2 * Copyright (c) 2016-present, Yann Collet, Facebook, Inc.
5 * This source code is licensed under both the BSD-style license (found in the
6 * LICENSE file in the root directory of this source tree) and the GPLv2 (found
7 * in the COPYING file in the root directory of this source tree).
8 * You may select, at your option, one of the above-listed licenses.
11 #ifndef DICTBUILDER_H_001
12 #define DICTBUILDER_H_001
14 #if defined (__cplusplus)
19 /*====== Dependencies ======*/
20 #include <stddef.h> /* size_t */
23 /* ===== ZDICTLIB_API : control library symbols visibility ===== */
24 #ifndef ZDICTLIB_VISIBILITY
25 # if defined(__GNUC__) && (__GNUC__ >= 4)
26 # define ZDICTLIB_VISIBILITY __attribute__ ((visibility ("default")))
28 # define ZDICTLIB_VISIBILITY
31 #if defined(ZSTD_DLL_EXPORT) && (ZSTD_DLL_EXPORT==1)
32 # define ZDICTLIB_API __declspec(dllexport) ZDICTLIB_VISIBILITY
33 #elif defined(ZSTD_DLL_IMPORT) && (ZSTD_DLL_IMPORT==1)
34 # define ZDICTLIB_API __declspec(dllimport) ZDICTLIB_VISIBILITY /* It isn't required but allows to generate better code, saving a function pointer load from the IAT and an indirect jump.*/
36 # define ZDICTLIB_API ZDICTLIB_VISIBILITY
40 /*! ZDICT_trainFromBuffer():
41 * Train a dictionary from an array of samples.
42 * Redirect towards ZDICT_optimizeTrainFromBuffer_fastCover() single-threaded, with d=8, steps=4,
44 * Samples must be stored concatenated in a single flat buffer `samplesBuffer`,
45 * supplied with an array of sizes `samplesSizes`, providing the size of each sample, in order.
46 * The resulting dictionary will be saved into `dictBuffer`.
47 * @return: size of dictionary stored into `dictBuffer` (<= `dictBufferCapacity`)
48 * or an error code, which can be tested with ZDICT_isError().
49 * Note: ZDICT_trainFromBuffer() requires about 9 bytes of memory for each input byte.
50 * Tips: In general, a reasonable dictionary has a size of ~ 100 KB.
51 * It's possible to select smaller or larger size, just by specifying `dictBufferCapacity`.
52 * In general, it's recommended to provide a few thousands samples, though this can vary a lot.
53 * It's recommended that total size of all samples be about ~x100 times the target size of dictionary.
55 ZDICTLIB_API size_t ZDICT_trainFromBuffer(void* dictBuffer, size_t dictBufferCapacity,
56 const void* samplesBuffer,
57 const size_t* samplesSizes, unsigned nbSamples);
60 /*====== Helper functions ======*/
61 ZDICTLIB_API unsigned ZDICT_getDictID(const void* dictBuffer, size_t dictSize); /**< extracts dictID; @return zero if error (not a valid dictionary) */
62 ZDICTLIB_API unsigned ZDICT_isError(size_t errorCode);
63 ZDICTLIB_API const char* ZDICT_getErrorName(size_t errorCode);
67 #ifdef ZDICT_STATIC_LINKING_ONLY
69 /* ====================================================================================
70 * The definitions in this section are considered experimental.
71 * They should never be used with a dynamic library, as they may change in the future.
72 * They are provided for advanced usages.
73 * Use them only in association with static linking.
74 * ==================================================================================== */
77 int compressionLevel; /* optimize for a specific zstd compression level; 0 means default */
78 unsigned notificationLevel; /* Write log to stderr; 0 = none (default); 1 = errors; 2 = progression; 3 = details; 4 = debug; */
79 unsigned dictID; /* force dictID value; 0 means auto mode (32-bits random value) */
82 /*! ZDICT_cover_params_t:
83 * k and d are the only required parameters.
84 * For others, value 0 means default.
87 unsigned k; /* Segment size : constraint: 0 < k : Reasonable range [16, 2048+] */
88 unsigned d; /* dmer size : constraint: 0 < d <= k : Reasonable range [6, 16] */
89 unsigned steps; /* Number of steps : Only used for optimization : 0 means default (40) : Higher means more parameters checked */
90 unsigned nbThreads; /* Number of threads : constraint: 0 < nbThreads : 1 means single-threaded : Only used for optimization : Ignored if ZSTD_MULTITHREAD is not defined */
91 double splitPoint; /* Percentage of samples used for training: Only used for optimization : the first nbSamples * splitPoint samples will be used to training, the last nbSamples * (1 - splitPoint) samples will be used for testing, 0 means default (1.0), 1.0 when all samples are used for both training and testing */
92 ZDICT_params_t zParams;
93 } ZDICT_cover_params_t;
96 unsigned k; /* Segment size : constraint: 0 < k : Reasonable range [16, 2048+] */
97 unsigned d; /* dmer size : constraint: 0 < d <= k : Reasonable range [6, 16] */
98 unsigned f; /* log of size of frequency array : constraint: 0 < f <= 31 : 1 means default(20)*/
99 unsigned steps; /* Number of steps : Only used for optimization : 0 means default (40) : Higher means more parameters checked */
100 unsigned nbThreads; /* Number of threads : constraint: 0 < nbThreads : 1 means single-threaded : Only used for optimization : Ignored if ZSTD_MULTITHREAD is not defined */
101 double splitPoint; /* Percentage of samples used for training: Only used for optimization : the first nbSamples * splitPoint samples will be used to training, the last nbSamples * (1 - splitPoint) samples will be used for testing, 0 means default (0.75), 1.0 when all samples are used for both training and testing */
102 unsigned accel; /* Acceleration level: constraint: 0 < accel <= 10, higher means faster and less accurate, 0 means default(1) */
103 ZDICT_params_t zParams;
104 } ZDICT_fastCover_params_t;
106 /*! ZDICT_trainFromBuffer_cover():
107 * Train a dictionary from an array of samples using the COVER algorithm.
108 * Samples must be stored concatenated in a single flat buffer `samplesBuffer`,
109 * supplied with an array of sizes `samplesSizes`, providing the size of each sample, in order.
110 * The resulting dictionary will be saved into `dictBuffer`.
111 * @return: size of dictionary stored into `dictBuffer` (<= `dictBufferCapacity`)
112 * or an error code, which can be tested with ZDICT_isError().
113 * Note: ZDICT_trainFromBuffer_cover() requires about 9 bytes of memory for each input byte.
114 * Tips: In general, a reasonable dictionary has a size of ~ 100 KB.
115 * It's possible to select smaller or larger size, just by specifying `dictBufferCapacity`.
116 * In general, it's recommended to provide a few thousands samples, though this can vary a lot.
117 * It's recommended that total size of all samples be about ~x100 times the target size of dictionary.
119 ZDICTLIB_API size_t ZDICT_trainFromBuffer_cover(
120 void *dictBuffer, size_t dictBufferCapacity,
121 const void *samplesBuffer, const size_t *samplesSizes, unsigned nbSamples,
122 ZDICT_cover_params_t parameters);
124 /*! ZDICT_optimizeTrainFromBuffer_cover():
125 * The same requirements as above hold for all the parameters except `parameters`.
126 * This function tries many parameter combinations and picks the best parameters.
127 * `*parameters` is filled with the best parameters found,
128 * dictionary constructed with those parameters is stored in `dictBuffer`.
130 * All of the parameters d, k, steps are optional.
131 * If d is non-zero then we don't check multiple values of d, otherwise we check d = {6, 8}.
132 * if steps is zero it defaults to its default value.
133 * If k is non-zero then we don't check multiple values of k, otherwise we check steps values in [50, 2000].
135 * @return: size of dictionary stored into `dictBuffer` (<= `dictBufferCapacity`)
136 * or an error code, which can be tested with ZDICT_isError().
137 * On success `*parameters` contains the parameters selected.
138 * Note: ZDICT_optimizeTrainFromBuffer_cover() requires about 8 bytes of memory for each input byte and additionally another 5 bytes of memory for each byte of memory for each thread.
140 ZDICTLIB_API size_t ZDICT_optimizeTrainFromBuffer_cover(
141 void* dictBuffer, size_t dictBufferCapacity,
142 const void* samplesBuffer, const size_t* samplesSizes, unsigned nbSamples,
143 ZDICT_cover_params_t* parameters);
145 /*! ZDICT_trainFromBuffer_fastCover():
146 * Train a dictionary from an array of samples using a modified version of COVER algorithm.
147 * Samples must be stored concatenated in a single flat buffer `samplesBuffer`,
148 * supplied with an array of sizes `samplesSizes`, providing the size of each sample, in order.
149 * d and k are required.
150 * All other parameters are optional, will use default values if not provided
151 * The resulting dictionary will be saved into `dictBuffer`.
152 * @return: size of dictionary stored into `dictBuffer` (<= `dictBufferCapacity`)
153 * or an error code, which can be tested with ZDICT_isError().
154 * Note: ZDICT_trainFromBuffer_fastCover() requires about 1 bytes of memory for each input byte and additionally another 6 * 2^f bytes of memory .
155 * Tips: In general, a reasonable dictionary has a size of ~ 100 KB.
156 * It's possible to select smaller or larger size, just by specifying `dictBufferCapacity`.
157 * In general, it's recommended to provide a few thousands samples, though this can vary a lot.
158 * It's recommended that total size of all samples be about ~x100 times the target size of dictionary.
160 ZDICTLIB_API size_t ZDICT_trainFromBuffer_fastCover(void *dictBuffer,
161 size_t dictBufferCapacity, const void *samplesBuffer,
162 const size_t *samplesSizes, unsigned nbSamples,
163 ZDICT_fastCover_params_t parameters);
165 /*! ZDICT_optimizeTrainFromBuffer_fastCover():
166 * The same requirements as above hold for all the parameters except `parameters`.
167 * This function tries many parameter combinations (specifically, k and d combinations)
168 * and picks the best parameters. `*parameters` is filled with the best parameters found,
169 * dictionary constructed with those parameters is stored in `dictBuffer`.
170 * All of the parameters d, k, steps, f, and accel are optional.
171 * If d is non-zero then we don't check multiple values of d, otherwise we check d = {6, 8}.
172 * if steps is zero it defaults to its default value.
173 * If k is non-zero then we don't check multiple values of k, otherwise we check steps values in [50, 2000].
174 * If f is zero, default value of 20 is used.
175 * If accel is zero, default value of 1 is used.
177 * @return: size of dictionary stored into `dictBuffer` (<= `dictBufferCapacity`)
178 * or an error code, which can be tested with ZDICT_isError().
179 * On success `*parameters` contains the parameters selected.
180 * Note: ZDICT_optimizeTrainFromBuffer_fastCover() requires about 1 byte of memory for each input byte and additionally another 6 * 2^f bytes of memory for each thread.
182 ZDICTLIB_API size_t ZDICT_optimizeTrainFromBuffer_fastCover(void* dictBuffer,
183 size_t dictBufferCapacity, const void* samplesBuffer,
184 const size_t* samplesSizes, unsigned nbSamples,
185 ZDICT_fastCover_params_t* parameters);
187 /*! ZDICT_finalizeDictionary():
188 * Given a custom content as a basis for dictionary, and a set of samples,
189 * finalize dictionary by adding headers and statistics.
191 * Samples must be stored concatenated in a flat buffer `samplesBuffer`,
192 * supplied with an array of sizes `samplesSizes`, providing the size of each sample in order.
194 * dictContentSize must be >= ZDICT_CONTENTSIZE_MIN bytes.
195 * maxDictSize must be >= dictContentSize, and must be >= ZDICT_DICTSIZE_MIN bytes.
197 * @return: size of dictionary stored into `dictBuffer` (<= `dictBufferCapacity`),
198 * or an error code, which can be tested by ZDICT_isError().
199 * Note: ZDICT_finalizeDictionary() will push notifications into stderr if instructed to, using notificationLevel>0.
200 * Note 2: dictBuffer and dictContent can overlap
202 #define ZDICT_CONTENTSIZE_MIN 128
203 #define ZDICT_DICTSIZE_MIN 256
204 ZDICTLIB_API size_t ZDICT_finalizeDictionary(void* dictBuffer, size_t dictBufferCapacity,
205 const void* dictContent, size_t dictContentSize,
206 const void* samplesBuffer, const size_t* samplesSizes, unsigned nbSamples,
207 ZDICT_params_t parameters);
210 unsigned selectivityLevel; /* 0 means default; larger => select more => larger dictionary */
211 ZDICT_params_t zParams;
212 } ZDICT_legacy_params_t;
214 /*! ZDICT_trainFromBuffer_legacy():
215 * Train a dictionary from an array of samples.
216 * Samples must be stored concatenated in a single flat buffer `samplesBuffer`,
217 * supplied with an array of sizes `samplesSizes`, providing the size of each sample, in order.
218 * The resulting dictionary will be saved into `dictBuffer`.
219 * `parameters` is optional and can be provided with values set to 0 to mean "default".
220 * @return: size of dictionary stored into `dictBuffer` (<= `dictBufferCapacity`)
221 * or an error code, which can be tested with ZDICT_isError().
222 * Tips: In general, a reasonable dictionary has a size of ~ 100 KB.
223 * It's possible to select smaller or larger size, just by specifying `dictBufferCapacity`.
224 * In general, it's recommended to provide a few thousands samples, though this can vary a lot.
225 * It's recommended that total size of all samples be about ~x100 times the target size of dictionary.
226 * Note: ZDICT_trainFromBuffer_legacy() will send notifications into stderr if instructed to, using notificationLevel>0.
228 ZDICTLIB_API size_t ZDICT_trainFromBuffer_legacy(
229 void *dictBuffer, size_t dictBufferCapacity,
230 const void *samplesBuffer, const size_t *samplesSizes, unsigned nbSamples,
231 ZDICT_legacy_params_t parameters);
233 /* Deprecation warnings */
234 /* It is generally possible to disable deprecation warnings from compiler,
235 for example with -Wno-deprecated-declarations for gcc
236 or _CRT_SECURE_NO_WARNINGS in Visual.
237 Otherwise, it's also possible to manually define ZDICT_DISABLE_DEPRECATE_WARNINGS */
238 #ifdef ZDICT_DISABLE_DEPRECATE_WARNINGS
239 # define ZDICT_DEPRECATED(message) ZDICTLIB_API /* disable deprecation warnings */
241 # define ZDICT_GCC_VERSION (__GNUC__ * 100 + __GNUC_MINOR__)
242 # if defined (__cplusplus) && (__cplusplus >= 201402) /* C++14 or greater */
243 # define ZDICT_DEPRECATED(message) [[deprecated(message)]] ZDICTLIB_API
244 # elif (ZDICT_GCC_VERSION >= 405) || defined(__clang__)
245 # define ZDICT_DEPRECATED(message) ZDICTLIB_API __attribute__((deprecated(message)))
246 # elif (ZDICT_GCC_VERSION >= 301)
247 # define ZDICT_DEPRECATED(message) ZDICTLIB_API __attribute__((deprecated))
248 # elif defined(_MSC_VER)
249 # define ZDICT_DEPRECATED(message) ZDICTLIB_API __declspec(deprecated(message))
251 # pragma message("WARNING: You need to implement ZDICT_DEPRECATED for this compiler")
252 # define ZDICT_DEPRECATED(message) ZDICTLIB_API
254 #endif /* ZDICT_DISABLE_DEPRECATE_WARNINGS */
256 ZDICT_DEPRECATED("use ZDICT_finalizeDictionary() instead")
257 size_t ZDICT_addEntropyTablesFromBuffer(void* dictBuffer, size_t dictContentSize, size_t dictBufferCapacity,
258 const void* samplesBuffer, const size_t* samplesSizes, unsigned nbSamples);
261 #endif /* ZDICT_STATIC_LINKING_ONLY */
263 #if defined (__cplusplus)
267 #endif /* DICTBUILDER_H_001 */