * xtc_mix_level(&p->mix, s->track, position + j, right ? 0U : 2U);
            if (right)
                right[j] += sample * xtc_mix_level(&p->mix, s->track, position + j, 1);
        }
    }
    for (i = 0; i < XTC_MIDI_VOICES; ++i) {
        voice *v = p->voices + i;
        double step = p->step[v->pitch];
        if (!v->active || step == 0.0)
            continue;
        for (j = 0; j < frames; ++j) {
            float sample = (float)(sin(TAU * v->phase) * 0.1 * v->velocity / 127.0);
            out[j] += sample;
            if (right)
                right[j] += sample;
            v->phase += step;
            if (v->phase >= 1)
                v->phase -= 1;
        }
    }
}

xtc_result xtc_midi_plan_render_stereo(xtc_midi_plan *p, uint64_t position, const xtc_midi_event *events, size_t count,
                                       float *out, float *right, size_t frames)
{
    size_t i, cursor = 0;
    if (!p || !out || (right && right == out) || count > XTC_MIDI_MAX_EVENTS || (count && !events) || frames > 65536U ||
        position > UINT64_MAX - frames)
        return XTC_ERR_INVALID_ARGUMENT;
    for (i = 0; i < count; ++i) {
        const xtc_midi_event *e = events + i;
        if (e->offset >= frames || (i && e->offset < events[i - 1].offset) || e->type < XTC_MIDI_NOTE_ON ||
            e->type > XTC_MIDI_ALL_OFF || e->pitch > 127 || e->velocity > 127 || e->channel > 15 ||
            (e->type == XTC_MIDI_NOTE_ON && !e->velocity))
            return XTC_ERR_INVALID_ARGUMENT;
    }
    if (!frames)
        return XTC_OK;
    if (p->next_position != position)
        xtc_midi_plan_reset(p);
    memset(out, 0, frames * sizeof(*out));
    if (right)
        memset(right, 0, frames * sizeof(*right));
    for (i = 0; i < count; ++i) {
        size_t next = events[i].offset;
        segment(p, position + cursor, out + cursor, right ? right + cursor : NULL, next - cursor);
        event(p, events + i, position + next);
        cursor = next;
    }
    segment(p, position + cursor, out + cursor, right ? right + cursor : NULL, frames - cursor);
    p->next_position = position + frames;
    return XTC_OK;
}

xtc_result xtc_midi_plan_render(xtc_midi_plan *p, uint64_t position, const xtc_midi_event *events, size_t count,
                                float *out, size_t frames)
{
    return xtc_midi_plan_render_stereo(p, position, events, count, out, NULL, frames);
}