int
lz_fx_compress(void *src, uint64_t srclen, void *dst, uint64_t *dstlen,
int level, uchar_t chdr, int btype, void *data)
{
int rv;
struct lzfx_params *lzdat = (struct lzfx_params *)data;
unsigned int _srclen = srclen;
unsigned int _dstlen = *dstlen;
/*
* Ignore compressed data in fast modes.
*/
if (level < 7 && PC_TYPE(btype) == TYPE_COMPRESSED)
return (-1);
rv = lzfx_compress(src, _srclen, dst, &_dstlen, lzdat->htab_bits);
if (rv != 0) {
if (rv != LZFX_ESIZE)
lz_fx_err(rv);
return (-1);
}
*dstlen = _dstlen;
return (0);
}
int
ppmd_compress(void *src, uint64_t srclen, void *dst,
uint64_t *dstlen, int level, uchar_t chdr, int btype, void *data)
{
CPpmd8 *_ppmd = (CPpmd8 *)data;
uchar_t *_src = (uchar_t *)src;
if (PC_TYPE(btype) == TYPE_COMPRESSED)
return (-1);
Ppmd8_RangeEnc_Init(_ppmd);
Ppmd8_Init(_ppmd, _ppmd->Order, PPMD8_RESTORE_METHOD_RESTART);
_ppmd->buf = (Byte *)dst;
_ppmd->bufLen = *dstlen;
_ppmd->bufUsed = 0;
_ppmd->overflow = 0;
Ppmd8_EncodeBuffer(_ppmd, _src, srclen);
Ppmd8_EncodeSymbol(_ppmd, -1);
Ppmd8_RangeEnc_FlushData(_ppmd);
if (_ppmd->overflow) return (-1);
*dstlen = _ppmd->bufUsed;
return (0);
}
int
zlib_compress(void *src, uint64_t srclen, void *dst, uint64_t *dstlen,
int level, uchar_t chdr, int btype, void *data)
{
int ret, ending;
unsigned int slen, dlen;
uint64_t _srclen = srclen;
uint64_t _dstlen = *dstlen;
uchar_t *dst1 = (uchar_t *)dst;
uchar_t *src1 = (uchar_t *)src;
z_stream *zs = (z_stream *)data;
/*
* If the data is known to be compressed then certain types less compressed data
* can be attempted to be compressed again for a possible gain. For others it is
* a waste of time.
*/
if (PC_TYPE(btype) == TYPE_COMPRESSED && level < 7) {
int subtype = PC_SUBTYPE(btype);
if (subtype != TYPE_COMPRESSED_LZW &&
subtype != TYPE_COMPRESSED_LZ && subtype != TYPE_COMPRESSED_LZO) {
return (-1);
}
}
ending = 0;
while (_srclen > 0) {
if (_srclen > SINGLE_CALL_MAX) {
slen = SINGLE_CALL_MAX;
} else {
slen = _srclen;
ending = 1;
}
if (_dstlen > SINGLE_CALL_MAX) {
dlen = SINGLE_CALL_MAX;
} else {
dlen = _dstlen;
}
zs->next_in = src1;
zs->avail_in = slen;
zs->next_out = dst1;
zs->avail_out = dlen;
if (!ending) {
ret = deflate(zs, Z_NO_FLUSH);
if (ret != Z_OK) {
deflateReset(zs);
zerr(ret, 1);
return (-1);
}
} else {
ret = deflate(zs, Z_FINISH);
if (ret != Z_STREAM_END) {
deflateReset(zs);
if (ret == Z_OK)
zerr(Z_BUF_ERROR, 1);
else
zerr(ret, 1);
return (-1);
}
}
dst1 += (dlen - zs->avail_out);
_dstlen -= (dlen - zs->avail_out);
src1 += slen;
_srclen -= slen;
}
*dstlen = *dstlen - _dstlen;
ret = deflateReset(zs);
if (ret != Z_OK) {
zerr(ret, 1);
return (-1);
}
return (0);
}
int
adapt_compress(void *src, uint64_t srclen, void *dst,
uint64_t *dstlen, int level, uchar_t chdr, int btype, void *data)
{
struct adapt_data *adat = (struct adapt_data *)(data);
uchar_t *src1 = (uchar_t *)src;
int rv = 0, bsc_type = 0;
if (btype == TYPE_UNKNOWN) {
uint64_t i, tot8b, tag1, tag2, tag3;
double tagcnt, pct_tag;
uchar_t cur_byte, prev_byte;
/*
* Count number of 8-bit binary bytes and XML tags in source.
*/
tot8b = 0;
tag1 = 0;
tag2 = 0;
tag3 = 0;
prev_byte = cur_byte = 0;
for (i = 0; i < srclen; i++) {
cur_byte = src1[i];
tot8b += (cur_byte & 0x80); // This way for possible auto-vectorization
tag1 += (cur_byte == '<');
tag2 += (cur_byte == '>');
tag3 += ((prev_byte == '<') & (cur_byte == '/'));
tag3 += ((prev_byte == '/') & (cur_byte == '>'));
if (cur_byte != ' ')
prev_byte = cur_byte;
}
tot8b /= 0x80;
tagcnt = tag1 + tag2 + tag3;
pct_tag = tagcnt / (double)srclen;
if (adat->adapt_mode == 2 && tot8b > FORTY_PCT(srclen)) {
btype = TYPE_BINARY;
} else if (adat->adapt_mode == 1 && tot8b > FIFTY_PCT(srclen)) {
btype = TYPE_BINARY;
} else {
btype = TYPE_TEXT;
if (tag1 > tag2 - 4 && tag1 < tag2 + 4 && tag3 > (double)tag1 * 0.40 &&
tagcnt > (double)srclen * 0.001)
btype |= TYPE_MARKUP;
}
}
/*
* Use PPMd if some percentage of source is 7-bit textual bytes, otherwise
* use Bzip2 or LZMA. For totally incompressible data we always use LZ4. There
* is no point trying to compress such data, like Jpegs. However some archive headers
* and zero paddings can exist which LZ4 can easily take care of very fast.
*/
#ifdef ENABLE_PC_LIBBSC
bsc_type = is_bsc_type(btype);
#endif
if (is_incompressible(btype)) {
rv = lz4_compress(src, srclen, dst, dstlen, level, chdr, btype, adat->lz4_data);
if (rv < 0)
return (rv);
rv = ADAPT_COMPRESS_LZ4;
lz4_count++;
} else if (adat->adapt_mode == 2 && PC_TYPE(btype) == TYPE_BINARY && !bsc_type) {
rv = lzma_compress(src, srclen, dst, dstlen, level, chdr, btype, adat->lzma_data);
if (rv < 0)
return (rv);
rv = ADAPT_COMPRESS_LZMA;
lzma_count++;
} else if (adat->adapt_mode == 1 && PC_TYPE(btype) == TYPE_BINARY && !bsc_type) {
rv = bzip2_compress(src, srclen, dst, dstlen, level, chdr, btype, NULL);
if (rv < 0)
return (rv);
rv = ADAPT_COMPRESS_BZIP2;
bzip2_count++;
} else {
#ifdef ENABLE_PC_LIBBSC
if (adat->bsc_data && bsc_type) {
rv = libbsc_compress(src, srclen, dst, dstlen, level, chdr, btype, adat->bsc_data);
if (rv < 0)
return (rv);
rv = ADAPT_COMPRESS_BSC;
bsc_count++;
} else {
#endif
rv = ppmd_alloc(adat->ppmd_data);
if (rv < 0)
return (rv);
rv = ppmd_compress(src, srclen, dst, dstlen, level, chdr, btype, adat->ppmd_data);
ppmd_free(adat->ppmd_data);
if (rv < 0)
return (rv);
rv = ADAPT_COMPRESS_PPMD;
ppmd_count++;
#ifdef ENABLE_PC_LIBBSC
}
#endif
}
return (rv);
}
