facebook
/
zstd
1
0
Fork 0
Code Issues Pull Requests Packages Projects Releases Wiki Activity

Merge pull request #1238 from jennifermliu/random 

Add random dictionary builder
dev
Nick Terrell 2018-07-19 13:52:15 -07:00 committed by GitHub
parent effa84c8d1 0c5eaef248
commit 4d1ad5cdb2
No known key found for this signature in database
GPG Key ID: 4AEE18F83AFDEB23
8 changed files with 780 additions and 0 deletions
Show Stats Download Patch File Download Diff File Expand all files Collapse all files

52
contrib/randomDictBuilder/Makefile Normal file
View File

@ -0,0 +1,52 @@
ARG :=
CC ?= gcc
CFLAGS ?= -O3
INCLUDES := -I ../../programs -I ../../lib/common -I ../../lib -I ../../lib/dictBuilder
TEST_INPUT := ../../lib
TEST_OUTPUT := randomDict
all: main run clean
.PHONY: test
test: main testrun testshell clean
.PHONY: run
run:
echo "Building a random dictionary with given arguments"
./main $(ARG)
main: main.o io.o random.o libzstd.a
$(CC) $(CFLAGS) main.o io.o random.o libzstd.a -o main
main.o: main.c
$(CC) $(CFLAGS) $(INCLUDES) -c main.c
random.o: random.c
$(CC) $(CFLAGS) $(INCLUDES) -c random.c
io.o: io.c
$(CC) $(CFLAGS) $(INCLUDES) -c io.c
libzstd.a:
$(MAKE) -C ../../lib libzstd.a
mv ../../lib/libzstd.a .
.PHONY: testrun
testrun: main
echo "Run with $(TEST_INPUT) and $(TEST_OUTPUT) "
./main in=$(TEST_INPUT) out=$(TEST_OUTPUT)
zstd -be3 -D $(TEST_OUTPUT) -r $(TEST_INPUT) -q
rm -f $(TEST_OUTPUT)
.PHONY: testshell
testshell: test.sh
sh test.sh
echo "Finish running test.sh"
.PHONY: clean
clean:
rm -f *.o main libzstd.a
$(MAKE) -C ../../lib clean
echo "Cleaning is completed"

20
contrib/randomDictBuilder/README.md Normal file
View File

@ -0,0 +1,20 @@
Random Dictionary Builder
### Permitted Arguments:
Input File/Directory (in=fileName): required; file/directory used to build dictionary; if directory, will operate recursively for files inside directory; can include multiple files/directories, each following "in="
Output Dictionary (out=dictName): if not provided, default to defaultDict
Dictionary ID (dictID=#): nonnegative number; if not provided, default to 0
Maximum Dictionary Size (maxdict=#): positive number; in bytes, if not provided, default to 110KB
Size of Randomly Selected Segment (k=#): positive number; in bytes; if not provided, default to 200
###Running Test:
make test
###Usage:
To build a random dictionary with the provided arguments: make ARG= followed by arguments
### Examples:
make ARG="in=../../lib/dictBuilder out=dict100 dictID=520"
make ARG="in=../../lib/dictBuilder in=../../lib/compress"

251
contrib/randomDictBuilder/io.c Normal file
View File

@ -0,0 +1,251 @@
#include <stdio.h> /* fprintf */
#include <stdlib.h> /* malloc, free, qsort */
#include <string.h> /* strcmp, strlen */
#include <errno.h> /* errno */
#include <ctype.h>
#include "io.h"
#include "fileio.h" /* stdinmark, stdoutmark, ZSTD_EXTENSION */
#include "platform.h" /* Large Files support */
#include "util.h"
#include "zdict.h"
/*-*************************************
* Console display
***************************************/
#define DISPLAY(...) fprintf(stderr, __VA_ARGS__)
#define DISPLAYLEVEL(l, ...) if (displayLevel>=l) { DISPLAY(__VA_ARGS__); }
static const U64 g_refreshRate = SEC_TO_MICRO / 6;
static UTIL_time_t g_displayClock = UTIL_TIME_INITIALIZER;
#define DISPLAYUPDATE(l, ...) { if (displayLevel>=l) { \
if ((UTIL_clockSpanMicro(g_displayClock) > g_refreshRate) || (displayLevel>=4)) \
{ g_displayClock = UTIL_getTime(); DISPLAY(__VA_ARGS__); \
if (displayLevel>=4) fflush(stderr); } } }
/*-*************************************
* Exceptions
***************************************/
#ifndef DEBUG
# define DEBUG 0
#endif
#define DEBUGOUTPUT(...) if (DEBUG) DISPLAY(__VA_ARGS__);
#define EXM_THROW(error, ...) \
{ \
DEBUGOUTPUT("Error defined at %s, line %i : \n", __FILE__, __LINE__); \
DISPLAY("Error %i : ", error); \
DISPLAY(__VA_ARGS__); \
DISPLAY("\n"); \
exit(error); \
}
/*-*************************************
* Constants
***************************************/
#define SAMPLESIZE_MAX (128 KB)
#define RANDOM_MAX_SAMPLES_SIZE (sizeof(size_t) == 8 ? ((U32)-1) : ((U32)1 GB))
#define RANDOM_MEMMULT 9
static const size_t g_maxMemory = (sizeof(size_t) == 4) ?
(2 GB - 64 MB) : ((size_t)(512 MB) << sizeof(size_t));
#define NOISELENGTH 32
/* ********************************************************
* File related operations
**********************************************************/
/** loadFiles() :
* load samples from files listed in fileNamesTable into buffer.
* works even if buffer is too small to load all samples.
* Also provides the size of each sample into sampleSizes table
* which must be sized correctly, using DiB_fileStats().
* @return : nb of samples effectively loaded into `buffer`
* *bufferSizePtr is modified, it provides the amount data loaded within buffer.
* sampleSizes is filled with the size of each sample.
*/
static unsigned loadFiles(void* buffer, size_t* bufferSizePtr, size_t* sampleSizes,
unsigned sstSize, const char** fileNamesTable, unsigned nbFiles,
size_t targetChunkSize, unsigned displayLevel) {
char* const buff = (char*)buffer;
size_t pos = 0;
unsigned nbLoadedChunks = 0, fileIndex;
for (fileIndex=0; fileIndex<nbFiles; fileIndex++) {
const char* const fileName = fileNamesTable[fileIndex];
unsigned long long const fs64 = UTIL_getFileSize(fileName);
unsigned long long remainingToLoad = (fs64 == UTIL_FILESIZE_UNKNOWN) ? 0 : fs64;
U32 const nbChunks = targetChunkSize ? (U32)((fs64 + (targetChunkSize-1)) / targetChunkSize) : 1;
U64 const chunkSize = targetChunkSize ? MIN(targetChunkSize, fs64) : fs64;
size_t const maxChunkSize = (size_t)MIN(chunkSize, SAMPLESIZE_MAX);
U32 cnb;
FILE* const f = fopen(fileName, "rb");
if (f==NULL) EXM_THROW(10, "zstd: dictBuilder: %s %s ", fileName, strerror(errno));
DISPLAYUPDATE(2, "Loading %s... \r", fileName);
for (cnb=0; cnb<nbChunks; cnb++) {
size_t const toLoad = (size_t)MIN(maxChunkSize, remainingToLoad);
if (toLoad > *bufferSizePtr-pos) break;
{ size_t const readSize = fread(buff+pos, 1, toLoad, f);
if (readSize != toLoad) EXM_THROW(11, "Pb reading %s", fileName);
pos += readSize;
sampleSizes[nbLoadedChunks++] = toLoad;
remainingToLoad -= targetChunkSize;
if (nbLoadedChunks == sstSize) { /* no more space left in sampleSizes table */
fileIndex = nbFiles; /* stop there */
break;
}
if (toLoad < targetChunkSize) {
fseek(f, (long)(targetChunkSize - toLoad), SEEK_CUR);
} } }
fclose(f);
}
DISPLAYLEVEL(2, "\r%79s\r", "");
*bufferSizePtr = pos;
DISPLAYLEVEL(4, "loaded : %u KB \n", (U32)(pos >> 10))
return nbLoadedChunks;
}
#define rotl32(x,r) ((x << r) | (x >> (32 - r)))
static U32 getRand(U32* src)
{
static const U32 prime1 = 2654435761U;
static const U32 prime2 = 2246822519U;
U32 rand32 = *src;
rand32 *= prime1;
rand32 ^= prime2;
rand32 = rotl32(rand32, 13);
*src = rand32;
return rand32 >> 5;
}
/* shuffle() :
* shuffle a table of file names in a semi-random way
* It improves dictionary quality by reducing "locality" impact, so if sample set is very large,
* it will load random elements from it, instead of just the first ones. */
static void shuffle(const char** fileNamesTable, unsigned nbFiles) {
U32 seed = 0xFD2FB528;
unsigned i;
for (i = nbFiles - 1; i > 0; --i) {
unsigned const j = getRand(&seed) % (i + 1);
const char* const tmp = fileNamesTable[j];
fileNamesTable[j] = fileNamesTable[i];
fileNamesTable[i] = tmp;
}
}
/*-********************************************************
* Dictionary training functions
**********************************************************/
static size_t findMaxMem(unsigned long long requiredMem) {
size_t const step = 8 MB;
void* testmem = NULL;
requiredMem = (((requiredMem >> 23) + 1) << 23);
requiredMem += step;
if (requiredMem > g_maxMemory) requiredMem = g_maxMemory;
while (!testmem) {
testmem = malloc((size_t)requiredMem);
requiredMem -= step;
}
free(testmem);
return (size_t)requiredMem;
}
void saveDict(const char* dictFileName,
const void* buff, size_t buffSize) {
FILE* const f = fopen(dictFileName, "wb");
if (f==NULL) EXM_THROW(3, "cannot open %s ", dictFileName);
{ size_t const n = fwrite(buff, 1, buffSize, f);
if (n!=buffSize) EXM_THROW(4, "%s : write error", dictFileName) }
{ size_t const n = (size_t)fclose(f);
if (n!=0) EXM_THROW(5, "%s : flush error", dictFileName) }
}
/*! getFileStats() :
* Given a list of files, and a chunkSize (0 == no chunk, whole files)
* provides the amount of data to be loaded and the resulting nb of samples.
* This is useful primarily for allocation purpose => sample buffer, and sample sizes table.
*/
static fileStats getFileStats(const char** fileNamesTable, unsigned nbFiles,
size_t chunkSize, unsigned displayLevel) {
fileStats fs;
unsigned n;
memset(&fs, 0, sizeof(fs));
for (n=0; n<nbFiles; n++) {
U64 const fileSize = UTIL_getFileSize(fileNamesTable[n]);
U64 const srcSize = (fileSize == UTIL_FILESIZE_UNKNOWN) ? 0 : fileSize;
U32 const nbSamples = (U32)(chunkSize ? (srcSize + (chunkSize-1)) / chunkSize : 1);
U64 const chunkToLoad = chunkSize ? MIN(chunkSize, srcSize) : srcSize;
size_t const cappedChunkSize = (size_t)MIN(chunkToLoad, SAMPLESIZE_MAX);
fs.totalSizeToLoad += cappedChunkSize * nbSamples;
fs.oneSampleTooLarge |= (chunkSize > 2*SAMPLESIZE_MAX);
fs.nbSamples += nbSamples;
}
DISPLAYLEVEL(4, "Preparing to load : %u KB \n", (U32)(fs.totalSizeToLoad >> 10));
return fs;
}
sampleInfo* getSampleInfo(const char** fileNamesTable, unsigned nbFiles, size_t chunkSize,
unsigned maxDictSize, const unsigned displayLevel) {
fileStats const fs = getFileStats(fileNamesTable, nbFiles, chunkSize, displayLevel);
size_t* const sampleSizes = (size_t*)malloc(fs.nbSamples * sizeof(size_t));
size_t const memMult = RANDOM_MEMMULT;
size_t const maxMem = findMaxMem(fs.totalSizeToLoad * memMult) / memMult;
size_t loadedSize = (size_t) MIN ((unsigned long long)maxMem, fs.totalSizeToLoad);
void* const srcBuffer = malloc(loadedSize+NOISELENGTH);
/* Checks */
if ((!sampleSizes) || (!srcBuffer))
EXM_THROW(12, "not enough memory for trainFromFiles"); /* should not happen */
if (fs.oneSampleTooLarge) {
DISPLAYLEVEL(2, "! Warning : some sample(s) are very large \n");
DISPLAYLEVEL(2, "! Note that dictionary is only useful for small samples. \n");
DISPLAYLEVEL(2, "! As a consequence, only the first %u bytes of each sample are loaded \n", SAMPLESIZE_MAX);
}
if (fs.nbSamples < 5) {
DISPLAYLEVEL(2, "! Warning : nb of samples too low for proper processing ! \n");
DISPLAYLEVEL(2, "! Please provide _one file per sample_. \n");
DISPLAYLEVEL(2, "! Alternatively, split files into fixed-size blocks representative of samples, with -B# \n");
EXM_THROW(14, "nb of samples too low"); /* we now clearly forbid this case */
}
if (fs.totalSizeToLoad < (unsigned long long)(8 * maxDictSize)) {
DISPLAYLEVEL(2, "! Warning : data size of samples too small for target dictionary size \n");
DISPLAYLEVEL(2, "! Samples should be about 100x larger than target dictionary size \n");
}
/* init */
if (loadedSize < fs.totalSizeToLoad)
DISPLAYLEVEL(1, "Not enough memory; training on %u MB only...\n", (unsigned)(loadedSize >> 20));
/* Load input buffer */
DISPLAYLEVEL(3, "Shuffling input files\n");
shuffle(fileNamesTable, nbFiles);
nbFiles = loadFiles(srcBuffer, &loadedSize, sampleSizes, fs.nbSamples,
fileNamesTable, nbFiles, chunkSize, displayLevel);
sampleInfo *info = (sampleInfo *)malloc(sizeof(sampleInfo));
info->nbSamples = fs.nbSamples;
info->samplesSizes = sampleSizes;
info->srcBuffer = srcBuffer;
return info;
}
void freeSampleInfo(sampleInfo *info) {
if (!info) return;
if (info->samplesSizes) free((void*)(info->samplesSizes));
if (info->srcBuffer) free((void*)(info->srcBuffer));
free(info);
}

50
contrib/randomDictBuilder/io.h Normal file
View File

@ -0,0 +1,50 @@
#include <stdio.h> /* fprintf */
#include <stdlib.h> /* malloc, free, qsort */
#include <string.h> /* strcmp, strlen */
#include <errno.h> /* errno */
#include <ctype.h>
#include "zstd_internal.h" /* includes zstd.h */
#include "fileio.h" /* stdinmark, stdoutmark, ZSTD_EXTENSION */
#include "platform.h" /* Large Files support */
#include "util.h"
#include "zdict.h"
/*-*************************************
* Structs
***************************************/
typedef struct {
U64 totalSizeToLoad;
unsigned oneSampleTooLarge;
unsigned nbSamples;
} fileStats;
typedef struct {
const void* srcBuffer;
const size_t *samplesSizes;
size_t nbSamples;
}sampleInfo;
/*! getSampleInfo():
* Load from input files and add samples to buffer
* @return: a sampleInfo struct containing infomation about buffer where samples are stored,
* size of each sample, and total number of samples
*/
sampleInfo* getSampleInfo(const char** fileNamesTable, unsigned nbFiles, size_t chunkSize,
unsigned maxDictSize, const unsigned displayLevel);
/*! freeSampleInfo():
* Free memory allocated for info
*/
void freeSampleInfo(sampleInfo *info);
/*! saveDict():
* Save data stored on buff to dictFileName
*/
void saveDict(const char* dictFileName, const void* buff, size_t buffSize);

201
contrib/randomDictBuilder/main.c Normal file
View File

@ -0,0 +1,201 @@
#include <stdio.h> /* fprintf */
#include <stdlib.h> /* malloc, free, qsort */
#include <string.h> /* strcmp, strlen */
#include <errno.h> /* errno */
#include <ctype.h>
#include "random.h"
#include "io.h"
#include "util.h"
#include "zdict.h"
/*-*************************************
* Console display
***************************************/
#define DISPLAY(...) fprintf(stderr, __VA_ARGS__)
#define DISPLAYLEVEL(l, ...) if (displayLevel>=l) { DISPLAY(__VA_ARGS__); }
static const U64 g_refreshRate = SEC_TO_MICRO / 6;
static UTIL_time_t g_displayClock = UTIL_TIME_INITIALIZER;
#define DISPLAYUPDATE(l, ...) { if (displayLevel>=l) { \
if ((UTIL_clockSpanMicro(g_displayClock) > g_refreshRate) || (displayLevel>=4)) \
{ g_displayClock = UTIL_getTime(); DISPLAY(__VA_ARGS__); \
if (displayLevel>=4) fflush(stderr); } } }
/*-*************************************
* Exceptions
***************************************/
#ifndef DEBUG
# define DEBUG 0
#endif
#define DEBUGOUTPUT(...) if (DEBUG) DISPLAY(__VA_ARGS__);
#define EXM_THROW(error, ...) \
{ \
DEBUGOUTPUT("Error defined at %s, line %i : \n", __FILE__, __LINE__); \
DISPLAY("Error %i : ", error); \
DISPLAY(__VA_ARGS__); \
DISPLAY("\n"); \
exit(error); \
}
/*-*************************************
* Constants
***************************************/
static const unsigned g_defaultMaxDictSize = 110 KB;
#define DEFAULT_CLEVEL 3
#define DEFAULT_k 200
#define DEFAULT_OUTPUTFILE "defaultDict"
#define DEFAULT_DICTID 0
/*-*************************************
* Commandline related functions
***************************************/
static unsigned readU32FromChar(const char** stringPtr){
const char errorMsg[] = "error: numeric value too large";
unsigned result = 0;
while ((**stringPtr >='0') && (**stringPtr <='9')) {
unsigned const max = (((unsigned)(-1)) / 10) - 1;
if (result > max) exit(1);
result *= 10, result += **stringPtr - '0', (*stringPtr)++ ;
}
if ((**stringPtr=='K') || (**stringPtr=='M')) {
unsigned const maxK = ((unsigned)(-1)) >> 10;
if (result > maxK) exit(1);
result <<= 10;
if (**stringPtr=='M') {
if (result > maxK) exit(1);
result <<= 10;
}
(*stringPtr)++; /* skip `K` or `M` */
if (**stringPtr=='i') (*stringPtr)++;
if (**stringPtr=='B') (*stringPtr)++;
}
return result;
}
/** longCommandWArg() :
* check if *stringPtr is the same as longCommand.
* If yes, @return 1 and advances *stringPtr to the position which immediately follows longCommand.
* @return 0 and doesn't modify *stringPtr otherwise.
*/
static unsigned longCommandWArg(const char** stringPtr, const char* longCommand){
size_t const comSize = strlen(longCommand);
int const result = !strncmp(*stringPtr, longCommand, comSize);
if (result) *stringPtr += comSize;
return result;
}
/*-*************************************
* RANDOM
***************************************/
int RANDOM_trainFromFiles(const char* dictFileName, sampleInfo *info,
unsigned maxDictSize,
ZDICT_random_params_t *params) {
unsigned const displayLevel = params->zParams.notificationLevel;
void* const dictBuffer = malloc(maxDictSize);
int result = 0;
/* Checks */
if (!dictBuffer)
EXM_THROW(12, "not enough memory for trainFromFiles"); /* should not happen */
{ size_t dictSize;
dictSize = ZDICT_trainFromBuffer_random(dictBuffer, maxDictSize, info->srcBuffer,
info->samplesSizes, info->nbSamples, *params);
DISPLAYLEVEL(2, "k=%u\n", params->k);
if (ZDICT_isError(dictSize)) {
DISPLAYLEVEL(1, "dictionary training failed : %s \n", ZDICT_getErrorName(dictSize)); /* should not happen */
result = 1;
goto _done;
}
/* save dict */
DISPLAYLEVEL(2, "Save dictionary of size %u into file %s \n", (U32)dictSize, dictFileName);
saveDict(dictFileName, dictBuffer, dictSize);
}
/* clean up */
_done:
free(dictBuffer);
return result;
}
int main(int argCount, const char* argv[])
{
int displayLevel = 2;
const char* programName = argv[0];
int operationResult = 0;
/* Initialize arguments to default values */
unsigned k = DEFAULT_k;
const char* outputFile = DEFAULT_OUTPUTFILE;
unsigned dictID = DEFAULT_DICTID;
unsigned maxDictSize = g_defaultMaxDictSize;
/* Initialize table to store input files */
const char** filenameTable = (const char**)malloc(argCount * sizeof(const char*));
unsigned filenameIdx = 0;
/* Parse arguments */
for (int i = 1; i < argCount; i++) {
const char* argument = argv[i];
if (longCommandWArg(&argument, "k=")) { k = readU32FromChar(&argument); continue; }
if (longCommandWArg(&argument, "dictID=")) { dictID = readU32FromChar(&argument); continue; }
if (longCommandWArg(&argument, "maxdict=")) { maxDictSize = readU32FromChar(&argument); continue; }
if (longCommandWArg(&argument, "in=")) {
filenameTable[filenameIdx] = argument;
filenameIdx++;
continue;
}
if (longCommandWArg(&argument, "out=")) {
outputFile = argument;
continue;
}
DISPLAYLEVEL(1, "Incorrect parameters\n");
operationResult = 1;
return operationResult;
}
char* fileNamesBuf = NULL;
unsigned fileNamesNb = filenameIdx;
int followLinks = 0; /* follow directory recursively */
const char** extendedFileList = NULL;
extendedFileList = UTIL_createFileList(filenameTable, filenameIdx, &fileNamesBuf,
&fileNamesNb, followLinks);
if (extendedFileList) {
unsigned u;
for (u=0; u<fileNamesNb; u++) DISPLAYLEVEL(4, "%u %s\n", u, extendedFileList[u]);
free((void*)filenameTable);
filenameTable = extendedFileList;
filenameIdx = fileNamesNb;
}
size_t blockSize = 0;
ZDICT_random_params_t params;
ZDICT_params_t zParams;
zParams.compressionLevel = DEFAULT_CLEVEL;
zParams.notificationLevel = displayLevel;
zParams.dictID = dictID;
params.zParams = zParams;
params.k = k;
sampleInfo* info= getSampleInfo(filenameTable,
filenameIdx, blockSize, maxDictSize, zParams.notificationLevel);
operationResult = RANDOM_trainFromFiles(outputFile, info, maxDictSize, &params);
/* Free allocated memory */
UTIL_freeFileList(extendedFileList, fileNamesBuf);
freeSampleInfo(info);
return operationResult;
}

163
contrib/randomDictBuilder/random.c Normal file
View File

@ -0,0 +1,163 @@
/*-*************************************
* Dependencies
***************************************/
#include <stdio.h> /* fprintf */
#include <stdlib.h> /* malloc, free, qsort */
#include <string.h> /* memset */
#include <time.h> /* clock */
#include "random.h"
#include "util.h" /* UTIL_getFileSize, UTIL_getTotalFileSize */
#ifndef ZDICT_STATIC_LINKING_ONLY
#define ZDICT_STATIC_LINKING_ONLY
#endif
#include "zdict.h"
/*-*************************************
* Console display
***************************************/
#define DISPLAY(...) fprintf(stderr, __VA_ARGS__)
#define DISPLAYLEVEL(l, ...) if (displayLevel>=l) { DISPLAY(__VA_ARGS__); }
#define LOCALDISPLAYUPDATE(displayLevel, l, ...) \
if (displayLevel >= l) { \
if ((clock() - g_time > refreshRate) || (displayLevel >= 4)) { \
g_time = clock(); \
DISPLAY(__VA_ARGS__); \
} \
}
#define DISPLAYUPDATE(l, ...) LOCALDISPLAYUPDATE(displayLevel, l, __VA_ARGS__)
static const clock_t refreshRate = CLOCKS_PER_SEC * 15 / 100;
static clock_t g_time = 0;
/* ********************************************************
* Random Dictionary Builder
**********************************************************/
/**
* Returns the sum of the sample sizes.
*/
static size_t RANDOM_sum(const size_t *samplesSizes, unsigned nbSamples) {
size_t sum = 0;
unsigned i;
for (i = 0; i < nbSamples; ++i) {
sum += samplesSizes[i];
}
return sum;
}
/**
* A segment is an inclusive range in the source.
*/
typedef struct {
U32 begin;
U32 end;
} RANDOM_segment_t;
/**
* Selects a random segment from totalSamplesSize - k + 1 possible segments
*/
static RANDOM_segment_t RANDOM_selectSegment(const size_t totalSamplesSize,
ZDICT_random_params_t parameters) {
const U32 k = parameters.k;
RANDOM_segment_t segment;
unsigned index;
/* Randomly generate a number from 0 to sampleSizes - k */
index = rand()%(totalSamplesSize - k + 1);
/* inclusive */
segment.begin = index;
segment.end = index + k - 1;
return segment;
}
/**
* Check the validity of the parameters.
* Returns non-zero if the parameters are valid and 0 otherwise.
*/
static int RANDOM_checkParameters(ZDICT_random_params_t parameters,
size_t maxDictSize) {
/* k is a required parameter */
if (parameters.k == 0) {
return 0;
}
/* k <= maxDictSize */
if (parameters.k > maxDictSize) {
return 0;
}
return 1;
}
/**
* Given the prepared context build the dictionary.
*/
static size_t RANDOM_buildDictionary(const size_t totalSamplesSize, const BYTE *samples,
void *dictBuffer, size_t dictBufferCapacity,
ZDICT_random_params_t parameters) {
BYTE *const dict = (BYTE *)dictBuffer;
size_t tail = dictBufferCapacity;
const int displayLevel = parameters.zParams.notificationLevel;
while (tail > 0) {
/* Select a segment */
RANDOM_segment_t segment = RANDOM_selectSegment(totalSamplesSize, parameters);
size_t segmentSize;
segmentSize = MIN(segment.end - segment.begin + 1, tail);
tail -= segmentSize;
memcpy(dict + tail, samples + segment.begin, segmentSize);
DISPLAYUPDATE(
2, "\r%u%% ",
(U32)(((dictBufferCapacity - tail) * 100) / dictBufferCapacity));
}
return tail;
}
ZDICTLIB_API size_t ZDICT_trainFromBuffer_random(
void *dictBuffer, size_t dictBufferCapacity,
const void *samplesBuffer, const size_t *samplesSizes, unsigned nbSamples,
ZDICT_random_params_t parameters) {
const int displayLevel = parameters.zParams.notificationLevel;
BYTE* const dict = (BYTE*)dictBuffer;
/* Checks */
if (!RANDOM_checkParameters(parameters, dictBufferCapacity)) {
DISPLAYLEVEL(1, "k is incorrect\n");
return ERROR(GENERIC);
}
if (nbSamples == 0) {
DISPLAYLEVEL(1, "Random must have at least one input file\n");
return ERROR(GENERIC);
}
if (dictBufferCapacity < ZDICT_DICTSIZE_MIN) {
DISPLAYLEVEL(1, "dictBufferCapacity must be at least %u\n",
ZDICT_DICTSIZE_MIN);
return ERROR(dstSize_tooSmall);
}
const size_t totalSamplesSize = RANDOM_sum(samplesSizes, nbSamples);
const BYTE *const samples = (const BYTE *)samplesBuffer;
DISPLAYLEVEL(2, "Building dictionary\n");
{
const size_t tail = RANDOM_buildDictionary(totalSamplesSize, samples,
dictBuffer, dictBufferCapacity, parameters);
const size_t dictSize = ZDICT_finalizeDictionary(
dict, dictBufferCapacity, dict + tail, dictBufferCapacity - tail,
samplesBuffer, samplesSizes, nbSamples, parameters.zParams);
if (!ZSTD_isError(dictSize)) {
DISPLAYLEVEL(2, "Constructed dictionary of size %u\n",
(U32)dictSize);
}
return dictSize;
}
}

29
contrib/randomDictBuilder/random.h Normal file
View File

@ -0,0 +1,29 @@
#include <stdio.h> /* fprintf */
#include <stdlib.h> /* malloc, free, qsort */
#include <string.h> /* memset */
#include <time.h> /* clock */
#include "zstd_internal.h" /* includes zstd.h */
#ifndef ZDICT_STATIC_LINKING_ONLY
#define ZDICT_STATIC_LINKING_ONLY
#endif
#include "zdict.h"
typedef struct {
unsigned k; /* Segment size : constraint: 0 < k : Reasonable range [16, 2048+]; Default to 200 */
ZDICT_params_t zParams;
} ZDICT_random_params_t;
/*! ZDICT_trainFromBuffer_random():
* Train a dictionary from an array of samples.
* Samples must be stored concatenated in a single flat buffer `samplesBuffer`,
* supplied with an array of sizes `samplesSizes`, providing the size of each sample, in order.
* The resulting dictionary will be saved into `dictBuffer`.
* @return: size of dictionary stored into `dictBuffer` (<= `dictBufferCapacity`)
* or an error code, which can be tested with ZDICT_isError().
*/
ZDICTLIB_API size_t ZDICT_trainFromBuffer_random( void *dictBuffer, size_t dictBufferCapacity,
const void *samplesBuffer, const size_t *samplesSizes, unsigned nbSamples,
ZDICT_random_params_t parameters);

14
contrib/randomDictBuilder/test.sh Normal file
View File

@ -0,0 +1,14 @@
echo "Building random dictionary with in=../../lib/common k=200 out=dict1"
./main in=../../lib/common k=200 out=dict1
zstd -be3 -D dict1 -r ../../lib/common -q
echo "Building random dictionary with in=../../lib/common k=500 out=dict2 dictID=100 maxdict=140000"
./main in=../../lib/common k=500 out=dict2 dictID=100 maxdict=140000
zstd -be3 -D dict2 -r ../../lib/common -q
echo "Building random dictionary with 2 sample sources"
./main in=../../lib/common in=../../lib/compress out=dict3
zstd -be3 -D dict3 -r ../../lib/common -q
echo "Removing dict1 dict2 dict3"
rm -f dict1 dict2 dict3
echo "Testing with invalid parameters, should fail"
! ./main r=10