// Part of JoeyLib, Copyright (c) 2026 Scott Duensing, under the MIT License.
// See LICENSE.md for license information.
// SPDX-License-Identifier: MIT
//
// 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 (a library contract). Use a
// tile-block draw if you need an opaque rectangle.
//
// Performance contract: jlSpriteDraw should be at least as fast as the
// IIgs reference (the design calls for runtime-compiled draw code per
// CPU). The machine ports have runtime emitters (IIgs 65816,
// Amiga/X68000 68k planar, ST 68k interleaved, DOS/DVX x86), and so does
// Win16 (x86, as DOS);
// jlSpriteCompile compiles the shift/op variants into the codegen arena
// and the draw/save/restore calls dispatch to the compiled routines,
// falling back to the interpreter for clipped draws or uncompiled
// variants. Nothing compiles on demand: compile while loading, or every
// draw interprets. The other hosted ports (JL_SPRITE_INTERPRETED) have no
// emitter and always interpret.

#ifndef JOEYLIB_SPRITE_H
#define JOEYLIB_SPRITE_H

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

// Sprite codegen emits per-shift variants so smooth horizontal motion
// at any pixel position uses pre-shifted routines without runtime
// bit-shifting. The variant count is per-platform (chunky 4bpp ports
// need only x & 1; Amiga planar needs x & 7; ST word-interleaved
// planar needs the full x & 15 intra-group phase) and lives with the
// dispatcher's index macro in src/core/spriteInternal.h so count and
// mask cannot drift. jlSpriteT is opaque to applications, so the
// count is not part of the public ABI.

typedef struct jlSpriteT jlSpriteT;

// jlSpriteBackupT 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 jlSpriteBackupT plus a caller-owned byte buffer keeps
// the runtime allocation-free.
typedef struct {
    jlSpriteT     *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;
} jlSpriteBackupT;

// Load up to maxCels cels from a baked .spr file produced by
// tools/assetbake/assetbake.c. Each cel becomes a freshly-allocated
// jlSpriteT (release with jlSpriteDestroy); the file's cellCount is
// capped at maxCels.
//
// If outPalette is non-NULL and the file embeds a 16-entry $0RGB
// LE16 palette, it is copied in; otherwise outPalette is unchanged.
//
// Returns the number of cels actually written into outCels[]. Returns
// 0 if the file is missing, the magic / target byte is wrong, or no
// cel could be allocated.
uint16_t jlSpriteBankLoad(const char *path, jlSpriteT **outCels, uint16_t maxCels, uint16_t *outPalette);

// Load up to maxCels cels from a baked .spc PRE-COMPILED sprite bank
// (tools/spritebake output for THIS platform). Identical contract to
// jlSpriteBankLoad, but each cel arrives already compiled: its native blit
// routines are copied straight into the codegen arena and the sprite draws at
// full compiled speed with NO runtime jlSpriteCompile. A cel that does not fit
// the arena is loaded uncompiled (interpreter fallback), never an error -- so
// this is a drop-in, faster-starting replacement for jlSpriteBankLoad + a
// jlSpriteCompile pass. Returns 0 if the file is missing, its magic /
// target / shift-count does not match this build, its cel size is out of
// range, or the 64 KB buffer a cel's code is read into cannot be allocated
// (the interpreting ports read it too, then drop the code).
uint16_t jlSpriteBankLoadPrecompiled(const char *path, jlSpriteT **outCels, uint16_t maxCels, uint16_t *outPalette);

// Write an already-compiled sprite bank to a .spc cache file (the runtime twin
// of the offline tools/spritebake). Call after jlSpriteBankLoad + a
// jlSpriteCompile pass so a later launch can jlSpriteBankLoadPrecompiled the
// cache and skip the JIT -- i.e. SHIP the small cross-platform .spr, compile
// once, and cache the (large) per-platform compiled bank on first run. `cels`
// are `count` sprites, all of one size (uncompiled ones are cached
// uncompiled, and nothing compiles them at load: jlSpriteCompile any whose
// jlSpriteCompiledSize is 0); `palette` (or NULL) is embedded. Returns false
// if the cels differ in size or the destination isn't writable (e.g. a
// write-protected disk), which is harmless -- the app just keeps compiling
// at startup.
bool jlSpriteBankSavePrecompiled(const char *path, jlSpriteT **cels, uint16_t count, const uint16_t *palette);

// Wrap a tile-data blob in a jlSpriteT. The tile data must outlive the
// jlSpriteT; we do not copy it. Returns NULL if tileData is NULL,
// widthTiles or heightTiles is 0, or memory runs out. The codegen arena
// is not touched until jlSpriteCompile.
jlSpriteT *jlSpriteCreate(const uint8_t *tileData, uint8_t widthTiles, uint8_t heightTiles);

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

// Compile the sprite's draw, save, and restore routines into the codegen
// arena. jlSpriteDraw never compiles on demand, so call this while
// loading (or load a pre-compiled bank). After it returns true the draw,
// save, and restore calls use the compiled routines that fit: a short
// arena keeps only the x-on-an-8-pixel-boundary draws on the Amiga, ST,
// and X68000, a routine over 16 KB is left out, and anything not compiled
// (and every clipped draw) interprets. Returns false if the arena is full
// even after compacting it (jlLastError says so), the sprite has no
// source tile data, the compiler's 64 KB working buffer cannot be
// allocated, no routine fits, or the port has no emitter
// (JL_SPRITE_INTERPRETED); the sprite still draws, interpreted.
//
// Idempotent: calling on a sprite that's already compiled is a
// no-op and returns true.
bool jlSpriteCompile(jlSpriteT *sp);

// Returns the number of bytes this sprite occupies in the codegen
// arena (sum of all compiled draw/save/restore shift variants).
// Returns 0 if the sprite has never been compiled or has no slot.
// Diagnostic only -- useful for tuning jlConfigT.codegenBytes against
// the actual per-cel cost of the current platform's emitter.
uint32_t jlSpriteCompiledSize(const jlSpriteT *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 jlSpriteDraw(jlSurfaceT *s, jlSpriteT *sp, int16_t x, int16_t y);

// TRUSTED draw (P7-1 family). Same trust contract as
// jlSpriteRestoreUnderTrusted: skips the per-call geometry validation
// for a caller that pre-compiles its sprites and guarantees the sprite
// is FULLY on-surface at (x,y); UB if violated (no clipping happens).
// Falls back to the fully-validated jlSpriteDraw for uncompiled
// sprites and on ports without a compiled fast path.
void jlSpriteDrawTrusted(jlSurfaceT *s, jlSpriteT *sp, int16_t x, int16_t y);

// How many _type elements a backup buffer needs for a widthTiles x
// heightTiles sprite at ANY x, on every port (rounded up), so that
//
//   static uint16_t gBacking[jlSpriteBackupLength(uint16_t, 4, 4)];
//
// is big enough, and jlSpriteBackupLength(uint8_t, w, h) is the size in
// bytes (for jlAlloc, say). The worst case is the 16-pixel-group-aligned
// region covering the sprite (x rounded DOWN to a 16-px boundary, width
// UP): one extra 16-px group, (ceil(w/16) + 1) * 8 bytes per row. Give
// the buffer 2-byte alignment (a uint16_t array has it): on the Amiga,
// ST, and X68000 an odd address sends save and restore down a much
// slower path, and it crashes the trusted variants.
#define jlSpriteBackupLength(_type, _wT, _hT) \
    ((uint16_t)((((((((uint16_t)(_wT) * 8u) + 15u) >> 4) + 1u) * 8u * ((uint16_t)(_hT) * 8u)) + sizeof(_type) - 1u) / sizeof(_type)))

// Capture the destination region a subsequent jlSpriteDraw at the same
// (x,y) would write to. backup->bytes must have room for
// jlSpriteBackupLength(uint8_t, widthTiles, heightTiles) bytes for
// fully in-bounds draws; for clipped draws only the visible bytes are
// stored. The captured region's exact geometry is reported in
// backup->x/y/width/height/sizeBytes: x is rounded down to even on every
// port, a save on the Amiga, ST, and X68000 takes the 16-px-aligned region
// containing the sprite (or its visible part), a compiled save on the IIgs
// only the rows holding an opaque pixel, and a clipped save the visible
// part.
// So the region a caller must not invalidate before restoring is the
// RECORDED region, not just the sprite rect. On the IIgs the buffer must
// not cross a 64 KB bank boundary. With backup->bytes NULL nothing is
// copied: the call only fills in the region a save would record.
void jlSpriteSaveUnder(const jlSurfaceT *s, jlSpriteT *sp, int16_t x, int16_t y, jlSpriteBackupT *backup);

// TRUSTED save-under (P7-1). Same trust contract as
// jlSpriteRestoreUnderTrusted (skips the per-call geometry validation
// for a caller that pre-compiles its sprites and keeps them on-screen;
// UB if violated). Falls back to the fully-validated jlSpriteSaveUnder
// for uncompiled sprites.
void jlSpriteSaveUnderTrusted(const jlSurfaceT *s, jlSpriteT *sp, int16_t x, int16_t y, jlSpriteBackupT *backup);

// Repaint the destination region from a jlSpriteBackupT captured by a
// prior jlSpriteSaveUnder. The backup must not have been invalidated
// by other writes that overlapped its captured region.
void jlSpriteRestoreUnder(jlSurfaceT *s, const jlSpriteBackupT *backup);

// TRUSTED fast-path restore (NATIVE-PERF Phase 7, P7-1). Identical
// effect to jlSpriteRestoreUnder, but SKIPS the geometry-validation
// chain (~11 range/alignment checks + the NULL and sprite-slot gates)
// that jlSpriteRestoreUnder runs every call. The caller MUST guarantee
// ALL of the following or behavior is UNDEFINED (out-of-bounds writes,
// crashes):
//   * s and backup (and backup->sprite, backup->bytes) are non-NULL,
//   * backup was produced by a jlSpriteSaveUnder that took the COMPILED
//     path (a "trusted" backup: 16-px-aligned x/width on planar ports,
//     even x/width on chunky, backup->height == the rows the compiled
//     save covers -- the sprite's pixel height, or on the IIgs the rows
//     holding an opaque pixel),
//   * on the Amiga, ST, and X68000, backup->bytes is at least 2-byte
//     aligned (the trusted save and save-and-draw need that too),
//   * the recorded region is fully on the destination surface.
// It is the expert inner-loop entry for a game that pre-compiles its
// sprites (jlSpriteCompile) and already clamps positions on-screen --
// use jlSpriteRestoreUnder anywhere those invariants are not statically
// guaranteed. Anything the trusted fast path does not recognize (an
// uncompiled shift, a non-stage / non-16-row region on the 68k and x86
// ports)
// falls back to the fully-validated jlSpriteRestoreUnder, so a
// conservative caller degrades safely rather than corrupting -- but the
// NULL / on-surface guarantees above are still the caller's.
void jlSpriteRestoreUnderTrusted(jlSurfaceT *s, const jlSpriteBackupT *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) per
// frame versus calling jlSpriteSaveUnder + jlSpriteDraw separately.
//
// Identical semantics to:
//     jlSpriteSaveUnder(s, sp, x, y, backup);
//     jlSpriteDraw(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 jlSpriteSaveAndDraw(jlSurfaceT *s, jlSpriteT *sp, int16_t x, int16_t y, jlSpriteBackupT *backup);

// TRUSTED save-and-draw (P7-1). The inner-loop companion to
// jlSpriteRestoreUnderTrusted -- same trust contract (skips the per-call
// geometry validation for a caller that pre-compiles its sprites and
// keeps them on-screen; UB if those guarantees are violated). The
// canonical pattern is a trusted save-and-draw this frame, a trusted
// restore next. Falls back to the fully-validated jlSpriteSaveAndDraw
// for anything the fast path does not recognize.
void jlSpriteSaveAndDrawTrusted(jlSurfaceT *s, jlSpriteT *sp, int16_t x, int16_t y, jlSpriteBackupT *backup);

// Snapshot an 8x8-aligned region of a jlSurfaceT into a new jlSpriteT.
// 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. Returns NULL if src is NULL, a dimension is 0, x or y is
// misaligned, the region is not wholly on the surface, or memory runs
// out.
jlSpriteT *jlSpriteCreateFromSurface(const jlSurfaceT *src, int16_t x, int16_t y, uint8_t widthTiles, uint8_t heightTiles);

// 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. jlSpriteT 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 jlSpriteCompact(void);

// Arena introspection. Used to gauge whether jlSpriteCompact is
// worth running, or whether the codegenBytes budget needs to grow.
// Free space is (Total - Used), but it may be fragmented across
// holes until jlSpriteCompact runs. Total is what jlInit was granted,
// which a machine short of memory can make less than codegenBytes; on the
// IIgs in blocks of at most 64 KB that compacting never merges (one
// sprite's code must fit in one). The interpreting ports have no arena:
// both are 0.
uint32_t jlSpriteCodegenBytesUsed(void);
uint32_t jlSpriteCodegenBytesTotal(void);

#endif