//  ================================================================
//
//    graphics.c
//
//  ----------------------------------------------------------------
//
//    (C) 2025 Mitya Selivanov <guattari.tech>
//
//    Any use of this code is prohibited.
//
//  ================================================================

#ifndef GRAPHICS_HEADER_GUARD_
#define GRAPHICS_HEADER_GUARD_

#ifdef GRAPHICS_HEADER
#define RUNTIME_HEADER
#endif

#include "runtime.c"

//  ================================================================
//
//    Options
//
//  ================================================================

#ifndef EPSILON
#define EPSILON (1e-9)
#endif

#ifndef GRAPHICS_CACHE_DEPTH
#define GRAPHICS_CACHE_DEPTH 4
#endif

#ifndef GRAPHICS_CACHE_SIZE
#define GRAPHICS_CACHE_SIZE (128 * GRAPHICS_CACHE_DEPTH)
#endif

//  ================================================================
//
//    Vector math
//
//  ================================================================

vec3_f32 vec3_from_vec4_f32(vec4_f32 v);
vec4_f32 vec4_from_vec3_f32(vec3_f32 v, f32 w);
vec3_f32 vec3_f32_lerp     (vec3_f32 a, vec3_f32 b, f32 t);
vec4_f32 vec4_f32_lerp     (vec4_f32 a, vec4_f32 b, f32 t);

//  ================================================================

//  NOTE: Dimensions are in pixels.

typedef struct {
  f64 x;
  f64 y;
  f64 width;
  f64 height;
} Box;

//  TODO:
// typedef struct {
//   b8  enabled;
//   i64 x;
//   i64 y;
//   i64 width;
//   i64 height;
// } Bounding_Box;

typedef struct {
  i64 width;
  i64 height;
  i64 stride;
  // Bounding_Box bounds;
  u8 *pixels;
} Stencil_Buffer;

typedef struct {
  i64       width;
  i64       height;
  i64       stride;
  // Bounding_Box bounds;
  vec4_f32 *pixels;
} Pixel_Buffer;

enum {
  GRAPHICS_DRAW_PIXELS,
  GRAPHICS_FILL_RECTANGLE,
  GRAPHICS_FILL_TRIANGLE,
  GRAPHICS_FILL_TRIANGLES,
  GRAPHICS_FILL_QUAD,
  GRAPHICS_FILL_ELLIPSE,
  GRAPHICS_FILL_LINE,
  GRAPHICS_DRAW_TEXT_AREA,
  GRAPHICS_DRAW_TEXT_CURSOR,
};

typedef struct graphics_request_ {
  u16 op;
  union {
    struct {
      Box          area;
      Pixel_Buffer src;
    } pixels;
    struct {
      vec4_f32 color;
      Box      area;
    } rectangle;
    struct {
      vec4_f32 color;
      vec2     vertices[3];
    } triangle;
    struct {
      vec4_f32 color;
      vec2     position;
      vec2     scale;
      i64      num_triangles;
      vec2 *   vertices;
    } triangles;
    struct {
      vec4_f32 color;
      vec2     vertices[4];
    } quad;
    struct {
      vec4_f32 color;
      Box      area;
    } ellipse;
    struct {
      vec4_f32 color;
      vec2     vertices[2];
      f64      width;
    } line;
    struct {
      i32      font;
      vec4_f32 color;
      Box      area;
      vec2     max_size;
      i64      num_chars;
      c32 *    text;
    } text_area;
    struct {
      i32      font;
      vec4_f32 color;
      Box      area;
      vec2     max_size;
      i64      num_chars;
      c32 *    text;
      i64      cursor;
      i64      selection;
    } text_cursor;
  };
} Graphics_Request;

typedef struct {
  b8           disable_cache;
  vec2         scale;
  Pixel_Buffer dst;
} Graphics_Context;

//  ================================================================
//
//    User program interface
//
//  ================================================================

Box font_text_area_dispatch(i32 font, vec2 scale, i64 num_chars, c32 *text);

void font_render_to_stencil_dispatch(Stencil_Buffer dst, i32 font, vec2 position, vec2 scale, i64 num_chars, c32 *text);

//  ================================================================

vec3_f32 lab_from_rgb(vec3_f32 rgb);
vec3_f32 rgb_from_lab(vec3_f32 lab);
vec3_f32 lab_from_lch(vec3_f32 lch);
vec3_f32 lch_from_lab(vec3_f32 lab);
vec3_f32 rgb_from_lch(vec3_f32 lch);
vec3_f32 lch_from_rgb(vec3_f32 rgb);

//  NOTE: General rule to work with gamma is removing it,
//        then doing computations like blending, interpolation, etc,
//        then adding it back.

vec3_f32 rgb_gamma_add    (vec3_f32 rgb);
vec3_f32 rgb_gamma_remove (vec3_f32 rgb);
vec4_f32 rgba_gamma_add   (vec4_f32 rgb);
vec4_f32 rgba_gamma_remove(vec4_f32 rgb);

vec3_f32 rgb_gamma_lerp (vec3_f32 a, vec3_f32 b, f32 t);
vec4_f32 rgba_gamma_lerp(vec4_f32 a, vec4_f32 b, f32 t);
vec3_f32 rgb_mix        (vec3_f32 a, vec3_f32 b, f32 t);
vec4_f32 rgba_mix       (vec4_f32 a, vec4_f32 b, f32 t);
vec3_f32 lch_mix        (vec3_f32 a, vec3_f32 b, f32 t);
vec3_f32 lch_lerp       (vec3_f32 a, vec3_f32 b, f32 t);

u32      rgb_u32_from_f32(vec3_f32 color);
vec3_f32 rgb_f32_from_u32(u32 color);
u32      rgba_u32_from_f32(vec4_f32 color);
vec4_f32 rgba_f32_from_u32(u32 color);

b8 hit_rectangle(Box area,                    vec2 point);
b8 hit_triangle (vec2 vertices[3],            vec2 point);
b8 hit_quad     (vec2 vertices[4],            vec2 point);
b8 hit_ellipse  (Box area,                    vec2 point);
b8 hit_line     (vec2 vertices[2], f64 width, vec2 point);

void draw_pixels_to_buffer     (Pixel_Buffer dst, Box area, Pixel_Buffer src);
void fill_rectangle_to_buffer  (Pixel_Buffer dst, vec4_f32 color, Box area);
void fill_triangle_to_buffer   (Pixel_Buffer dst, vec4_f32 color, vec2 vertices[3]);
void fill_triangles_to_buffer  (Pixel_Buffer dst, vec4_f32 color, vec2 position, vec2 scale, i64 num_triangles, vec2 *vertices);
void fill_quad_to_buffer       (Pixel_Buffer dst, vec4_f32 color, vec2 vertices[4]);
void fill_ellipse_to_buffer    (Pixel_Buffer dst, vec4_f32 color, Box area);
void fill_line_to_buffer       (Pixel_Buffer dst, vec4_f32 color, vec2 vertices[2], f64 width);
void draw_text_area_to_buffer  (Pixel_Buffer dst, i32 font, vec4_f32 color, Box area, vec2 max_size, i64 num_chars, c32 *text);
void draw_text_cursor_to_buffer(Pixel_Buffer dst, i32 font, vec4_f32 color, Box area, vec2 max_size, i64 num_chars, c32 *text, i64 cursor, i64 selection);

void draw_pixels_cached     (Box area, Pixel_Buffer src);
void fill_rectangle_cached  (vec4_f32 color, Box area);
void fill_triangle_cached   (vec4_f32 color, vec2 vertices[3]);
void fill_triangles_cached  (vec4_f32 color, vec2 position, vec2 scale, i64 num_triangles, vec2 *vertices);
void fill_quad_cached       (vec4_f32 color, vec2 vertices[4]);
void fill_ellipse_cached    (vec4_f32 color, Box area);
void fill_line_cached       (vec4_f32 color, vec2 vertices[2], f64 width);
void draw_text_area_cached  (i32 font, vec4_f32 color, Box area, vec2 max_size, i64 num_chars, c32 *text);
void draw_text_cursor_cached(i32 font, vec4_f32 color, Box area, vec2 max_size, i64 num_chars, c32 *text, i64 cursor, i64 selection);

void draw_pixels     (Box area, Pixel_Buffer src);
void fill_rectangle  (vec4_f32 color, Box area);
void fill_triangle   (vec4_f32 color, vec2 vertices[3]);
void fill_triangles  (vec4_f32 color, vec2 position, vec2 scale, i64 num_triangles, vec2 *vertices);
void fill_quad       (vec4_f32 color, vec2 vertices[4]);
void fill_ellipse    (vec4_f32 color, Box area);
void fill_line       (vec4_f32 color, vec2 vertices[2], f64 width);
void draw_text_area  (i32 font, vec4_f32 color, Box area, vec2 max_size, i64 num_chars, c32 *text);
void draw_text_cursor(i32 font, vec4_f32 color, Box area, vec2 max_size, i64 num_chars, c32 *text, i64 cursor, i64 selection);

void perform_graphics_request(Graphics_Context context, Graphics_Request req);
void clean_graphics_requests_cache(i64 amount);

#endif // GRAPHICS_HEADER_GUARD_

//  ================================================================

#ifndef GRAPHICS_HEADER
#ifndef GRAPHICS_IMPL_GUARD_
#define GRAPHICS_IMPL_GUARD_

//  FIXME, PERF: Re-encode for faster rendering.
static u64 _bitfont[] = {
  0xbc0000000000, 0xc00300000, 0x5fd5040093f24fc9, 0xa00a2c2a1a280105, 0xc000415e6f, 0x400000020be0000, 0x1c38a8400000007d, 0x40002043e1020215, 0x408102000000010, 0x9800000000020002, 0xf913e00000033, 0x53200000207c8800, 0x3654880000099, 0x54b800000f840e00, 0xe953c000001a, 0x953e000000674080, 0x1e54b800000f, 0x490000000000240, 0x88a08000000, 0x20a220050a142850, 0x6520800000, 0x912f801eab260be, 0x800034952bf0001f, 0xc850bf0000921427, 0xf00010a54afc0003, 0xd29427800002142b, 0x840007e1023f0000, 0x7d09100000217e, 0x3f000188a08fc000, 0xc30c0cfc00000810, 0x27803f101013f00f, 0xc244bf0000f214, 0x4bf0002f21427800, 0xc254a480006c24, 0x407c00102fc08100, 0xf208080f0000fa0, 0x531007d81c607c0, 0xc208288c031141, 0x83fc00046954b10, 0x180e03000000, 0x41040000000ff04, 0x8102040810000404, 0x2a54600000000101, 0x309123e0000e, 0xc912180000a22447, 0x8000062a54700007, 0xe52a4300000029f0, 0xa0000602043e0001, 0x1d48000002074, 0x1f000003610f8000, 0x13e04f800000010, 0x470000780813e00f, 0x184893e0000e224, 0x23e0001f12243000, 0x82a54100000008, 0x40780000009f0200, 0xe208080e0001f20, 0xa22007981860780, 0x82082888022282, 0x16c200004ca95320, 0x7f000004, 0x408200000086d04, 0x8204,
};

enum {
  CHAR_NUM_BITS_X_ = 6,
  CHAR_NUM_BITS_Y_ = 7,
  CHAR_NUM_BITS_   = CHAR_NUM_BITS_X_ * CHAR_NUM_BITS_Y_,
};

static void put_pixel_(vec4_f32 *dst, vec4_f32 color) {
  if (color.w == 1.f)
    *dst = color;
  else {
    f32 dst_amount = 1. - color.w;

    *dst = (vec4_f32) {
      .x = dst->x * dst_amount + color.x * color.w,
      .y = dst->y * dst_amount + color.y * color.w,
      .z = dst->z * dst_amount + color.z * color.w,
      .w = max2_f32_(dst->w, color.w),
    };
  }
}

void draw_pixels_to_buffer(Pixel_Buffer dst, Box area, Pixel_Buffer src) {
  if (area.width < EPSILON || area.height < EPSILON)
    return;

  PROFILER_begin(PROFILE_DRAW_PIXELS);

  i64 i0 = (i64) floor(area.x + 0.5);
  i64 i1 = (i64) floor(area.x + area.width + 0.5);
  i64 j0 = (i64) floor(area.y + 0.5);
  i64 j1 = (i64) floor(area.y + area.height + 0.5);

  if (i0 < 0)          i0 = 0;
  if (i1 > dst.width)  i1 = dst.width;
  if (j0 < 0)          j0 = 0;
  if (j1 > dst.height) j1 = dst.height;

  //  FIXME, PERF: Use integer arithmetic.

  f64 di = src.width  / area.width;
  f64 dj = src.height / area.height;
  f64 jj = (j0 - area.y) * dj + dj * .5;
  for (i64 j = j0; j < j1; ++j, jj += dj) {
    if (jj < 0 || jj >= src.height) continue;
    vec4_f32 *d     = dst.pixels + j * dst.stride + i0;
    vec4_f32 *d_end = d + i1 - i0;
    vec4_f32 *s     = src.pixels + (i64) jj * src.stride;
    f64       ii    = (i0 - area.x) * di + di * .5;
    if (ii < 0 || ii >= src.width) continue;
    for (; d < d_end; ++d, ii += di)
      put_pixel_(d, s[(i64) ii]);
  }

  PROFILER_end(PROFILE_DRAW_PIXELS);
}

void fill_rectangle_to_buffer(Pixel_Buffer dst, vec4_f32 color, Box area) {
  PROFILER_begin(PROFILE_FILL_RECTANGLE);

  i64 i0 = (i64) floor(area.x + 0.5);
  i64 i1 = (i64) floor(area.x + area.width + 0.5);
  i64 j0 = (i64) floor(area.y + 0.5);
  i64 j1 = (i64) floor(area.y + area.height + 0.5);

  if (i0 < 0)          i0 = 0;
  if (i1 > dst.width)  i1 = dst.width;
  if (j0 < 0)          j0 = 0;
  if (j1 > dst.height) j1 = dst.height;

  for (i64 j = j0; j < j1; ++j) {
    vec4_f32 *p     = dst.pixels + j * dst.stride + i0;
    vec4_f32 *p_end = dst.pixels + j * dst.stride + i1;
    for (; p < p_end; ++p)
      put_pixel_(p, color);
  }

  PROFILER_end(PROFILE_FILL_RECTANGLE);
}

void fill_triangle_to_buffer(Pixel_Buffer dst, vec4_f32 color, vec2 vertices[3]) {
  PROFILER_begin(PROFILE_FILL_TRIANGLE);

  f64 x0 = vertices[0].x;
  f64 y0 = vertices[0].y;
  f64 x1 = vertices[1].x;
  f64 y1 = vertices[1].y;
  f64 x2 = vertices[2].x;
  f64 y2 = vertices[2].y;

  i64 i0 = (i64) floor(min3_(x0, x1, x2));
  i64 j0 = (i64) floor(min3_(y0, y1, y2));
  i64 i1 = (i64) ceil (max3_(x0, x1, x2));
  i64 j1 = (i64) ceil (max3_(y0, y1, y2));

  if (i0 < 0)           i0 = 0;
  if (i1 >= dst.width)  i1 = dst.width  - 1;
  if (j0 < 0)           j0 = 0;
  if (j1 >= dst.height) j1 = dst.height - 1;

  for (i64 j = j0; j <= j1; ++j) {
    //  FIXME, PERF: Calculate the exact intersection.

    i64 left  = i1;
    i64 right = i0;

    for (i64 i = i0; i <= i1; ++i)
      if (hit_triangle(vertices, (vec2) { (f64) i, (f64) j, })) {
        left = i;
        break;
      }

    for (i64 i = i1; i >= i0; --i)
      if (hit_triangle(vertices, (vec2) { (f64) i, (f64) j, })) {
        right = i;
        break;
      }

    vec4_f32 *p     = dst.pixels + j * dst.stride + left;
    vec4_f32 *p_end = dst.pixels + j * dst.stride + right + 1;
    for (; p < p_end; ++p)
      put_pixel_(p, color);
  }

  PROFILER_end(PROFILE_FILL_TRIANGLE);
}

static void triangles_area_(f64 *x0, f64 *y0, f64 *x1, f64 *y1, vec2 p, vec2 s, i64 num, vec2 *v) {
  if (num <= 0)
    return;

  *x0 = min3_(p.x + v[0].x * s.x, p.x + v[1].x * s.x, p.x + v[2].x * s.x);
  *y0 = min3_(p.y + v[0].y * s.y, p.y + v[1].y * s.y, p.y + v[2].y * s.y);
  *x1 = max3_(p.x + v[0].x * s.x, p.x + v[1].x * s.x, p.x + v[2].x * s.x);
  *y1 = max3_(p.y + v[0].y * s.y, p.y + v[1].y * s.y, p.y + v[2].y * s.y);

  for (i64 i = 1; i < num; ++i) {
    *x0 = min4_(*x0, p.x + v[i * 3].x * s.x, p.x + v[i * 3 + 1].x * s.x, p.x + v[i * 3 + 2].x * s.x);
    *y0 = min4_(*y0, p.y + v[i * 3].y * s.y, p.y + v[i * 3 + 1].y * s.y, p.y + v[i * 3 + 2].y * s.y);
    *x1 = max4_(*x1, p.x + v[i * 3].x * s.x, p.x + v[i * 3 + 1].x * s.x, p.x + v[i * 3 + 2].x * s.x);
    *y1 = max4_(*y1, p.y + v[i * 3].y * s.y, p.y + v[i * 3 + 1].y * s.y, p.y + v[i * 3 + 2].y * s.y);
  }
}

void fill_triangles_to_buffer(Pixel_Buffer dst, vec4_f32 color, vec2 position, vec2 scale, i64 num_triangles, vec2 *vertices) {
  PROFILER_begin(PROFILE_FILL_TRIANGLES);

  f64 x0, y0, x1, y1;
  triangles_area_(&x0, &y0, &x1, &y1, position, scale, num_triangles, vertices);

  i64 i0 = (i64) floor(x0);
  i64 j0 = (i64) floor(y0);
  i64 i1 = (i64) ceil (x1);
  i64 j1 = (i64) ceil (y1);

  if (i0 < 0)           i0 = 0;
  if (i1 >= dst.width)  i1 = dst.width  - 1;
  if (j0 < 0)           j0 = 0;
  if (j1 >= dst.height) j1 = dst.height - 1;

  //  FIXME, PERF: Use stencil buffer.

  for (i64 j = j0; j <= j1; ++j)
    for (i64 i = i0; i <= i1; ++i)
      for (i64 k = 0; k < num_triangles; ++k)
        if (hit_triangle(vertices + 3 * k, (vec2) { (f64) i, (f64) j, })) {
          put_pixel_(dst.pixels + j * dst.stride + i, color);
          break;
        }

  PROFILER_end(PROFILE_FILL_TRIANGLES);
}

void fill_quad_to_buffer(Pixel_Buffer dst, vec4_f32 color, vec2 vertices[4]) {
  PROFILER_begin(PROFILE_FILL_QUAD);

  f64 x0 = vertices[0].x;
  f64 y0 = vertices[0].y;
  f64 x1 = vertices[1].x;
  f64 y1 = vertices[1].y;
  f64 x2 = vertices[2].x;
  f64 y2 = vertices[2].y;
  f64 x3 = vertices[3].x;
  f64 y3 = vertices[3].y;

  i64 i0 = (i64) floor(min4_(x0, x1, x2, x3));
  i64 j0 = (i64) floor(min4_(y0, y1, y2, y3));
  i64 i1 = (i64) ceil (max4_(x0, x1, x2, x3));
  i64 j1 = (i64) ceil (max4_(y0, y1, y2, y3));

  if (i0 < 0)           i0 = 0;
  if (i1 >= dst.width)  i1 = dst.width  - 1;
  if (j0 < 0)           j0 = 0;
  if (j1 >= dst.height) j1 = dst.height - 1;

  for (i64 j = j0; j <= j1; ++j) {
    //  FIXME, PERF: Calculate the exact intersection.

    i64 left  = i1;
    i64 right = i0;

    for (i64 i = i0; i <= i1; ++i)
      if (hit_quad(vertices, (vec2) { (f64) i, (f64) j, })) {
        left = i;
        break;
      }

    for (i64 i = i1; i >= i0; --i)
      if (hit_quad(vertices, (vec2) { (f64) i, (f64) j, })) {
        right = i;
        break;
      }

    vec4_f32 *p     = dst.pixels + j * dst.stride + left;
    vec4_f32 *p_end = dst.pixels + j * dst.stride + right + 1;
    for (; p < p_end; ++p)
      put_pixel_(p, color);
  }

  PROFILER_end(PROFILE_FILL_QUAD);
}

void fill_ellipse_to_buffer(Pixel_Buffer dst, vec4_f32 color, Box area) {
  PROFILER_begin(PROFILE_FILL_ELLIPSE);

  f64 y1 = area.y + area.height;

  i64 j0 = (i64) floor(area.y);
  i64 j1 = (i64) ceil (y1);

  if (j0 < 0)           j0 = 0;
  if (j1 >= dst.height) j1 = dst.height - 1;

  f64 half_w = area.width  / 2.0;
  f64 half_h = area.height / 2.0;

  if (half_h < EPSILON)
    goto _finish;
  f64 inv_half_h = 1.0 / half_h;

  f64 cx = area.x + half_w;
  f64 cy = area.y + half_h;

  for (i64 j = j0; j <= j1; ++j) {
    f64 ry = (j - cy) * inv_half_h;
    if (ry > 1.0 - EPSILON || ry < -1.0 + EPSILON)
      continue;

    f64 dx = half_w * sqrt(1.0 - ry * ry);

    i64 left  = (i64) ceil (cx - dx - EPSILON);
    i64 right = (i64) floor(cx + dx + EPSILON);

    if (left  <  0)         left  = 0;
    if (right >= dst.width) right = dst.width - 1;

    vec4_f32 *p     = dst.pixels + j * dst.stride + left;
    vec4_f32 *p_end = dst.pixels + j * dst.stride + right + 1;
    for (; p < p_end; ++p)
      put_pixel_(p, color);
  }

_finish:
  PROFILER_end(PROFILE_FILL_ELLIPSE);
}

static b8 quad_from_line_(vec2 *dst, vec2 vertices[2], f64 width) {
  f64 x0 = vertices[0].x;
  f64 y0 = vertices[0].y;
  f64 x1 = vertices[1].x;
  f64 y1 = vertices[1].y;

  f64 dx = x1 - x0;
  f64 dy = y1 - y0;

  //  Tangent
  //
  f64 tx = -dy;
  f64 ty =  dx;
  f64 tl = sqrt(tx * tx + ty * ty);

  if (tl < EPSILON)
    return 0;

  f64 k = .5 / tl;
  tx *= width * k;
  ty *= width * k;

  dst[0] = (vec2) { x0 - tx, y0 - ty };
  dst[1] = (vec2) { x0 + tx, y0 + ty };
  dst[2] = (vec2) { x1 + tx, y1 + ty };
  dst[3] = (vec2) { x1 - tx, y1 - ty };
  return 1;
}

void fill_line_to_buffer(Pixel_Buffer dst, vec4_f32 color, vec2 vertices[2], f64 width) {
  if (width < EPSILON) return;
  PROFILER_begin(PROFILE_FILL_LINE);

  vec2 quad[4];
  if (!quad_from_line_(quad, vertices, width)) return;
  fill_quad_to_buffer(dst, color, quad);

  //  TODO: Also render line ends as circles, use stencil buffer.
  PROFILER_end(PROFILE_FILL_LINE);
}

static i64 char_column_offset_(c32 c, i64 column_index) {
  if (column_index < 0 || column_index >= CHAR_NUM_BITS_X_)
    return -1;
  return (c - 32) * CHAR_NUM_BITS_ + column_index * CHAR_NUM_BITS_Y_;
}

static b8 char_bit_(i64 column_offset, i64 row_index) {
  if (column_offset < 0 || row_index < 0 || row_index >= CHAR_NUM_BITS_Y_)
    return 0;

  i64 bit_index   = column_offset + row_index;
  i64 qword_index = bit_index / 64;
  if (qword_index < 0 || qword_index >= (i64) (sizeof _bitfont / sizeof *_bitfont))
    return 0;
  u64 mask        = 1ull << (bit_index % 64);
  return !!(_bitfont[qword_index] & mask);
}

static u64 char_column_convolved_(c32 c, i64 column_index) {
  if (column_index < 0 || column_index >= CHAR_NUM_BITS_X_)
    return 0;

  u64 column = 0;
  i64 offset = char_column_offset_(c, column_index);

  for (i64 y = 0; y < CHAR_NUM_BITS_Y_; ++y)
    if (char_bit_(offset, y))
      column |= 3ull << y;

  return column;
}

static b8 char_column_empty_(c32 c, i64 column_index) {
  if (column_index < 0 || column_index >= CHAR_NUM_BITS_X_)
    return 1;

  i64 offset = char_column_offset_(c, column_index);

  for (i64 y = 0; y < CHAR_NUM_BITS_Y_; ++y)
    if (char_bit_(offset, y))
      return 0;

  return 1;
}

static i64 char_width_(c32 c) {
  if (c < 32)
    return 0;
  if (c == ' ' || c > 127)
    return 4;

  i64 width = 0;

  for (; width < CHAR_NUM_BITS_X_; ++width)
    if (char_column_empty_(c, width) && char_column_empty_(c, width + 1))
      break;

  return width;
}

static i64 char_spacing_(i64 num_chars, c32 *text, i64 index) {
  if (text == NULL)
    return 0;

  if (index < 0 || index + 1 >= num_chars)
    return 0;

  u64 a = char_column_convolved_(text[index], char_width_(text[index]) - 1);
  u64 b = char_column_convolved_(text[index + 1], 0);

  if (!!(a & b))
    return 1;

  return 0;
}

static i64 text_cursor_(i64 num_chars, c32 *text) {
  if (text == NULL)
    return 0;

  i64 cursor = 0;

  for (i64 i = 0; i < num_chars; ++i) {
    if (text[i] <= ' ') {
      if (text[i] == '\n')
        cursor = 0;
      else if (text[i] == '\b' && i > 0)
        cursor -= char_width_(text[i - 1]) + char_spacing_(num_chars, text, i - 1);
      else if (text[i] == '\r')
        cursor = 0;
      else
        cursor += char_width_(' ') + char_spacing_(num_chars, text, i);
      continue;
    }
    cursor += char_width_(text[i]) + char_spacing_(num_chars, text, i);
  }

  return cursor;
}

static i64 enum_text_columns_(i64 num_chars, c32 *text) {
  if (text == NULL)
    return 0;

  i64 cols = 0;
  i64 n    = 0;

  for (i64 i = 0; i < num_chars; ++i) {
    if (text[i] <= ' ') {
      if (text[i] == '\n') {
        if (cols < n)
          cols = n;
        n = 0;
      } else if (text[i] == '\b' && i > 0) {
        if (cols < n)
          cols = n;
        n -= char_width_(text[i - 1]) + char_spacing_(num_chars, text, i - 1);
      } else if (text[i] == '\r') {
        if (cols < n)
          cols = n;
        n = 0;
      } else
        n += char_width_(' ') + char_spacing_(num_chars, text, i);
      continue;
    }
    n += char_width_(text[i]) + char_spacing_(num_chars, text, i);
  }

  if (cols < n)
    cols = n;

  return cols;
}

static i64 enum_text_rows_(i64 num_chars, c32 *text) {
  if (text == NULL)
    return 0;

  i64 rows = 0;

  for (i64 i = 0; i <= num_chars; ++i)
    if (i == num_chars || text[i] == '\n') {
      if (rows > 0)
        ++rows;
      rows += CHAR_NUM_BITS_Y_;
    }

  return rows;
}

static void draw_text_(Pixel_Buffer dst, vec4_f32 color, f64 x0, f64 y0, f64 scale_x, f64 scale_y, i64 num_chars, c32 *text) {
  f64 kx = scale_x;
  f64 h  = scale_y * CHAR_NUM_BITS_Y_;

  if (h < EPSILON)
    return;

  f64 x = x0;
  f64 y = y0;

  for (i64 n = 0; n < num_chars; ++n) {
    if (text[n] <= ' ') {
      if (text[n] == '\n') {
        x  = x0;
        y += scale_y * (CHAR_NUM_BITS_Y_ + 1);
      }
      else if (text[n] == '\b' && n > 0)
        x -= kx * (char_width_(text[n - 1]) + char_spacing_(num_chars, text, n - 1));
      else if (text[n] == '\r')
        x  = x0;
      else
        x += kx * (char_width_(' ') + char_spacing_(num_chars, text, n));
      continue;
    }

    i64 num_cols = char_width_(text[n]);
    f64 w        = num_cols * kx;

    if (w >= EPSILON) {
      i64 i0 = (i64) floor(x);
      i64 i1 = (i64) ceil (x + w);
      i64 j0 = (i64) floor(y);
      i64 j1 = (i64) ceil (y + h);

      if (i0 <  0)          i0 = 0;
      if (i1 >= dst.width)  i1 = dst.width - 1;
      if (j0 <  0)          j0 = 0;
      if (j1 >= dst.height) j1 = dst.height - 1;

      f64 w_inv = 1. / w;
      f64 h_inv = 1. / h;

      for (i64 i = i0; i <= i1; ++i) {
        i64 column = (i64) floor(((i - x + .5) * num_cols) * w_inv);
        i64 offset = char_column_offset_(text[n], column);

        for (i64 j = j0; j <= j1; ++j) {
          i64 row = (i64) floor(((j - y + .5) * CHAR_NUM_BITS_Y_) * h_inv);
          if (char_bit_(offset, row))
            put_pixel_(dst.pixels + j * dst.stride + i, color);
        }
      }
    }

    x += kx * (num_cols + char_spacing_(num_chars, text, n));
  }
}

#if !defined(GRAPHICS_ENABLE_FONT_CUSTOM_DISPATCH)
Box font_text_area_dispatch(i32 font, vec2 scale, i64 num_chars, c32 *text) {
  //  TODO

  (void) font;
  (void) scale;
  (void) num_chars;
  (void) text;

  return (Box) {0};
}

void font_render_to_stencil_dispatch(Stencil_Buffer dst, i32 font, vec2 position, vec2 scale, i64 num_chars, c32 *text) {
  //  TODO

  (void) dst;
  (void) font;
  (void) position;
  (void) scale;
  (void) num_chars;
  (void) text;
}
#endif // !defined(GRAPHICS_ENABLE_FONT_CUSTOM_DISPATCH)

void draw_text_area_to_buffer(Pixel_Buffer dst, i32 font, vec4_f32 color, Box area, vec2 max_size, i64 num_chars, c32 *text) {
  (void) font;

  if (text == NULL || num_chars <= 0 || max_size.x < EPSILON || max_size.y < EPSILON)
    return;

  PROFILER_begin(PROFILE_DRAW_TEXT_AREA);

  i64 num_columns = enum_text_columns_(num_chars, text);
  i64 num_rows    = enum_text_rows_(num_chars, text);

  f64 scale_x = area.width  / num_columns;
  f64 scale_y = area.height / num_rows;

  f64 kx = scale_x / max_size.x;
  f64 ky = scale_y / max_size.y;

  f64 k = kx < ky ? kx : ky;

  kx = k * max_size.x;
  ky = k * max_size.y;

  draw_text_(dst, color, area.x, area.y, kx, ky, num_chars, text);

  PROFILER_end(PROFILE_DRAW_TEXT_AREA);
}

void draw_text_cursor_to_buffer(Pixel_Buffer dst, i32 font, vec4_f32 color, Box area, vec2 max_size, i64 num_chars, c32 *text, i64 cursor, i64 selection) {
  (void) font;

  if (max_size.x < EPSILON || max_size.y < EPSILON)
    return;

  PROFILER_begin(PROFILE_DRAW_TEXT_CURSOR);

  i64 num_columns = enum_text_columns_(num_chars, text);
  i64 num_rows    = enum_text_rows_(num_chars, text);
  i64 cursor_x    = text_cursor_(cursor, text);
  i64 cursor_y    = enum_text_rows_(cursor, text);

  f64 scale_x = area.width  / num_columns;
  f64 scale_y = area.height / num_rows;

  f64 kx = scale_x / max_size.x;
  f64 ky = scale_y / max_size.y;

  f64 k = kx < ky ? kx : ky;

  kx = k * max_size.x;
  ky = k * max_size.y;

  if (selection != 0) {
    i64 selection_x, selection_y;

    if (selection > 0) {
      selection_x = text_cursor_(cursor + selection, text);
      selection_y = enum_text_rows_(cursor + selection, text);
    } else {
      selection_x = cursor_x;
      selection_y = cursor_y;
      cursor_x    = text_cursor_(cursor + selection, text);
      cursor_y    = enum_text_rows_(cursor + selection, text);
    }

    if (cursor_y == selection_y)
      fill_rectangle_to_buffer(
        dst,
        color,
        (Box) {
          .x      = area.x + kx * cursor_x,
          .y      = area.y + ky * cursor_y - ky * (CHAR_NUM_BITS_Y_ + 1),
          .width  = kx * (selection_x - cursor_x),
          .height = ky * (CHAR_NUM_BITS_Y_ + 1),
        }
      );
    else {
      fill_rectangle_to_buffer(
        dst,
        color,
        (Box) {
          .x      = area.x + kx * cursor_x,
          .y      = area.y + ky * cursor_y - ky * (CHAR_NUM_BITS_Y_ + 1),
          .width  = kx * (num_columns - cursor_x),
          .height = ky * (CHAR_NUM_BITS_Y_ + 1),
        }
      );
      for (i64 j = cursor_y + CHAR_NUM_BITS_Y_ + 1; j < selection_y; j += CHAR_NUM_BITS_Y_ + 1)
        fill_rectangle_to_buffer(
          dst,
          color,
          (Box) {
            .x      = area.x,
            .y      = area.y + ky * j - ky * (CHAR_NUM_BITS_Y_ + 1),
            .width  = kx * num_columns,
            .height = ky * (CHAR_NUM_BITS_Y_ + 1),
          }
        );
      fill_rectangle_to_buffer(
        dst,
        color,
        (Box) {
          .x      = area.x,
          .y      = area.y + ky * selection_y - ky * (CHAR_NUM_BITS_Y_ + 1),
          .width  = kx * selection_x,
          .height = ky * (CHAR_NUM_BITS_Y_ + 1),
        }
      );
    }
  } else
    fill_rectangle_to_buffer(
      dst,
      color,
      (Box) {
        .x      = area.x + kx * cursor_x,
        .y      = area.y + ky * cursor_y - ky * CHAR_NUM_BITS_Y_,
        .width  = kx * .5,
        .height = ky * (CHAR_NUM_BITS_Y_ - 1),
      }
    );

  PROFILER_end(PROFILE_DRAW_TEXT_CURSOR);
}

//  ================================================================

vec3_f32 vec3_from_vec4_f32(vec4_f32 v) {
  return (vec3_f32) {
    .x = v.x,
    .y = v.y,
    .z = v.z,
  };
}

vec4_f32 vec4_from_vec3_f32(vec3_f32 v, f32 w) {
  return (vec4_f32) {
    .x = v.x,
    .y = v.y,
    .z = v.z,
    .w = w,
  };
}

vec3_f32 vec3_f32_lerp(vec3_f32 a, vec3_f32 b, f32 t) {
  return (vec3_f32) {
    .x = a.x + (b.x - a.x) * t,
    .y = a.y + (b.y - a.y) * t,
    .z = a.z + (b.z - a.z) * t,
  };
}

vec4_f32 vec4_f32_lerp(vec4_f32 a, vec4_f32 b, f32 t) {
  return (vec4_f32) {
    .x = a.x + (b.x - a.x) * t,
    .y = a.y + (b.y - a.y) * t,
    .z = a.z + (b.z - a.z) * t,
    .w = a.w + (b.w - a.w) * t,
  };
}

//  ================================================================

vec3_f32 rgb_gamma_add(vec3_f32 rgb) {
  return (vec3_f32) {
    .x = gamma_(rgb.x),
    .y = gamma_(rgb.y),
    .z = gamma_(rgb.z),
  };
}

vec3_f32 rgb_gamma_remove(vec3_f32 rgb) {
  return (vec3_f32) {
    .x = gamma_re_(rgb.x),
    .y = gamma_re_(rgb.y),
    .z = gamma_re_(rgb.z),
  };
}

vec4_f32 rgba_gamma_add(vec4_f32 rgba) {
  return (vec4_f32) {
    .x = gamma_(rgba.x),
    .y = gamma_(rgba.y),
    .z = gamma_(rgba.z),
    .w = gamma_(rgba.w),
  };
}

vec4_f32 rgba_gamma_remove(vec4_f32 rgba) {
  return (vec4_f32) {
    .x = gamma_re_(rgba.x),
    .y = gamma_re_(rgba.y),
    .z = gamma_re_(rgba.z),
    .w = gamma_re_(rgba.w),
  };
}

vec3_f32 lab_from_rgb(vec3_f32 rgb) {
  f64 l = 0.4122214708 * rgb.x + 0.5363325363 * rgb.y + 0.0514459929 * rgb.z;
  f64 m = 0.2119034982 * rgb.x + 0.6806995451 * rgb.y + 0.1073969566 * rgb.z;
  f64 s = 0.0883024619 * rgb.x + 0.2817188376 * rgb.y + 0.6299787005 * rgb.z;

  f64 l_ = cbrt(l);
  f64 m_ = cbrt(m);
  f64 s_ = cbrt(s);

  return (vec3_f32) {
    .x = (f32) (0.2104542553 * l_ + 0.7936177850 * m_ - 0.0040720468 * s_),
    .y = (f32) (1.9779984951 * l_ - 2.4285922050 * m_ + 0.4505937099 * s_),
    .z = (f32) (0.0259040371 * l_ + 0.7827717662 * m_ - 0.8086757660 * s_),
  };
}

vec3_f32 rgb_from_lab(vec3_f32 lab) {
  f64 l_ = lab.x + 0.3963377774 * lab.y + 0.2158037573 * lab.z;
  f64 m_ = lab.x - 0.1055613458 * lab.y - 0.0638541728 * lab.z;
  f64 s_ = lab.x - 0.0894841775 * lab.y - 1.2914855480 * lab.z;

  f64 l = l_ * l_ * l_;
  f64 m = m_ * m_ * m_;
  f64 s = s_ * s_ * s_;

  return (vec3_f32) {
    .x = (f32) (+4.0767416621 * l - 3.3077115913 * m + 0.2309699292 * s),
    .y = (f32) (-1.2684380046 * l + 2.6097574011 * m - 0.3413193965 * s),
    .z = (f32) (-0.0041960863 * l - 0.7034186147 * m + 1.7076147010 * s),
  };
}

vec3_f32 lch_from_lab(vec3_f32 lab) {
  f64 a = lab.y;
  f64 b = lab.z;

  return (vec3_f32) {
    .x = lab.x,
    .y = (f32) sqrt (a * a + b * b),
    .z = (f32) atan2(b, a),
  };
}

vec3_f32 lab_from_lch(vec3_f32 lch) {
  return (vec3_f32) {
    .x = lch.x,
    .y = (f32) (lch.y * cos(lch.z)),
    .z = (f32) (lch.y * sin(lch.z)),
  };
}

vec3_f32 rgb_from_lch(vec3_f32 lch) {
  return rgb_from_lab(lab_from_lch(lch));
}

vec3_f32 lch_from_rgb(vec3_f32 rgb) {
  return lch_from_lab(lab_from_rgb(rgb));
}

vec3_f32 rgb_gamma_lerp (vec3_f32 a, vec3_f32 b, f32 t) {
  return rgb_gamma_add(vec3_f32_lerp(
    rgb_gamma_remove(a),
    rgb_gamma_remove(b),
    t
  ));
}

vec4_f32 rgba_gamma_lerp(vec4_f32 a, vec4_f32 b, f32 t) {
  return rgba_gamma_add(vec4_f32_lerp(
    rgba_gamma_remove(a),
    rgba_gamma_remove(b),
    t
  ));
}

vec3_f32 rgb_mix(vec3_f32 a, vec3_f32 b, f32 t) {
  vec3_f32 a_lch = lch_from_rgb(a);
  vec3_f32 b_lch = lch_from_rgb(b);

  return rgb_from_lch((vec3_f32) {
    a_lch.x + (b_lch.x - a_lch.x) * t,
    a_lch.y + (b_lch.y - a_lch.y) * t,
    lch_from_rgb(rgb_gamma_lerp(a, b, t)).z,
  });
}

vec4_f32 rgba_mix(vec4_f32 a, vec4_f32 b, f32 t) {
  vec3_f32 a_lch = lch_from_rgb(vec3_from_vec4_f32(a));
  vec3_f32 b_lch = lch_from_rgb(vec3_from_vec4_f32(b));
  vec4_f32 ab    = rgba_gamma_lerp(a, b, t);

  return vec4_from_vec3_f32(rgb_from_lch((vec3_f32) {
    a_lch.x + (b_lch.x - a_lch.x) * t,
    a_lch.y + (b_lch.y - a_lch.y) * t,
    lch_from_rgb(vec3_from_vec4_f32(ab)).z,
  }), ab.w);
}

vec3_f32 lch_mix(vec3_f32 a, vec3_f32 b, f32 t) {
  return (vec3_f32) {
    a.x + (b.x - a.x) * t,
    a.y + (b.y - a.y) * t,
    lch_from_rgb(rgb_gamma_lerp(rgb_from_lch(a), rgb_from_lch(b), t)).z,
  };
}

vec3_f32 lch_lerp(vec3_f32 a, vec3_f32 b, f32 t) {
  f32 delta_hue = b.z - a.z;
  if (delta_hue >  M_PI) delta_hue -= M_PI * 2;
  if (delta_hue < -M_PI) delta_hue += M_PI * 2;

  return (vec3_f32) {
    .x = a.x + (b.x - a.x) * t,
    .y = a.y + (b.y - a.y) * t,
    .z = a.z +  delta_hue  * t,
  };
}

u32 rgb_u32_from_f32(vec3_f32 color) {
  i32 ir = (i32) floor(color.x * 255. + .5);
  i32 ig = (i32) floor(color.y * 255. + .5);
  i32 ib = (i32) floor(color.z * 255. + .5);

  u32 r = ir < 0 ? 0u : ir > 255 ? 255u : (u32) ir;
  u32 g = ig < 0 ? 0u : ig > 255 ? 255u : (u32) ig;
  u32 b = ib < 0 ? 0u : ib > 255 ? 255u : (u32) ib;

  return (r << 16) | (g << 8) | b;
}

vec3_f32 rgb_f32_from_u32(u32 color) {
  return (vec3_f32) {
    .x = ((color & 0xff0000) >> 16) / 255.f,
    .y = ((color & 0x00ff00) >>  8) / 255.f,
    .z =  (color & 0x0000ff)        / 255.f,
  };
}

u32 rgba_u32_from_f32(vec4_f32 color) {
  i32 ir = (i32) floor(color.x * 255. + .5);
  i32 ig = (i32) floor(color.y * 255. + .5);
  i32 ib = (i32) floor(color.z * 255. + .5);
  i32 ia = (i32) floor(color.w * 255. + .5);

  u32 r = ir < 0 ? 0u : ir > 255 ? 255u : (u32) ir;
  u32 g = ig < 0 ? 0u : ig > 255 ? 255u : (u32) ig;
  u32 b = ib < 0 ? 0u : ib > 255 ? 255u : (u32) ib;
  u32 a = ia < 0 ? 0u : ia > 255 ? 255u : (u32) ia;

  return (a << 24) | (r << 16) | (g << 8) | b;
}

vec4_f32 rgba_f32_from_u32(u32 color) {
  return (vec4_f32) {
    .x = ((color & 0x00ff0000) >> 16) / 255.f,
    .y = ((color & 0x0000ff00) >>  8) / 255.f,
    .z =  (color & 0x000000ff)        / 255.f,
    .w = ((color & 0xff000000) >> 24) / 255.f,
  };
}

b8 hit_rectangle(Box area, vec2 point) {
  return point.x >= area.x && point.x < area.x + area.width && point.y >= area.y && point.y < area.y + area.height;
}

static b8 same_sign_(f64 a, f64 b) {
  if (a >=  EPSILON && b <= -EPSILON) return 0;
  if (a <= -EPSILON && b >=  EPSILON) return 0;
  return 1;
}

b8 hit_triangle(vec2 vertices[3], vec2 point) {
  f64 x0 = vertices[0].x;
  f64 y0 = vertices[0].y;
  f64 x1 = vertices[1].x;
  f64 y1 = vertices[1].y;
  f64 x2 = vertices[2].x;
  f64 y2 = vertices[2].y;
  f64 px = point.x;
  f64 py = point.y;

  if ( (px < x0 && px < x1 && px < x2)
    || (px > x0 && px > x1 && px > x2)
    || (py < y0 && py < y1 && py < y2)
    || (py > y0 && py > y1 && py > y2) )
    return 0;

  f64 x10 = x1 - x0;
  f64 x21 = x2 - x1;
  f64 x02 = x0 - x2;

  f64 y10 = y1 - y0;
  f64 y21 = y2 - y1;
  f64 y02 = y0 - y2;

  //  Z-components of cross-products

  f64 z0  = x02 * y10 - x10 * y02;
  f64 z1  = x10 * y21 - x21 * y10;
  f64 z2  = x21 * y02 - x02 * y21;

  f64 pz0 = x02 * (py - y0) - (px - x0) * y02;
  f64 pz1 = x10 * (py - y1) - (px - x1) * y10;
  f64 pz2 = x21 * (py - y2) - (px - x2) * y21;

  return same_sign_(z0, pz0)
      && same_sign_(z1, pz1)
      && same_sign_(z2, pz2);
}

b8 hit_quad(vec2 vertices[4], vec2 point) {
  f64 x0 = vertices[0].x;
  f64 y0 = vertices[0].y;
  f64 x1 = vertices[1].x;
  f64 y1 = vertices[1].y;
  f64 x2 = vertices[2].x;
  f64 y2 = vertices[2].y;
  f64 x3 = vertices[3].x;
  f64 y3 = vertices[3].y;

  return hit_triangle((vec2[3]) { { x0, y0 }, { x1, y1 }, { x2, y2 }, }, point)
      || hit_triangle((vec2[3]) { { x0, y0 }, { x2, y2 }, { x3, y3 }, }, point);
}

b8 hit_ellipse(Box area, vec2 point) {
  f64 dw = area.width  / 2;
  f64 dh = area.height / 2;

  if (dw < EPSILON || dh < EPSILON)
    return 0;

  f64 cx = area.x + dw;
  f64 cy = area.y + dh;
  f64 kx = 1. / dw;
  f64 ky = 1. / dh;

  f64 dx = (point.x - cx) * kx;
  f64 dy = (point.y - cy) * ky;

  return dx * dx + dy * dy - 1.0 < EPSILON;
}

b8 hit_line(vec2 vertices[2], f64 width, vec2 point) {
  f64 x0 = vertices[0].x;
  f64 y0 = vertices[0].y;
  f64 x1 = vertices[1].x;
  f64 y1 = vertices[1].y;

  f64 dx = x1 - x0;
  f64 dy = y1 - y0;

  //  Tangent
  //
  f64 tx = -dy;
  f64 ty = dx;
  f64 tl = sqrt(tx * tx + ty * ty);
  if (tl >= EPSILON) {
    tx /= tl;
    ty /= tl;
  }
  tx *= width * .5;
  ty *= width * .5;

  return hit_quad((vec2[4]) {
    { x0 - tx, y0 - ty },
    { x0 + tx, y0 + ty },
    { x1 + tx, y1 + ty },
    { x1 - tx, y1 - ty },
  }, point);
}

void draw_pixels_cached(Box area, Pixel_Buffer src) {
  perform_graphics_request(
    (Graphics_Context) {0},
    (Graphics_Request) {
      .op = GRAPHICS_DRAW_PIXELS,
      .pixels = {
        .area = area,
        .src  = src,
      },
    }
  );
}

void fill_rectangle_cached(vec4_f32 color, Box area) {
  perform_graphics_request(
    (Graphics_Context) {0},
    (Graphics_Request) {
      .op = GRAPHICS_FILL_RECTANGLE,
      .rectangle = {
        .color = color,
        .area  = area,
      },
    }
  );
}

void fill_triangle_cached(vec4_f32 color, vec2 vertices[3]) {
  perform_graphics_request(
    (Graphics_Context) {0},
    (Graphics_Request) {
      .op = GRAPHICS_FILL_TRIANGLE,
      .triangle = {
        .color = color,
        .vertices = {
          vertices[0],
          vertices[1],
          vertices[2],
        },
      },
    }
  );
}

void fill_triangles_cached(vec4_f32 color, vec2 position, vec2 scale, i64 num_triangles, vec2 *vertices) {
  perform_graphics_request(
    (Graphics_Context) {0},
    (Graphics_Request) {
      .op        = GRAPHICS_FILL_TRIANGLES,
      .triangles = {
        .color         = color,
        .position      = position,
        .scale         = scale,
        .num_triangles = num_triangles,
        .vertices      = vertices,
      },
    }
  );
}

void fill_quad_cached(vec4_f32 color, vec2 vertices[4]) {
  perform_graphics_request(
    (Graphics_Context) {0},
    (Graphics_Request) {
      .op = GRAPHICS_FILL_QUAD,
      .quad = {
        .color = color,
        .vertices = {
          vertices[0],
          vertices[1],
          vertices[2],
          vertices[3],
        },
      },
    }
  );
}

void fill_ellipse_cached(vec4_f32 color, Box area) {
  perform_graphics_request(
    (Graphics_Context) {0},
    (Graphics_Request) {
      .op = GRAPHICS_FILL_ELLIPSE,
      .ellipse = {
        .color = color,
        .area  = area,
      },
    }
  );
}

void fill_line_cached(vec4_f32 color, vec2 vertices[2], f64 width) {
  perform_graphics_request(
    (Graphics_Context) {0},
    (Graphics_Request) {
      .op = GRAPHICS_FILL_LINE,
      .line = {
        .color    = color,
        .vertices = { vertices[0], vertices[1], },
        .width    = width,
      },
    }
  );
}

void draw_text_area_cached(i32 font, vec4_f32 color, Box area, vec2 max_size, i64 num_chars, c32 *text) {
  perform_graphics_request(
    (Graphics_Context) {0},
    (Graphics_Request) {
      .op = GRAPHICS_DRAW_TEXT_AREA,
      .text_area = {
        .font      = font,
        .color     = color,
        .area      = area,
        .max_size  = max_size,
        .num_chars = num_chars,
        .text      = text,
      },
    }
  );
}

void draw_text_cursor_cached(i32 font, vec4_f32 color, Box area, vec2 max_size, i64 num_chars, c32 *text, i64 cursor, i64 selection) {
  perform_graphics_request(
    (Graphics_Context) {0},
    (Graphics_Request) {
      .op = GRAPHICS_DRAW_TEXT_CURSOR,
      .text_cursor = {
        .font      = font,
        .color     = color,
        .area      = area,
        .max_size  = max_size,
        .num_chars = num_chars,
        .text      = text,
        .cursor    = cursor,
        .selection = selection,
      },
    }
  );
}

void draw_pixels(Box area, Pixel_Buffer src) {
  perform_graphics_request(
    (Graphics_Context) { .disable_cache = 1, },
    (Graphics_Request) {
      .op = GRAPHICS_DRAW_PIXELS,
      .pixels = {
        .area = area,
        .src  = src,
      },
    }
  );
}

void fill_rectangle(vec4_f32 color, Box area) {
  perform_graphics_request(
    (Graphics_Context) { .disable_cache = 1, },
    (Graphics_Request) {
      .op = GRAPHICS_FILL_RECTANGLE,
      .rectangle = {
        .color = color,
        .area  = area,
      },
    }
  );
}

void fill_triangle(vec4_f32 color, vec2 vertices[3]) {
  perform_graphics_request(
    (Graphics_Context) { .disable_cache = 1, },
    (Graphics_Request) {
      .op = GRAPHICS_FILL_TRIANGLE,
      .triangle = {
        .color = color,
        .vertices = {
          vertices[0],
          vertices[1],
          vertices[2],
        },
      },
    }
  );
}

void fill_triangles(vec4_f32 color, vec2 position, vec2 scale, i64 num_triangles, vec2 *vertices) {
  perform_graphics_request(
    (Graphics_Context) { .disable_cache = 1, },
    (Graphics_Request) {
      .op        = GRAPHICS_FILL_TRIANGLES,
      .triangles = {
        .color         = color,
        .position      = position,
        .scale         = scale,
        .num_triangles = num_triangles,
        .vertices      = vertices,
      },
    }
  );
}

void fill_quad(vec4_f32 color, vec2 vertices[4]) {
  perform_graphics_request(
    (Graphics_Context) { .disable_cache = 1, },
    (Graphics_Request) {
      .op = GRAPHICS_FILL_QUAD,
      .quad = {
        .color = color,
        .vertices = {
          vertices[0],
          vertices[1],
          vertices[2],
          vertices[3],
        },
      },
    }
  );
}

void fill_ellipse(vec4_f32 color, Box area) {
  perform_graphics_request(
    (Graphics_Context) { .disable_cache = 1, },
    (Graphics_Request) {
      .op = GRAPHICS_FILL_ELLIPSE,
      .ellipse = {
        .color = color,
        .area  = area,
      },
    }
  );
}

void fill_line(vec4_f32 color, vec2 vertices[2], f64 width) {
  perform_graphics_request(
    (Graphics_Context) { .disable_cache = 1, },
    (Graphics_Request) {
      .op = GRAPHICS_FILL_LINE,
      .line = {
        .color    = color,
        .vertices = { vertices[0], vertices[1], },
        .width    = width,
      },
    }
  );
}

void draw_text_area(i32 font, vec4_f32 color, Box area, vec2 max_size, i64 num_chars, c32 *text) {
  perform_graphics_request(
    (Graphics_Context) { .disable_cache = 1, },
    (Graphics_Request) {
      .op = GRAPHICS_DRAW_TEXT_AREA,
      .text_area = {
        .font      = font,
        .color     = color,
        .area      = area,
        .max_size  = max_size,
        .num_chars = num_chars,
        .text      = text,
      },
    }
  );
}

void draw_text_cursor(i32 font, vec4_f32 color, Box area, vec2 max_size, i64 num_chars, c32 *text, i64 cursor, i64 selection) {
  perform_graphics_request(
    (Graphics_Context) { .disable_cache = 1, },
    (Graphics_Request) {
      .op = GRAPHICS_DRAW_TEXT_CURSOR,
      .text_cursor = {
        .font      = font,
        .color     = color,
        .area      = area,
        .max_size  = max_size,
        .num_chars = num_chars,
        .text      = text,
        .cursor    = cursor,
        .selection = selection,
      },
    }
  );
}

//  ================================================================

typedef struct {
  Graphics_Request req;
  i64 x;
  i64 y;
  i64 width;
  i64 height;
} Graphics_Request_Norm_;

typedef union {
  u8  hash[BLAKE2B_OUTBYTES];
  u64 index;
} Graphics_Request_Id_;

typedef struct {
  b8                   occupied;
  Graphics_Request_Id_ id;
  Graphics_Request     req;
  i64                  width;
  i64                  height;
  i64                  buffer_len;
  vec4_f32 *           buffer;
} Graphics_Cache_Slot_;

static Graphics_Cache_Slot_ _graphics_cache[GRAPHICS_CACHE_SIZE] = {0};

static i64 _graphics_cache_gc_index = 0;

static void graphics_request_hash_(Blake2b_State *S, Graphics_Request req) {
  switch (req.op) {
    case GRAPHICS_FILL_TRIANGLES:
      blake2b_update(S, (u8 *) &req.triangles.color, sizeof req.triangles.color);
      blake2b_update(S, (u8 *) &req.triangles.position, sizeof req.triangles.position);
      blake2b_update(S, (u8 *) &req.triangles.scale, sizeof req.triangles.scale);
      blake2b_update(S, (u8 *) &req.triangles.num_triangles, sizeof req.triangles.num_triangles);
      blake2b_update(S, (u8 *) req.triangles.vertices, req.triangles.num_triangles * 3 * sizeof *req.triangles.vertices);
      break;

    case GRAPHICS_DRAW_PIXELS:
      blake2b_update(S, (u8 *) &req.pixels.area, sizeof req.pixels.area);
      blake2b_update(S, (u8 *) &req.pixels.src.width, sizeof req.pixels.src.width);
      blake2b_update(S, (u8 *) &req.pixels.src.height, sizeof req.pixels.src.height);
      blake2b_update(S, (u8 *) &req.pixels.src.stride, sizeof req.pixels.src.stride);
      for (i64 j = 0; j < req.pixels.src.height; ++j)
        blake2b_update(
          S,
          (u8 *) (req.pixels.src.pixels + j * req.pixels.src.stride),
          req.pixels.src.width * sizeof *req.pixels.src.pixels
        );
      break;

    case GRAPHICS_DRAW_TEXT_AREA:
      blake2b_update(S, (u8 *) &req.text_area.font, sizeof req.text_area.font);
      blake2b_update(S, (u8 *) &req.text_area.color, sizeof req.text_area.color);
      blake2b_update(S, (u8 *) &req.text_area.area, sizeof req.text_area.area);
      blake2b_update(S, (u8 *) &req.text_area.max_size, sizeof req.text_area.max_size);
      blake2b_update(S, (u8 *) &req.text_area.num_chars, sizeof req.text_area.num_chars);
      blake2b_update(
        S,
        (u8 *) req.text_area.text,
        req.text_area.num_chars * sizeof *req.text_area.text
      );
      break;

    case GRAPHICS_DRAW_TEXT_CURSOR:
      blake2b_update(S, (u8 *) &req.text_cursor.font, sizeof req.text_cursor.font);
      blake2b_update(S, (u8 *) &req.text_cursor.color, sizeof req.text_cursor.color);
      blake2b_update(S, (u8 *) &req.text_cursor.area, sizeof req.text_cursor.area);
      blake2b_update(S, (u8 *) &req.text_cursor.max_size, sizeof req.text_cursor.max_size);
      blake2b_update(S, (u8 *) &req.text_cursor.num_chars, sizeof req.text_cursor.num_chars);
      blake2b_update(S, (u8 *) req.text_cursor.text, req.text_cursor.num_chars * sizeof *req.text_cursor.text);
      blake2b_update(S, (u8 *) &req.text_cursor.cursor, sizeof req.text_cursor.cursor);
      blake2b_update(
        S,
        (u8 *) &req.text_cursor.selection,
        sizeof req.text_cursor.selection
      );
      break;

    default:
      blake2b_update(S, (u8 *) &req, sizeof req);
  }
}

static Graphics_Request_Id_ graphics_request_id_(Graphics_Request req) {
  Blake2b_State S[1];
  blake2b_init(S, BLAKE2B_OUTBYTES);
  graphics_request_hash_(S, req);
  Graphics_Request_Id_ id = {0};
  blake2b_final(S, id.hash, BLAKE2B_OUTBYTES);
  return id;
}

static Graphics_Request graphics_request_scaled_(Graphics_Request req, vec2 scale) {
  Graphics_Request scaled = req;

  switch (req.op) {
    case GRAPHICS_DRAW_PIXELS:
      scaled.pixels.area.x      *= scale.x;
      scaled.pixels.area.y      *= scale.y;
      scaled.pixels.area.width  *= scale.x;
      scaled.pixels.area.height *= scale.y;
      break;

    case GRAPHICS_FILL_RECTANGLE:
      scaled.rectangle.area.x      *= scale.x;
      scaled.rectangle.area.y      *= scale.y;
      scaled.rectangle.area.width  *= scale.x;
      scaled.rectangle.area.height *= scale.y;
      break;

    case GRAPHICS_FILL_TRIANGLE:
      scaled.triangle.vertices[0].x *= scale.x;
      scaled.triangle.vertices[0].y *= scale.y;
      scaled.triangle.vertices[1].x *= scale.x;
      scaled.triangle.vertices[1].y *= scale.y;
      scaled.triangle.vertices[2].x *= scale.x;
      scaled.triangle.vertices[2].y *= scale.y;
      break;

    case GRAPHICS_FILL_TRIANGLES:
      scaled.triangles.position.x *= scale.x;
      scaled.triangles.position.y *= scale.y;
      scaled.triangles.scale.x    *= scale.x;
      scaled.triangles.scale.y    *= scale.y;
      break;

    case GRAPHICS_FILL_QUAD:
      scaled.quad.vertices[0].x *= scale.x;
      scaled.quad.vertices[0].y *= scale.y;
      scaled.quad.vertices[1].x *= scale.x;
      scaled.quad.vertices[1].y *= scale.y;
      scaled.quad.vertices[2].x *= scale.x;
      scaled.quad.vertices[2].y *= scale.y;
      scaled.quad.vertices[3].x *= scale.x;
      scaled.quad.vertices[3].y *= scale.y;
      break;

    case GRAPHICS_FILL_ELLIPSE:
      scaled.ellipse.area.x      *= scale.x;
      scaled.ellipse.area.y      *= scale.y;
      scaled.ellipse.area.width  *= scale.x;
      scaled.ellipse.area.height *= scale.y;
      break;

    case GRAPHICS_FILL_LINE:
      scaled.line.vertices[0].x *= scale.x;
      scaled.line.vertices[0].y *= scale.y;
      scaled.line.vertices[1].x *= scale.x;
      scaled.line.vertices[1].y *= scale.y;

      f64 dx = req.line.vertices[1].x - req.line.vertices[0].x;
      f64 dy = req.line.vertices[1].y - req.line.vertices[0].y;

      f64 l = sqrt(dx * dx + dy * dy);
      if (l > EPSILON) {
        dx /= l;
        dy /= l;

        f64 tx =  dy * scale.x;
        f64 ty = -dx * scale.y;

        scaled.line.width *= sqrt(tx * tx + ty * ty);
      } else
        scaled.line.width *= (scale.x + scale.y) / 2.0;

      break;

    case GRAPHICS_DRAW_TEXT_AREA:
      scaled.text_area.area.x      *= scale.x;
      scaled.text_area.area.y      *= scale.y;
      scaled.text_area.area.width  *= scale.x;
      scaled.text_area.area.height *= scale.y;
      scaled.text_area.max_size.x  *= scale.x;
      scaled.text_area.max_size.y  *= scale.y;
      break;

    case GRAPHICS_DRAW_TEXT_CURSOR:
      scaled.text_cursor.area.x      *= scale.x;
      scaled.text_cursor.area.y      *= scale.y;
      scaled.text_cursor.area.width  *= scale.x;
      scaled.text_cursor.area.height *= scale.y;
      scaled.text_cursor.max_size.x  *= scale.x;
      scaled.text_cursor.max_size.y  *= scale.y;
      break;

    default:;
  }

  return scaled;
}

static Graphics_Request graphics_request_moved_(Graphics_Request req, vec2 offset) {
  Graphics_Request moved = req;

  switch (req.op) {
    case GRAPHICS_DRAW_PIXELS:
      moved.pixels.area.x += offset.x;
      moved.pixels.area.y += offset.y;
      break;

    case GRAPHICS_FILL_RECTANGLE:
      moved.rectangle.area.x += offset.x;
      moved.rectangle.area.y += offset.y;
      break;

    case GRAPHICS_FILL_TRIANGLE:
      moved.triangle.vertices[0].x += offset.x;
      moved.triangle.vertices[0].y += offset.y;
      moved.triangle.vertices[1].x += offset.x;
      moved.triangle.vertices[1].y += offset.y;
      moved.triangle.vertices[2].x += offset.x;
      moved.triangle.vertices[2].y += offset.y;
      break;

    case GRAPHICS_FILL_TRIANGLES:
      moved.triangles.position.x += offset.x;
      moved.triangles.position.y += offset.y;
      break;

    case GRAPHICS_FILL_QUAD:
      moved.quad.vertices[0].x += offset.x;
      moved.quad.vertices[0].y += offset.y;
      moved.quad.vertices[1].x += offset.x;
      moved.quad.vertices[1].y += offset.y;
      moved.quad.vertices[2].x += offset.x;
      moved.quad.vertices[2].y += offset.y;
      moved.quad.vertices[3].x += offset.x;
      moved.quad.vertices[3].y += offset.y;
      break;

    case GRAPHICS_FILL_ELLIPSE:
      moved.ellipse.area.x += offset.x;
      moved.ellipse.area.y += offset.y;
      break;

    case GRAPHICS_FILL_LINE:
      moved.line.vertices[0].x += offset.x;
      moved.line.vertices[0].y += offset.y;
      moved.line.vertices[1].x += offset.x;
      moved.line.vertices[1].y += offset.y;
      break;

    case GRAPHICS_DRAW_TEXT_AREA:
      moved.text_area.area.x += offset.x;
      moved.text_area.area.y += offset.y;
      break;

    case GRAPHICS_DRAW_TEXT_CURSOR:
      moved.text_cursor.area.x += offset.x;
      moved.text_cursor.area.y += offset.y;
      break;

    default:;
  }

  return moved;
}

static Box graphics_request_area_(Graphics_Request req) {
  Box area = {0};

  switch (req.op) {
    case GRAPHICS_DRAW_PIXELS:      area = req.pixels     .area; break;
    case GRAPHICS_FILL_RECTANGLE:   area = req.rectangle  .area; break;
    case GRAPHICS_FILL_ELLIPSE:     area = req.ellipse    .area; break;
    case GRAPHICS_DRAW_TEXT_AREA:   area = req.text_area  .area; break;
    case GRAPHICS_DRAW_TEXT_CURSOR: area = req.text_cursor.area; break;

    case GRAPHICS_FILL_TRIANGLE: {
      vec2 *v     = req.triangle.vertices;
      area.x      = min3_(v[0].x, v[1].x, v[2].x);
      area.y      = min3_(v[0].y, v[1].y, v[2].y);
      area.width  = max3_(v[0].x, v[1].x, v[2].x) - area.x;
      area.height = max3_(v[0].y, v[1].y, v[2].y) - area.y;
    } break;

    case GRAPHICS_FILL_TRIANGLES:
      if (req.triangles.num_triangles > 0) {
        f64 x0, y0, x1, y1;
        triangles_area_(&x0, &y0, &x1, &y1, req.triangles.position, req.triangles.scale, req.triangles.num_triangles, req.triangles.vertices);

        area.x      = x0;
        area.y      = y0;
        area.width  = x1 - x0;
        area.height = y1 - y0;
      }
      break;

    case GRAPHICS_FILL_QUAD:
    case GRAPHICS_FILL_LINE: {
      vec2 quad[4] = {0};

      if (req.op == GRAPHICS_FILL_LINE)
        quad_from_line_(quad, req.line.vertices, req.line.width);
      else {
        quad[0] = req.quad.vertices[0];
        quad[1] = req.quad.vertices[1];
        quad[2] = req.quad.vertices[2];
        quad[3] = req.quad.vertices[3];
      }

      area.x      = min4_(quad[0].x, quad[1].x, quad[2].x, quad[3].x);
      area.y      = min4_(quad[0].y, quad[1].y, quad[2].y, quad[3].y);
      area.width  = max4_(quad[0].x, quad[1].x, quad[2].x, quad[3].x) - area.x;
      area.height = max4_(quad[0].y, quad[1].y, quad[2].y, quad[3].y) - area.y;
    } break;

    default:;
  }

  return area;
}

static Graphics_Request_Norm_ normalize_graphics_request_(Graphics_Request req) {
  Box area = graphics_request_area_(req);
  i64 x    = (i64) floor(area.x - .5);
  i64 y    = (i64) floor(area.y - .5);
  i64 w    = (i64) ceil (area.x + area.width  + .5) - x;
  i64 h    = (i64) ceil (area.y + area.height + .5) - y;

  return (Graphics_Request_Norm_) {
    .req    = graphics_request_moved_(req, (vec2) { -x, -y }),
    .x      = x,
    .y      = y,
    .width  = w,
    .height = h,
  };
}

static void perform_graphics_request_to_buffer_(Pixel_Buffer dst, Graphics_Request req) {
  switch (req.op) {
    case GRAPHICS_DRAW_PIXELS:
      draw_pixels_to_buffer(
        dst,
        req.pixels.area,
        req.pixels.src
      );
      break;

    case GRAPHICS_FILL_RECTANGLE:
      fill_rectangle_to_buffer(
        dst,
        req.rectangle.color,
        req.rectangle.area
      );
      break;

    case GRAPHICS_FILL_TRIANGLE:
      fill_triangle_to_buffer(
        dst,
        req.triangle.color,
        req.triangle.vertices
      );
      break;

    case GRAPHICS_FILL_TRIANGLES:
      fill_triangles_to_buffer(
        dst,
        req.triangles.color,
        req.triangles.position,
        req.triangles.scale,
        req.triangles.num_triangles,
        req.triangles.vertices
      );
      break;

    case GRAPHICS_FILL_QUAD:
      fill_quad_to_buffer(
        dst,
        req.quad.color,
        req.quad.vertices
      );
      break;

    case GRAPHICS_FILL_ELLIPSE:
      fill_ellipse_to_buffer(
        dst,
        req.ellipse.color,
        req.ellipse.area
      );
      break;

    case GRAPHICS_FILL_LINE:
      fill_line_to_buffer(
        dst,
        req.line.color,
        req.line.vertices,
        req.line.width
      );
      break;

    case GRAPHICS_DRAW_TEXT_AREA:
      draw_text_area_to_buffer(
        dst,
        req.text_area.font,
        req.text_area.color,
        req.text_area.area,
        req.text_area.max_size,
        req.text_area.num_chars,
        req.text_area.text
      );
      break;

    case GRAPHICS_DRAW_TEXT_CURSOR:
      draw_text_cursor_to_buffer(
        dst,
        req.text_cursor.font,
        req.text_cursor.color,
        req.text_cursor.area,
        req.text_cursor.max_size,
        req.text_cursor.num_chars,
        req.text_cursor.text,
        req.text_cursor.cursor,
        req.text_cursor.selection
      );
      break;

    default:;
  }
}

static b8 mem_eq_(i64 size, void *x, void *y) {
  for (i64 i = 0; i < size; ++i)
    if (((u8 *) x)[i] != ((u8 *) y)[i])
      return 0;
  return 1;
}

static void copy_pixels_quick_(Pixel_Buffer dst, i64 x, i64 y, Pixel_Buffer src) {
  for (i64 j = 0; j < src.height; ++j) {
    vec4_f32 *s     = src.pixels + j * src.stride;
    vec4_f32 *d     = dst.pixels + (y + j) * dst.stride + x;
    vec4_f32 *d_end = d + src.width;

    for (; d < d_end; ++d, ++s) {
      d->x = d->x * (1.f - s->w) + s->x * s->w;
      d->y = d->y * (1.f - s->w) + s->y * s->w;
      d->z = d->z * (1.f - s->w) + s->z * s->w;
    }
  }
}

static void perform_graphics_request_cached_(Pixel_Buffer dst, Graphics_Request req) {
  Graphics_Request_Norm_ norm = normalize_graphics_request_(req);
  if (norm.width <= 0 || norm.height <= 0)
    return;

  Graphics_Request_Id_ id = graphics_request_id_(norm.req);

  b8  reset_slot = 1;
  i64 slot       = -1;

  i64 n = (i64) (id.index % (GRAPHICS_CACHE_SIZE / GRAPHICS_CACHE_DEPTH)) * GRAPHICS_CACHE_DEPTH;

  for (i64 i = 0; i < GRAPHICS_CACHE_DEPTH; ++i) {
    if (!_graphics_cache[n + i].occupied)
      continue;
    if (!mem_eq_(BLAKE2B_OUTBYTES, _graphics_cache[n + i].id.hash, id.hash))
      continue;
    if (!mem_eq_(sizeof req, &_graphics_cache[n + i].req, &norm.req))
      continue;
    reset_slot = 0;
    slot       = n + i;
    break;
  }

  if (slot == -1)
    for (i64 i = 0; i < GRAPHICS_CACHE_DEPTH; ++i)
      if (!_graphics_cache[n + i].occupied) {
        slot = n + i;
        break;
      }

  if (slot == -1) {
    //  Cache overflow.
    slot = n;
  }

  Graphics_Cache_Slot_ *s = _graphics_cache + slot;

  i64 num_pixels = norm.width * norm.height;

  if (reset_slot) {
    s->occupied = 1;
    s->id       = id;
    s->req      = norm.req;
    s->width    = norm.width;
    s->height   = norm.height;

    if (num_pixels > s->buffer_len)
      resize_dynamic_array_exact(&s->buffer_len, (void **) &s->buffer, sizeof *s->buffer, num_pixels);

    if (s->buffer_len < num_pixels)
      //  Out of memory
      s->occupied = 0;
    else {
      mem_set_(s->buffer, 0, s->buffer_len * sizeof *s->buffer);
      perform_graphics_request_to_buffer_((Pixel_Buffer) {
        .width  = s->width,
        .height = s->height,
        .stride = s->width,
        .pixels = s->buffer,
      }, s->req);
    }
  }

  if (s->occupied)
    copy_pixels_quick_(dst, norm.x, norm.y, (Pixel_Buffer) {
      .width  = s->width,
      .height = s->height,
      .stride = s->width,
      .pixels = s->buffer,
    });
  else
    perform_graphics_request_to_buffer_(dst, req);
}

static void scale_and_perform_graphics_request_(Graphics_Context context, Graphics_Request req) {
  Graphics_Request scaled = graphics_request_scaled_(req, context.scale);

  if (context.disable_cache)
    perform_graphics_request_to_buffer_(context.dst, scaled);
  else
    perform_graphics_request_cached_(context.dst, scaled);
}

static Graphics_Context graphics_context_defaults_(Graphics_Context context) {
  if (context.dst.pixels == NULL) {
    i64 max_height = g_platform.num_pixels / g_platform.frame_width;

    context.dst.width  = g_platform.frame_width;
    context.dst.height = min2_i64_(max_height, g_platform.frame_height);
    context.dst.stride = g_platform.frame_width;
    context.dst.pixels = g_platform.pixels;

    if (context.scale.x == 0. && context.scale.y == 0.)
      context.scale = (vec2) {
        .x = ((f64) g_platform.frame_width)  / g_platform.real_width,
        .y = ((f64) g_platform.frame_height) / g_platform.real_height,
      };
  }

  return context;
}

//  ================================================================

void perform_graphics_request(Graphics_Context context, Graphics_Request req) {
  scale_and_perform_graphics_request_(graphics_context_defaults_(context), req);
}

void clean_graphics_requests_cache(i64 amount) {
  for (; amount > 0; --amount) {
    Graphics_Cache_Slot_ *s = _graphics_cache + _graphics_cache_gc_index;

    i64 required_len = !s->occupied ? 0 : s->width * s->height;

    if (s->buffer_len > required_len)
      resize_dynamic_array_exact(&s->buffer_len, (void **) &s->buffer, sizeof *s->buffer, required_len);

    _graphics_cache_gc_index = (_graphics_cache_gc_index + 1) % GRAPHICS_CACHE_SIZE;
  }
}

//  ================================================================
//
//  Test suite
//
//  ================================================================

#if ENABLE_TESTING

#define TEST_FILE graphics
#include "test.c"

TEST("colors") {
  REQUIRE_EQ(rgb_u32_from_f32((vec3_f32) { 1.f, 0.f, 0.f, }), 0xff0000);
  REQUIRE_EQ(rgb_u32_from_f32((vec3_f32) { 0.f, 1.f, 0.f, }), 0x00ff00);
  REQUIRE_EQ(rgb_u32_from_f32((vec3_f32) { 0.f, 0.f, 1.f, }), 0x0000ff);

  REQUIRE_EQ(rgb_u32_from_f32((vec3_f32) { .5f, 0.f, 0.f, }), 0x800000);
  REQUIRE_EQ(rgb_u32_from_f32((vec3_f32) { 0.f, .5f, 0.f, }), 0x008000);
  REQUIRE_EQ(rgb_u32_from_f32((vec3_f32) { 0.f, 0.f, .5f, }), 0x000080);

  REQUIRE_EQ(rgba_u32_from_f32((vec4_f32) { 1.f, 0.f, 0.f, 0.f, }), 0x00ff0000);
  REQUIRE_EQ(rgba_u32_from_f32((vec4_f32) { 0.f, 1.f, 0.f, 0.f, }), 0x0000ff00);
  REQUIRE_EQ(rgba_u32_from_f32((vec4_f32) { 0.f, 0.f, 1.f, 0.f, }), 0x000000ff);
  REQUIRE_EQ(rgba_u32_from_f32((vec4_f32) { 0.f, 0.f, 0.f, 1.f, }), 0xff000000);

  REQUIRE_EQ(rgba_u32_from_f32((vec4_f32) { .5f, 0.f, 0.f, 0.f, }), 0x00800000);
  REQUIRE_EQ(rgba_u32_from_f32((vec4_f32) { 0.f, .5f, 0.f, 0.f, }), 0x00008000);
  REQUIRE_EQ(rgba_u32_from_f32((vec4_f32) { 0.f, 0.f, .5f, 0.f, }), 0x00000080);
  REQUIRE_EQ(rgba_u32_from_f32((vec4_f32) { 0.f, 0.f, 0.f, .5f, }), 0x80000000);

  REQUIRE_EQ(lch_from_rgb((vec3_f32) { 0.f, 0.f, 0.f, }).x * 100,   0);
  REQUIRE_EQ(lch_from_rgb((vec3_f32) { 1.f, 1.f, 1.f, }).x * 100, 100);
  REQUIRE_EQ(lch_from_rgb((vec3_f32) { 1.f, 0.f, 0.f, }).x * 100,  62);
  REQUIRE_EQ(lch_from_rgb((vec3_f32) { 0.f, 1.f, 0.f, }).x * 100,  86);
  REQUIRE_EQ(lch_from_rgb((vec3_f32) { 0.f, 0.f, 1.f, }).x * 100,  45);

  REQUIRE_EQ(lch_from_rgb((vec3_f32) { .5f, .5f, .5f, }).y * 100,  0);
  REQUIRE_EQ(lch_from_rgb((vec3_f32) { 1.f, 0.f, 0.f, }).y * 100, 25);
  REQUIRE_EQ(lch_from_rgb((vec3_f32) { 0.f, 1.f, 0.f, }).y * 100, 29);
  REQUIRE_EQ(lch_from_rgb((vec3_f32) { 0.f, 0.f, 1.f, }).y * 100, 31);

  REQUIRE_EQ(lch_from_rgb((vec3_f32) { 1.f, 0.f, 0.f, }).z * 100,   51);
  REQUIRE_EQ(lch_from_rgb((vec3_f32) { 0.f, 1.f, 0.f, }).z * 100,  248);
  REQUIRE_EQ(lch_from_rgb((vec3_f32) { 0.f, 0.f, 1.f, }).z * 100, -167);

  vec3_f32 white = rgb_from_lch(lch_from_rgb((vec3_f32) { 1.f, 1.f, 1.f, }));
  REQUIRE_EQ(white.x * 100, 100);
  REQUIRE_EQ(white.y * 100, 100);
  REQUIRE_EQ((white.z + 6e-7) * 100, 100);

  vec3_f32 black = rgb_from_lch(lch_from_rgb((vec3_f32) { 0.f, 0.f, 0.f, }));
  REQUIRE_EQ(black.x * 100, 0);
  REQUIRE_EQ(black.y * 100, 0);
  REQUIRE_EQ(black.z * 100, 0);

  vec3_f32 grey = rgb_from_lch(lch_from_rgb((vec3_f32) { .5f, .5f, .5f, }));
  REQUIRE_EQ(grey.x * 100, 50);
  REQUIRE_EQ((grey.y + 3e-8) * 100, 50);
  REQUIRE_EQ((grey.z + 3e-7) * 100, 50);

  vec3_f32 red = rgb_from_lch(lch_from_rgb((vec3_f32) { 1.f, 0.f, 0.f, }));
  REQUIRE_EQ(red.x * 100, 100);
  REQUIRE_EQ(red.y * 100,   0);
  REQUIRE_EQ(red.z * 100,   0);

  vec3_f32 green = rgb_from_lch(lch_from_rgb((vec3_f32) { 0.f, 1.f, 0.f, }));
  REQUIRE_EQ(green.x * 100,   0);
  REQUIRE_EQ((green.y + 3e-7) * 100, 100);
  REQUIRE_EQ(green.z * 100,   0);

  vec3_f32 blue = rgb_from_lch(lch_from_rgb((vec3_f32) { 0.f, 0.f, 1.f, }));
  REQUIRE_EQ(blue.x * 100, 0);
  REQUIRE_EQ(blue.y * 100, 0);
  REQUIRE_EQ((blue.z + 3e-7) * 100, 100);

  vec3_f32 yellow = rgb_from_lch(lch_from_rgb((vec3_f32) { 1.f, 1.f, 0.f, }));
  REQUIRE_EQ(yellow.x * 100, 100);
  REQUIRE_EQ((yellow.y + 6e-8) * 100, 100);
  REQUIRE_EQ(yellow.z * 100,   0);

  vec3_f32 cyan = rgb_from_lch(lch_from_rgb((vec3_f32) { 0.f, 1.f, 1.f, }));
  REQUIRE_EQ(cyan.x * 100,   0);
  REQUIRE_EQ((cyan.y + 3e-7) * 100, 100);
  REQUIRE_EQ((cyan.z + 3e-7) * 100, 100);

  vec3_f32 purple = rgb_from_lch(lch_from_rgb((vec3_f32) { 1.f, 0.f, 1.f, }));
  REQUIRE_EQ(purple.x * 100, 100);
  REQUIRE_EQ(purple.y * 100, 0);
  REQUIRE_EQ((purple.z + 3e-7) * 100, 100);
}

TEST("rectangle corners") {
  i64 w = 10;
  i64 h = 10;

  vec4_f32 pixels[100] = {0};

  fill_rectangle_to_buffer(
    (Pixel_Buffer) { .width = w, .height = h, .stride = w, .pixels = pixels },
    (vec4_f32)     { 1.0f, 1.0f, 1.0f, 1.0f, },
    (Box)          { .x = -1, .y = -1, .width = 2, .height = 2, }
  );

  fill_rectangle_to_buffer(
    (Pixel_Buffer) { .width = w, .height = h, .stride = w, .pixels = pixels },
    (vec4_f32)     { 1.0f, 1.0f, 1.0f, 1.0f, },
    (Box)          { .x = 9, .y = -1, .width = 2, .height = 2, }
  );

  fill_rectangle_to_buffer(
    (Pixel_Buffer) { .width = w, .height = h, .stride = w, .pixels = pixels },
    (vec4_f32)     { 1.0f, 1.0f, 1.0f, 1.0f, },
    (Box)          { .x = -1, .y = 9, .width = 2, .height = 2, }
  );

  fill_rectangle_to_buffer(
    (Pixel_Buffer) { .width = w, .height = h, .stride = w, .pixels = pixels },
    (vec4_f32)     { 1.0f, 1.0f, 1.0f, 1.0f, },
    (Box)          { .x = 9, .y = 9, .width = 2, .height = 2, }
  );

  b8 expect_bits[100] = {
    1, 0, 0, 0, 0, 0, 0, 0, 0, 1,
    0, 0, 0, 0, 0, 0, 0, 0, 0, 0,
    0, 0, 0, 0, 0, 0, 0, 0, 0, 0,
    0, 0, 0, 0, 0, 0, 0, 0, 0, 0,
    0, 0, 0, 0, 0, 0, 0, 0, 0, 0,
    0, 0, 0, 0, 0, 0, 0, 0, 0, 0,
    0, 0, 0, 0, 0, 0, 0, 0, 0, 0,
    0, 0, 0, 0, 0, 0, 0, 0, 0, 0,
    0, 0, 0, 0, 0, 0, 0, 0, 0, 0,
    1, 0, 0, 0, 0, 0, 0, 0, 0, 1,
  };

  b8 ok = 1;

  for (i64 i = 0; i < w * h; ++i) {
    if (expect_bits[i] != (pixels[i].x > 0.5) ||
        expect_bits[i] != (pixels[i].y > 0.5) ||
        expect_bits[i] != (pixels[i].z > 0.5) ||
        expect_bits[i] != (pixels[i].w > 0.5)) {
      ok = 0;
      break;
    }
  }

  REQUIRE(ok);
}

TEST("rectangle middle") {
  i64 w = 10;
  i64 h = 10;

  vec4_f32 pixels[100] = {0};

  fill_rectangle_to_buffer(
    (Pixel_Buffer) { .width = w, .height = h, .stride = w, .pixels = pixels },
    (vec4_f32)     { 1.0f, 1.0f, 1.0f, 1.0f, },
    (Box)          { .x = 1, .y = 1, .width = 8, .height = 8, }
  );

  b8 expect_bits[100] = {
    0, 0, 0, 0, 0, 0, 0, 0, 0, 0,
    0, 1, 1, 1, 1, 1, 1, 1, 1, 0,
    0, 1, 1, 1, 1, 1, 1, 1, 1, 0,
    0, 1, 1, 1, 1, 1, 1, 1, 1, 0,
    0, 1, 1, 1, 1, 1, 1, 1, 1, 0,
    0, 1, 1, 1, 1, 1, 1, 1, 1, 0,
    0, 1, 1, 1, 1, 1, 1, 1, 1, 0,
    0, 1, 1, 1, 1, 1, 1, 1, 1, 0,
    0, 1, 1, 1, 1, 1, 1, 1, 1, 0,
    0, 0, 0, 0, 0, 0, 0, 0, 0, 0,
  };

  b8 ok = 1;

  for (i64 i = 0; i < w * h; ++i) {
    if (expect_bits[i] != (pixels[i].x > 0.5) ||
        expect_bits[i] != (pixels[i].y > 0.5) ||
        expect_bits[i] != (pixels[i].z > 0.5) ||
        expect_bits[i] != (pixels[i].w > 0.5)) {
      ok = 0;
      break;
    }
  }

  REQUIRE(ok);
}

TEST("rectangle aliasing") {
  i64 w = 10;
  i64 h = 10;

  vec4_f32 pixels[100] = {0};

  fill_rectangle_to_buffer(
    (Pixel_Buffer) { .width = w, .height = h, .stride = w, .pixels = pixels },
    (vec4_f32)     { 1.0f, 1.0f, 1.0f, 1.0f, },
    (Box)          { .x = 1.4, .y = 1.4, .width = 2.0, .height = 2.0, }
  );

  fill_rectangle_to_buffer(
    (Pixel_Buffer) { .width = w, .height = h, .stride = w, .pixels = pixels },
    (vec4_f32)     { 1.0f, 1.0f, 1.0f, 1.0f, },
    (Box)          { .x = 4.0, .y = 1.0, .width = 2.4, .height = 2.4, }
  );

  fill_rectangle_to_buffer(
    (Pixel_Buffer) { .width = w, .height = h, .stride = w, .pixels = pixels },
    (vec4_f32)     { 1.0f, 1.0f, 1.0f, 1.0f, },
    (Box)          { .x = 6.6, .y = 1.0, .width = 2.0, .height = 2.0, }
  );

  fill_rectangle_to_buffer(
    (Pixel_Buffer) { .width = w, .height = h, .stride = w, .pixels = pixels },
    (vec4_f32)     { 1.0f, 1.0f, 1.0f, 1.0f, },
    (Box)          { .x = 1.0, .y = 3.6, .width = 2.0, .height = 2.0, }
  );

  fill_rectangle_to_buffer(
    (Pixel_Buffer) { .width = w, .height = h, .stride = w, .pixels = pixels },
    (vec4_f32)     { 1.0f, 1.0f, 1.0f, 1.0f, },
    (Box)          { .x = 4.0, .y = 4.0, .width = 2.4, .height = 2.4, }
  );

  fill_rectangle_to_buffer(
    (Pixel_Buffer) { .width = w, .height = h, .stride = w, .pixels = pixels },
    (vec4_f32)     { 1.0f, 1.0f, 1.0f, 1.0f, },
    (Box)          { .x = 7.4, .y = 4.4, .width = 1.2, .height = 1.2, }
  );

  b8 expect_bits[100] = {
    0, 0, 0, 0, 0, 0, 0, 0, 0, 0,
    0, 1, 1, 0, 1, 1, 0, 1, 1, 0,
    0, 1, 1, 0, 1, 1, 0, 1, 1, 0,
    0, 0, 0, 0, 0, 0, 0, 0, 0, 0,
    0, 1, 1, 0, 1, 1, 0, 1, 1, 0,
    0, 1, 1, 0, 1, 1, 0, 1, 1, 0,
    0, 0, 0, 0, 0, 0, 0, 0, 0, 0,
    0, 0, 0, 0, 0, 0, 0, 0, 0, 0,
    0, 0, 0, 0, 0, 0, 0, 0, 0, 0,
    0, 0, 0, 0, 0, 0, 0, 0, 0, 0,
  };

  b8 ok = 1;

  for (i64 i = 0; i < w * h; ++i) {
    if (expect_bits[i] != (pixels[i].x > 0.5) ||
        expect_bits[i] != (pixels[i].y > 0.5) ||
        expect_bits[i] != (pixels[i].z > 0.5) ||
        expect_bits[i] != (pixels[i].w > 0.5)) {
      ok = 0;
      break;
    }
  }

  REQUIRE(ok);
}

TEST("text letter A") {
  i64 w = 10;
  i64 h = 10;

  vec4_f32 pixels[100] = {0};

  draw_text_area_to_buffer(
    (Pixel_Buffer) {
      .width  = w,
      .height = h,
      .stride = w,
      .pixels = pixels,
    },
    0,
    (vec4_f32) { 1.0f, 1.0f, 1.0f, 1.0f, },
    (Box)      { .x = 0, .y = 0, .width = 10, .height = 10, },
    (vec2)     { .x = 10.0, .y = 10.0, },
    1,
    (c32[1])   { 'A', }
  );

  b8 expect_bits[100] = {
    0, 1, 1, 1, 0, 0, 0, 0, 0, 0,
    1, 0, 0, 0, 1, 1, 0, 0, 0, 0,
    1, 0, 0, 0, 1, 1, 0, 0, 0, 0,
    1, 0, 0, 0, 1, 1, 0, 0, 0, 0,
    1, 1, 1, 1, 1, 1, 0, 0, 0, 0,
    1, 1, 1, 1, 1, 1, 0, 0, 0, 0,
    1, 0, 0, 0, 1, 1, 0, 0, 0, 0,
    1, 0, 0, 0, 1, 1, 0, 0, 0, 0,
    1, 0, 0, 0, 1, 1, 0, 0, 0, 0,
    0, 0, 0, 0, 0, 0, 0, 0, 0, 0,
  };

  b8 ok = 1;

  for (i64 i = 0; i < w * h; ++i) {
    if (expect_bits[i] != (pixels[i].x > 0.5) ||
        expect_bits[i] != (pixels[i].y > 0.5) ||
        expect_bits[i] != (pixels[i].z > 0.5) ||
        expect_bits[i] != (pixels[i].w > 0.5)) {
      ok = 0;
      break;
    }
  }

  REQUIRE(ok);
}

TEST("text letter z") {
  i64 w = 10;
  i64 h = 10;

  vec4_f32 pixels[100] = {0};

  draw_text_area_to_buffer(
    (Pixel_Buffer) {
      .width  = w,
      .height = h,
      .stride = w,
      .pixels = pixels,
    },
    0,
    (vec4_f32) { 1.0f, 1.0f, 1.0f, 1.0f, },
    (Box)      { .x = 0, .y = 0, .width = 10, .height = 10, },
    (vec2)     { .x = 10.0, .y = 10.0, },
    1,
    (c32[1])   { 'z', }
  );

  b8 expect_bits[100] = {
    0, 0, 0, 0, 0, 0, 0, 0, 0, 0,
    1, 1, 1, 1, 1, 1, 0, 0, 0, 0,
    1, 1, 1, 1, 1, 1, 0, 0, 0, 0,
    0, 0, 0, 0, 1, 1, 0, 0, 0, 0,
    0, 1, 1, 1, 0, 0, 0, 0, 0, 0,
    0, 1, 1, 1, 0, 0, 0, 0, 0, 0,
    1, 0, 0, 0, 0, 0, 0, 0, 0, 0,
    1, 1, 1, 1, 1, 1, 0, 0, 0, 0,
    1, 1, 1, 1, 1, 1, 0, 0, 0, 0,
    0, 0, 0, 0, 0, 0, 0, 0, 0, 0,
  };

  b8 ok = 1;

  for (i64 i = 0; i < w * h; ++i) {
    if (expect_bits[i] != (pixels[i].x > 0.5) ||
        expect_bits[i] != (pixels[i].y > 0.5) ||
        expect_bits[i] != (pixels[i].z > 0.5) ||
        expect_bits[i] != (pixels[i].w > 0.5)) {
      ok = 0;
      break;
    }
  }

  REQUIRE(ok);
}

TEST("text digits 42") {
  i64 w = 10;
  i64 h = 10;

  vec4_f32 pixels[100] = {0};

  draw_text_area_to_buffer(
    (Pixel_Buffer) {
      .width  = w,
      .height = h,
      .stride = w,
      .pixels = pixels,
    },
    0,
    (vec4_f32) { 1.0f, 1.0f, 1.0f, 1.0f, },
    (Box)      { .x = 0, .y = 0, .width = 10, .height = 10, },
    (vec2)     { .x = 10.0, .y = 10.0, },
    2,
    (c32[2])   { '4', '2', }
  );

  b8 expect_bits[100] = {
    0, 0, 0, 0, 0, 0, 0, 0, 0, 0,
    1, 0, 0, 1, 0, 0, 1, 1, 1, 1,
    1, 0, 0, 1, 0, 0, 1, 1, 1, 1,
    1, 0, 0, 1, 0, 0, 0, 0, 0, 1,
    1, 1, 1, 1, 0, 0, 0, 1, 1, 0,
    1, 1, 1, 1, 0, 0, 0, 1, 1, 0,
    0, 0, 0, 1, 0, 0, 1, 0, 0, 0,
    0, 0, 0, 1, 0, 0, 1, 1, 1, 1,
    0, 0, 0, 1, 0, 0, 1, 1, 1, 1,
    0, 0, 0, 0, 0, 0, 0, 0, 0, 0,
  };

  b8 ok = 1;

  for (i64 i = 0; i < w * h; ++i) {
    if (expect_bits[i] != (pixels[i].x > 0.5) ||
        expect_bits[i] != (pixels[i].y > 0.5) ||
        expect_bits[i] != (pixels[i].z > 0.5) ||
        expect_bits[i] != (pixels[i].w > 0.5)) {
      ok = 0;
      break;
    }
  }

  REQUIRE(ok);
}

#undef TEST_FILE
#endif // ENABLE_TESTING

//  ================================================================

#endif // GRAPHICS_IMPL_GUARD_
#endif // GRAPHICS_HEADER