1 // 2016 and later: Unicode, Inc. and others. 2 // License & terms of use: http://www.unicode.org/copyright.html 3 /* 4 ***************************************************************************** 5 * Copyright (C) 1996-2014, International Business Machines Corporation and others. 6 * All Rights Reserved. 7 ***************************************************************************** 8 * 9 * File sortkey.h 10 * 11 * Created by: Helena Shih 12 * 13 * Modification History: 14 * 15 * Date Name Description 16 * 17 * 6/20/97 helena Java class name change. 18 * 8/18/97 helena Added internal API documentation. 19 * 6/26/98 erm Changed to use byte arrays and memcmp. 20 ***************************************************************************** 21 */ 22 23 #ifndef SORTKEY_H 24 #define SORTKEY_H 25 26 #include "unicode/utypes.h" 27 28 /** 29 * \file 30 * \brief C++ API: Keys for comparing strings multiple times. 31 */ 32 33 #if !UCONFIG_NO_COLLATION 34 35 #include "unicode/uobject.h" 36 #include "unicode/unistr.h" 37 #include "unicode/coll.h" 38 39 U_NAMESPACE_BEGIN 40 41 /* forward declaration */ 42 class RuleBasedCollator; 43 class CollationKeyByteSink; 44 45 /** 46 * 47 * Collation keys are generated by the Collator class. Use the CollationKey objects 48 * instead of Collator to compare strings multiple times. A CollationKey 49 * preprocesses the comparison information from the Collator object to 50 * make the comparison faster. If you are not going to comparing strings 51 * multiple times, then using the Collator object is generally faster, 52 * since it only processes as much of the string as needed to make a 53 * comparison. 54 * <p> For example (with strength == tertiary) 55 * <p>When comparing "Abernathy" to "Baggins-Smythworthy", Collator 56 * only needs to process a couple of characters, while a comparison 57 * with CollationKeys will process all of the characters. On the other hand, 58 * if you are doing a sort of a number of fields, it is much faster to use 59 * CollationKeys, since you will be comparing strings multiple times. 60 * <p>Typical use of CollationKeys are in databases, where you store a CollationKey 61 * in a hidden field, and use it for sorting or indexing. 62 * 63 * <p>Example of use: 64 * <pre> 65 * \code 66 * UErrorCode success = U_ZERO_ERROR; 67 * Collator* myCollator = Collator::createInstance(success); 68 * CollationKey* keys = new CollationKey [3]; 69 * myCollator->getCollationKey("Tom", keys[0], success ); 70 * myCollator->getCollationKey("Dick", keys[1], success ); 71 * myCollator->getCollationKey("Harry", keys[2], success ); 72 * 73 * // Inside body of sort routine, compare keys this way: 74 * CollationKey tmp; 75 * if(keys[0].compareTo( keys[1] ) > 0 ) { 76 * tmp = keys[0]; keys[0] = keys[1]; keys[1] = tmp; 77 * } 78 * //... 79 * \endcode 80 * </pre> 81 * <p>Because Collator::compare()'s algorithm is complex, it is faster to sort 82 * long lists of words by retrieving collation keys with Collator::getCollationKey(). 83 * You can then cache the collation keys and compare them using CollationKey::compareTo(). 84 * <p> 85 * <strong>Note:</strong> <code>Collator</code>s with different Locale, 86 * CollationStrength and DecompositionMode settings will return different 87 * CollationKeys for the same set of strings. Locales have specific 88 * collation rules, and the way in which secondary and tertiary differences 89 * are taken into account, for example, will result in different CollationKeys 90 * for same strings. 91 * <p> 92 93 * @see Collator 94 * @see RuleBasedCollator 95 * @version 1.3 12/18/96 96 * @author Helena Shih 97 * @stable ICU 2.0 98 */ 99 class U_I18N_API CollationKey : public UObject { 100 public: 101 /** 102 * This creates an empty collation key based on the null string. An empty 103 * collation key contains no sorting information. When comparing two empty 104 * collation keys, the result is Collator::EQUAL. Comparing empty collation key 105 * with non-empty collation key is always Collator::LESS. 106 * @stable ICU 2.0 107 */ 108 CollationKey(); 109 110 111 /** 112 * Creates a collation key based on the collation key values. 113 * @param values the collation key values 114 * @param count number of collation key values, including trailing nulls. 115 * @stable ICU 2.0 116 */ 117 CollationKey(const uint8_t* values, 118 int32_t count); 119 120 /** 121 * Copy constructor. 122 * @param other the object to be copied. 123 * @stable ICU 2.0 124 */ 125 CollationKey(const CollationKey& other); 126 127 /** 128 * Sort key destructor. 129 * @stable ICU 2.0 130 */ 131 virtual ~CollationKey(); 132 133 /** 134 * Assignment operator 135 * @param other the object to be copied. 136 * @stable ICU 2.0 137 */ 138 const CollationKey& operator=(const CollationKey& other); 139 140 /** 141 * Compare if two collation keys are the same. 142 * @param source the collation key to compare to. 143 * @return Returns true if two collation keys are equal, false otherwise. 144 * @stable ICU 2.0 145 */ 146 UBool operator==(const CollationKey& source) const; 147 148 /** 149 * Compare if two collation keys are not the same. 150 * @param source the collation key to compare to. 151 * @return Returns TRUE if two collation keys are different, FALSE otherwise. 152 * @stable ICU 2.0 153 */ 154 UBool operator!=(const CollationKey& source) const; 155 156 157 /** 158 * Test to see if the key is in an invalid state. The key will be in an 159 * invalid state if it couldn't allocate memory for some operation. 160 * @return Returns TRUE if the key is in an invalid, FALSE otherwise. 161 * @stable ICU 2.0 162 */ 163 UBool isBogus(void) const; 164 165 /** 166 * Returns a pointer to the collation key values. The storage is owned 167 * by the collation key and the pointer will become invalid if the key 168 * is deleted. 169 * @param count the output parameter of number of collation key values, 170 * including any trailing nulls. 171 * @return a pointer to the collation key values. 172 * @stable ICU 2.0 173 */ 174 const uint8_t* getByteArray(int32_t& count) const; 175 176 #ifdef U_USE_COLLATION_KEY_DEPRECATES 177 /** 178 * Extracts the collation key values into a new array. The caller owns 179 * this storage and should free it. 180 * @param count the output parameter of number of collation key values, 181 * including any trailing nulls. 182 * @obsolete ICU 2.6. Use getByteArray instead since this API will be removed in that release. 183 */ 184 uint8_t* toByteArray(int32_t& count) const; 185 #endif 186 187 #ifndef U_HIDE_DEPRECATED_API 188 /** 189 * Convenience method which does a string(bit-wise) comparison of the 190 * two collation keys. 191 * @param target target collation key to be compared with 192 * @return Returns Collator::LESS if sourceKey < targetKey, 193 * Collator::GREATER if sourceKey > targetKey and Collator::EQUAL 194 * otherwise. 195 * @deprecated ICU 2.6 use the overload with error code 196 */ 197 Collator::EComparisonResult compareTo(const CollationKey& target) const; 198 #endif /* U_HIDE_DEPRECATED_API */ 199 200 /** 201 * Convenience method which does a string(bit-wise) comparison of the 202 * two collation keys. 203 * @param target target collation key to be compared with 204 * @param status error code 205 * @return Returns UCOL_LESS if sourceKey < targetKey, 206 * UCOL_GREATER if sourceKey > targetKey and UCOL_EQUAL 207 * otherwise. 208 * @stable ICU 2.6 209 */ 210 UCollationResult compareTo(const CollationKey& target, UErrorCode &status) const; 211 212 /** 213 * Creates an integer that is unique to the collation key. NOTE: this 214 * is not the same as String.hashCode. 215 * <p>Example of use: 216 * <pre> 217 * . UErrorCode status = U_ZERO_ERROR; 218 * . Collator *myCollation = Collator::createInstance(Locale::US, status); 219 * . if (U_FAILURE(status)) return; 220 * . CollationKey key1, key2; 221 * . UErrorCode status1 = U_ZERO_ERROR, status2 = U_ZERO_ERROR; 222 * . myCollation->getCollationKey("abc", key1, status1); 223 * . if (U_FAILURE(status1)) { delete myCollation; return; } 224 * . myCollation->getCollationKey("ABC", key2, status2); 225 * . if (U_FAILURE(status2)) { delete myCollation; return; } 226 * . // key1.hashCode() != key2.hashCode() 227 * </pre> 228 * @return the hash value based on the string's collation order. 229 * @see UnicodeString#hashCode 230 * @stable ICU 2.0 231 */ 232 int32_t hashCode(void) const; 233 234 /** 235 * ICU "poor man's RTTI", returns a UClassID for the actual class. 236 * @stable ICU 2.2 237 */ 238 virtual UClassID getDynamicClassID() const; 239 240 /** 241 * ICU "poor man's RTTI", returns a UClassID for this class. 242 * @stable ICU 2.2 243 */ 244 static UClassID U_EXPORT2 getStaticClassID(); 245 246 private: 247 /** 248 * Replaces the current bytes buffer with a new one of newCapacity 249 * and copies length bytes from the old buffer to the new one. 250 * @return the new buffer, or NULL if the allocation failed 251 */ 252 uint8_t *reallocate(int32_t newCapacity, int32_t length); 253 /** 254 * Set a new length for a new sort key in the existing fBytes. 255 */ 256 void setLength(int32_t newLength); 257 258 uint8_t *getBytes() { 259 return (fFlagAndLength >= 0) ? fUnion.fStackBuffer : fUnion.fFields.fBytes; 260 } 261 const uint8_t *getBytes() const { 262 return (fFlagAndLength >= 0) ? fUnion.fStackBuffer : fUnion.fFields.fBytes; 263 } 264 int32_t getCapacity() const { 265 return (fFlagAndLength >= 0) ? (int32_t)sizeof(fUnion) : fUnion.fFields.fCapacity; 266 } 267 int32_t getLength() const { return fFlagAndLength & 0x7fffffff; } 268 269 /** 270 * Set the CollationKey to a "bogus" or invalid state 271 * @return this CollationKey 272 */ 273 CollationKey& setToBogus(void); 274 /** 275 * Resets this CollationKey to an empty state 276 * @return this CollationKey 277 */ 278 CollationKey& reset(void); 279 280 /** 281 * Allow private access to RuleBasedCollator 282 */ 283 friend class RuleBasedCollator; 284 friend class CollationKeyByteSink; 285 286 // Class fields. sizeof(CollationKey) is intended to be 48 bytes 287 // on a machine with 64-bit pointers. 288 // We use a union to maximize the size of the internal buffer, 289 // similar to UnicodeString but not as tight and complex. 290 291 // (implicit) *vtable; 292 /** 293 * Sort key length and flag. 294 * Bit 31 is set if the buffer is heap-allocated. 295 * Bits 30..0 contain the sort key length. 296 */ 297 int32_t fFlagAndLength; 298 /** 299 * Unique hash value of this CollationKey. 300 * Special value 2 if the key is bogus. 301 */ 302 mutable int32_t fHashCode; 303 /** 304 * fUnion provides 32 bytes for the internal buffer or for 305 * pointer+capacity. 306 */ 307 union StackBufferOrFields { 308 /** fStackBuffer is used iff fFlagAndLength>=0, else fFields is used */ 309 uint8_t fStackBuffer[32]; 310 struct { 311 uint8_t *fBytes; 312 int32_t fCapacity; 313 } fFields; 314 } fUnion; 315 }; 316 317 inline UBool 318 CollationKey::operator!=(const CollationKey& other) const 319 { 320 return !(*this == other); 321 } 322 323 inline UBool 324 CollationKey::isBogus() const 325 { 326 return fHashCode == 2; // kBogusHashCode 327 } 328 329 inline const uint8_t* 330 CollationKey::getByteArray(int32_t &count) const 331 { 332 count = getLength(); 333 return getBytes(); 334 } 335 336 U_NAMESPACE_END 337 338 #endif /* #if !UCONFIG_NO_COLLATION */ 339 340 #endif 341