andreacavalli/rocksdb
0
0
Fork 0
Code Issues Pull Requests Projects Releases Wiki Activity

Support using ZDICT_finalizeDictionary to generate zstd dictionary (#9857)

Browse Source 
Summary:
An untrained dictionary is currently simply the concatenation of several samples. The ZSTD API, ZDICT_finalizeDictionary(), can improve such a dictionary's effectiveness at low cost. This PR changes how dictionary is created by calling the ZSTD ZDICT_finalizeDictionary() API instead of creating raw content dictionary (when max_dict_buffer_bytes > 0), and pass in all buffered uncompressed data blocks as samples.

Pull Request resolved: https://github.com/facebook/rocksdb/pull/9857

Test Plan:
#### db_bench test for cpu/memory of compression+decompression and space saving on synthetic data:
Set up: change the parameter [here](fb9a167a55/tools/db_bench_tool.cc (L1766)) to 16384 to make synthetic data more compressible.
```
# linked local ZSTD with version 1.5.2
# DEBUG_LEVEL=0 ROCKSDB_NO_FBCODE=1 ROCKSDB_DISABLE_ZSTD=1  EXTRA_CXXFLAGS="-DZSTD_STATIC_LINKING_ONLY -DZSTD -I/data/users/changyubi/install/include/" EXTRA_LDFLAGS="-L/data/users/changyubi/install/lib/ -l:libzstd.a" make -j32 db_bench

dict_bytes=16384
train_bytes=1048576
echo "========== No Dictionary =========="
TEST_TMPDIR=/dev/shm ./db_bench -benchmarks=filluniquerandom,compact -num=10000000 -compression_type=zstd -compression_max_dict_bytes=0 -block_size=4096 -max_background_jobs=24 -memtablerep=vector -allow_concurrent_memtable_write=false -disable_wal=true -max_write_buffer_number=8 >/dev/null 2>&1
TEST_TMPDIR=/dev/shm /usr/bin/time ./db_bench -use_existing_db=true -benchmarks=compact -compression_type=zstd -compression_max_dict_bytes=0 -block_size=4096 2>&1 | grep elapsed
du -hc /dev/shm/dbbench/*sst | grep total

echo "========== Raw Content Dictionary =========="
TEST_TMPDIR=/dev/shm ./db_bench_main -benchmarks=filluniquerandom,compact -num=10000000 -compression_type=zstd -compression_max_dict_bytes=$dict_bytes -block_size=4096 -max_background_jobs=24 -memtablerep=vector -allow_concurrent_memtable_write=false -disable_wal=true -max_write_buffer_number=8 >/dev/null 2>&1
TEST_TMPDIR=/dev/shm /usr/bin/time ./db_bench_main -use_existing_db=true -benchmarks=compact -compression_type=zstd -compression_max_dict_bytes=$dict_bytes -block_size=4096 2>&1 | grep elapsed
du -hc /dev/shm/dbbench/*sst | grep total

echo "========== FinalizeDictionary =========="
TEST_TMPDIR=/dev/shm ./db_bench -benchmarks=filluniquerandom,compact -num=10000000 -compression_type=zstd -compression_max_dict_bytes=$dict_bytes -compression_zstd_max_train_bytes=$train_bytes -compression_use_zstd_dict_trainer=false -block_size=4096 -max_background_jobs=24 -memtablerep=vector -allow_concurrent_memtable_write=false -disable_wal=true -max_write_buffer_number=8 >/dev/null 2>&1
TEST_TMPDIR=/dev/shm /usr/bin/time ./db_bench -use_existing_db=true -benchmarks=compact -compression_type=zstd -compression_max_dict_bytes=$dict_bytes -compression_zstd_max_train_bytes=$train_bytes -compression_use_zstd_dict_trainer=false -block_size=4096 2>&1 | grep elapsed
du -hc /dev/shm/dbbench/*sst | grep total

echo "========== TrainDictionary =========="
TEST_TMPDIR=/dev/shm ./db_bench -benchmarks=filluniquerandom,compact -num=10000000 -compression_type=zstd -compression_max_dict_bytes=$dict_bytes -compression_zstd_max_train_bytes=$train_bytes -block_size=4096 -max_background_jobs=24 -memtablerep=vector -allow_concurrent_memtable_write=false -disable_wal=true -max_write_buffer_number=8 >/dev/null 2>&1
TEST_TMPDIR=/dev/shm /usr/bin/time ./db_bench -use_existing_db=true -benchmarks=compact -compression_type=zstd -compression_max_dict_bytes=$dict_bytes -compression_zstd_max_train_bytes=$train_bytes -block_size=4096 2>&1 | grep elapsed
du -hc /dev/shm/dbbench/*sst | grep total

# Result: TrainDictionary is much better on space saving, but FinalizeDictionary seems to use less memory.
# before compression data size: 1.2GB
dict_bytes=16384
max_dict_buffer_bytes =  1048576
                    space   cpu/memory
No Dictionary       468M    14.93user 1.00system 0:15.92elapsed 100%CPU (0avgtext+0avgdata 23904maxresident)k
Raw Dictionary      251M    15.81user 0.80system 0:16.56elapsed 100%CPU (0avgtext+0avgdata 156808maxresident)k
FinalizeDictionary  236M    11.93user 0.64system 0:12.56elapsed 100%CPU (0avgtext+0avgdata 89548maxresident)k
TrainDictionary     84M     7.29user 0.45system 0:07.75elapsed 100%CPU (0avgtext+0avgdata 97288maxresident)k
```

#### Benchmark on 10 sample SST files for spacing saving and CPU time on compression:
FinalizeDictionary is comparable to TrainDictionary in terms of space saving, and takes less time in compression.
```
dict_bytes=16384
train_bytes=1048576

for sst_file in `ls ../temp/myrock-sst/`
do
  echo "********** $sst_file **********"
  echo "========== No Dictionary =========="
  ./sst_dump --file="../temp/myrock-sst/$sst_file" --command=recompress --compression_level_from=6 --compression_level_to=6 --compression_types=kZSTD

  echo "========== Raw Content Dictionary =========="
  ./sst_dump --file="../temp/myrock-sst/$sst_file" --command=recompress --compression_level_from=6 --compression_level_to=6 --compression_types=kZSTD --compression_max_dict_bytes=$dict_bytes

  echo "========== FinalizeDictionary =========="
  ./sst_dump --file="../temp/myrock-sst/$sst_file" --command=recompress --compression_level_from=6 --compression_level_to=6 --compression_types=kZSTD --compression_max_dict_bytes=$dict_bytes --compression_zstd_max_train_bytes=$train_bytes --compression_use_zstd_finalize_dict

  echo "========== TrainDictionary =========="
  ./sst_dump --file="../temp/myrock-sst/$sst_file" --command=recompress --compression_level_from=6 --compression_level_to=6 --compression_types=kZSTD --compression_max_dict_bytes=$dict_bytes --compression_zstd_max_train_bytes=$train_bytes
done

                         010240.sst (Size/Time) 011029.sst              013184.sst              021552.sst              185054.sst              185137.sst              191666.sst              7560381.sst             7604174.sst             7635312.sst
No Dictionary           28165569 / 2614419      32899411 / 2976832      32977848 / 3055542      31966329 / 2004590      33614351 / 1755877      33429029 / 1717042      33611933 / 1776936      33634045 / 2771417      33789721 / 2205414      33592194 / 388254
Raw Content Dictionary  28019950 / 2697961      33748665 / 3572422      33896373 / 3534701      26418431 / 2259658      28560825 / 1839168      28455030 / 1846039      28494319 / 1861349      32391599 / 3095649      33772142 / 2407843      33592230 / 474523
FinalizeDictionary      27896012 / 2650029      33763886 / 3719427      33904283 / 3552793      26008225 / 2198033      28111872 / 1869530      28014374 / 1789771      28047706 / 1848300      32296254 / 3204027      33698698 / 2381468      33592344 / 517433
TrainDictionary         28046089 / 2740037      33706480 / 3679019      33885741 / 3629351      25087123 / 2204558      27194353 / 1970207      27234229 / 1896811      27166710 / 1903119      32011041 / 3322315      32730692 / 2406146      33608631 / 570593
```

#### Decompression/Read test:
With FinalizeDictionary/TrainDictionary, some data structure used for decompression are in stored in dictionary, so they are expected to be faster in terms of decompression/reads.
```
dict_bytes=16384
train_bytes=1048576
echo "No Dictionary"
TEST_TMPDIR=/dev/shm/ ./db_bench -benchmarks=filluniquerandom,compact -compression_type=zstd -compression_max_dict_bytes=0 > /dev/null 2>&1
TEST_TMPDIR=/dev/shm/ ./db_bench -use_existing_db=true -benchmarks=readrandom -cache_size=0 -compression_type=zstd -compression_max_dict_bytes=0 2>&1 | grep MB/s

echo "Raw Dictionary"
TEST_TMPDIR=/dev/shm/ ./db_bench -benchmarks=filluniquerandom,compact -compression_type=zstd -compression_max_dict_bytes=$dict_bytes > /dev/null 2>&1
TEST_TMPDIR=/dev/shm/ ./db_bench -use_existing_db=true -benchmarks=readrandom -cache_size=0 -compression_type=zstd  -compression_max_dict_bytes=$dict_bytes 2>&1 | grep MB/s

echo "FinalizeDict"
TEST_TMPDIR=/dev/shm/ ./db_bench -benchmarks=filluniquerandom,compact -compression_type=zstd -compression_max_dict_bytes=$dict_bytes -compression_zstd_max_train_bytes=$train_bytes -compression_use_zstd_dict_trainer=false  > /dev/null 2>&1
TEST_TMPDIR=/dev/shm/ ./db_bench -use_existing_db=true -benchmarks=readrandom -cache_size=0 -compression_type=zstd -compression_max_dict_bytes=$dict_bytes -compression_zstd_max_train_bytes=$train_bytes -compression_use_zstd_dict_trainer=false 2>&1 | grep MB/s

echo "Train Dictionary"
TEST_TMPDIR=/dev/shm/ ./db_bench -benchmarks=filluniquerandom,compact -compression_type=zstd -compression_max_dict_bytes=$dict_bytes -compression_zstd_max_train_bytes=$train_bytes > /dev/null 2>&1
TEST_TMPDIR=/dev/shm/ ./db_bench -use_existing_db=true -benchmarks=readrandom -cache_size=0 -compression_type=zstd -compression_max_dict_bytes=$dict_bytes -compression_zstd_max_train_bytes=$train_bytes 2>&1 | grep MB/s

No Dictionary
readrandom   :      12.183 micros/op 82082 ops/sec 12.183 seconds 1000000 operations;    9.1 MB/s (1000000 of 1000000 found)
Raw Dictionary
readrandom   :      12.314 micros/op 81205 ops/sec 12.314 seconds 1000000 operations;    9.0 MB/s (1000000 of 1000000 found)
FinalizeDict
readrandom   :       9.787 micros/op 102180 ops/sec 9.787 seconds 1000000 operations;   11.3 MB/s (1000000 of 1000000 found)
Train Dictionary
readrandom   :       9.698 micros/op 103108 ops/sec 9.699 seconds 1000000 operations;   11.4 MB/s (1000000 of 1000000 found)
```

Reviewed By: ajkr

Differential Revision: D35720026

Pulled By: cbi42

fbshipit-source-id: 24d230fdff0fd28a1bb650658798f00dfcfb2a1f
This commit is contained in:
Changyu Bi 2022-05-20 12:09:09 -07:00 committed by Facebook GitHub Bot
parent 6255ac7223
commit cc23b46da1
22 changed files with 274 additions and 25 deletions
Show Stats Download Patch File Download Diff File Expand all files Collapse all files

1
HISTORY.md
View File

@ -26,6 +26,7 @@
* Changed `GetUniqueIdFromTableProperties` to return a 128-bit unique identifier, which will be the standard size now. The old functionality (192-bit) is available from `GetExtendedUniqueIdFromTableProperties`. Both functions are no longer "experimental" and are ready for production use.
* In IOOptions, mark `prio` as deprecated for future removal.
* In `file_system.h`, mark `IOPriority` as deprecated for future removal.
* Add an option, `CompressionOptions::use_zstd_dict_trainer`, to indicate whether zstd dictionary trainer should be used for generating zstd compression dictionaries. The default value of this option is true for backward compatibility. When this option is set to false, zstd API `ZDICT_finalizeDictionary` is used to generate compression dictionaries.
### Bug Fixes
* RocksDB calls FileSystem::Poll API during FilePrefetchBuffer destruction which impacts performance as it waits for read requets completion which is not needed anymore. Calling FileSystem::AbortIO to abort those requests instead fixes that performance issue.

24
db/c.cc
View File

@ -2855,6 +2855,20 @@ void rocksdb_options_set_bottommost_compression_options_zstd_max_train_bytes(
opt->rep.bottommost_compression_opts.enabled = enabled;
}
void rocksdb_options_set_bottommost_compression_options_use_zstd_dict_trainer(
rocksdb_options_t* opt, unsigned char use_zstd_dict_trainer,
unsigned char enabled) {
opt->rep.bottommost_compression_opts.use_zstd_dict_trainer =
use_zstd_dict_trainer;
opt->rep.bottommost_compression_opts.enabled = enabled;
}
unsigned char
rocksdb_options_get_bottommost_compression_options_use_zstd_dict_trainer(
rocksdb_options_t* opt) {
return opt->rep.bottommost_compression_opts.use_zstd_dict_trainer;
}
void rocksdb_options_set_bottommost_compression_options_max_dict_buffer_bytes(
rocksdb_options_t* opt, uint64_t max_dict_buffer_bytes,
unsigned char enabled) {
@ -2882,6 +2896,16 @@ int rocksdb_options_get_compression_options_zstd_max_train_bytes(
return opt->rep.compression_opts.zstd_max_train_bytes;
}
void rocksdb_options_set_compression_options_use_zstd_dict_trainer(
rocksdb_options_t* opt, unsigned char use_zstd_dict_trainer) {
opt->rep.compression_opts.use_zstd_dict_trainer = use_zstd_dict_trainer;
}
unsigned char rocksdb_options_get_compression_options_use_zstd_dict_trainer(
rocksdb_options_t* opt) {
return opt->rep.compression_opts.use_zstd_dict_trainer;
}
void rocksdb_options_set_compression_options_parallel_threads(
rocksdb_options_t* opt, int value) {
opt->rep.compression_opts.parallel_threads = value;

3
db/c_test.c
View File

@ -2539,6 +2539,9 @@ int main(int argc, char** argv) {
200 ==
rocksdb_options_get_compression_options_max_dict_buffer_bytes(co));
rocksdb_options_set_compression_options_use_zstd_dict_trainer(co, 0);
CheckCondition(
0 == rocksdb_options_get_compression_options_use_zstd_dict_trainer(co));
rocksdb_options_destroy(co);
}

6
db/column_family.cc
View File

@ -136,11 +136,17 @@ Status CheckCompressionSupported(const ColumnFamilyOptions& cf_options) {
}
}
if (cf_options.compression_opts.zstd_max_train_bytes > 0) {
if (cf_options.compression_opts.use_zstd_dict_trainer) {
if (!ZSTD_TrainDictionarySupported()) {
return Status::InvalidArgument(
"zstd dictionary trainer cannot be used because ZSTD 1.1.3+ "
"is not linked with the binary.");
}
} else if (!ZSTD_FinalizeDictionarySupported()) {
return Status::InvalidArgument(
"zstd finalizeDictionary cannot be used because ZSTD 1.4.5+ "
"is not linked with the binary.");
}
if (cf_options.compression_opts.max_dict_bytes == 0) {
return Status::InvalidArgument(
"The dictionary size limit (`CompressionOptions::max_dict_bytes`) "

20
db/db_test2.cc
View File

@ -1296,6 +1296,7 @@ TEST_F(DBTest2, PresetCompressionDict) {
enum DictionaryTypes : int {
kWithoutDict,
kWithDict,
kWithZSTDfinalizeDict,
kWithZSTDTrainedDict,
kDictEnd,
};
@ -1304,6 +1305,7 @@ TEST_F(DBTest2, PresetCompressionDict) {
options.compression = compression_type;
size_t bytes_without_dict = 0;
size_t bytes_with_dict = 0;
size_t bytes_with_zstd_finalize_dict = 0;
size_t bytes_with_zstd_trained_dict = 0;
for (int i = kWithoutDict; i < kDictEnd; i++) {
// First iteration: compress without preset dictionary
@ -1323,12 +1325,21 @@ TEST_F(DBTest2, PresetCompressionDict) {
options.compression_opts.max_dict_bytes = kBlockSizeBytes;
options.compression_opts.zstd_max_train_bytes = 0;
break;
case kWithZSTDfinalizeDict:
if (compression_type != kZSTD) {
continue;
}
options.compression_opts.max_dict_bytes = kBlockSizeBytes;
options.compression_opts.zstd_max_train_bytes = kL0FileBytes;
options.compression_opts.use_zstd_dict_trainer = false;
break;
case kWithZSTDTrainedDict:
if (compression_type != kZSTD) {
continue;
}
options.compression_opts.max_dict_bytes = kBlockSizeBytes;
options.compression_opts.zstd_max_train_bytes = kL0FileBytes;
options.compression_opts.use_zstd_dict_trainer = true;
break;
default:
assert(false);
@ -1365,6 +1376,8 @@ TEST_F(DBTest2, PresetCompressionDict) {
bytes_without_dict = total_sst_bytes;
} else if (i == kWithDict) {
bytes_with_dict = total_sst_bytes;
} else if (i == kWithZSTDfinalizeDict) {
bytes_with_zstd_finalize_dict = total_sst_bytes;
} else if (i == kWithZSTDTrainedDict) {
bytes_with_zstd_trained_dict = total_sst_bytes;
}
@ -1375,6 +1388,13 @@ TEST_F(DBTest2, PresetCompressionDict) {
}
if (i == kWithDict) {
ASSERT_GT(bytes_without_dict, bytes_with_dict);
} else if (i == kWithZSTDTrainedDict) {
// In zstd compression, it is sometimes possible that using a finalized
// dictionary does not get as good a compression ratio as raw content
// dictionary. But using a dictionary should always get better
// compression ratio than not using one.
ASSERT_TRUE(bytes_with_dict > bytes_with_zstd_finalize_dict ||
bytes_without_dict > bytes_with_zstd_finalize_dict);
} else if (i == kWithZSTDTrainedDict) {
// In zstd compression, it is sometimes possible that using a trained
// dictionary does not get as good a compression ratio as without

1
db_stress_tool/db_stress_common.h
View File

@ -221,6 +221,7 @@ DECLARE_int32(compression_max_dict_bytes);
DECLARE_int32(compression_zstd_max_train_bytes);
DECLARE_int32(compression_parallel_threads);
DECLARE_uint64(compression_max_dict_buffer_bytes);
DECLARE_bool(compression_use_zstd_dict_trainer);
DECLARE_string(checksum_type);
DECLARE_string(env_uri);
DECLARE_string(fs_uri);

7
db_stress_tool/db_stress_gflags.cc
View File

@ -752,6 +752,13 @@ DEFINE_uint64(compression_max_dict_buffer_bytes, 0,
"Buffering limit for SST file data to sample for dictionary "
"compression.");
DEFINE_bool(
compression_use_zstd_dict_trainer, true,
"Use zstd's trainer to generate dictionary. If the options is false, "
"zstd's finalizeDictionary() API is used to generate dictionary. "
"ZSTD 1.4.5+ is required. If ZSTD 1.4.5+ is not linked with the binary, "
"this flag will have the default value true.");
DEFINE_string(bottommost_compression_type, "disable",
"Algorithm to use to compress bottommost level of the database. "
"\"disable\" means disabling the feature");

11
db_stress_tool/db_stress_test_base.cc
View File

@ -8,6 +8,7 @@
// found in the LICENSE file. See the AUTHORS file for names of contributors.
//
#include "util/compression.h"
#ifdef GFLAGS
#include "db_stress_tool/db_stress_common.h"
#include "db_stress_tool/db_stress_compaction_filter.h"
@ -2915,6 +2916,16 @@ void InitializeOptionsFromFlags(
FLAGS_compression_parallel_threads;
options.compression_opts.max_dict_buffer_bytes =
FLAGS_compression_max_dict_buffer_bytes;
if (ZSTD_FinalizeDictionarySupported()) {
options.compression_opts.use_zstd_dict_trainer =
FLAGS_compression_use_zstd_dict_trainer;
} else if (!FLAGS_compression_use_zstd_dict_trainer) {
fprintf(
stderr,
"WARNING: use_zstd_dict_trainer is false but zstd finalizeDictionary "
"cannot be used because ZSTD 1.4.5+ is not linked with the binary."
" zstd dictionary trainer will be used.\n");
}
options.max_manifest_file_size = FLAGS_max_manifest_file_size;
options.inplace_update_support = FLAGS_in_place_update;
options.max_subcompactions = static_cast<uint32_t>(FLAGS_subcompactions);

28
include/rocksdb/advanced_options.h
View File

@ -100,8 +100,9 @@ struct CompressionOptions {
//
// The dictionary is created by sampling the SST file data. If
// `zstd_max_train_bytes` is nonzero, the samples are passed through zstd's
// dictionary generator. Otherwise, the random samples are used directly as
// the dictionary.
// dictionary generator (see comments for option `use_zstd_dict_trainer` for
// detail on dictionary generator). If `zstd_max_train_bytes` is zero, the
// random samples are used directly as the dictionary.
//
// When compression dictionary is disabled, we compress and write each block
// before buffering data for the next one. When compression dictionary is
@ -173,6 +174,20 @@ struct CompressionOptions {
// Default: 0 (unlimited)
uint64_t max_dict_buffer_bytes;
// Use zstd trainer to generate dictionaries. When this option is set to true,
// zstd_max_train_bytes of training data sampled from max_dict_buffer_bytes
// buffered data will be passed to zstd dictionary trainer to generate a
// dictionary of size max_dict_bytes.
//
// When this option is false, zstd's API ZDICT_finalizeDictionary() will be
// called to generate dictionaries. zstd_max_train_bytes of training sampled
// data will be passed to this API. Using this API should save CPU time on
// dictionary training, but the compression ratio may not be as good as using
// a dictionary trainer.
//
// Default: true
bool use_zstd_dict_trainer;
CompressionOptions()
: window_bits(-14),
level(kDefaultCompressionLevel),
@ -181,11 +196,13 @@ struct CompressionOptions {
zstd_max_train_bytes(0),
parallel_threads(1),
enabled(false),
max_dict_buffer_bytes(0) {}
max_dict_buffer_bytes(0),
use_zstd_dict_trainer(true) {}
CompressionOptions(int wbits, int _lev, int _strategy,
uint32_t _max_dict_bytes, uint32_t _zstd_max_train_bytes,
uint32_t _parallel_threads, bool _enabled,
uint64_t _max_dict_buffer_bytes)
uint64_t _max_dict_buffer_bytes,
bool _use_zstd_dict_trainer)
: window_bits(wbits),
level(_lev),
strategy(_strategy),
@ -193,7 +210,8 @@ struct CompressionOptions {
zstd_max_train_bytes(_zstd_max_train_bytes),
parallel_threads(_parallel_threads),
enabled(_enabled),
max_dict_buffer_bytes(_max_dict_buffer_bytes) {}
max_dict_buffer_bytes(_max_dict_buffer_bytes),
use_zstd_dict_trainer(_use_zstd_dict_trainer) {}
};
// Temperature of a file. Used to pass to FileSystem for a different

12
include/rocksdb/c.h
View File

@ -1041,6 +1041,12 @@ extern ROCKSDB_LIBRARY_API int
rocksdb_options_get_compression_options_zstd_max_train_bytes(
rocksdb_options_t* opt);
extern ROCKSDB_LIBRARY_API void
rocksdb_options_set_compression_options_use_zstd_dict_trainer(
rocksdb_options_t*, unsigned char);
extern ROCKSDB_LIBRARY_API unsigned char
rocksdb_options_get_compression_options_use_zstd_dict_trainer(
rocksdb_options_t* opt);
extern ROCKSDB_LIBRARY_API void
rocksdb_options_set_compression_options_parallel_threads(rocksdb_options_t*,
int);
extern ROCKSDB_LIBRARY_API int
@ -1059,6 +1065,12 @@ extern ROCKSDB_LIBRARY_API void
rocksdb_options_set_bottommost_compression_options_zstd_max_train_bytes(
rocksdb_options_t*, int, unsigned char);
extern ROCKSDB_LIBRARY_API void
rocksdb_options_set_bottommost_compression_options_use_zstd_dict_trainer(
rocksdb_options_t*, unsigned char, unsigned char);
extern ROCKSDB_LIBRARY_API unsigned char
rocksdb_options_get_bottommost_compression_options_use_zstd_dict_trainer(
rocksdb_options_t* opt);
extern ROCKSDB_LIBRARY_API void
rocksdb_options_set_bottommost_compression_options_max_dict_buffer_bytes(
rocksdb_options_t*, uint64_t, unsigned char);
extern ROCKSDB_LIBRARY_API void rocksdb_options_set_prefix_extractor(

24
java/rocksjni/compression_options.cc
View File

@ -154,6 +154,30 @@ jlong Java_org_rocksdb_CompressionOptions_maxDictBufferBytes(JNIEnv*, jobject,
auto* opt = reinterpret_cast<ROCKSDB_NAMESPACE::CompressionOptions*>(jhandle);
return static_cast<jlong>(opt->max_dict_buffer_bytes);
}
/*
* Class: org_rocksdb_CompressionOptions
* Method: setZstdMaxTrainBytes
* Signature: (JZ)V
*/
void Java_org_rocksdb_CompressionOptions_setUseZstdDictTrainer(
JNIEnv*, jobject, jlong jhandle, jboolean juse_zstd_dict_trainer) {
auto* opt = reinterpret_cast<ROCKSDB_NAMESPACE::CompressionOptions*>(jhandle);
opt->use_zstd_dict_trainer = juse_zstd_dict_trainer == JNI_TRUE;
}
/*
* Class: org_rocksdb_CompressionOptions
* Method: zstdMaxTrainBytes
* Signature: (J)Z
*/
jboolean Java_org_rocksdb_CompressionOptions_useZstdDictTrainer(JNIEnv*,
jobject,
jlong jhandle) {
auto* opt = reinterpret_cast<ROCKSDB_NAMESPACE::CompressionOptions*>(jhandle);
return static_cast<bool>(opt->use_zstd_dict_trainer);
}
/*
* Class: org_rocksdb_CompressionOptions
* Method: setEnabled

13
options/cf_options.cc
View File

@ -116,6 +116,15 @@ static Status ParseCompressionOptions(const std::string& value,
compression_opts.max_dict_buffer_bytes = ParseUint64(field);
}
// use_zstd_dict_trainer is optional for backwards compatibility
if (!field_stream.eof()) {
if (!std::getline(field_stream, field, kDelimiter)) {
return Status::InvalidArgument(
"unable to parse the specified CF option " + name);
}
compression_opts.use_zstd_dict_trainer = ParseBoolean("", field);
}
if (!field_stream.eof()) {
return Status::InvalidArgument("unable to parse the specified CF option " +
name);
@ -156,6 +165,10 @@ static std::unordered_map<std::string, OptionTypeInfo>
{offsetof(struct CompressionOptions, max_dict_buffer_bytes),
OptionType::kUInt64T, OptionVerificationType::kNormal,
OptionTypeFlags::kMutable}},
{"use_zstd_dict_trainer",
{offsetof(struct CompressionOptions, use_zstd_dict_trainer),
OptionType::kBoolean, OptionVerificationType::kNormal,
OptionTypeFlags::kMutable}},
};
static std::unordered_map<std::string, OptionTypeInfo>

7
options/options.cc
View File

@ -211,6 +211,10 @@ void ColumnFamilyOptions::Dump(Logger* log) const {
" Options.bottommost_compression_opts.max_dict_buffer_bytes: "
"%" PRIu64,
bottommost_compression_opts.max_dict_buffer_bytes);
ROCKS_LOG_HEADER(
log,
" Options.bottommost_compression_opts.use_zstd_dict_trainer: %s",
bottommost_compression_opts.use_zstd_dict_trainer ? "true" : "false");
ROCKS_LOG_HEADER(log, " Options.compression_opts.window_bits: %d",
compression_opts.window_bits);
ROCKS_LOG_HEADER(log, " Options.compression_opts.level: %d",
@ -225,6 +229,9 @@ void ColumnFamilyOptions::Dump(Logger* log) const {
" Options.compression_opts.zstd_max_train_bytes: "
"%" PRIu32,
compression_opts.zstd_max_train_bytes);
ROCKS_LOG_HEADER(
log, " Options.compression_opts.use_zstd_dict_trainer: %s",
compression_opts.use_zstd_dict_trainer ? "true" : "false");
ROCKS_LOG_HEADER(log,
" Options.compression_opts.parallel_threads: "
"%" PRIu32,

4
options/options_settable_test.cc
View File

@ -481,8 +481,8 @@ TEST_F(OptionsSettableTest, ColumnFamilyOptionsAllFieldsSettable) {
"max_bytes_for_level_multiplier=60;"
"memtable_factory=SkipListFactory;"
"compression=kNoCompression;"
"compression_opts=5:6:7:8:9:10:true:11;"
"bottommost_compression_opts=4:5:6:7:8:9:true:10;"
"compression_opts=5:6:7:8:9:10:true:11:false;"
"bottommost_compression_opts=4:5:6:7:8:9:true:10:true;"
"bottommost_compression=kDisableCompressionOption;"
"level0_stop_writes_trigger=33;"
"num_levels=99;"

49
options/options_test.cc
View File

@ -68,7 +68,7 @@ TEST_F(OptionsTest, GetOptionsFromMapTest) {
"kZSTDNotFinalCompression"},
{"bottommost_compression", "kLZ4Compression"},
{"bottommost_compression_opts", "5:6:7:8:10:true"},
{"compression_opts", "4:5:6:7:8:true"},
{"compression_opts", "4:5:6:7:8:2:true:100:false"},
{"num_levels", "8"},
{"level0_file_num_compaction_trigger", "8"},
{"level0_slowdown_writes_trigger", "9"},
@ -191,9 +191,10 @@ TEST_F(OptionsTest, GetOptionsFromMapTest) {
ASSERT_EQ(new_cf_opt.compression_opts.strategy, 6);
ASSERT_EQ(new_cf_opt.compression_opts.max_dict_bytes, 7u);
ASSERT_EQ(new_cf_opt.compression_opts.zstd_max_train_bytes, 8u);
ASSERT_EQ(new_cf_opt.compression_opts.parallel_threads,
CompressionOptions().parallel_threads);
ASSERT_EQ(new_cf_opt.compression_opts.parallel_threads, 2u);
ASSERT_EQ(new_cf_opt.compression_opts.enabled, true);
ASSERT_EQ(new_cf_opt.compression_opts.max_dict_buffer_bytes, 100u);
ASSERT_EQ(new_cf_opt.compression_opts.use_zstd_dict_trainer, false);
ASSERT_EQ(new_cf_opt.bottommost_compression, kLZ4Compression);
ASSERT_EQ(new_cf_opt.bottommost_compression_opts.window_bits, 5);
ASSERT_EQ(new_cf_opt.bottommost_compression_opts.level, 6);
@ -203,6 +204,8 @@ TEST_F(OptionsTest, GetOptionsFromMapTest) {
ASSERT_EQ(new_cf_opt.bottommost_compression_opts.parallel_threads,
CompressionOptions().parallel_threads);
ASSERT_EQ(new_cf_opt.bottommost_compression_opts.enabled, true);
ASSERT_EQ(new_cf_opt.bottommost_compression_opts.use_zstd_dict_trainer,
CompressionOptions().use_zstd_dict_trainer);
ASSERT_EQ(new_cf_opt.num_levels, 8);
ASSERT_EQ(new_cf_opt.level0_file_num_compaction_trigger, 8);
ASSERT_EQ(new_cf_opt.level0_slowdown_writes_trigger, 9);
@ -604,6 +607,8 @@ TEST_F(OptionsTest, CompressionOptionsFromString) {
ASSERT_EQ(base_cf_opt.compression_opts.parallel_threads,
dflt.parallel_threads);
ASSERT_EQ(base_cf_opt.compression_opts.enabled, dflt.enabled);
ASSERT_EQ(base_cf_opt.compression_opts.use_zstd_dict_trainer,
dflt.use_zstd_dict_trainer);
ASSERT_EQ(base_cf_opt.bottommost_compression_opts.window_bits, 4);
ASSERT_EQ(base_cf_opt.bottommost_compression_opts.level, 5);
ASSERT_EQ(base_cf_opt.bottommost_compression_opts.strategy, 6);
@ -613,10 +618,12 @@ TEST_F(OptionsTest, CompressionOptionsFromString) {
ASSERT_EQ(base_cf_opt.bottommost_compression_opts.parallel_threads,
dflt.parallel_threads);
ASSERT_EQ(base_cf_opt.bottommost_compression_opts.enabled, dflt.enabled);
ASSERT_EQ(base_cf_opt.bottommost_compression_opts.use_zstd_dict_trainer,
dflt.use_zstd_dict_trainer);
ASSERT_OK(GetColumnFamilyOptionsFromString(
config_options, ColumnFamilyOptions(),
"compression_opts=4:5:6:7:8:9:true; "
"compression_opts=4:5:6:7:8:9:true:10:false; "
"bottommost_compression_opts=5:6:7:8:9:false",
&base_cf_opt));
ASSERT_EQ(base_cf_opt.compression_opts.window_bits, 4);
@ -626,6 +633,8 @@ TEST_F(OptionsTest, CompressionOptionsFromString) {
ASSERT_EQ(base_cf_opt.compression_opts.zstd_max_train_bytes, 8u);
ASSERT_EQ(base_cf_opt.compression_opts.parallel_threads, 9u);
ASSERT_EQ(base_cf_opt.compression_opts.enabled, true);
ASSERT_EQ(base_cf_opt.compression_opts.max_dict_buffer_bytes, 10u);
ASSERT_EQ(base_cf_opt.compression_opts.use_zstd_dict_trainer, false);
ASSERT_EQ(base_cf_opt.bottommost_compression_opts.window_bits, 5);
ASSERT_EQ(base_cf_opt.bottommost_compression_opts.level, 6);
ASSERT_EQ(base_cf_opt.bottommost_compression_opts.strategy, 7);
@ -634,6 +643,8 @@ TEST_F(OptionsTest, CompressionOptionsFromString) {
ASSERT_EQ(base_cf_opt.bottommost_compression_opts.parallel_threads,
dflt.parallel_threads);
ASSERT_EQ(base_cf_opt.bottommost_compression_opts.enabled, false);
ASSERT_EQ(base_cf_opt.bottommost_compression_opts.use_zstd_dict_trainer,
dflt.use_zstd_dict_trainer);
ASSERT_OK(
GetStringFromColumnFamilyOptions(config_options, base_cf_opt, &opts_str));
@ -646,6 +657,8 @@ TEST_F(OptionsTest, CompressionOptionsFromString) {
ASSERT_EQ(new_cf_opt.compression_opts.zstd_max_train_bytes, 8u);
ASSERT_EQ(new_cf_opt.compression_opts.parallel_threads, 9u);
ASSERT_EQ(new_cf_opt.compression_opts.enabled, true);
ASSERT_EQ(base_cf_opt.compression_opts.max_dict_buffer_bytes, 10u);
ASSERT_EQ(base_cf_opt.compression_opts.use_zstd_dict_trainer, false);
ASSERT_EQ(new_cf_opt.bottommost_compression_opts.window_bits, 5);
ASSERT_EQ(new_cf_opt.bottommost_compression_opts.level, 6);
ASSERT_EQ(new_cf_opt.bottommost_compression_opts.strategy, 7);
@ -654,15 +667,18 @@ TEST_F(OptionsTest, CompressionOptionsFromString) {
ASSERT_EQ(new_cf_opt.bottommost_compression_opts.parallel_threads,
dflt.parallel_threads);
ASSERT_EQ(new_cf_opt.bottommost_compression_opts.enabled, false);
ASSERT_EQ(base_cf_opt.bottommost_compression_opts.use_zstd_dict_trainer,
dflt.use_zstd_dict_trainer);
// Test as struct values
ASSERT_OK(GetColumnFamilyOptionsFromString(
config_options, ColumnFamilyOptions(),
"compression_opts={window_bits=5; level=6; strategy=7; max_dict_bytes=8;"
"zstd_max_train_bytes=9;parallel_threads=10;enabled=true}; "
"zstd_max_train_bytes=9;parallel_threads=10;enabled=true;use_zstd_dict_"
"trainer=false}; "
"bottommost_compression_opts={window_bits=4; level=5; strategy=6;"
" max_dict_bytes=7;zstd_max_train_bytes=8;parallel_threads=9;"
"enabled=false}; ",
"enabled=false;use_zstd_dict_trainer=true}; ",
&new_cf_opt));
ASSERT_EQ(new_cf_opt.compression_opts.window_bits, 5);
ASSERT_EQ(new_cf_opt.compression_opts.level, 6);
@ -671,6 +687,7 @@ TEST_F(OptionsTest, CompressionOptionsFromString) {
ASSERT_EQ(new_cf_opt.compression_opts.zstd_max_train_bytes, 9u);
ASSERT_EQ(new_cf_opt.compression_opts.parallel_threads, 10u);
ASSERT_EQ(new_cf_opt.compression_opts.enabled, true);
ASSERT_EQ(new_cf_opt.compression_opts.use_zstd_dict_trainer, false);
ASSERT_EQ(new_cf_opt.bottommost_compression_opts.window_bits, 4);
ASSERT_EQ(new_cf_opt.bottommost_compression_opts.level, 5);
ASSERT_EQ(new_cf_opt.bottommost_compression_opts.strategy, 6);
@ -678,6 +695,7 @@ TEST_F(OptionsTest, CompressionOptionsFromString) {
ASSERT_EQ(new_cf_opt.bottommost_compression_opts.zstd_max_train_bytes, 8u);
ASSERT_EQ(new_cf_opt.bottommost_compression_opts.parallel_threads, 9u);
ASSERT_EQ(new_cf_opt.bottommost_compression_opts.enabled, false);
ASSERT_EQ(new_cf_opt.bottommost_compression_opts.use_zstd_dict_trainer, true);
ASSERT_OK(GetColumnFamilyOptionsFromString(
config_options, base_cf_opt,
@ -709,6 +727,8 @@ TEST_F(OptionsTest, CompressionOptionsFromString) {
base_cf_opt.bottommost_compression_opts.parallel_threads);
ASSERT_EQ(new_cf_opt.bottommost_compression_opts.enabled,
base_cf_opt.bottommost_compression_opts.enabled);
ASSERT_EQ(new_cf_opt.bottommost_compression_opts.use_zstd_dict_trainer,
base_cf_opt.bottommost_compression_opts.use_zstd_dict_trainer);
// Test a few individual struct values
ASSERT_OK(GetColumnFamilyOptionsFromString(
@ -1351,6 +1371,7 @@ TEST_F(OptionsTest, GetOptionsFromStringTest) {
ASSERT_EQ(new_options.compression_opts.zstd_max_train_bytes, 0u);
ASSERT_EQ(new_options.compression_opts.parallel_threads, 1u);
ASSERT_EQ(new_options.compression_opts.enabled, false);
ASSERT_EQ(new_options.compression_opts.use_zstd_dict_trainer, true);
ASSERT_EQ(new_options.bottommost_compression, kDisableCompressionOption);
ASSERT_EQ(new_options.bottommost_compression_opts.window_bits, 5);
ASSERT_EQ(new_options.bottommost_compression_opts.level, 6);
@ -1359,6 +1380,8 @@ TEST_F(OptionsTest, GetOptionsFromStringTest) {
ASSERT_EQ(new_options.bottommost_compression_opts.zstd_max_train_bytes, 0u);
ASSERT_EQ(new_options.bottommost_compression_opts.parallel_threads, 1u);
ASSERT_EQ(new_options.bottommost_compression_opts.enabled, false);
ASSERT_EQ(new_options.bottommost_compression_opts.use_zstd_dict_trainer,
true);
ASSERT_EQ(new_options.write_buffer_size, 10U);
ASSERT_EQ(new_options.max_write_buffer_number, 16);
const auto new_bbto =
@ -2263,7 +2286,7 @@ TEST_F(OptionsOldApiTest, GetOptionsFromMapTest) {
"kZSTDNotFinalCompression"},
{"bottommost_compression", "kLZ4Compression"},
{"bottommost_compression_opts", "5:6:7:8:9:true"},
{"compression_opts", "4:5:6:7:8:true"},
{"compression_opts", "4:5:6:7:8:9:true:10:false"},
{"num_levels", "8"},
{"level0_file_num_compaction_trigger", "8"},
{"level0_slowdown_writes_trigger", "9"},
@ -2381,9 +2404,10 @@ TEST_F(OptionsOldApiTest, GetOptionsFromMapTest) {
ASSERT_EQ(new_cf_opt.compression_opts.strategy, 6);
ASSERT_EQ(new_cf_opt.compression_opts.max_dict_bytes, 7u);
ASSERT_EQ(new_cf_opt.compression_opts.zstd_max_train_bytes, 8u);
ASSERT_EQ(new_cf_opt.compression_opts.parallel_threads,
CompressionOptions().parallel_threads);
ASSERT_EQ(new_cf_opt.compression_opts.parallel_threads, 9u);
ASSERT_EQ(new_cf_opt.compression_opts.enabled, true);
ASSERT_EQ(new_cf_opt.compression_opts.max_dict_buffer_bytes, 10u);
ASSERT_EQ(new_cf_opt.compression_opts.use_zstd_dict_trainer, false);
ASSERT_EQ(new_cf_opt.bottommost_compression, kLZ4Compression);
ASSERT_EQ(new_cf_opt.bottommost_compression_opts.window_bits, 5);
ASSERT_EQ(new_cf_opt.bottommost_compression_opts.level, 6);
@ -2393,6 +2417,10 @@ TEST_F(OptionsOldApiTest, GetOptionsFromMapTest) {
ASSERT_EQ(new_cf_opt.bottommost_compression_opts.parallel_threads,
CompressionOptions().parallel_threads);
ASSERT_EQ(new_cf_opt.bottommost_compression_opts.enabled, true);
ASSERT_EQ(new_cf_opt.bottommost_compression_opts.max_dict_buffer_bytes,
CompressionOptions().max_dict_buffer_bytes);
ASSERT_EQ(new_cf_opt.bottommost_compression_opts.use_zstd_dict_trainer,
CompressionOptions().use_zstd_dict_trainer);
ASSERT_EQ(new_cf_opt.num_levels, 8);
ASSERT_EQ(new_cf_opt.level0_file_num_compaction_trigger, 8);
ASSERT_EQ(new_cf_opt.level0_slowdown_writes_trigger, 9);
@ -3055,6 +3083,7 @@ TEST_F(OptionsOldApiTest, GetOptionsFromStringTest) {
ASSERT_EQ(new_options.compression_opts.zstd_max_train_bytes, 0u);
ASSERT_EQ(new_options.compression_opts.parallel_threads, 1u);
ASSERT_EQ(new_options.compression_opts.enabled, false);
ASSERT_EQ(new_options.compression_opts.use_zstd_dict_trainer, true);
ASSERT_EQ(new_options.bottommost_compression, kDisableCompressionOption);
ASSERT_EQ(new_options.bottommost_compression_opts.window_bits, 5);
ASSERT_EQ(new_options.bottommost_compression_opts.level, 6);
@ -3063,6 +3092,8 @@ TEST_F(OptionsOldApiTest, GetOptionsFromStringTest) {
ASSERT_EQ(new_options.bottommost_compression_opts.zstd_max_train_bytes, 0u);
ASSERT_EQ(new_options.bottommost_compression_opts.parallel_threads, 1u);
ASSERT_EQ(new_options.bottommost_compression_opts.enabled, false);
ASSERT_EQ(new_options.bottommost_compression_opts.use_zstd_dict_trainer,
true);
ASSERT_EQ(new_options.write_buffer_size, 10U);
ASSERT_EQ(new_options.max_write_buffer_number, 16);

7
table/block_based/block_based_table_builder.cc
View File

@ -1896,9 +1896,15 @@ void BlockBasedTableBuilder::EnterUnbuffered() {
// OK if compression_dict_samples is empty, we'll just get empty dictionary.
std::string dict;
if (r->compression_opts.zstd_max_train_bytes > 0) {
if (r->compression_opts.use_zstd_dict_trainer) {
dict = ZSTD_TrainDictionary(compression_dict_samples,
compression_dict_sample_lens,
r->compression_opts.max_dict_bytes);
} else {
dict = ZSTD_FinalizeDictionary(
compression_dict_samples, compression_dict_sample_lens,
r->compression_opts.max_dict_bytes, r->compression_opts.level);
}
} else {
dict = std::move(compression_dict_samples);
}
@ -1934,7 +1940,6 @@ void BlockBasedTableBuilder::EnterUnbuffered() {
}
auto& data_block = r->data_block_buffers[i];
if (r->IsParallelCompressionEnabled()) {
Slice first_key_in_next_block;
const Slice* first_key_in_next_block_ptr = &first_key_in_next_block;

3
table/sst_file_dumper.cc
View File

@ -253,7 +253,7 @@ Status SstFileDumper::ShowAllCompressionSizes(
compression_types,
int32_t compress_level_from, int32_t compress_level_to,
uint32_t max_dict_bytes, uint32_t zstd_max_train_bytes,
uint64_t max_dict_buffer_bytes) {
uint64_t max_dict_buffer_bytes, bool use_zstd_dict_trainer) {
fprintf(stdout, "Block Size: %" ROCKSDB_PRIszt "\n", block_size);
for (auto& i : compression_types) {
if (CompressionTypeSupported(i.first)) {
@ -262,6 +262,7 @@ Status SstFileDumper::ShowAllCompressionSizes(
compress_opt.max_dict_bytes = max_dict_bytes;
compress_opt.zstd_max_train_bytes = zstd_max_train_bytes;
compress_opt.max_dict_buffer_bytes = max_dict_buffer_bytes;
compress_opt.use_zstd_dict_trainer = use_zstd_dict_trainer;
for (int32_t j = compress_level_from; j <= compress_level_to; j++) {
fprintf(stdout, "Compression level: %d", j);
compress_opt.level = j;

2
table/sst_file_dumper.h
View File

@ -44,7 +44,7 @@ class SstFileDumper {
compression_types,
int32_t compress_level_from, int32_t compress_level_to,
uint32_t max_dict_bytes, uint32_t zstd_max_train_bytes,
uint64_t max_dict_buffer_bytes);
uint64_t max_dict_buffer_bytes, bool use_zstd_dict_trainer);
Status ShowCompressionSize(size_t block_size, CompressionType compress_type,
const CompressionOptions& compress_opt);

7
tools/db_bench_tool.cc
View File

@ -1216,6 +1216,11 @@ DEFINE_uint64(compression_max_dict_buffer_bytes,
ROCKSDB_NAMESPACE::CompressionOptions().max_dict_buffer_bytes,
"Maximum bytes to buffer to collect samples for dictionary.");
DEFINE_bool(compression_use_zstd_dict_trainer,
ROCKSDB_NAMESPACE::CompressionOptions().use_zstd_dict_trainer,
"If true, use ZSTD_TrainDictionary() to create dictionary, else"
"use ZSTD_FinalizeDictionary() to create dictionary");
static bool ValidateTableCacheNumshardbits(const char* flagname,
int32_t value) {
if (0 >= value || value >= 20) {
@ -3976,6 +3981,8 @@ class Benchmark {
FLAGS_compression_parallel_threads;
options.compression_opts.max_dict_buffer_bytes =
FLAGS_compression_max_dict_buffer_bytes;
options.compression_opts.use_zstd_dict_trainer =
FLAGS_compression_use_zstd_dict_trainer;
options.max_open_files = FLAGS_open_files;
if (FLAGS_cost_write_buffer_to_cache || FLAGS_db_write_buffer_size != 0) {

1
tools/db_crashtest.py
View File

@ -58,6 +58,7 @@ default_params = {
# lambda: random.choice([1] * 9 + [4])
"compression_parallel_threads": 1,
"compression_max_dict_buffer_bytes": lambda: (1 << random.randint(0, 40)) - 1,
"compression_use_zstd_dict_trainer": lambda: random.randint(0, 1),
"clear_column_family_one_in": 0,
"compact_files_one_in": 1000000,
"compact_range_one_in": 1000000,

10
tools/sst_dump_tool.cc
View File

@ -122,6 +122,9 @@ void print_help(bool to_stderr) {
--compression_max_dict_buffer_bytes=<int64_t>
Limit on buffer size from which we collect samples for dictionary generation.
--compression_use_zstd_finalize_dict
Use zstd's finalizeDictionary() API instead of zstd's dictionary trainer to generate dictionary.
)",
supported_compressions.c_str());
}
@ -188,6 +191,8 @@ int SSTDumpTool::Run(int argc, char const* const* argv, Options options) {
ROCKSDB_NAMESPACE::CompressionOptions().zstd_max_train_bytes;
uint64_t compression_max_dict_buffer_bytes =
ROCKSDB_NAMESPACE::CompressionOptions().max_dict_buffer_bytes;
bool compression_use_zstd_finalize_dict =
!ROCKSDB_NAMESPACE::CompressionOptions().use_zstd_dict_trainer;
int64_t tmp_val;
@ -311,6 +316,8 @@ int SSTDumpTool::Run(int argc, char const* const* argv, Options options) {
return 1;
}
compression_max_dict_buffer_bytes = static_cast<uint64_t>(tmp_val);
} else if (strcmp(argv[i], "--compression_use_zstd_finalize_dict") == 0) {
compression_use_zstd_finalize_dict = true;
} else if (strcmp(argv[i], "--help") == 0) {
print_help(/*to_stderr*/ false);
return 0;
@ -439,7 +446,8 @@ int SSTDumpTool::Run(int argc, char const* const* argv, Options options) {
set_block_size ? block_size : 16384,
compression_types.empty() ? kCompressions : compression_types,
compress_level_from, compress_level_to, compression_max_dict_bytes,
compression_zstd_max_train_bytes, compression_max_dict_buffer_bytes);
compression_zstd_max_train_bytes, compression_max_dict_buffer_bytes,
!compression_use_zstd_finalize_dict);
if (!st.ok()) {
fprintf(stderr, "Failed to recompress: %s\n", st.ToString().c_str());
exit(1);

49
util/compression.h
View File

@ -653,6 +653,9 @@ inline std::string CompressionOptionsToString(
result.append("max_dict_buffer_bytes=")
.append(std::to_string(compression_options.max_dict_buffer_bytes))
.append("; ");
result.append("use_zstd_dict_trainer=")
.append(std::to_string(compression_options.use_zstd_dict_trainer))
.append("; ");
return result;
}
@ -1482,6 +1485,52 @@ inline std::string ZSTD_TrainDictionary(const std::string& samples,
#endif // ZSTD_VERSION_NUMBER >= 10103
}
inline bool ZSTD_FinalizeDictionarySupported() {
#ifdef ZSTD
// ZDICT_finalizeDictionary API is stable since v1.4.5
return (ZSTD_versionNumber() >= 10405);
#else
return false;
#endif
}
inline std::string ZSTD_FinalizeDictionary(
const std::string& samples, const std::vector<size_t>& sample_lens,
size_t max_dict_bytes, int level) {
// ZDICT_finalizeDictionary is stable since version v1.4.5
#if ZSTD_VERSION_NUMBER >= 10405 // v1.4.5+
assert(samples.empty() == sample_lens.empty());
if (samples.empty()) {
return "";
}
if (level == CompressionOptions::kDefaultCompressionLevel) {
// 3 is the value of ZSTD_CLEVEL_DEFAULT (not exposed publicly), see
// https://github.com/facebook/zstd/issues/1148
level = 3;
}
std::string dict_data(max_dict_bytes, '\0');
size_t dict_len = ZDICT_finalizeDictionary(
dict_data.data(), max_dict_bytes, samples.data(),
std::min(static_cast<size_t>(samples.size()), max_dict_bytes),
samples.data(), sample_lens.data(),
static_cast<unsigned>(sample_lens.size()),
{level, 0 /* notificationLevel */, 0 /* dictID */});
if (ZDICT_isError(dict_len)) {
return "";
} else {
assert(dict_len <= max_dict_bytes);
dict_data.resize(dict_len);
return dict_data;
}
#else // up to v1.4.4
(void)samples;
(void)sample_lens;
(void)max_dict_bytes;
(void)level;
return "";
#endif // ZSTD_VERSION_NUMBER >= 10405
}
inline bool CompressData(const Slice& raw,
const CompressionInfo& compression_info,
uint32_t compress_format_version,