#include "common.h"
#include "world.h"
#include "display.h"
#include "time.h"
#include "draw.h"
#include "frustum.h"

static CVECTOR colors[] = 
{
    { 255, 0, 0 },
    { 0, 255, 0 },
    { 0, 0, 255 },
    { 255, 255, 0 },
    { 255, 0, 255 },
    { 0, 255, 255 },
    { 128, 255, 0 },
    { 255, 128, 0 },
    { 128, 0, 255 },
    { 255, 0, 128 },
    { 0, 128, 255 },
    { 0, 255, 128 },
};
static const int numColors = sizeof(colors) / sizeof(CVECTOR);

// Set data pointers directly from an in-memory byte buffer
#define LOAD_CHUNK(type, dst, num, src, entry)   \
    dst = (type*)((src) + (entry).offset);  \
    num = (entry).size / sizeof(type);

// Allocate memory per chunk and copy data from a byte buffer
// #define LOAD_CHUNK(type, dst, num, src, entry)  \
//     dst = (type*)malloc((entry).size);  \
//     memcpy(dst, (src) + (entry).offset, (entry).size);  \
//     num = (entry).size / sizeof(type);

void world_load(const u_long *data, world_t *world)
{
    const char *bytes = (const char*)data;

    ps1bsp_header_t* header = (ps1bsp_header_t*)bytes;

    LOAD_CHUNK(ps1bsp_texture_t, world->textures, world->numTextures, bytes, header->textures);
    LOAD_CHUNK(ps1bsp_vertex_t, world->vertices, world->numVertices, bytes, header->vertices);
    LOAD_CHUNK(ps1bsp_polygon_t, world->polygons, world->numPolygons, bytes, header->polygons);
    LOAD_CHUNK(ps1bsp_polyvertex_t, world->polyVertices, world->numPolyVertices, bytes, header->polyVertices);
    LOAD_CHUNK(ps1bsp_face_t, world->faces, world->numFaces, bytes, header->faces);
    LOAD_CHUNK(ps1bsp_facevertex_t, world->faceVertices, world->numFaceVertices, bytes, header->faceVertices);
    LOAD_CHUNK(ps1bsp_plane_t, world->planes, world->numPlanes, bytes, header->planes);
    LOAD_CHUNK(ps1bsp_node_t, world->nodes, world->numNodes, bytes, header->nodes);
    LOAD_CHUNK(ps1bsp_leaf_t, world->leaves, world->numLeaves, bytes, header->leaves);
    LOAD_CHUNK(u_short, world->leafFaces, world->numLeafFaces, bytes, header->leafFaces);
    LOAD_CHUNK(u_char, world->visData, world->numVisData, bytes, header->visData);
}

static INLINE short world_pointPlaneDist(const VECTOR *point, const ps1bsp_plane_t *plane)
{
    // Make use of axis-aligned planes to skip the need for a dot product
    if (plane->type < 3)
        return (short)(((int*)point)[plane->type] - plane->dist);
    
    return (short)m_pointPlaneDist2(*point, plane->normal, plane->dist);
}

static INLINE short world_planeDot(const SVECTOR *dir, const ps1bsp_plane_t *plane)
{
    // Make use of axis-aligned planes to skip the need for a dot product
    if (plane->type < 3)
        return ((short*)dir)[plane->type];

    return (short)m_dot12(*dir, plane->normal);
}

static INLINE char world_cull_backface(const world_t *world, const ps1bsp_face_t *face, short *dot)
{
    // Backface culling using the face's plane and center point
    // This eliminates the need for normal clipping per polygon
    SVECTOR cam_vec;
    cam_vec.vx = face->center.vx - cam_pos.vx;
    cam_vec.vy = face->center.vy - cam_pos.vy;
    cam_vec.vz = face->center.vz - cam_pos.vz;

    // Check if the face's plane points towards the camera
    const ps1bsp_plane_t *plane = &world->planes[face->planeId];
    *dot = world_planeDot(&cam_vec, plane);
    if ((*dot >= 0) ^ face->side)
        return 1;

    // Check if the face is behind the camera 
    // TODO: this may cull faces close to the camera, use VectorNormalS to normalize and check for angles < -60 degrees
    // TODO: or, just check against all corners of the face...
    short camDot = m_dot12(cam_vec, cam_dir);
    return camDot < 0;
}

static void world_drawface_fast(const world_t *world, const ps1bsp_face_t *face)
{
    // TODO: early primitive buffer check (POLY_G4)

    short dot;
    if (world_cull_backface(world, face, &dot))
        return;

    // Draw untextured, vertex colored faces, skipping the entire polygon tessellation step
    SVECTOR *verts = (SVECTOR*)(scratchpad + 256);
    SVECTOR *curVert = verts;
    const ps1bsp_facevertex_t *faceVertex = &world->faceVertices[face->firstFaceVertex];
    for (u_char vertIdx = 0; vertIdx < face->numFaceVertices; ++vertIdx, ++faceVertex, ++curVert)
    {
        const ps1bsp_vertex_t *vert = &world->vertices[faceVertex->index];
        curVert->vx = vert->x;
        curVert->vy = vert->y;
        curVert->vz = vert->z;
        curVert->pad = (unsigned short)(size_t)face & 0xFF; // TODO: apply averaged light * color value
    }

    if (face->numFaceVertices == 3)
        draw_trianglefan_lit(verts, 3);
    else
        draw_quadstrip_lit(verts, face->numFaceVertices);
}

static void world_drawface_lit(const world_t *world, const ps1bsp_face_t *face)
{
    // TODO: early primitive buffer check (POLY_G4)

    short dot;
    if (world_cull_backface(world, face, &dot))
        return;

    // Draw untextured, vertex colored polygons
    SVECTOR *verts = (SVECTOR*)(scratchpad + 256);
    const ps1bsp_polygon_t* poly = &world->polygons[face->firstPolygon];
    for (u_char polyIdx = 0; polyIdx < face->numPolygons; ++polyIdx, ++poly)
    {
        ps1bsp_polyvertex_t *polyVertex = &world->polyVertices[poly->firstPolyVertex];
        SVECTOR *curVert = verts;
        for (u_char vertIdx = 0; vertIdx < poly->numPolyVertices; ++vertIdx, ++polyVertex, ++curVert)
        {
            const ps1bsp_vertex_t *vert = &world->vertices[polyVertex->index];
            curVert->vx = vert->x;
            curVert->vy = vert->y;
            curVert->vz = vert->z;
            curVert->pad = polyVertex->light;
        }

        if (poly->numPolyVertices == 3)
            draw_trianglefan_lit(verts, 3);
        else
            draw_quadstrip_lit(verts, poly->numPolyVertices);
    }
}

static void world_drawface_textured(const world_t *world, const ps1bsp_face_t *face)
{
    // NOTE: this value could be REALLY useful for determining the tessellation subdivisions. It has camera distance *and* angle in it.
    // Just include the face size/area for an approximate screen size. Maybe also separate x/y/z for angle-dependent tessellation.
    short dot;
    if (world_cull_backface(world, face, &dot))
        return;

    // TODO: do an early primitive buffer check here (POLY_GT4), so we can skip vertex copying if it's already full
    // When doing tessellation, we need the above dot product to decide how many polys to draw, so we can only safely check the primbuffer size here
    // Though since we're drawing front-to-back and tessellation will only happen close to the camera, I suppose we could get away with just checking for non-tessellated polycounts

    // Draw textured, vertex colored polygons
    STVECTOR *verts = (STVECTOR*)(scratchpad + 256);
    ps1bsp_texture_t *faceTexture = &world->textures[face->textureId];
    const ps1bsp_polygon_t* poly = &world->polygons[face->firstPolygon];
    for (u_char polyIdx = 0; polyIdx < face->numPolygons; ++polyIdx, ++poly)
    {
        ps1bsp_polyvertex_t *polyVertex = &world->polyVertices[poly->firstPolyVertex];
        STVECTOR *curVert = verts;
        for (u_char vertIdx = 0; vertIdx < poly->numPolyVertices; ++vertIdx, ++polyVertex, ++curVert)
        {
            const ps1bsp_vertex_t *vert = &world->vertices[polyVertex->index];
            curVert->vx = vert->x;
            curVert->vy = vert->y;
            curVert->vz = vert->z;
            curVert->u = (u_short)polyVertex->u;
            curVert->v = (u_short)polyVertex->v;
            curVert->pad = polyVertex->light;
        }

        if (face->flags & SURF_DRAWWATER)
            draw_quadstrip_water(verts, poly->numPolyVertices, faceTexture->tpage);
        else
            draw_quadstrip_textured(verts, poly->numPolyVertices, faceTexture->tpage);
    }
}

static void (*world_drawface)(const world_t*, const ps1bsp_face_t*) = &world_drawface_fast;

static void world_drawnode(const world_t *world, short nodeIdx, u_char *pvs)
{
    if (nodeIdx < 0)    // Leaf node
    {
        // Check if this leaf is visible from the current camera position
        u_short test = ~nodeIdx - 1;
        if ((pvs[test >> 3] & (1 << (test & 0x7))) == 0)
            return;

        const ps1bsp_leaf_t *leaf = &world->leaves[~nodeIdx];
        // if (!frustum_sphereInside(&leaf->boundingSphere))    // TODO: not sure if it's actually faster to do all these additional frustum checks
        //     return;

        u_long frameNum = time_getFrameNumber();
        const u_short *leafFace = &world->leafFaces[leaf->firstLeafFace];

        for (u_short leafFaceIdx = 0; leafFaceIdx < leaf->numLeafFaces; ++leafFaceIdx, ++leafFace)
        {
            ps1bsp_face_t *face = &world->faces[*leafFace];

            // Check if we've already drawn this face on the current frame
            // NOTE: this can cause sorting issues when rendering front-to-back with leaf-based depth
            if (face->drawFrame == frameNum)
                continue;

            world_drawface(world, face);
            face->drawFrame = frameNum;
        }

        return;
    }

    // Perform frustum culling
    const ps1bsp_node_t *node = &world->nodes[nodeIdx];
    if (!frustum_sphereInside(&node->boundingSphere))
        return;

    const ps1bsp_plane_t *plane = &world->planes[node->planeId];
    short dist = world_pointPlaneDist(&cam_pos, plane);
    
    // Draw child nodes in front-to-back order; adding faces to the OT will reverse the drawing order
    // PLAN: traverse back-to-front to determine leaf order & depth, then draw faces front-to-back with primbuf checks
    char order = dist < 0;
    world_drawnode(world, node->children[order], pvs);
    world_drawnode(world, node->children[order ^ 1], pvs);
}

// Decompress PVS data for the given leaf ID and store it in RAM at the given buffer pointer location.
// Returns the memory location of decompressed PVS data, and moves the buffer pointer forward.
static u_char *world_loadVisData(const world_t *world, u_short leafIdx, u_char **buffer)
{
    u_char *head = *buffer;
    u_char *tail = head;

    const ps1bsp_leaf_t *leaf = &world->leaves[leafIdx];
    
    const u_char *v = &world->visData[leaf->vislist];
    for (int l = 1; l < world->numLeaves; )
    {
        u_char bits = *v++;
        if (bits)
        {
            *tail++ = bits;
            l += 8;
        }
        else
        {
            u_char skip = *v++;
            for (u_char i = 0; i < skip; ++i, l += 8)
            {
                *tail++ = 0;
            }
        }
    }

    *buffer = tail;
    return head;
}

static u_char *world_noVisData(const world_t *world, u_char **buffer)
{
    u_char *head = *buffer;
    u_char *tail = head;

    for (int l = 1; l < world->numLeaves; l += 8)
    {
        *tail++ = 0xFF;
    }

    *buffer = tail;
    return head;
}

static u_short world_leafAtPoint(const world_t *world, const VECTOR *point)
{
    short nodeIdx = 0;
    while (nodeIdx >= 0)
    {
        const ps1bsp_node_t *node = &world->nodes[nodeIdx];
        const ps1bsp_plane_t *plane = &world->planes[node->planeId];

        short dist = world_pointPlaneDist(point, plane);
        nodeIdx = node->children[(dist > 0) ^ 1];
    }

    return ~nodeIdx;
}

void world_draw(const world_t *world)
{
    // The world doesn't move, so we just set the camera view-projection matrix
    gte_SetRotMatrix(&vp_matrix);
	gte_SetTransMatrix(&vp_matrix);

    cam_leaf = world_leafAtPoint(world, &cam_pos);

    u_char *pvsbuf = scratchpad;
    u_char *pvs = world_loadVisData(world, cam_leaf, &pvsbuf);
    //u_char *pvs = world_noVisData(world, &pvsbuf);

    if (enableTexturing)
        world_drawface = &world_drawface_textured;
    else
        world_drawface = &world_drawface_lit;

    world_drawnode(world, 0, pvs);
}