diff options
author | Andrei Karas <akaras@inbox.ru> | 2013-09-03 00:02:25 +0300 |
---|---|---|
committer | Andrei Karas <akaras@inbox.ru> | 2013-09-03 00:02:25 +0300 |
commit | dcc5c7ee607aa811f3c9893710c895abb9dd2c2c (patch) | |
tree | fe353899385aa258edb956cf2dba6f52b04ab610 /src/sdl2gfx | |
parent | 937e8601f3091c72c23e4b9621662dc5239209a0 (diff) | |
download | manaplus-dcc5c7ee607aa811f3c9893710c895abb9dd2c2c.tar.gz manaplus-dcc5c7ee607aa811f3c9893710c895abb9dd2c2c.tar.bz2 manaplus-dcc5c7ee607aa811f3c9893710c895abb9dd2c2c.tar.xz manaplus-dcc5c7ee607aa811f3c9893710c895abb9dd2c2c.zip |
add missing files.
Diffstat (limited to 'src/sdl2gfx')
-rw-r--r-- | src/sdl2gfx/SDL_rotozoom.c | 1651 | ||||
-rw-r--r-- | src/sdl2gfx/SDL_rotozoom.h | 123 |
2 files changed, 1774 insertions, 0 deletions
diff --git a/src/sdl2gfx/SDL_rotozoom.c b/src/sdl2gfx/SDL_rotozoom.c new file mode 100644 index 000000000..6313f4008 --- /dev/null +++ b/src/sdl2gfx/SDL_rotozoom.c @@ -0,0 +1,1651 @@ +/* + +SDL2_rotozoom.c: rotozoomer, zoomer and shrinker for 32bit or 8bit surfaces + +Copyright (C) 2012 Andreas Schiffler + +This software is provided 'as-is', without any express or implied +warranty. In no event will the authors be held liable for any damages +arising from the use of this software. + +Permission is granted to anyone to use this software for any purpose, +including commercial applications, and to alter it and redistribute it +freely, subject to the following restrictions: + +1. The origin of this software must not be misrepresented; you must not +claim that you wrote the original software. If you use this software +in a product, an acknowledgment in the product documentation would be +appreciated but is not required. + +2. Altered source versions must be plainly marked as such, and must not be +misrepresented as being the original software. + +3. This notice may not be removed or altered from any source +distribution. + +Andreas Schiffler -- aschiffler at ferzkopp dot net + +*/ + +#ifdef WIN32 +#include <windows.h> +#endif + +#include <stdlib.h> +#include <string.h> + +#include "SDL_rotozoom.h" + +/* ---- Internally used structures */ + +/*! +\brief A 32 bit RGBA pixel. +*/ +typedef struct tColorRGBA { + Uint8 r; + Uint8 g; + Uint8 b; + Uint8 a; +} tColorRGBA; + +/*! +\brief A 8bit Y/palette pixel. +*/ +typedef struct tColorY { + Uint8 y; +} tColorY; + +/*! +\brief Returns maximum of two numbers a and b. +*/ +#define MAX(a,b) (((a) > (b)) ? (a) : (b)) + +/*! +\brief Number of guard rows added to destination surfaces. + +This is a simple but effective workaround for observed issues. +These rows allocate extra memory and are then hidden from the surface. +Rows are added to the end of destination surfaces when they are allocated. +This catches any potential overflows which seem to happen with +just the right src image dimensions and scale/rotation and can lead +to a situation where the program can segfault. +*/ +#define GUARD_ROWS (2) + +/*! +\brief Lower limit of absolute zoom factor or rotation degrees. +*/ +#define VALUE_LIMIT 0.001 + +/*! +\brief Returns colorkey info for a surface +*/ +Uint32 _colorkey(SDL_Surface *src) +{ + Uint32 key = 0; + SDL_GetColorKey(src, &key); + return key; +} + + +/*! +\brief Internal 32 bit integer-factor averaging Shrinker. + +Shrinks 32 bit RGBA/ABGR 'src' surface to 'dst' surface. +Averages color and alpha values values of src pixels to calculate dst pixels. +Assumes src and dst surfaces are of 32 bit depth. +Assumes dst surface was allocated with the correct dimensions. + +\param src The surface to shrink (input). +\param dst The shrunken surface (output). +\param factorx The horizontal shrinking ratio. +\param factory The vertical shrinking ratio. + +\return 0 for success or -1 for error. +*/ +int _shrinkSurfaceRGBA(SDL_Surface * src, SDL_Surface * dst, int factorx, int factory) +{ + int x, y, dx, dy, sgap, dgap, ra, ga, ba, aa; + int n_average; + tColorRGBA *sp, *osp, *oosp; + tColorRGBA *dp; + + /* + * Averaging integer shrink + */ + + /* Precalculate division factor */ + n_average = factorx*factory; + + /* + * Scan destination + */ + sp = (tColorRGBA *) src->pixels; + sgap = src->pitch - src->w * 4; + + dp = (tColorRGBA *) dst->pixels; + dgap = dst->pitch - dst->w * 4; + + for (y = 0; y < dst->h; y++) { + + osp=sp; + for (x = 0; x < dst->w; x++) { + + /* Trace out source box and accumulate */ + oosp=sp; + ra=ga=ba=aa=0; + for (dy=0; dy < factory; dy++) { + for (dx=0; dx < factorx; dx++) { + ra += sp->r; + ga += sp->g; + ba += sp->b; + aa += sp->a; + + sp++; + } + /* src dx loop */ + sp = (tColorRGBA *)((Uint8*)sp + (src->pitch - 4*factorx)); // next y + } + /* src dy loop */ + + /* next box-x */ + sp = (tColorRGBA *)((Uint8*)oosp + 4*factorx); + + /* Store result in destination */ + dp->r = ra/n_average; + dp->g = ga/n_average; + dp->b = ba/n_average; + dp->a = aa/n_average; + + /* + * Advance destination pointer + */ + dp++; + } + /* dst x loop */ + + /* next box-y */ + sp = (tColorRGBA *)((Uint8*)osp + src->pitch*factory); + + /* + * Advance destination pointers + */ + dp = (tColorRGBA *) ((Uint8 *) dp + dgap); + } + /* dst y loop */ + + return (0); +} + +/*! +\brief Internal 8 bit integer-factor averaging shrinker. + +Shrinks 8bit Y 'src' surface to 'dst' surface. +Averages color (brightness) values values of src pixels to calculate dst pixels. +Assumes src and dst surfaces are of 8 bit depth. +Assumes dst surface was allocated with the correct dimensions. + +\param src The surface to shrink (input). +\param dst The shrunken surface (output). +\param factorx The horizontal shrinking ratio. +\param factory The vertical shrinking ratio. + +\return 0 for success or -1 for error. +*/ +int _shrinkSurfaceY(SDL_Surface * src, SDL_Surface * dst, int factorx, int factory) +{ + int x, y, dx, dy, sgap, dgap, a; + int n_average; + Uint8 *sp, *osp, *oosp; + Uint8 *dp; + + /* + * Averaging integer shrink + */ + + /* Precalculate division factor */ + n_average = factorx*factory; + + /* + * Scan destination + */ + sp = (Uint8 *) src->pixels; + sgap = src->pitch - src->w; + + dp = (Uint8 *) dst->pixels; + dgap = dst->pitch - dst->w; + + for (y = 0; y < dst->h; y++) { + + osp=sp; + for (x = 0; x < dst->w; x++) { + + /* Trace out source box and accumulate */ + oosp=sp; + a=0; + for (dy=0; dy < factory; dy++) { + for (dx=0; dx < factorx; dx++) { + a += (*sp); + /* next x */ + sp++; + } + /* end src dx loop */ + /* next y */ + sp = (Uint8 *)((Uint8*)sp + (src->pitch - factorx)); + } + /* end src dy loop */ + + /* next box-x */ + sp = (Uint8 *)((Uint8*)oosp + factorx); + + /* Store result in destination */ + *dp = a/n_average; + + /* + * Advance destination pointer + */ + dp++; + } + /* end dst x loop */ + + /* next box-y */ + sp = (Uint8 *)((Uint8*)osp + src->pitch*factory); + + /* + * Advance destination pointers + */ + dp = (Uint8 *)((Uint8 *)dp + dgap); + } + /* end dst y loop */ + + return (0); +} + +/*! +\brief Internal 32 bit Zoomer with optional anti-aliasing by bilinear interpolation. + +Zooms 32 bit RGBA/ABGR 'src' surface to 'dst' surface. +Assumes src and dst surfaces are of 32 bit depth. +Assumes dst surface was allocated with the correct dimensions. + +\param src The surface to zoom (input). +\param dst The zoomed surface (output). +\param flipx Flag indicating if the image should be horizontally flipped. +\param flipy Flag indicating if the image should be vertically flipped. +\param smooth Antialiasing flag; set to SMOOTHING_ON to enable. + +\return 0 for success or -1 for error. +*/ +int _zoomSurfaceRGBA(SDL_Surface * src, SDL_Surface * dst, int flipx, int flipy, int smooth) +{ + int x, y, sx, sy, ssx, ssy, *sax, *say, *csax, *csay, *salast, csx, csy, ex, ey, cx, cy, sstep, sstepx, sstepy; + tColorRGBA *c00, *c01, *c10, *c11; + tColorRGBA *sp, *csp, *dp; + int spixelgap, spixelw, spixelh, dgap, t1, t2; + + /* + * Allocate memory for row/column increments + */ + if ((sax = (int *) malloc((dst->w + 1) * sizeof(Uint32))) == NULL) { + return (-1); + } + if ((say = (int *) malloc((dst->h + 1) * sizeof(Uint32))) == NULL) { + free(sax); + return (-1); + } + + /* + * Precalculate row increments + */ + spixelw = (src->w - 1); + spixelh = (src->h - 1); + if (smooth) { + sx = (int) (65536.0 * (float) spixelw / (float) (dst->w - 1)); + sy = (int) (65536.0 * (float) spixelh / (float) (dst->h - 1)); + } else { + sx = (int) (65536.0 * (float) (src->w) / (float) (dst->w)); + sy = (int) (65536.0 * (float) (src->h) / (float) (dst->h)); + } + + /* Maximum scaled source size */ + ssx = (src->w << 16) - 1; + ssy = (src->h << 16) - 1; + + /* Precalculate horizontal row increments */ + csx = 0; + csax = sax; + for (x = 0; x <= dst->w; x++) { + *csax = csx; + csax++; + csx += sx; + + /* Guard from overflows */ + if (csx > ssx) { + csx = ssx; + } + } + + /* Precalculate vertical row increments */ + csy = 0; + csay = say; + for (y = 0; y <= dst->h; y++) { + *csay = csy; + csay++; + csy += sy; + + /* Guard from overflows */ + if (csy > ssy) { + csy = ssy; + } + } + + sp = (tColorRGBA *) src->pixels; + dp = (tColorRGBA *) dst->pixels; + dgap = dst->pitch - dst->w * 4; + spixelgap = src->pitch/4; + + if (flipx) sp += spixelw; + if (flipy) sp += (spixelgap * spixelh); + + /* + * Switch between interpolating and non-interpolating code + */ + if (smooth) { + + /* + * Interpolating Zoom + */ + csay = say; + for (y = 0; y < dst->h; y++) { + csp = sp; + csax = sax; + for (x = 0; x < dst->w; x++) { + /* + * Setup color source pointers + */ + ex = (*csax & 0xffff); + ey = (*csay & 0xffff); + cx = (*csax >> 16); + cy = (*csay >> 16); + sstepx = cx < spixelw; + sstepy = cy < spixelh; + c00 = sp; + c01 = sp; + c10 = sp; + if (sstepy) { + if (flipy) { + c10 -= spixelgap; + } else { + c10 += spixelgap; + } + } + c11 = c10; + if (sstepx) { + if (flipx) { + c01--; + c11--; + } else { + c01++; + c11++; + } + } + + /* + * Draw and interpolate colors + */ + t1 = ((((c01->r - c00->r) * ex) >> 16) + c00->r) & 0xff; + t2 = ((((c11->r - c10->r) * ex) >> 16) + c10->r) & 0xff; + dp->r = (((t2 - t1) * ey) >> 16) + t1; + t1 = ((((c01->g - c00->g) * ex) >> 16) + c00->g) & 0xff; + t2 = ((((c11->g - c10->g) * ex) >> 16) + c10->g) & 0xff; + dp->g = (((t2 - t1) * ey) >> 16) + t1; + t1 = ((((c01->b - c00->b) * ex) >> 16) + c00->b) & 0xff; + t2 = ((((c11->b - c10->b) * ex) >> 16) + c10->b) & 0xff; + dp->b = (((t2 - t1) * ey) >> 16) + t1; + t1 = ((((c01->a - c00->a) * ex) >> 16) + c00->a) & 0xff; + t2 = ((((c11->a - c10->a) * ex) >> 16) + c10->a) & 0xff; + dp->a = (((t2 - t1) * ey) >> 16) + t1; + /* + * Advance source pointer x + */ + salast = csax; + csax++; + sstep = (*csax >> 16) - (*salast >> 16); + if (flipx) { + sp -= sstep; + } else { + sp += sstep; + } + + /* + * Advance destination pointer x + */ + dp++; + } + /* + * Advance source pointer y + */ + salast = csay; + csay++; + sstep = (*csay >> 16) - (*salast >> 16); + sstep *= spixelgap; + if (flipy) { + sp = csp - sstep; + } else { + sp = csp + sstep; + } + + /* + * Advance destination pointer y + */ + dp = (tColorRGBA *) ((Uint8 *) dp + dgap); + } + } else { + /* + * Non-Interpolating Zoom + */ + csay = say; + for (y = 0; y < dst->h; y++) { + csp = sp; + csax = sax; + for (x = 0; x < dst->w; x++) { + /* + * Draw + */ + *dp = *sp; + + /* + * Advance source pointer x + */ + salast = csax; + csax++; + sstep = (*csax >> 16) - (*salast >> 16); + if (flipx) sstep = -sstep; + sp += sstep; + + /* + * Advance destination pointer x + */ + dp++; + } + /* + * Advance source pointer y + */ + salast = csay; + csay++; + sstep = (*csay >> 16) - (*salast >> 16); + sstep *= spixelgap; + if (flipy) sstep = -sstep; + sp = csp + sstep; + + /* + * Advance destination pointer y + */ + dp = (tColorRGBA *) ((Uint8 *) dp + dgap); + } + } + + /* + * Remove temp arrays + */ + free(sax); + free(say); + + return (0); +} + +/*! + +\brief Internal 8 bit Zoomer without smoothing. + +Zooms 8bit palette/Y 'src' surface to 'dst' surface. +Assumes src and dst surfaces are of 8 bit depth. +Assumes dst surface was allocated with the correct dimensions. + +\param src The surface to zoom (input). +\param dst The zoomed surface (output). +\param flipx Flag indicating if the image should be horizontally flipped. +\param flipy Flag indicating if the image should be vertically flipped. + +\return 0 for success or -1 for error. +*/ +int _zoomSurfaceY(SDL_Surface * src, SDL_Surface * dst, int flipx, int flipy) +{ + int x, y; + Uint32 *sax, *say, *csax, *csay; + int csx, csy; + Uint8 *sp, *dp, *csp; + int dgap; + + /* + * Allocate memory for row increments + */ + if ((sax = (Uint32 *) malloc((dst->w + 1) * sizeof(Uint32))) == NULL) { + return (-1); + } + if ((say = (Uint32 *) malloc((dst->h + 1) * sizeof(Uint32))) == NULL) { + free(sax); + return (-1); + } + + /* + * Pointer setup + */ + sp = csp = (Uint8 *) src->pixels; + dp = (Uint8 *) dst->pixels; + dgap = dst->pitch - dst->w; + + if (flipx) csp += (src->w-1); + if (flipy) csp = ( (Uint8*)csp + src->pitch*(src->h-1) ); + + /* + * Precalculate row increments + */ + csx = 0; + csax = sax; + for (x = 0; x < dst->w; x++) { + csx += src->w; + *csax = 0; + while (csx >= dst->w) { + csx -= dst->w; + (*csax)++; + } + (*csax) = (*csax) * (flipx ? -1 : 1); + csax++; + } + csy = 0; + csay = say; + for (y = 0; y < dst->h; y++) { + csy += src->h; + *csay = 0; + while (csy >= dst->h) { + csy -= dst->h; + (*csay)++; + } + (*csay) = (*csay) * (flipy ? -1 : 1); + csay++; + } + + /* + * Draw + */ + csay = say; + for (y = 0; y < dst->h; y++) { + csax = sax; + sp = csp; + for (x = 0; x < dst->w; x++) { + /* + * Draw + */ + *dp = *sp; + /* + * Advance source pointers + */ + sp += (*csax); + csax++; + /* + * Advance destination pointer + */ + dp++; + } + /* + * Advance source pointer (for row) + */ + csp += ((*csay) * src->pitch); + csay++; + + /* + * Advance destination pointers + */ + dp += dgap; + } + + /* + * Remove temp arrays + */ + free(sax); + free(say); + + return (0); +} + +/*! +\brief Internal 32 bit rotozoomer with optional anti-aliasing. + +Rotates and zooms 32 bit RGBA/ABGR 'src' surface to 'dst' surface based on the control +parameters by scanning the destination surface and applying optionally anti-aliasing +by bilinear interpolation. +Assumes src and dst surfaces are of 32 bit depth. +Assumes dst surface was allocated with the correct dimensions. + +\param src Source surface. +\param dst Destination surface. +\param cx Horizontal center coordinate. +\param cy Vertical center coordinate. +\param isin Integer version of sine of angle. +\param icos Integer version of cosine of angle. +\param flipx Flag indicating horizontal mirroring should be applied. +\param flipy Flag indicating vertical mirroring should be applied. +\param smooth Flag indicating anti-aliasing should be used. +*/ +void _transformSurfaceRGBA(SDL_Surface * src, SDL_Surface * dst, int cx, int cy, int isin, int icos, int flipx, int flipy, int smooth) +{ + int x, y, t1, t2, dx, dy, xd, yd, sdx, sdy, ax, ay, ex, ey, sw, sh; + tColorRGBA c00, c01, c10, c11, cswap; + tColorRGBA *pc, *sp; + int gap; + + /* + * Variable setup + */ + xd = ((src->w - dst->w) << 15); + yd = ((src->h - dst->h) << 15); + ax = (cx << 16) - (icos * cx); + ay = (cy << 16) - (isin * cx); + sw = src->w - 1; + sh = src->h - 1; + pc = (tColorRGBA*) dst->pixels; + gap = dst->pitch - dst->w * 4; + + /* + * Switch between interpolating and non-interpolating code + */ + if (smooth) { + for (y = 0; y < dst->h; y++) { + dy = cy - y; + sdx = (ax + (isin * dy)) + xd; + sdy = (ay - (icos * dy)) + yd; + for (x = 0; x < dst->w; x++) { + dx = (sdx >> 16); + dy = (sdy >> 16); + if (flipx) dx = sw - dx; + if (flipy) dy = sh - dy; + if ((dx > -1) && (dy > -1) && (dx < (src->w-1)) && (dy < (src->h-1))) { + sp = (tColorRGBA *)src->pixels;; + sp += ((src->pitch/4) * dy); + sp += dx; + c00 = *sp; + sp += 1; + c01 = *sp; + sp += (src->pitch/4); + c11 = *sp; + sp -= 1; + c10 = *sp; + if (flipx) { + cswap = c00; c00=c01; c01=cswap; + cswap = c10; c10=c11; c11=cswap; + } + if (flipy) { + cswap = c00; c00=c10; c10=cswap; + cswap = c01; c01=c11; c11=cswap; + } + /* + * Interpolate colors + */ + ex = (sdx & 0xffff); + ey = (sdy & 0xffff); + t1 = ((((c01.r - c00.r) * ex) >> 16) + c00.r) & 0xff; + t2 = ((((c11.r - c10.r) * ex) >> 16) + c10.r) & 0xff; + pc->r = (((t2 - t1) * ey) >> 16) + t1; + t1 = ((((c01.g - c00.g) * ex) >> 16) + c00.g) & 0xff; + t2 = ((((c11.g - c10.g) * ex) >> 16) + c10.g) & 0xff; + pc->g = (((t2 - t1) * ey) >> 16) + t1; + t1 = ((((c01.b - c00.b) * ex) >> 16) + c00.b) & 0xff; + t2 = ((((c11.b - c10.b) * ex) >> 16) + c10.b) & 0xff; + pc->b = (((t2 - t1) * ey) >> 16) + t1; + t1 = ((((c01.a - c00.a) * ex) >> 16) + c00.a) & 0xff; + t2 = ((((c11.a - c10.a) * ex) >> 16) + c10.a) & 0xff; + pc->a = (((t2 - t1) * ey) >> 16) + t1; + } + sdx += icos; + sdy += isin; + pc++; + } + pc = (tColorRGBA *) ((Uint8 *) pc + gap); + } + } else { + for (y = 0; y < dst->h; y++) { + dy = cy - y; + sdx = (ax + (isin * dy)) + xd; + sdy = (ay - (icos * dy)) + yd; + for (x = 0; x < dst->w; x++) { + dx = (short) (sdx >> 16); + dy = (short) (sdy >> 16); + if (flipx) dx = (src->w-1)-dx; + if (flipy) dy = (src->h-1)-dy; + if ((dx >= 0) && (dy >= 0) && (dx < src->w) && (dy < src->h)) { + sp = (tColorRGBA *) ((Uint8 *) src->pixels + src->pitch * dy); + sp += dx; + *pc = *sp; + } + sdx += icos; + sdy += isin; + pc++; + } + pc = (tColorRGBA *) ((Uint8 *) pc + gap); + } + } +} + +/*! + +\brief Rotates and zooms 8 bit palette/Y 'src' surface to 'dst' surface without smoothing. + +Rotates and zooms 8 bit RGBA/ABGR 'src' surface to 'dst' surface based on the control +parameters by scanning the destination surface. +Assumes src and dst surfaces are of 8 bit depth. +Assumes dst surface was allocated with the correct dimensions. + +\param src Source surface. +\param dst Destination surface. +\param cx Horizontal center coordinate. +\param cy Vertical center coordinate. +\param isin Integer version of sine of angle. +\param icos Integer version of cosine of angle. +\param flipx Flag indicating horizontal mirroring should be applied. +\param flipy Flag indicating vertical mirroring should be applied. +*/ +void transformSurfaceY(SDL_Surface * src, SDL_Surface * dst, int cx, int cy, int isin, int icos, int flipx, int flipy) +{ + int x, y, dx, dy, xd, yd, sdx, sdy, ax, ay, sw, sh; + tColorY *pc, *sp; + int gap; + + /* + * Variable setup + */ + xd = ((src->w - dst->w) << 15); + yd = ((src->h - dst->h) << 15); + ax = (cx << 16) - (icos * cx); + ay = (cy << 16) - (isin * cx); + sw = src->w - 1; + sh = src->h - 1; + pc = (tColorY*) dst->pixels; + gap = dst->pitch - dst->w; + /* + * Clear surface to colorkey + */ + memset(pc, (int)(_colorkey(src) & 0xff), dst->pitch * dst->h); + /* + * Iterate through destination surface + */ + for (y = 0; y < dst->h; y++) { + dy = cy - y; + sdx = (ax + (isin * dy)) + xd; + sdy = (ay - (icos * dy)) + yd; + for (x = 0; x < dst->w; x++) { + dx = (short) (sdx >> 16); + dy = (short) (sdy >> 16); + if (flipx) dx = (src->w-1)-dx; + if (flipy) dy = (src->h-1)-dy; + if ((dx >= 0) && (dy >= 0) && (dx < src->w) && (dy < src->h)) { + sp = (tColorY *) (src->pixels); + sp += (src->pitch * dy + dx); + *pc = *sp; + } + sdx += icos; + sdy += isin; + pc++; + } + pc += gap; + } +} + +/*! +\brief Rotates a 32 bit surface in increments of 90 degrees. + +Specialized 90 degree rotator which rotates a 'src' surface in 90 degree +increments clockwise returning a new surface. Faster than rotozoomer since +not scanning or interpolation takes place. Input surface must be 32 bit. +(code contributed by J. Schiller, improved by C. Allport and A. Schiffler) + +\param src Source surface to rotate. +\param numClockwiseTurns Number of clockwise 90 degree turns to apply to the source. + +\returns The new, rotated surface; or NULL for surfaces with incorrect input format. +*/ +SDL_Surface* rotateSurface90Degrees(SDL_Surface* src, int numClockwiseTurns) +{ + int row, col, newWidth, newHeight; + int bpp, src_ipr, dst_ipr; + SDL_Surface* dst; + Uint32* srcBuf; + Uint32* dstBuf; + + /* Has to be a valid surface pointer and only 32-bit surfaces (for now) */ + if (!src || src->format->BitsPerPixel != 32) { return NULL; } + + /* normalize numClockwiseTurns */ + while(numClockwiseTurns < 0) { numClockwiseTurns += 4; } + numClockwiseTurns = (numClockwiseTurns % 4); + + /* if it's even, our new width will be the same as the source surface */ + newWidth = (numClockwiseTurns % 2) ? (src->h) : (src->w); + newHeight = (numClockwiseTurns % 2) ? (src->w) : (src->h); + dst = SDL_CreateRGBSurface( src->flags, newWidth, newHeight, src->format->BitsPerPixel, + src->format->Rmask, + src->format->Gmask, + src->format->Bmask, + src->format->Amask); + if(!dst) { + return NULL; + } + + if (SDL_MUSTLOCK(dst)) { + SDL_LockSurface(dst); + } + if (SDL_MUSTLOCK(dst)) { + SDL_LockSurface(dst); + } + + /* Calculate int-per-row */ + bpp = src->format->BitsPerPixel / 8; + src_ipr = src->pitch / bpp; + dst_ipr = dst->pitch / bpp; + + switch(numClockwiseTurns) { + case 0: /* Make a copy of the surface */ + { + /* Unfortunately SDL_BlitSurface cannot be used to make a copy of the surface + since it does not preserve alpha. */ + + if (src->pitch == dst->pitch) { + /* If the pitch is the same for both surfaces, the memory can be copied all at once. */ + memcpy(dst->pixels, src->pixels, (src->h * src->pitch)); + } + else + { + /* If the pitch differs, copy each row separately */ + srcBuf = (Uint32*)(src->pixels); + dstBuf = (Uint32*)(dst->pixels); + for (row = 0; row < src->h; row++) { + memcpy(dstBuf, srcBuf, dst->w * bpp); + srcBuf += src_ipr; + dstBuf += dst_ipr; + } /* end for(col) */ + } /* end for(row) */ + } + break; + + /* rotate clockwise */ + case 1: /* rotated 90 degrees clockwise */ + { + for (row = 0; row < src->h; ++row) { + srcBuf = (Uint32*)(src->pixels) + (row * src_ipr); + dstBuf = (Uint32*)(dst->pixels) + (dst->w - row - 1); + for (col = 0; col < src->w; ++col) { + *dstBuf = *srcBuf; + ++srcBuf; + dstBuf += dst_ipr; + } + /* end for(col) */ + } + /* end for(row) */ + } + break; + + case 2: /* rotated 180 degrees clockwise */ + { + for (row = 0; row < src->h; ++row) { + srcBuf = (Uint32*)(src->pixels) + (row * src_ipr); + dstBuf = (Uint32*)(dst->pixels) + ((dst->h - row - 1) * dst_ipr) + (dst->w - 1); + for (col = 0; col < src->w; ++col) { + *dstBuf = *srcBuf; + ++srcBuf; + --dstBuf; + } + } + } + break; + + case 3: + { + for (row = 0; row < src->h; ++row) { + srcBuf = (Uint32*)(src->pixels) + (row * src_ipr); + dstBuf = (Uint32*)(dst->pixels) + row + ((dst->h - 1) * dst_ipr); + for (col = 0; col < src->w; ++col) { + *dstBuf = *srcBuf; + ++srcBuf; + dstBuf -= dst_ipr; + } + } + } + break; + } + /* end switch */ + + if (SDL_MUSTLOCK(src)) { + SDL_UnlockSurface(src); + } + if (SDL_MUSTLOCK(dst)) { + SDL_UnlockSurface(dst); + } + + return dst; +} + + +/*! +\brief Internal target surface sizing function for rotozooms with trig result return. + +\param width The source surface width. +\param height The source surface height. +\param angle The angle to rotate in degrees. +\param zoomx The horizontal scaling factor. +\param zoomy The vertical scaling factor. +\param dstwidth The calculated width of the destination surface. +\param dstheight The calculated height of the destination surface. +\param canglezoom The sine of the angle adjusted by the zoom factor. +\param sanglezoom The cosine of the angle adjusted by the zoom factor. + +*/ +void _rotozoomSurfaceSizeTrig(int width, int height, double angle, double zoomx, double zoomy, + int *dstwidth, int *dstheight, + double *canglezoom, double *sanglezoom) +{ + double x, y, cx, cy, sx, sy; + double radangle; + int dstwidthhalf, dstheighthalf; + + /* + * Determine destination width and height by rotating a centered source box + */ + radangle = angle * (M_PI / 180.0); + *sanglezoom = sin(radangle); + *canglezoom = cos(radangle); + *sanglezoom *= zoomx; + *canglezoom *= zoomx; + x = (double)(width / 2); + y = (double)(height / 2); + cx = *canglezoom * x; + cy = *canglezoom * y; + sx = *sanglezoom * x; + sy = *sanglezoom * y; + + dstwidthhalf = MAX((int) + ceil(MAX(MAX(MAX(fabs(cx + sy), fabs(cx - sy)), fabs(-cx + sy)), fabs(-cx - sy))), 1); + dstheighthalf = MAX((int) + ceil(MAX(MAX(MAX(fabs(sx + cy), fabs(sx - cy)), fabs(-sx + cy)), fabs(-sx - cy))), 1); + *dstwidth = 2 * dstwidthhalf; + *dstheight = 2 * dstheighthalf; +} + +/*! +\brief Returns the size of the resulting target surface for a rotozoomSurfaceXY() call. + +\param width The source surface width. +\param height The source surface height. +\param angle The angle to rotate in degrees. +\param zoomx The horizontal scaling factor. +\param zoomy The vertical scaling factor. +\param dstwidth The calculated width of the rotozoomed destination surface. +\param dstheight The calculated height of the rotozoomed destination surface. +*/ +void rotozoomSurfaceSizeXY(int width, int height, double angle, double zoomx, double zoomy, int *dstwidth, int *dstheight) +{ + double dummy_sanglezoom, dummy_canglezoom; + + _rotozoomSurfaceSizeTrig(width, height, angle, zoomx, zoomy, dstwidth, dstheight, &dummy_sanglezoom, &dummy_canglezoom); +} + +/*! +\brief Returns the size of the resulting target surface for a rotozoomSurface() call. + +\param width The source surface width. +\param height The source surface height. +\param angle The angle to rotate in degrees. +\param zoom The scaling factor. +\param dstwidth The calculated width of the rotozoomed destination surface. +\param dstheight The calculated height of the rotozoomed destination surface. +*/ +void rotozoomSurfaceSize(int width, int height, double angle, double zoom, int *dstwidth, int *dstheight) +{ + double dummy_sanglezoom, dummy_canglezoom; + + _rotozoomSurfaceSizeTrig(width, height, angle, zoom, zoom, dstwidth, dstheight, &dummy_sanglezoom, &dummy_canglezoom); +} + +/*! +\brief Rotates and zooms a surface and optional anti-aliasing. + +Rotates and zoomes a 32bit or 8bit 'src' surface to newly created 'dst' surface. +'angle' is the rotation in degrees and 'zoom' a scaling factor. If 'smooth' is set +then the destination 32bit surface is anti-aliased. If the surface is not 8bit +or 32bit RGBA/ABGR it will be converted into a 32bit RGBA format on the fly. + +\param src The surface to rotozoom. +\param angle The angle to rotate in degrees. +\param zoom The scaling factor. +\param smooth Antialiasing flag; set to SMOOTHING_ON to enable. + +\return The new rotozoomed surface. +*/ +SDL_Surface *rotozoomSurface(SDL_Surface * src, double angle, double zoom, int smooth) +{ + return rotozoomSurfaceXY(src, angle, zoom, zoom, smooth); +} + +/*! +\brief Rotates and zooms a surface with different horizontal and vertival scaling factors and optional anti-aliasing. + +Rotates and zooms a 32bit or 8bit 'src' surface to newly created 'dst' surface. +'angle' is the rotation in degrees, 'zoomx and 'zoomy' scaling factors. If 'smooth' is set +then the destination 32bit surface is anti-aliased. If the surface is not 8bit +or 32bit RGBA/ABGR it will be converted into a 32bit RGBA format on the fly. + +\param src The surface to rotozoom. +\param angle The angle to rotate in degrees. +\param zoomx The horizontal scaling factor. +\param zoomy The vertical scaling factor. +\param smooth Antialiasing flag; set to SMOOTHING_ON to enable. + +\return The new rotozoomed surface. +*/ +SDL_Surface *rotozoomSurfaceXY(SDL_Surface * src, double angle, double zoomx, double zoomy, int smooth) +{ + SDL_Surface *rz_src; + SDL_Surface *rz_dst; + double zoominv; + double sanglezoom, canglezoom, sanglezoominv, canglezoominv; + int dstwidthhalf, dstwidth, dstheighthalf, dstheight; + int is32bit; + int i, src_converted; + int flipx,flipy; + + /* + * Sanity check + */ + if (src == NULL) { + return (NULL); + } + + /* + * Determine if source surface is 32bit or 8bit + */ + is32bit = (src->format->BitsPerPixel == 32); + if ((is32bit) || (src->format->BitsPerPixel == 8)) { + /* + * Use source surface 'as is' + */ + rz_src = src; + src_converted = 0; + } else { + /* + * New source surface is 32bit with a defined RGBA ordering + */ + rz_src = + SDL_CreateRGBSurface(SDL_SWSURFACE, src->w, src->h, 32, +#if SDL_BYTEORDER == SDL_LIL_ENDIAN + 0x000000ff, 0x0000ff00, 0x00ff0000, 0xff000000 +#else + 0xff000000, 0x00ff0000, 0x0000ff00, 0x000000ff +#endif + ); + + SDL_BlitSurface(src, NULL, rz_src, NULL); + + src_converted = 1; + is32bit = 1; + } + + /* + * Sanity check zoom factor + */ + flipx = (zoomx<0.0); + if (flipx) zoomx=-zoomx; + flipy = (zoomy<0.0); + if (flipy) zoomy=-zoomy; + if (zoomx < VALUE_LIMIT) zoomx = VALUE_LIMIT; + if (zoomy < VALUE_LIMIT) zoomy = VALUE_LIMIT; + zoominv = 65536.0 / (zoomx * zoomx); + + /* + * Check if we have a rotozoom or just a zoom + */ + if (fabs(angle) > VALUE_LIMIT) { + + /* + * Angle!=0: full rotozoom + */ + /* + * ----------------------- + */ + + /* Determine target size */ + _rotozoomSurfaceSizeTrig(rz_src->w, rz_src->h, angle, zoomx, zoomy, &dstwidth, &dstheight, &canglezoom, &sanglezoom); + + /* + * Calculate target factors from sin/cos and zoom + */ + sanglezoominv = sanglezoom; + canglezoominv = canglezoom; + sanglezoominv *= zoominv; + canglezoominv *= zoominv; + + /* Calculate half size */ + dstwidthhalf = dstwidth / 2; + dstheighthalf = dstheight / 2; + + /* + * Alloc space to completely contain the rotated surface + */ + rz_dst = NULL; + if (is32bit) { + /* + * Target surface is 32bit with source RGBA/ABGR ordering + */ + rz_dst = + SDL_CreateRGBSurface(SDL_SWSURFACE, dstwidth, dstheight + GUARD_ROWS, 32, + rz_src->format->Rmask, rz_src->format->Gmask, + rz_src->format->Bmask, rz_src->format->Amask); + } else { + /* + * Target surface is 8bit + */ + rz_dst = SDL_CreateRGBSurface(SDL_SWSURFACE, dstwidth, dstheight + GUARD_ROWS, 8, 0, 0, 0, 0); + } + + /* Check target */ + if (rz_dst == NULL) + return NULL; + + /* Adjust for guard rows */ + rz_dst->h = dstheight; + + /* + * Lock source surface + */ + if (SDL_MUSTLOCK(rz_src)) { + SDL_LockSurface(rz_src); + } + + /* + * Check which kind of surface we have + */ + if (is32bit) { + /* + * Call the 32bit transformation routine to do the rotation (using alpha) + */ + _transformSurfaceRGBA(rz_src, rz_dst, dstwidthhalf, dstheighthalf, + (int) (sanglezoominv), (int) (canglezoominv), + flipx, flipy, + smooth); + } else { + /* + * Copy palette and colorkey info + */ + for (i = 0; i < rz_src->format->palette->ncolors; i++) { + rz_dst->format->palette->colors[i] = rz_src->format->palette->colors[i]; + } + rz_dst->format->palette->ncolors = rz_src->format->palette->ncolors; + /* + * Call the 8bit transformation routine to do the rotation + */ + transformSurfaceY(rz_src, rz_dst, dstwidthhalf, dstheighthalf, + (int) (sanglezoominv), (int) (canglezoominv), + flipx, flipy); + } + /* + * Unlock source surface + */ + if (SDL_MUSTLOCK(rz_src)) { + SDL_UnlockSurface(rz_src); + } + + } else { + + /* + * Angle=0: Just a zoom + */ + /* + * -------------------- + */ + + /* + * Calculate target size + */ + zoomSurfaceSize(rz_src->w, rz_src->h, zoomx, zoomy, &dstwidth, &dstheight); + + /* + * Alloc space to completely contain the zoomed surface + */ + rz_dst = NULL; + if (is32bit) { + /* + * Target surface is 32bit with source RGBA/ABGR ordering + */ + rz_dst = + SDL_CreateRGBSurface(SDL_SWSURFACE, dstwidth, dstheight + GUARD_ROWS, 32, + rz_src->format->Rmask, rz_src->format->Gmask, + rz_src->format->Bmask, rz_src->format->Amask); + } else { + /* + * Target surface is 8bit + */ + rz_dst = SDL_CreateRGBSurface(SDL_SWSURFACE, dstwidth, dstheight + GUARD_ROWS, 8, 0, 0, 0, 0); + } + + /* Check target */ + if (rz_dst == NULL) + return NULL; + + /* Adjust for guard rows */ + rz_dst->h = dstheight; + + /* + * Lock source surface + */ + if (SDL_MUSTLOCK(rz_src)) { + SDL_LockSurface(rz_src); + } + + /* + * Check which kind of surface we have + */ + if (is32bit) { + /* + * Call the 32bit transformation routine to do the zooming (using alpha) + */ + _zoomSurfaceRGBA(rz_src, rz_dst, flipx, flipy, smooth); + + } else { + /* + * Copy palette and colorkey info + */ + for (i = 0; i < rz_src->format->palette->ncolors; i++) { + rz_dst->format->palette->colors[i] = rz_src->format->palette->colors[i]; + } + rz_dst->format->palette->ncolors = rz_src->format->palette->ncolors; + + /* + * Call the 8bit transformation routine to do the zooming + */ + _zoomSurfaceY(rz_src, rz_dst, flipx, flipy); + } + + /* + * Unlock source surface + */ + if (SDL_MUSTLOCK(rz_src)) { + SDL_UnlockSurface(rz_src); + } + } + + /* + * Cleanup temp surface + */ + if (src_converted) { + SDL_FreeSurface(rz_src); + } + + /* + * Return destination surface + */ + return (rz_dst); +} + +/*! +\brief Calculates the size of the target surface for a zoomSurface() call. + +The minimum size of the target surface is 1. The input factors can be positive or negative. + +\param width The width of the source surface to zoom. +\param height The height of the source surface to zoom. +\param zoomx The horizontal zoom factor. +\param zoomy The vertical zoom factor. +\param dstwidth Pointer to an integer to store the calculated width of the zoomed target surface. +\param dstheight Pointer to an integer to store the calculated height of the zoomed target surface. +*/ +void zoomSurfaceSize(int width, int height, double zoomx, double zoomy, int *dstwidth, int *dstheight) +{ + /* + * Make zoom factors positive + */ + int flipx, flipy; + flipx = (zoomx<0.0); + if (flipx) zoomx = -zoomx; + flipy = (zoomy<0.0); + if (flipy) zoomy = -zoomy; + + /* + * Sanity check zoom factors + */ + if (zoomx < VALUE_LIMIT) { + zoomx = VALUE_LIMIT; + } + if (zoomy < VALUE_LIMIT) { + zoomy = VALUE_LIMIT; + } + + /* + * Calculate target size + */ + *dstwidth = (int) floor(((double) width * zoomx) + 0.5); + *dstheight = (int) floor(((double) height * zoomy) + 0.5); + if (*dstwidth < 1) { + *dstwidth = 1; + } + if (*dstheight < 1) { + *dstheight = 1; + } +} + +/*! +\brief Zoom a surface by independent horizontal and vertical factors with optional smoothing. + +Zooms a 32bit or 8bit 'src' surface to newly created 'dst' surface. +'zoomx' and 'zoomy' are scaling factors for width and height. If 'smooth' is on +then the destination 32bit surface is anti-aliased. If the surface is not 8bit +or 32bit RGBA/ABGR it will be converted into a 32bit RGBA format on the fly. +If zoom factors are negative, the image is flipped on the axes. + +\param src The surface to zoom. +\param zoomx The horizontal zoom factor. +\param zoomy The vertical zoom factor. +\param smooth Antialiasing flag; set to SMOOTHING_ON to enable. + +\return The new, zoomed surface. +*/ +SDL_Surface *zoomSurface(SDL_Surface * src, double zoomx, double zoomy, int smooth) +{ + SDL_Surface *rz_src; + SDL_Surface *rz_dst; + int dstwidth, dstheight; + int is32bit; + int i, src_converted; + int flipx, flipy; + + /* + * Sanity check + */ + if (src == NULL) + return (NULL); + + /* + * Determine if source surface is 32bit or 8bit + */ + is32bit = (src->format->BitsPerPixel == 32); + if ((is32bit) || (src->format->BitsPerPixel == 8)) { + /* + * Use source surface 'as is' + */ + rz_src = src; + src_converted = 0; + } else { + /* + * New source surface is 32bit with a defined RGBA ordering + */ + rz_src = + SDL_CreateRGBSurface(SDL_SWSURFACE, src->w, src->h, 32, +#if SDL_BYTEORDER == SDL_LIL_ENDIAN + 0x000000ff, 0x0000ff00, 0x00ff0000, 0xff000000 +#else + 0xff000000, 0x00ff0000, 0x0000ff00, 0x000000ff +#endif + ); + if (rz_src == NULL) { + return NULL; + } + SDL_BlitSurface(src, NULL, rz_src, NULL); + src_converted = 1; + is32bit = 1; + } + + flipx = (zoomx<0.0); + if (flipx) zoomx = -zoomx; + flipy = (zoomy<0.0); + if (flipy) zoomy = -zoomy; + + /* Get size if target */ + zoomSurfaceSize(rz_src->w, rz_src->h, zoomx, zoomy, &dstwidth, &dstheight); + + /* + * Alloc space to completely contain the zoomed surface + */ + rz_dst = NULL; + if (is32bit) { + /* + * Target surface is 32bit with source RGBA/ABGR ordering + */ + rz_dst = + SDL_CreateRGBSurface(SDL_SWSURFACE, dstwidth, dstheight + GUARD_ROWS, 32, + rz_src->format->Rmask, rz_src->format->Gmask, + rz_src->format->Bmask, rz_src->format->Amask); + } else { + /* + * Target surface is 8bit + */ + rz_dst = SDL_CreateRGBSurface(SDL_SWSURFACE, dstwidth, dstheight + GUARD_ROWS, 8, 0, 0, 0, 0); + } + + /* Check target */ + if (rz_dst == NULL) { + /* + * Cleanup temp surface + */ + if (src_converted) { + SDL_FreeSurface(rz_src); + } + return NULL; + } + + /* Adjust for guard rows */ + rz_dst->h = dstheight; + + /* + * Lock source surface + */ + if (SDL_MUSTLOCK(rz_src)) { + SDL_LockSurface(rz_src); + } + + /* + * Check which kind of surface we have + */ + if (is32bit) { + /* + * Call the 32bit transformation routine to do the zooming (using alpha) + */ + _zoomSurfaceRGBA(rz_src, rz_dst, flipx, flipy, smooth); + } else { + /* + * Copy palette and colorkey info + */ + for (i = 0; i < rz_src->format->palette->ncolors; i++) { + rz_dst->format->palette->colors[i] = rz_src->format->palette->colors[i]; + } + rz_dst->format->palette->ncolors = rz_src->format->palette->ncolors; + /* + * Call the 8bit transformation routine to do the zooming + */ + _zoomSurfaceY(rz_src, rz_dst, flipx, flipy); + } + /* + * Unlock source surface + */ + if (SDL_MUSTLOCK(rz_src)) { + SDL_UnlockSurface(rz_src); + } + + /* + * Cleanup temp surface + */ + if (src_converted) { + SDL_FreeSurface(rz_src); + } + + /* + * Return destination surface + */ + return (rz_dst); +} + +/*! +\brief Shrink a surface by an integer ratio using averaging. + +Shrinks a 32bit or 8bit 'src' surface to a newly created 'dst' surface. +'factorx' and 'factory' are the shrinking ratios (i.e. 2=1/2 the size, +3=1/3 the size, etc.) The destination surface is antialiased by averaging +the source box RGBA or Y information. If the surface is not 8bit +or 32bit RGBA/ABGR it will be converted into a 32bit RGBA format on the fly. +The input surface is not modified. The output surface is newly allocated. + +\param src The surface to shrink. +\param factorx The horizontal shrinking ratio. +\param factory The vertical shrinking ratio. + +\return The new, shrunken surface. +*/ +/*@null@*/ +SDL_Surface *shrinkSurface(SDL_Surface *src, int factorx, int factory) +{ + int result; + SDL_Surface *rz_src; + SDL_Surface *rz_dst = NULL; + int dstwidth, dstheight; + int is32bit; + int i, src_converted; + int haveError = 0; + + /* + * Sanity check + */ + if (src == NULL) { + return (NULL); + } + + /* + * Determine if source surface is 32bit or 8bit + */ + is32bit = (src->format->BitsPerPixel == 32); + if ((is32bit) || (src->format->BitsPerPixel == 8)) { + /* + * Use source surface 'as is' + */ + rz_src = src; + src_converted = 0; + } else { + /* + * New source surface is 32bit with a defined RGBA ordering + */ + rz_src = SDL_CreateRGBSurface(SDL_SWSURFACE, src->w, src->h, 32, +#if SDL_BYTEORDER == SDL_LIL_ENDIAN + 0x000000ff, 0x0000ff00, 0x00ff0000, 0xff000000 +#else + 0xff000000, 0x00ff0000, 0x0000ff00, 0x000000ff +#endif + ); + if (rz_src==NULL) { + haveError = 1; + goto exitShrinkSurface; + } + + SDL_BlitSurface(src, NULL, rz_src, NULL); + src_converted = 1; + is32bit = 1; + } + + /* + * Lock the surface + */ + if (SDL_MUSTLOCK(rz_src)) { + if (SDL_LockSurface(rz_src) < 0) { + haveError = 1; + goto exitShrinkSurface; + } + } + + /* Get size for target */ + dstwidth=rz_src->w/factorx; + while (dstwidth*factorx>rz_src->w) { dstwidth--; } + dstheight=rz_src->h/factory; + while (dstheight*factory>rz_src->h) { dstheight--; } + + /* + * Alloc space to completely contain the shrunken surface + * (with added guard rows) + */ + if (is32bit==1) { + /* + * Target surface is 32bit with source RGBA/ABGR ordering + */ + rz_dst = + SDL_CreateRGBSurface(SDL_SWSURFACE, dstwidth, dstheight + GUARD_ROWS, 32, + rz_src->format->Rmask, rz_src->format->Gmask, + rz_src->format->Bmask, rz_src->format->Amask); + } else { + /* + * Target surface is 8bit + */ + rz_dst = SDL_CreateRGBSurface(SDL_SWSURFACE, dstwidth, dstheight + GUARD_ROWS, 8, 0, 0, 0, 0); + } + + /* Check target */ + if (rz_dst == NULL) { + haveError = 1; + goto exitShrinkSurface; + } + + /* Adjust for guard rows */ + rz_dst->h = dstheight; + + /* + * Check which kind of surface we have + */ + if (is32bit==1) { + /* + * Call the 32bit transformation routine to do the shrinking (using alpha) + */ + result = _shrinkSurfaceRGBA(rz_src, rz_dst, factorx, factory); + if ((result!=0) || (rz_dst==NULL)) { + haveError = 1; + goto exitShrinkSurface; + } + } else { + /* + * Copy palette and colorkey info + */ + for (i = 0; i < rz_src->format->palette->ncolors; i++) { + rz_dst->format->palette->colors[i] = rz_src->format->palette->colors[i]; + } + rz_dst->format->palette->ncolors = rz_src->format->palette->ncolors; + /* + * Call the 8bit transformation routine to do the shrinking + */ + result = _shrinkSurfaceY(rz_src, rz_dst, factorx, factory); + if (result!=0) { + haveError = 1; + goto exitShrinkSurface; + } + } + +exitShrinkSurface: + if (rz_src!=NULL) { + /* + * Unlock source surface + */ + if (SDL_MUSTLOCK(rz_src)) { + SDL_UnlockSurface(rz_src); + } + + /* + * Cleanup temp surface + */ + if (src_converted==1) { + SDL_FreeSurface(rz_src); + } + } + + /* Check error state; maybe need to cleanup destination */ + if (haveError==1) { + if (rz_dst!=NULL) { + SDL_FreeSurface(rz_dst); + } + rz_dst=NULL; + } + + /* + * Return destination surface + */ + return (rz_dst); +} diff --git a/src/sdl2gfx/SDL_rotozoom.h b/src/sdl2gfx/SDL_rotozoom.h new file mode 100644 index 000000000..f4c236ba4 --- /dev/null +++ b/src/sdl2gfx/SDL_rotozoom.h @@ -0,0 +1,123 @@ +/* + +SDL2_rotozoom.c: rotozoomer, zoomer and shrinker for 32bit or 8bit surfaces + +Copyright (C) 2001-2012 Andreas Schiffler + +This software is provided 'as-is', without any express or implied +warranty. In no event will the authors be held liable for any damages +arising from the use of this software. + +Permission is granted to anyone to use this software for any purpose, +including commercial applications, and to alter it and redistribute it +freely, subject to the following restrictions: + +1. The origin of this software must not be misrepresented; you must not +claim that you wrote the original software. If you use this software +in a product, an acknowledgment in the product documentation would be +appreciated but is not required. + +2. Altered source versions must be plainly marked as such, and must not be +misrepresented as being the original software. + +3. This notice may not be removed or altered from any source +distribution. + +Andreas Schiffler -- aschiffler at ferzkopp dot net + +*/ + +#ifndef _SDL2_rotozoom_h +#define _SDL2_rotozoom_h + +#include <math.h> + +/* Set up for C function definitions, even when using C++ */ +#ifdef __cplusplus +extern "C" { +#endif + +#ifndef M_PI +#define M_PI 3.1415926535897932384626433832795 +#endif + +#include "SDL.h" + + /* ---- Defines */ + + /*! + \brief Disable anti-aliasing (no smoothing). + */ +#define SMOOTHING_OFF 0 + + /*! + \brief Enable anti-aliasing (smoothing). + */ +#define SMOOTHING_ON 1 + + /* ---- Function Prototypes */ + +#ifdef _MSC_VER +# if defined(DLL_EXPORT) && !defined(LIBSDL2_GFX_DLL_IMPORT) +# define SDL2_ROTOZOOM_SCOPE __declspec(dllexport) +# else +# ifdef LIBSDL2_GFX_DLL_IMPORT +# define SDL2_ROTOZOOM_SCOPE __declspec(dllimport) +# endif +# endif +#endif +#ifndef SDL2_ROTOZOOM_SCOPE +# define SDL2_ROTOZOOM_SCOPE extern +#endif + + /* + + Rotozoom functions + + */ + + SDL2_ROTOZOOM_SCOPE SDL_Surface *rotozoomSurface(SDL_Surface * src, double angle, double zoom, int smooth); + + SDL2_ROTOZOOM_SCOPE SDL_Surface *rotozoomSurfaceXY + (SDL_Surface * src, double angle, double zoomx, double zoomy, int smooth); + + + SDL2_ROTOZOOM_SCOPE void rotozoomSurfaceSize(int width, int height, double angle, double zoom, int *dstwidth, + int *dstheight); + + SDL2_ROTOZOOM_SCOPE void rotozoomSurfaceSizeXY + (int width, int height, double angle, double zoomx, double zoomy, + int *dstwidth, int *dstheight); + + /* + + Zooming functions + + */ + + SDL2_ROTOZOOM_SCOPE SDL_Surface *zoomSurface(SDL_Surface * src, double zoomx, double zoomy, int smooth); + + SDL2_ROTOZOOM_SCOPE void zoomSurfaceSize(int width, int height, double zoomx, double zoomy, int *dstwidth, int *dstheight); + + /* + + Shrinking functions + + */ + + SDL2_ROTOZOOM_SCOPE SDL_Surface *shrinkSurface(SDL_Surface * src, int factorx, int factory); + + /* + + Specialized rotation functions + + */ + + SDL2_ROTOZOOM_SCOPE SDL_Surface* rotateSurface90Degrees(SDL_Surface* src, int numClockwiseTurns); + + /* Ends C function definitions when using C++ */ +#ifdef __cplusplus +} +#endif + +#endif /* _SDL2_rotozoom_h */ |