Merge pull request #1760 from bimbashrestha/extract_sequences_api

Adding api for extracting sequences from seqstore
dev
Yann Collet 2019-10-10 13:11:18 -07:00 committed by GitHub
commit fb77afc626
No known key found for this signature in database
GPG Key ID: 4AEE18F83AFDEB23
4 changed files with 160 additions and 3 deletions

View File

@ -2265,6 +2265,77 @@ static size_t ZSTD_buildSeqStore(ZSTD_CCtx* zc, const void* src, size_t srcSize)
return ZSTDbss_compress;
}
static void ZSTD_copyBlockSequences(ZSTD_CCtx* zc)
{
const seqStore_t* seqStore = ZSTD_getSeqStore(zc);
const seqDef* seqs = seqStore->sequencesStart;
size_t seqsSize = seqStore->sequences - seqs;
ZSTD_Sequence* outSeqs = &zc->seqCollector.seqStart[zc->seqCollector.seqIndex];
size_t i; size_t position; int repIdx;
assert(zc->seqCollector.seqIndex + 1 < zc->seqCollector.maxSequences);
for (i = 0, position = 0; i < seqsSize; ++i) {
outSeqs[i].offset = seqs[i].offset;
outSeqs[i].litLength = seqs[i].litLength;
outSeqs[i].matchLength = seqs[i].matchLength + MINMATCH;
if (i == seqStore->longLengthPos) {
if (seqStore->longLengthID == 1) {
outSeqs[i].litLength += 0x10000;
} else if (seqStore->longLengthID == 2) {
outSeqs[i].matchLength += 0x10000;
}
}
if (outSeqs[i].offset <= ZSTD_REP_NUM) {
outSeqs[i].rep = outSeqs[i].offset;
repIdx = (unsigned int)i - outSeqs[i].offset;
if (outSeqs[i].litLength == 0) {
if (outSeqs[i].offset < 3) {
--repIdx;
} else {
repIdx = (unsigned int)i - 1;
}
++outSeqs[i].rep;
}
assert(repIdx >= -3);
outSeqs[i].offset = repIdx >= 0 ? outSeqs[repIdx].offset : repStartValue[-repIdx - 1];
if (outSeqs[i].rep == 4) {
--outSeqs[i].offset;
}
} else {
outSeqs[i].offset -= ZSTD_REP_NUM;
}
position += outSeqs[i].litLength;
outSeqs[i].matchPos = (unsigned int)position;
position += outSeqs[i].matchLength;
}
zc->seqCollector.seqIndex += seqsSize;
}
size_t ZSTD_getSequences(ZSTD_CCtx* zc, ZSTD_Sequence* outSeqs,
size_t outSeqsSize, const void* src, size_t srcSize)
{
const size_t dstCapacity = ZSTD_compressBound(srcSize);
void* dst = ZSTD_malloc(dstCapacity, ZSTD_defaultCMem);
SeqCollector seqCollector;
RETURN_ERROR_IF(dst == NULL, memory_allocation);
seqCollector.collectSequences = 1;
seqCollector.seqStart = outSeqs;
seqCollector.seqIndex = 0;
seqCollector.maxSequences = outSeqsSize;
zc->seqCollector = seqCollector;
ZSTD_compress2(zc, dst, dstCapacity, src, srcSize);
ZSTD_free(dst, ZSTD_defaultCMem);
return zc->seqCollector.seqIndex;
}
/* Returns true if the given block is a RLE block */
static int ZSTD_isRLE(const BYTE *ip, size_t length) {
size_t i;
@ -2296,6 +2367,11 @@ static size_t ZSTD_compressBlock_internal(ZSTD_CCtx* zc,
if (bss == ZSTDbss_noCompress) { cSize = 0; goto out; }
}
if (zc->seqCollector.collectSequences) {
ZSTD_copyBlockSequences(zc);
return 0;
}
/* encode sequences and literals */
cSize = ZSTD_compressSequences(&zc->seqStore,
&zc->blockState.prevCBlock->entropy, &zc->blockState.nextCBlock->entropy,
@ -2360,7 +2436,6 @@ static void ZSTD_overflowCorrectIfNeeded(ZSTD_matchState_t* ms,
}
}
/*! ZSTD_compress_frameChunk() :
* Compress a chunk of data into one or multiple blocks.
* All blocks will be terminated, all input will be consumed.
@ -2405,7 +2480,6 @@ static size_t ZSTD_compress_frameChunk (ZSTD_CCtx* cctx,
op+ZSTD_blockHeaderSize, dstCapacity-ZSTD_blockHeaderSize,
ip, blockSize, 1 /* frame */);
FORWARD_IF_ERROR(cSize);
if (cSize == 0) { /* block is not compressible */
cSize = ZSTD_noCompressBlock(op, dstCapacity, ip, blockSize, lastBlock);
FORWARD_IF_ERROR(cSize);

View File

@ -193,6 +193,13 @@ typedef struct {
size_t capacity; /* The capacity starting from `seq` pointer */
} rawSeqStore_t;
typedef struct {
int collectSequences;
ZSTD_Sequence* seqStart;
size_t seqIndex;
size_t maxSequences;
} SeqCollector;
struct ZSTD_CCtx_params_s {
ZSTD_format_e format;
ZSTD_compressionParameters cParams;
@ -240,6 +247,7 @@ struct ZSTD_CCtx_s {
XXH64_state_t xxhState;
ZSTD_customMem customMem;
size_t staticSize;
SeqCollector seqCollector;
int isFirstBlock;
seqStore_t seqStore; /* sequences storage ptrs */

View File

@ -1077,6 +1077,24 @@ ZSTDLIB_API size_t ZSTD_sizeof_DDict(const ZSTD_DDict* ddict);
typedef struct ZSTD_CCtx_params_s ZSTD_CCtx_params;
typedef struct {
unsigned int matchPos; /* Match pos in dst */
/* If seqDef.offset > 3, then this is seqDef.offset - 3
* If seqDef.offset < 3, then this is the corresponding repeat offset
* But if seqDef.offset < 3 and litLength == 0, this is the
* repeat offset before the corresponding repeat offset
* And if seqDef.offset == 3 and litLength == 0, this is the
* most recent repeat offset - 1
*/
unsigned int offset;
unsigned int litLength; /* Literal length */
unsigned int matchLength; /* Match length */
/* 0 when seq not rep and seqDef.offset otherwise
* when litLength == 0 this will be <= 4, otherwise <= 3 like normal
*/
unsigned int rep;
} ZSTD_Sequence;
typedef struct {
unsigned windowLog; /**< largest match distance : larger == more compression, more memory needed during decompression */
unsigned chainLog; /**< fully searched segment : larger == more compression, slower, more memory (useless for fast) */
@ -1215,6 +1233,15 @@ ZSTDLIB_API unsigned long long ZSTD_decompressBound(const void* src, size_t srcS
* or an error code (if srcSize is too small) */
ZSTDLIB_API size_t ZSTD_frameHeaderSize(const void* src, size_t srcSize);
/*! ZSTD_getSequences() :
* Extract sequences from the sequence store
* zc can be used to insert custom compression params.
* This function invokes ZSTD_compress2
* @return : number of sequences extracted
*/
ZSTDLIB_API size_t ZSTD_getSequences(ZSTD_CCtx* zc, ZSTD_Sequence* outSeqs,
size_t outSeqsSize, const void* src, size_t srcSize);
/***************************************
* Memory management

View File

@ -304,6 +304,28 @@ static int FUZ_mallocTests(unsigned seed, double compressibility, unsigned part)
#endif
static void FUZ_decodeSequences(BYTE* dst, ZSTD_Sequence* seqs, size_t seqsSize, BYTE* src, size_t size)
{
size_t i;
size_t j;
for(i = 0; i < seqsSize - 1; ++i) {
assert(dst + seqs[i].litLength + seqs[i].matchLength < dst + size);
assert(src + seqs[i].litLength + seqs[i].matchLength < src + size);
memcpy(dst, src, seqs[i].litLength);
dst += seqs[i].litLength;
src += seqs[i].litLength;
size -= seqs[i].litLength;
for (j = 0; j < seqs[i].matchLength; ++j)
dst[j] = dst[j - seqs[i].offset];
dst += seqs[i].matchLength;
src += seqs[i].matchLength;
size -= seqs[i].matchLength;
}
memcpy(dst, src, size);
}
/*=============================================
* Unit tests
=============================================*/
@ -1960,6 +1982,33 @@ static int basicUnitTests(U32 const seed, double compressibility)
DISPLAYLEVEL(3, "OK \n");
}
DISPLAYLEVEL(3, "test%3i : ZSTD_getSequences decode from sequences test : ", testNb++);
{
size_t srcSize = 100 KB;
BYTE* src = (BYTE*)CNBuffer;
BYTE* decoded = (BYTE*)compressedBuffer;
ZSTD_CCtx* cctx = ZSTD_createCCtx();
ZSTD_Sequence* seqs = (ZSTD_Sequence*)malloc(srcSize * sizeof(ZSTD_Sequence));
size_t seqsSize;
if (seqs == NULL) goto _output_error;
assert(cctx != NULL);
/* Populate src with random data */
RDG_genBuffer(CNBuffer, srcSize, compressibility, 0., seed);
/* get the sequences */
seqsSize = ZSTD_getSequences(cctx, seqs, srcSize, src, srcSize);
/* "decode" and compare the sequences */
FUZ_decodeSequences(decoded, seqs, seqsSize, src, srcSize);
assert(!memcmp(CNBuffer, compressedBuffer, srcSize));
ZSTD_freeCCtx(cctx);
free(seqs);
}
/* Multiple blocks of zeros test */
#define LONGZEROSLENGTH 1000000 /* 1MB of zeros */
DISPLAYLEVEL(3, "test%3i : compress %u zeroes : ", testNb++, LONGZEROSLENGTH);
@ -1972,7 +2021,6 @@ static int basicUnitTests(U32 const seed, double compressibility)
if (r != LONGZEROSLENGTH) goto _output_error; }
DISPLAYLEVEL(3, "OK \n");
/* All zeroes test (test bug #137) */
#define ZEROESLENGTH 100
DISPLAYLEVEL(3, "test%3i : compress %u zeroes : ", testNb++, ZEROESLENGTH);