#include "private/rendering.h"
#include "context.h"
#include "textures.h"

#include <SDL2/SDL.h>
#include <stb_ds.h>
#include <glad/glad.h>

#include <stddef.h>
#include <stdlib.h>
#include <tgmath.h>

/* http://www.swiftless.com/opengltuts.html */

void render_queue_clear(void) {
    /* since i don't intend to free the queues, */
    /* it's faster and simpler to just "start over" */
    /* and start overwriting the existing data */
    arrsetlen(ctx.render_queue_sprites, 0);
    arrsetlen(ctx.render_queue_rectangles, 0);
    arrsetlen(ctx.render_queue_circles, 0);

    for (size_t i = 0; i < hmlenu(ctx.uncolored_mesh_batches); ++i)
        arrsetlen(ctx.uncolored_mesh_batches[i].value.primitives, 0);
}


/*
 * an implementation note:
 * try to avoid doing expensive work in the push functions,
 * because they will be called multiple times in the main loop
 * before anything is really rendered
 */

/* sprite */
void push_sprite(char *path, t_frect rect) {
    struct sprite_primitive sprite = {
        .rect = rect,
        .color = (t_color) { 255, 255, 255, 255 },
        .rotation = 0.0,
        .blend_mode = SDL_BLENDMODE_BLEND,
        .texture_key = textures_get_key(&ctx.texture_cache, path),
        .layer = 0,
        .flip_x = false,
        .flip_y = false,
    };

    arrput(ctx.render_queue_sprites, sprite);
}


void push_sprite_ex(t_frect rect, t_push_sprite_args args) {
    struct sprite_primitive sprite = {
        .rect = rect,
        .color = args.color,
        .rotation = args.rotation,
        .blend_mode = args.blend_mode,
        .texture_key =  textures_get_key(&ctx.texture_cache, args.path),
        .layer = args.layer,
        .flip_x = args.flip_x,
        .flip_y = args.flip_y,
    };

    arrput(ctx.render_queue_sprites, sprite);
}


/* rectangle */
void push_rectangle(t_frect rect, t_color color) {
    struct rect_primitive rectangle = {
        .rect = rect,
        .color = color,
    };

    arrput(ctx.render_queue_rectangles, rectangle);
}


/* circle */
void push_circle(t_fvec2 position, float radius, t_color color) {
    struct circle_primitive circle = {
        .radius = radius,
        .color = color,
        .position = position,
    };

    arrput(ctx.render_queue_circles, circle);
}


/* TODO: automatic handling of repeating textures */
/*       for that we could allocate a loner texture */
void unfurl_triangle(const char *path,
                     t_fvec3 v0,
                     t_fvec3 v1,
                     t_fvec3 v2,
                     t_shvec2 uv0,
                     t_shvec2 uv1,
                     t_shvec2 uv2)
{
    const t_texture_key texture_key = textures_get_key(&ctx.texture_cache, path);

    struct mesh_batch_item *batch_p = hmgetp_null(ctx.uncolored_mesh_batches, texture_key);
    if (!batch_p) {
        struct mesh_batch item = {0};
        hmput(ctx.uncolored_mesh_batches, texture_key, item);
        batch_p = &ctx.uncolored_mesh_batches[hmlenu(ctx.uncolored_mesh_batches) - 1]; /* TODO: can last index be used? */
    }

    union uncolored_space_triangle triangle = { .primitive = {
        .v0 = v0,
        .v1 = v1,
        .v2 = v2,
        .uv1 = m_to_fvec2(uv1),
        .uv0 = m_to_fvec2(uv0),
        .uv2 = m_to_fvec2(uv2),
    }};

    union uncolored_space_triangle *triangles =
        (union uncolored_space_triangle *)batch_p->value.primitives;
    arrpush(triangles, triangle);
    batch_p->value.primitives = (uint8_t *)triangles;
}


/* compare functions for the sort in render_sprites */
static int cmp_atlases(const void *a, const void *b) {
    int index_a = ((const struct sprite_primitive *)a)->texture_key.id;
    int index_b = ((const struct sprite_primitive *)b)->texture_key.id;

    return (index_a > index_b) - (index_a < index_b);
}


static int cmp_blend_modes(const void *a, const void *b) {
    SDL_BlendMode mode_a = ((const struct sprite_primitive *)a)->blend_mode;
    SDL_BlendMode mode_b = ((const struct sprite_primitive *)b)->blend_mode;

    return (mode_a > mode_b) - (mode_a < mode_b);
}


static int cmp_colors(const void *a, const void *b) {
    t_color color_a = ((const struct sprite_primitive *)a)->color;
    t_color color_b = ((const struct sprite_primitive *)b)->color;

    /* check reds */
    if (color_a.r < color_b.r)
        return -1;
    else if (color_a.r > color_b.r)
        return 1;

    /* reds were equal, check greens */
    else if (color_a.g < color_b.g)
        return -1;
    else if (color_a.g > color_b.g)
        return 1;

    /* greens were equal, check blues */
    else if (color_a.b < color_b.b)
        return -1;
    else if (color_a.b > color_b.b)
        return 1;

    /* blues were equal, check alphas */
    else if (color_a.a < color_b.a)
        return -1;
    else if (color_a.a > color_b.a)
        return 1;

    /* entirely equal */
    else
        return 0;
}


static int cmp_layers(const void *a, const void *b) {
    int layer_a = ((const struct sprite_primitive *)a)->layer;
    int layer_b = ((const struct sprite_primitive *)b)->layer;

    return (layer_a > layer_b) - (layer_a < layer_b);
}


/* TODO: not that we're SDL free we need to implement batching ourselves */
/* necessary to allow the renderer to draw in batches in the best case */
static void sort_sprites(struct sprite_primitive *sprites) {
    size_t sprites_len = arrlenu(sprites);
    qsort(sprites, sprites_len, sizeof *sprites, cmp_atlases);
    qsort(sprites, sprites_len, sizeof *sprites, cmp_blend_modes);
    qsort(sprites, sprites_len, sizeof *sprites, cmp_colors);
    qsort(sprites, sprites_len, sizeof *sprites, cmp_layers);
}


static void upload_quad_vertices(t_frect rect) {
    /* client memory needs to be reachable on glDraw*, so */
    static float vertices[6 * 2];

    vertices[0]  = rect.x;          vertices[1]  = rect.y;
    vertices[2]  = rect.x;          vertices[3]  = rect.y + rect.h;
    vertices[4]  = rect.x + rect.w; vertices[5]  = rect.y + rect.h;
    vertices[6]  = rect.x + rect.w; vertices[7]  = rect.y + rect.h;
    vertices[8]  = rect.x + rect.w; vertices[9]  = rect.y;
    vertices[10] = rect.x;          vertices[11] = rect.y;

    glVertexPointer(2, GL_FLOAT, 0, (void *)&vertices);
}


/* TODO: texture flipping */
/* assumes that orthogonal matrix setup is done already */
static void render_sprite(struct sprite_primitive *sprite) {
    glEnableClientState(GL_TEXTURE_COORD_ARRAY);
    glClientActiveTexture(GL_TEXTURE0);

    textures_bind(&ctx.texture_cache, sprite->texture_key, GL_TEXTURE_2D);

    t_rect srcrect = textures_get_srcrect(&ctx.texture_cache, sprite->texture_key);
    t_rect dims    = textures_get_dims(&ctx.texture_cache, sprite->texture_key);

    glTexCoordPointer(2,
                      GL_FLOAT,
                      0,
                      /* TODO: try using shorts */
                      (void *)(float[6 * 2]) {
                        (float)srcrect.x               / (float)dims.w,
                        (float)srcrect.y               / (float)dims.h,
                        (float)srcrect.x               / (float)dims.w,
                        (float)(srcrect.y + srcrect.h) / (float)dims.h,
                        (float)(srcrect.x + srcrect.w) / (float)dims.w,
                        (float)(srcrect.y + srcrect.h) / (float)dims.h,
                        (float)(srcrect.x + srcrect.w) / (float)dims.w,
                        (float)(srcrect.y + srcrect.h) / (float)dims.h,
                        (float)(srcrect.x + srcrect.w) / (float)dims.w,
                        (float)srcrect.y               / (float)dims.h,
                        (float)srcrect.x               / (float)dims.w,
                        (float)srcrect.y               / (float)dims.h });

    glColor4ub(sprite->color.r, sprite->color.g, sprite->color.b, sprite->color.a);

    // SDL_SetTextureBlendMode(texture, sprite->blend_mode);

    glEnableClientState(GL_VERTEX_ARRAY);
    upload_quad_vertices(sprite->rect);

    glDrawArrays(GL_TRIANGLES, 0, 6);

    glDisableClientState(GL_TEXTURE_COORD_ARRAY);
    glDisableClientState(GL_VERTEX_ARRAY);
    glBindTexture(GL_TEXTURE_2D, 0);
}


static void render_rectangle(struct rect_primitive *rectangle) {
    glColor4ub(rectangle->color.r, rectangle->color.g,
               rectangle->color.b, rectangle->color.a);

    glEnableClientState(GL_VERTEX_ARRAY);
    upload_quad_vertices(rectangle->rect);

    glDrawArrays(GL_TRIANGLES, 0, 6);
    glDisableClientState(GL_VERTEX_ARRAY);
}


/* vertices_out and indices_out MUST BE FREED */
static void create_circle_geometry(t_fvec2 position,
                                   t_color color,
                                   float radius,
                                   size_t num_vertices,
                                   SDL_Vertex **vertices_out,
                                   int **indices_out)
{
    SDL_Vertex *vertices = cmalloc(sizeof *vertices * (num_vertices + 1));
    int *indices = cmalloc(sizeof *indices * (num_vertices * 3));

    /* the angle (in radians) to rotate by on each iteration */
    float seg_rotation_angle = (360.0f / (float)num_vertices) * ((float)M_PI / 180);
    
    vertices[0].position.x = (float)position.x;
    vertices[0].position.y = (float)position.y;
    vertices[0].color.r = color.r;
    vertices[0].color.g = color.g;
    vertices[0].color.b = color.b;
    vertices[0].color.a = color.a;
    vertices[0].tex_coord = (SDL_FPoint){ 0, 0 };

    /* this point will rotate around the center */
    float start_x = 0.0f - radius;
    float start_y = 0.0f;

    for (size_t i = 1; i < num_vertices + 1; ++i) {
        float final_seg_rotation_angle = (float)i * seg_rotation_angle;

        vertices[i].position.x =
            cos(final_seg_rotation_angle) * start_x -
            sin(final_seg_rotation_angle) * start_y;
        vertices[i].position.y =
            cos(final_seg_rotation_angle) * start_y +
            sin(final_seg_rotation_angle) * start_x;

        vertices[i].position.x += position.x;
        vertices[i].position.y += position.y;

        vertices[i].color.r = color.r;
        vertices[i].color.g = color.g;
        vertices[i].color.b = color.b;
        vertices[i].color.a = color.a;

        vertices[i].tex_coord = (SDL_FPoint){ 0, 0 };


        size_t triangle_offset = 3 * (i - 1);

        /* center point index */
        indices[triangle_offset] = 0;
        /* generated point index */
        indices[triangle_offset + 1] = (int)i;

        size_t index = (i + 1) % num_vertices;
        if (index == 0)
            index = num_vertices;
        indices[triangle_offset + 2] = (int)index;
    }

    *vertices_out = vertices;
    *indices_out = indices;
}


static void render_circle(struct circle_primitive *circle) {
    SDL_Vertex *vertices = NULL;
    int *indices = NULL;
    int num_vertices = (int)circle->radius;

    create_circle_geometry(circle->position,
                           circle->color,
                           circle->radius,
                           num_vertices,
                           &vertices,
                           &indices);

    glEnableClientState(GL_VERTEX_ARRAY);
    glVertexPointer(2,
                    GL_FLOAT,
                    sizeof (SDL_Vertex),
                    &vertices->position);

    glEnableClientState(GL_COLOR_ARRAY);
    glColorPointer(4,
                   GL_UNSIGNED_BYTE,
                   sizeof (SDL_Vertex),
                   &vertices->color);

    glDrawElements(GL_TRIANGLES,
                   num_vertices * 3,
                   GL_UNSIGNED_INT,
                   indices);

    glDisableClientState(GL_COLOR_ARRAY);
    glDisableClientState(GL_VERTEX_ARRAY);

    free(vertices);
    free(indices);
}


static void render_sprites(void) {
    sort_sprites(ctx.render_queue_sprites);

    for (size_t i = 0; i < arrlenu(ctx.render_queue_sprites); ++i) {
        render_sprite(&ctx.render_queue_sprites[i]);
    }
}


static void render_rectangles(void) {
    for (size_t i = 0; i < arrlenu(ctx.render_queue_rectangles); ++i) {
        render_rectangle(&ctx.render_queue_rectangles[i]);
    }
}


static void render_circles(void) {
    for (size_t i = 0; i < arrlenu(ctx.render_queue_circles); ++i) {
        render_circle(&ctx.render_queue_circles[i]);
    }
}


static void draw_uncolored_space_traingle_batch(struct mesh_batch *batch,
                                                t_texture_key texture_key)
{
    size_t primitives_len = arrlenu(batch->primitives);

    if (primitives_len == 0)
        return;

    /* create vertex array object */
    if (batch->buffer == 0)
        glGenBuffers(1, &batch->buffer);

    /* TODO: try using mapped buffers while building batches instead? */
    /*       this way we could skip client side copy that is kept until commitment */
    /*       alternatively we could commit glBufferSubData based on a threshold */

    /* update pixel-based uvs to correspond with texture atlases */
    for (size_t i = 0; i < primitives_len; ++i) {
        struct uncolored_space_triangle_payload *payload =
            &((union uncolored_space_triangle *)batch->primitives)[i].payload;

        t_rect srcrect = textures_get_srcrect(&ctx.texture_cache, texture_key);
        t_rect dims = textures_get_dims(&ctx.texture_cache, texture_key);

        const float wr = (float)srcrect.w / (float)dims.w;
        const float hr = (float)srcrect.h / (float)dims.h;
        const float xr = (float)srcrect.x / (float)dims.w;
        const float yr = (float)srcrect.y / (float)dims.h;

        payload->uv0.x = xr + ((float)payload->uv0.x / (float)srcrect.w) * wr;
        payload->uv0.y = yr + ((float)payload->uv0.y / (float)srcrect.h) * hr;
        payload->uv1.x = xr + ((float)payload->uv1.x / (float)srcrect.w) * wr;
        payload->uv1.y = yr + ((float)payload->uv1.y / (float)srcrect.h) * hr;
        payload->uv2.x = xr + ((float)payload->uv2.x / (float)srcrect.w) * wr;
        payload->uv2.y = yr + ((float)payload->uv2.y / (float)srcrect.h) * hr;
    }

    textures_bind(&ctx.texture_cache, texture_key, GL_TEXTURE_2D);

    glBindBuffer(GL_ARRAY_BUFFER, batch->buffer);

    /* upload batched data */
    glBufferData(GL_ARRAY_BUFFER,
                 primitives_len * sizeof (struct uncolored_space_triangle_payload),
                 batch->primitives,
                 GL_STREAM_DRAW);

    /* vertex specification*/
    glEnableClientState(GL_VERTEX_ARRAY);
    glVertexPointer(3,
                    GL_FLOAT,
                    offsetof(struct uncolored_space_triangle_payload, v1),
                    (void *)offsetof(struct uncolored_space_triangle_payload, v0));

    glEnableClientState(GL_TEXTURE_COORD_ARRAY);
    glClientActiveTexture(GL_TEXTURE0);
    glTexCoordPointer(2,
                      GL_FLOAT,
                      offsetof(struct uncolored_space_triangle_payload, v1),
                      (void *)offsetof(struct uncolored_space_triangle_payload, uv0));

    /* commit for drawing */
    glDrawArrays(GL_TRIANGLES, 0, 3 * (int)primitives_len);

    glDisableClientState(GL_TEXTURE_COORD_ARRAY);
    glDisableClientState(GL_VERTEX_ARRAY);

    /* invalidate the buffer immediately */
    glBufferData(GL_ARRAY_BUFFER, 0, NULL, GL_STREAM_DRAW);
    glBindBuffer(GL_ARRAY_BUFFER, 0);
}


static void render_space(void) {
    glEnable(GL_TEXTURE_2D);
    glActiveTexture(GL_TEXTURE0);

    /* solid white, no modulation */
    glColor4f(1.0f, 1.0f, 1.0f, 1.0f);

    for (size_t i = 0; i < hmlenu(ctx.uncolored_mesh_batches); ++i) {
        draw_uncolored_space_traingle_batch(&ctx.uncolored_mesh_batches[i].value,
                                            ctx.uncolored_mesh_batches[i].key);
    }
}


void render(void) {
    textures_update_atlas(&ctx.texture_cache);

    /* fit rendering context onto the resizable screen */
    if ((float)ctx.window_w / (float)ctx.window_h > RENDER_BASE_RATIO) {
        float ratio = (float)ctx.window_h / (float)RENDER_BASE_HEIGHT;
        int w = (int)((float)RENDER_BASE_WIDTH * ratio);
        glViewport(
            ctx.window_w / 2 - w / 2,
            0,
            w,
            ctx.window_h
        );
    } else {
        float ratio = (float)ctx.window_w / (float)RENDER_BASE_WIDTH;
        int h = (int)((float)RENDER_BASE_HEIGHT * ratio);
        glViewport(
            0,
            ctx.window_h / 2 - h / 2,
            ctx.window_w,
            h
        );
    }

    glEnable(GL_DEPTH_TEST);
    glEnable(GL_ALPHA_TEST);

    glClearColor((1.0f / 255) * 230,
                 (1.0f / 255) * 230,
                 (1.0f / 255) * 230, 1);

    glClear(GL_COLOR_BUFFER_BIT |
            GL_DEPTH_BUFFER_BIT |
            GL_STENCIL_BUFFER_BIT);

    {
        glDisable(GL_CULL_FACE);
        glDepthFunc(GL_ALWAYS); /* fill depth buffer with ones, 2d view is always in front */
        glEnable(GL_BLEND);
        glBlendFunc(GL_SRC_ALPHA, GL_ONE_MINUS_SRC_ALPHA);
    
        glMatrixMode(GL_PROJECTION);
        glLoadIdentity();
        glOrtho(0, RENDER_BASE_WIDTH, RENDER_BASE_HEIGHT, 0, -1, 1);

        glMatrixMode(GL_MODELVIEW);
        glLoadIdentity();

        glEnable(GL_TEXTURE_2D);
    
        render_sprites();
        render_rectangles();
        render_circles();
    }

    {
        glMatrixMode(GL_PROJECTION);
        glLoadIdentity();

        glMatrixMode(GL_MODELVIEW);
        glLoadIdentity();
    
        glEnable(GL_CULL_FACE);
        glDepthFunc(GL_LESS);
        glEnable(GL_BLEND);
        glBlendFunc(GL_SRC_ALPHA, GL_ONE_MINUS_SRC_ALPHA);

        /* TODO: use depth test to optimize gui regions away */
        render_space();
    }

    SDL_GL_SwapWindow(ctx.window);
}