summaryrefslogtreecommitdiffstats
path: root/generic
diff options
context:
space:
mode:
authorericm <ericm>2000-01-27 16:58:42 (GMT)
committerericm <ericm>2000-01-27 16:58:42 (GMT)
commit5f3b08cd9452e26e67ee02c9ce0ce128afd8bb31 (patch)
tree85a5e3202a3a40eeb9bc170df8e8987b79a29a16 /generic
parent366fd9beda032b13d073e0f6331831adcbb02559 (diff)
downloadtk-5f3b08cd9452e26e67ee02c9ce0ce128afd8bb31.zip
tk-5f3b08cd9452e26e67ee02c9ce0ce128afd8bb31.tar.gz
tk-5f3b08cd9452e26e67ee02c9ce0ce128afd8bb31.tar.bz2
* generic/tkImgGIF.c: Additional code cleanup (now we only have
one decoder! neat!)
Diffstat (limited to 'generic')
-rw-r--r--generic/tkImgGIF.c207
1 files changed, 44 insertions, 163 deletions
diff --git a/generic/tkImgGIF.c b/generic/tkImgGIF.c
index 52ba2d7..87a90c7 100644
--- a/generic/tkImgGIF.c
+++ b/generic/tkImgGIF.c
@@ -29,7 +29,7 @@
* | provided "as is" without express or implied warranty. |
* +-------------------------------------------------------------------+
*
- * RCS: @(#) $Id: tkImgGIF.c,v 1.9 2000/01/26 21:11:00 ericm Exp $
+ * RCS: @(#) $Id: tkImgGIF.c,v 1.10 2000/01/27 16:58:42 ericm Exp $
*/
/*
@@ -131,8 +131,6 @@ static int GetCode _ANSI_ARGS_((Tcl_Channel chan, int code_size,
int flag));
static int GetDataBlock _ANSI_ARGS_((Tcl_Channel chan,
unsigned char *buf));
-static int LWZReadByte _ANSI_ARGS_((Tcl_Channel chan, int flag,
- int input_code_size));
static int ReadColorMap _ANSI_ARGS_((Tcl_Channel chan, int number,
unsigned char buffer[MAXCOLORMAPSIZE][4]));
static int ReadGIFHeader _ANSI_ARGS_((Tcl_Channel chan,
@@ -229,6 +227,7 @@ FileReadGIF(interp, chan, fileName, format, imageHandle, destX, destY,
Tcl_Obj **objv;
Tk_PhotoImageBlock block;
unsigned char buf[100];
+ unsigned char *trashBuffer = NULL;
int bitPixel;
unsigned char colorMap[MAXCOLORMAPSIZE][4];
int transparent = -1;
@@ -361,37 +360,49 @@ FileReadGIF(interp, chan, fileName, format, imageHandle, destX, destY,
bitPixel = 1<<((buf[8]&0x07)+1);
if (index--) {
- int x,y;
- unsigned char c;
/* this is not the image we want to read: skip it. */
-
if (BitSet(buf[8], LOCALCOLORMAP)) {
- if (!ReadColorMap(chan, bitPixel, 0)) {
+ if (!ReadColorMap(chan, bitPixel, colorMap)) {
Tcl_AppendResult(interp,
"error reading color map", (char *) NULL);
goto error;
}
}
- /* read data */
- if (!ReadOK(chan,&c,1)) {
- goto error;
+ /* If we've not yet allocated a trash buffer, do so now */
+ if (trashBuffer == NULL) {
+ nBytes = fileWidth * fileHeight * 3;
+ trashBuffer =
+ (unsigned char *) ckalloc((unsigned int) nBytes);
}
- LWZReadByte(chan, 1, c);
-
- for (y=0; y<fileHeight; y++) {
- for (x=0; x<fileWidth; x++) {
- if (LWZReadByte(chan, 0, c) < 0) {
- Tcl_AppendResult(interp,
- "error reading image data", (char *) NULL);
- goto error;
- }
- }
+ /*
+ * Slurp! Process the data for this image and stuff it in a
+ * trash buffer.
+ *
+ * Yes, it might be more efficient here to *not* store the data
+ * (we're just going to throw it away later). However, I elected
+ * to implement it this way for good reasons. First, I wanted to
+ * avoid duplicating the (fairly complex) LWZ decoder in ReadImage.
+ * Fine, you say, why didn't you just modify it to allow the use of
+ * a NULL specifier for the output buffer? I tried that, but it
+ * negatively impacted the performance of what I think will be the
+ * common case: reading the first image in the file. Rather than
+ * marginally improve the speed of the less frequent case, I chose
+ * to maintain high performance for the common case.
+ */
+ if (ReadImage(interp, trashBuffer, chan, fileWidth,
+ fileHeight, colorMap, 0, 0, 0, 0, 0, -1) != TCL_OK) {
+ goto error;
}
continue;
}
+ /* If a trash buffer has been allocated, free it now */
+ if (trashBuffer != NULL) {
+ ckfree((char *)trashBuffer);
+ trashBuffer = NULL;
+ }
if (BitSet(buf[8], LOCALCOLORMAP)) {
if (!ReadColorMap(chan, bitPixel, colorMap)) {
Tcl_AppendResult(interp, "error reading color map",
@@ -730,18 +741,11 @@ GetDataBlock(chan, buf)
* which is (c) 2000 ImageMagick Studio.
*
* Some thoughts on our implementation:
- * We currently have implementations for two separate GIF decoders in
- * this source. One is here, in ReadImage; the other is in LWZReadByte.
- * This is an artifact of our original implementation. Ideally, we
- * should remove all vestiges of LWZReadByte, which is slow and poorly
- * documented. However, that would require some additional effort to
- * collapse the calling layers of this system: Instead of
- * FileReadGIF -> ReadImage, it would change to FileReadGIF.
+ * It sure would be nice if ReadImage didn't take 11 parameters! I think
+ * that if we were smarter, we could avoid doing that.
*
* Possible further optimizations: we could pull the GetCode function
- * directly into ReadImage, which would improve our speed. The reason
- * I have not yet done so is because GetCode is still used by LWZReadByte,
- * and I don't want to duplicate the code.
+ * directly into ReadImage, which would improve our speed.
*
* Results:
* Processes a GIF image and loads the pixel data into a memory array.
@@ -768,19 +772,18 @@ ReadImage(interp, imagePtr, chan, len, rows, cmap,
unsigned char initialCodeSize;
int v;
int xpos = 0, ypos = 0, pass = 0, i;
- char *pixelPtr;
+ register char *pixelPtr;
const static int interlaceStep[] = { 8, 8, 4, 2 };
const static int interlaceStart[] = { 0, 4, 2, 1 };
unsigned short prefix[(1 << MAX_LWZ_BITS)];
unsigned char append[(1 << MAX_LWZ_BITS)];
unsigned char stack[(1 << MAX_LWZ_BITS)*2];
- unsigned char *top;
+ register unsigned char *top;
int codeSize, clearCode, inCode, endCode, oldCode, maxCode,
code, firstCode;
-
/*
- * Initialize the decompression routines
+ * Initialize the decoder
*/
if (! ReadOK(chan, &initialCodeSize, 1)) {
Tcl_AppendResult(interp, "error reading GIF image: ",
@@ -920,6 +923,12 @@ ReadImage(interp, imagePtr, chan, len, rows, cmap,
if (v < 0) {
return TCL_OK;
}
+
+ /*
+ * If pixelPtr is null, we're skipping this image (presumably
+ * there are more in the file and we will be called to read
+ * one of them later)
+ */
*pixelPtr++ = cmap[v][CM_RED];
*pixelPtr++ = cmap[v][CM_GREEN];
*pixelPtr++ = cmap[v][CM_BLUE];
@@ -927,6 +936,7 @@ ReadImage(interp, imagePtr, chan, len, rows, cmap,
*pixelPtr++ = cmap[v][CM_ALPHA];
}
xpos++;
+
}
/* If interlacing, the next ypos is not just +1 */
@@ -947,135 +957,6 @@ ReadImage(interp, imagePtr, chan, len, rows, cmap,
return TCL_OK;
}
-static int
-LWZReadByte(chan, flag, input_code_size)
- Tcl_Channel chan;
- int flag;
- int input_code_size;
-{
- static int fresh = 0;
- int code, incode;
- static int code_size, set_code_size;
- static int max_code, max_code_size;
- static int firstcode, oldcode;
- static int clear_code, end_code;
- static int prefix[(1 << MAX_LWZ_BITS)];
- static int append[(1 << MAX_LWZ_BITS)];
- static int stack[(1<<(MAX_LWZ_BITS))*2], *sp;
- register int i;
-
- if (flag) {
- set_code_size = input_code_size;
- code_size = set_code_size+1;
- clear_code = 1 << set_code_size ;
- end_code = clear_code + 1;
- max_code_size = 2*clear_code;
- max_code = clear_code+2;
-
- /* Init the lwz unpacker */
- GetCode(chan, 0, 1);
-
- fresh = 1;
-
- /* Initialize prefix and append tables */
- memset((void *)prefix, 0, (1 << MAX_LWZ_BITS) * sizeof(int));
- memset((void *)append, 0, (1 << MAX_LWZ_BITS) * sizeof(int));
- for (i = 0; i < clear_code; ++i) {
- append[i] = i;
- }
- sp = stack;
-
- return 0;
- } else if (fresh) {
- /*
- * find the first "real" code and return that. Because of
- * how LWZ works, it will necessarily refer to a single char,
- * so we need not bother with the stack
- */
- fresh = 0;
- do {
- firstcode = GetCode(chan, code_size, 0);
- } while (firstcode == clear_code);
- oldcode = firstcode;
- return firstcode;
- }
-
- if (sp > stack) {
- return *--sp;
- }
-
- while ((code = GetCode(chan, code_size, 0)) >= 0) {
- if (code == clear_code) {
- memset((void *)prefix, 0, (1 << MAX_LWZ_BITS) * sizeof(int));
- memset((void *)append, 0, (1 << MAX_LWZ_BITS) * sizeof(int));
- for (i = 0; i < clear_code; ++i) {
- append[i] = i;
- }
-
- code_size = set_code_size+1;
- max_code_size = 2*clear_code;
- max_code = clear_code+2;
- sp = stack;
- firstcode = oldcode = GetCode(chan, code_size, 0);
- return firstcode;
-
- } else if (code == end_code) {
- int count;
- unsigned char buf[260];
-
- if (ZeroDataBlock) {
- return -2;
- }
-
- while ((count = GetDataBlock(chan, buf)) > 0)
- /* Empty body */;
-
- if (count != 0) {
- return -2;
- }
- }
-
- incode = code;
-
- if (code >= max_code) {
- *sp++ = firstcode;
- code = oldcode;
- }
-
- while (code >= clear_code) {
- *sp++ = append[code];
- if (code == prefix[code]) {
- return -2;
-
- /*
- * Used to be this instead, Steve Ball suggested
- * the change to just return.
- printf("circular table entry BIG ERROR\n");
- */
- }
- code = prefix[code];
- }
-
- *sp++ = firstcode = append[code];
-
- if ((code = max_code) <(1<<MAX_LWZ_BITS)) {
- prefix[code] = oldcode;
- append[code] = firstcode;
- ++max_code;
- if ((max_code>=max_code_size) && (max_code_size < (1<<MAX_LWZ_BITS))) {
- max_code_size *= 2;
- ++code_size;
- }
- }
-
- oldcode = incode;
-
- if (sp > stack)
- return *--sp;
- }
- return code;
-}
-
/*
*----------------------------------------------------------------------