// Sprites: rectangles of 8x8 tiles drawn at arbitrary pixel positions
// with color-0 transparency.
//
// A sprite's pixel data is `widthTiles * heightTiles * 32` bytes,
// 4bpp packed, laid out as a flat blob of 8x8 tiles. Tile order is
// row-major (tile (0,0), tile (1,0), ..., tile (widthTiles-1,0),
// tile (0,1), ...). Within each tile, rows are top-to-bottom and
// each row is 4 bytes (8 pixels at 4bpp packed; high nibble = left
// pixel).
//
// Color 0 is always transparent on draw (DESIGN.md contract). Use a
// tile-block draw if you need an opaque rectangle.
//
// Performance contract: spriteDraw should be at least as fast as the
// IIgs reference (the DESIGN.md spec calls for runtime-compiled draw
// code per CPU). v1 ships with an interpreted fallback that gives
// correct output everywhere; codegen lands per platform after the
// API stabilizes. spritePrewarm is a hint that the application is
// about to draw the sprite repeatedly -- a future codegen-enabled
// build will use it to compile shift variants ahead of the first
// draw. With the interpreter it is a no-op.

#ifndef JOEYLIB_SPRITE_H
#define JOEYLIB_SPRITE_H

#include "platform.h"
#include "surface.h"
#include "types.h"

// Sprite codegen emits per-shift variants. Chunky 4bpp ports (DOS,
// IIgs, Atari ST) only need 2 shifts -- pixel offset 0 (sprite/dest
// byte boundaries align) and offset 1 (every dest byte combines two
// sprite bytes' nibbles). Planar ports (Amiga -- 8 px per plane byte)
// need 8 shifts: one for each x % 8 alignment, so smooth horizontal
// motion at any pixel position uses pre-shifted source bytes without
// runtime bit-shifting. Allocate the max so routineOffsets[] has
// slots for every variant; chunky ports leave shifts 2..7 as
// SPRITE_NOT_COMPILED, planar ports use all 8.
#define JOEY_SPRITE_SHIFT_COUNT 8

typedef enum {
    SPRITE_FLAGS_NONE = 0
} SpriteFlagsE;

typedef struct SpriteT SpriteT;

// SpriteBackupT holds the destination bytes that lived under a sprite
// before it was drawn, so the application can restore them after the
// sprite moves. Sized for the largest backup the app will need; a
// stack-allocated SpriteBackupT plus a caller-owned byte buffer keeps
// the runtime allocation-free.
typedef struct {
    SpriteT       *sprite;
    int16_t        x;
    int16_t        y;
    uint16_t       width;       // pixels
    uint16_t       height;      // pixels
    uint8_t       *bytes;       // caller-owned, capacity >= sizeBytes
    uint16_t       sizeBytes;
} SpriteBackupT;

// Wrap a tile-data blob in a SpriteT. The tile data must outlive the
// SpriteT; we do not copy it. Returns NULL if widthTiles or
// heightTiles is 0, or if the codegen arena cannot fit a placeholder
// entry for this sprite.
SpriteT *spriteCreate(const uint8_t *tileData, uint8_t widthTiles, uint8_t heightTiles, SpriteFlagsE flags);

// Release a SpriteT and any codegen entries cached for it. The tile
// data the sprite was constructed from is NOT freed -- the caller
// owns that buffer.
void spriteDestroy(SpriteT *sp);

// Compile the sprite's draw routines into the codegen arena. After
// this returns true, spriteDraw uses the compiled fast path on
// platforms where the emitter is wired (currently x86/DOS). Returns
// false if the arena is full (caller may run spriteCompact and
// retry), the platform doesn't have a real emitter yet, or the
// sprite has no source tile data (e.g., it was loaded already
// compiled via spriteLoadFile).
//
// Idempotent: calling on a sprite that's already compiled is a
// no-op and returns true.
bool spriteCompile(SpriteT *sp);

// Hint that this sprite will be drawn soon. Currently a wrapper
// around spriteCompile that ignores the return value, kept for API
// symmetry with the rest of the library and for callers that don't
// care about compile success.
void spritePrewarm(SpriteT *sp);

// Draw the sprite at pixel (x,y) on the destination surface. Pixels
// equal to color 0 in the sprite source are skipped (transparent).
// Off-surface portions are clipped.
void spriteDraw(SurfaceT *s, SpriteT *sp, int16_t x, int16_t y);

// Capture the destination region a subsequent spriteDraw at the same
// (x,y) would write to. backup->bytes must have at least
// (widthTiles*4) * (heightTiles*8) bytes of capacity for fully
// in-bounds draws; for clipped draws only the visible bytes are
// stored. The captured region's exact size is reported in
// backup->sizeBytes.
void spriteSaveUnder(const SurfaceT *s, SpriteT *sp, int16_t x, int16_t y, SpriteBackupT *backup);

// Repaint the destination region from a SpriteBackupT captured by a
// prior spriteSaveUnder. The backup must not have been invalidated
// by other writes that overlapped its captured region.
void spriteRestoreUnder(SurfaceT *s, const SpriteBackupT *backup);

// Combined save-then-draw entry point. The common animation pattern
// captures the destination bytes about to be overwritten, then draws
// the sprite. Both ops share validation, the destination ptr is
// computed once, and a single dirty-rect mark covers both. Saves
// roughly one full dispatcher chain (~150 cyc on IIgs ORCA-C) per
// frame versus calling spriteSaveUnder + spriteDraw separately.
//
// Identical semantics to:
//     spriteSaveUnder(s, sp, x, y, backup);
//     spriteDraw(s, sp, x, y);
// modulo: the dirty rect is marked once for the union (which here is
// just the draw rect, since save doesn't write).
void spriteSaveAndDraw(SurfaceT *s, SpriteT *sp, int16_t x, int16_t y, SpriteBackupT *backup);

// Snapshot an 8x8-aligned region of a SurfaceT into a new SpriteT.
// The captured pixel data is copied into a sprite-owned buffer so
// the source surface can be modified afterwards. Width and height
// are in TILES (each tile = 8x8 pixels). x and y are in pixels and
// must be aligned to a tile boundary (multiple of 8) on the source
// surface; misaligned coordinates return NULL.
SpriteT *spriteCreateFromSurface(const SurfaceT *src, int16_t x, int16_t y,
                                 uint8_t widthTiles, uint8_t heightTiles, SpriteFlagsE flags);

// Load a sprite from a `.spr` file produced by the host-side
// joeysprite tool or by spriteSaveFile. Format is target-native for
// the compiled-code section:
//   byte  0     widthTiles
//   byte  1     heightTiles
//   bytes 2-3   codeSize  (LE16)
//   bytes 4-5   tileBytes (LE16) = widthTiles*heightTiles*32
//   ...         offsets table (JOEY_SPRITE_SHIFT_COUNT *
//               SPRITE_OP_COUNT * uint16_t LE)
//   ...         compiled code (codeSize bytes)
//   ...         raw tile data (tileBytes bytes; tile-major 4bpp)
// The runtime keeps both the compiled bytes (fast-path draws) and
// the tile data (interpreter clip path), so loaded sprites work for
// partially-off-surface draws without crashing.
SpriteT *spriteLoadFile(const char *path, SpriteFlagsE flags);

// Same as spriteLoadFile but parses bytes already in memory.
SpriteT *spriteFromCompiledMem(const uint8_t *data, uint32_t length, SpriteFlagsE flags);

// Persist a sprite to disk in `.spr` format. Sprites created via
// spriteCreate / spriteCreateFromSurface that have not been
// compiled yet are force-compiled here (so the resulting file can
// always be loaded back via spriteLoadFile). Returns false if the
// codegen arena is full or the platform's emitter is not yet
// implemented.
bool spriteSaveFile(SpriteT *sp, const char *path);

// Defragment the codegen arena. Walks live sprite slots and
// memmoves them down to consolidate free space; any holes left by
// destroyed sprites are reclaimed. Costs O(arena_used_bytes); call
// between levels rather than per frame. SpriteT pointers held by
// the application are NOT invalidated -- internal indirection
// through the slot record means draw calls automatically pick up
// the new code address on the next call.
void spriteCompact(void);

// Arena introspection. Used to gauge whether spriteCompact is
// worth running, or whether the codegenBytes budget needs to grow.
// Free space is (Total - Used), but it may be fragmented across
// holes until spriteCompact runs.
uint32_t spriteCodegenBytesUsed(void);
uint32_t spriteCodegenBytesTotal(void);

#endif