1
0
Fork 0
mirror of https://github.com/notwa/mm synced 2024-11-05 06:39:02 -08:00
mm/z64yaz0.new.c

338 lines
8.9 KiB
C
Raw Normal View History

2019-06-24 04:07:24 -07:00
#include <stdio.h>
#include <stdlib.h>
#include <string.h>
2019-07-09 06:09:12 -07:00
#include <errno.h>
2019-06-24 04:07:24 -07:00
//version 1.0 (20050707) by shevious
//Thanks to thakis for yaz0dec 1.0.
#define lament(...) fprintf(stderr, __VA_ARGS__)
#define error_when(cond, ...) do { \
if ((cond) || errno) { \
lament(__VA_ARGS__); \
lament(": %s\n", strerror(errno)); \
goto error; \
} \
} while (0)
typedef unsigned char u8;
enum {
max_runlen = 0xFF + 0x12
};
// simple and straight encoding scheme for Yaz0
static long
simpleEnc(u8 *src, int size, int pos, long *pMatchPos)
{
int startPos = pos - 0x1000;
long numBytes = 1;
long matchPos = 0;
2019-07-09 06:09:12 -07:00
int i;
2019-06-24 04:07:24 -07:00
int end = size - pos;
// maximum runlength for 3 byte encoding
if (end > max_runlen)
end = max_runlen;
if (startPos < 0)
startPos = 0;
2019-07-09 06:09:12 -07:00
for (i = startPos; i < pos; i++) {
2019-06-24 04:07:24 -07:00
int j;
for (j = 0; j < end; j++) {
if (src[i + j] != src[j + pos])
break;
}
if (j > numBytes) {
numBytes = j;
matchPos = i;
}
}
*pMatchPos = matchPos;
if (numBytes == 2)
numBytes = 1;
return numBytes;
}
// a lookahead encoding scheme for ngc Yaz0
static long
nintendoEnc(u8 *src, int size, int pos, long *pMatchPos)
{
long numBytes = 1;
static long numBytes1;
static long matchPos;
static int prevFlag = 0;
// if prevFlag is set, it means that the previous position
// was determined by look-ahead try.
// so just use it. this is not the best optimization,
// but nintendo's choice for speed.
if (prevFlag == 1) {
*pMatchPos = matchPos;
prevFlag = 0;
return numBytes1;
}
prevFlag = 0;
numBytes = simpleEnc(src, size, pos, &matchPos);
*pMatchPos = matchPos;
// if this position is RLE encoded, then compare to copying 1 byte and next position(pos+1) encoding
if (numBytes >= 3) {
numBytes1 = simpleEnc(src, size, pos + 1, &matchPos);
// if the next position encoding is +2 longer than current position, choose it.
// this does not guarantee the best optimization, but fairly good optimization with speed.
if (numBytes1 >= numBytes + 2) {
numBytes = 1;
prevFlag = 1;
}
}
return numBytes;
}
static int
encodeYaz0(u8 *src, u8 *dst, int srcSize)
{
int srcPos = 0;
int dstPos = 0;
int bufPos = 0;
u8 buf[24]; // 8 codes * 3 bytes maximum
long validBitCount = 0; // number of valid bits left in "code" byte
u8 currCodeByte = 0; // a bitfield, set bits meaning copy, unset meaning RLE
while (srcPos < srcSize) {
long numBytes;
long matchPos;
numBytes = nintendoEnc(src, srcSize, srcPos, &matchPos);
if (numBytes < 3) {
// straight copy
buf[bufPos] = src[srcPos];
bufPos++;
srcPos++;
//set flag for straight copy
currCodeByte |= (0x80 >> validBitCount);
} else {
//RLE part
long dist = srcPos - matchPos - 1;
u8 byte1, byte2, byte3;
if (numBytes >= 0x12) { // 3 byte encoding
byte1 = 0 | (dist >> 8);
byte2 = dist & 0xFF;
buf[bufPos++] = byte1;
buf[bufPos++] = byte2;
// maximum runlength for 3 byte encoding
if (numBytes > max_runlen)
numBytes = max_runlen;
byte3 = numBytes - 0x12;
buf[bufPos++] = byte3;
} else { // 2 byte encoding
byte1 = ((numBytes - 2) << 4) | (dist >> 8);
byte2 = dist & 0xFF;
buf[bufPos++] = byte1;
buf[bufPos++] = byte2;
}
srcPos += numBytes;
}
validBitCount++;
// write eight codes
if (validBitCount == 8) {
2019-07-09 06:09:12 -07:00
int j;
2019-06-24 04:07:24 -07:00
dst[dstPos++] = currCodeByte;
2019-07-09 06:09:12 -07:00
for (j = 0; j < bufPos; j++)
2019-06-24 04:07:24 -07:00
dst[dstPos++] = buf[j];
currCodeByte = 0;
validBitCount = 0;
bufPos = 0;
}
}
if (validBitCount > 0) {
2019-07-09 06:09:12 -07:00
int j;
2019-06-24 04:07:24 -07:00
dst[dstPos++] = currCodeByte;
2019-07-09 06:09:12 -07:00
for (j = 0; j < bufPos; j++)
2019-06-24 04:07:24 -07:00
dst[dstPos++] = buf[j];
currCodeByte = 0;
validBitCount = 0;
bufPos = 0;
}
return dstPos;
}
static void
decompress(u8 *src, u8 *dst, int uncompressedSize)
{
int srcPlace = 0, dstPlace = 0; // current read/write positions
long validBitCount = 0; // number of valid bits left in "code" byte
u8 currCodeByte = 0;
while (dstPlace < uncompressedSize) {
// read new "code" byte if the current one is used up
if (validBitCount == 0) {
currCodeByte = src[srcPlace++];
validBitCount = 8;
}
if ((currCodeByte & 0x80) != 0) {
// straight copy
dst[dstPlace++] = src[srcPlace++];
} else {
// RLE part
u8 byte1 = src[srcPlace++];
u8 byte2 = src[srcPlace++];
long dist = ((byte1 & 0xF) << 8) | byte2;
long copySource = dstPlace - (dist + 1);
2019-07-09 06:09:12 -07:00
int i;
2019-06-24 04:07:24 -07:00
long numBytes = byte1 >> 4;
if (numBytes == 0) {
numBytes = src[srcPlace++] + 0x12;
} else {
numBytes += 2;
}
// copy run
2019-07-09 06:09:12 -07:00
for(i = 0; i < numBytes; ++i) {
2019-06-24 04:07:24 -07:00
dst[dstPlace++] = dst[copySource++];
}
}
// use next bit from "code" byte
currCodeByte <<= 1;
validBitCount--;
}
}
int
process(const char *fpi, const char *fpo)
{
u8 *bufi = NULL;
u8 *bufo = NULL;
FILE *fi = NULL;
FILE *fo = NULL;
long isize;
long csize;
long outsize;
long i;
fi = fopen(fpi, "rb");
error_when(fi == NULL, "Error opening file for reading: %s", fpi);
error_when(fseek(fi, 0, SEEK_END) != 0, "Error seeking in file: %s", fpi);
isize = ftell(fi);
error_when(isize < 0, "Error telling in file: %s", fpi);
error_when(fseek(fi, 0, SEEK_SET) != 0, "Error seeking in file: %s", fpi);
if (isize > 0) {
bufi = malloc(isize);
error_when(bufi == NULL, "Error allocating %li bytes", isize);
error_when(fread(bufi, 1, isize, fi) != (size_t)isize,
"Error reading %li bytes from file: %s", isize, fpi);
}
error_when(fclose(fi) != 0, "Error closing file: %s", fpi);
if (isize < 5) {
// FIXME: encodeYaz0 segfaults in this case.
lament("Error: input file must be at least 5 bytes.\n");
goto error;
}
if (isize > 0x10
&& bufi[0] == 'Y'
&& bufi[1] == 'a'
&& bufi[2] == 'z'
&& bufi[3] == '0') {
outsize = (bufi[4] << 24)
| (bufi[5] << 16)
| (bufi[6] << 8)
| bufi[7];
bufo = malloc(outsize);
error_when(bufo == NULL, "Error allocating %li bytes", outsize);
decompress(bufi + 16, bufo, outsize);
} else {
// we don't know how big the "compressed" file could get,
// so over-allocate!
// modern systems have more RAM than the largest Yaz0 file, so...
csize = 0x10 + isize * 2;
bufo = malloc(csize);
error_when(bufo == NULL, "Error allocating %li bytes", csize);
// write 4 bytes yaz0 header
bufo[0] = 'Y';
bufo[1] = 'a';
bufo[2] = 'z';
bufo[3] = '0';
// write 4 bytes uncompressed size
bufo[4] = (isize >> 24) & 0xFF;
bufo[5] = (isize >> 16) & 0xFF;
bufo[6] = (isize >> 8) & 0xFF;
bufo[7] = (isize >> 0) & 0xFF;
// write 8 bytes unused dummy
bufo[8] = 0;
bufo[9] = 0;
bufo[10] = 0;
bufo[11] = 0;
bufo[12] = 0;
bufo[13] = 0;
bufo[14] = 0;
bufo[15] = 0;
csize = encodeYaz0(bufi, bufo + 16, isize) + 16;
// pad compressed file to be a multiple of 16 bytes.
outsize = (csize + 15) & ~0xF;
for (i = csize; i < outsize; i++) bufo[i] = 0;
}
fo = fopen(fpo, "wb");
error_when(fo == NULL, "Error opening file for writing: %s", fpo);
error_when(fwrite(bufo, 1, outsize, fo) != (size_t)outsize,
"Error writing %li bytes to file: %s", outsize, fpo);
error_when(fclose(fo) != 0, "Error closing file: %s", fpo);
free(bufi);
free(bufo);
return 0;
error:
if (bufi != NULL) free(bufi);
if (bufo != NULL) free(bufo);
if (fi != NULL) fclose(fi);
if (fo != NULL) fclose(fo);
return 1;
}
int
main(int argc, char *argv[])
{
if (argc <= 0 || argv == NULL || argv[0] == NULL) {
lament("You've met with a terrible fate.\n");
return 1;
}
if (argc != 3) {
lament("usage: %s {input file} {output file}\n", argv[0]);
return 1;
}
return process(argv[1], argv[2]);
}