[apple/icu.git] / icuSources / i18n / unicode / sortkey.h

/*
 *****************************************************************************
 * Copyright (C) 1996-2014, International Business Machines Corporation and others.
 * All Rights Reserved.
 *****************************************************************************
 *
 * File sortkey.h
 *
 * Created by: Helena Shih
 *
 * Modification History:
 *
 *  Date         Name          Description
 *
 *  6/20/97     helena      Java class name change.
 *  8/18/97     helena      Added internal API documentation.
 *  6/26/98     erm         Changed to use byte arrays and memcmp.
 *****************************************************************************
 */

#ifndef SORTKEY_H
#define SORTKEY_H

#include "unicode/utypes.h"

/**
 * \file 
 * \brief C++ API: Keys for comparing strings multiple times. 
 */
 
#if !UCONFIG_NO_COLLATION

#include "unicode/uobject.h"
#include "unicode/unistr.h"
#include "unicode/coll.h"

U_NAMESPACE_BEGIN

/* forward declaration */
class RuleBasedCollator;
class CollationKeyByteSink;

/**
 *
 * Collation keys are generated by the Collator class.  Use the CollationKey objects
 * instead of Collator to compare strings multiple times.  A CollationKey
 * preprocesses the comparison information from the Collator object to
 * make the comparison faster.  If you are not going to comparing strings
 * multiple times, then using the Collator object is generally faster,
 * since it only processes as much of the string as needed to make a
 * comparison.
 * <p> For example (with strength == tertiary)
 * <p>When comparing "Abernathy" to "Baggins-Smythworthy", Collator
 * only needs to process a couple of characters, while a comparison
 * with CollationKeys will process all of the characters.  On the other hand,
 * if you are doing a sort of a number of fields, it is much faster to use
 * CollationKeys, since you will be comparing strings multiple times.
 * <p>Typical use of CollationKeys are in databases, where you store a CollationKey
 * in a hidden field, and use it for sorting or indexing.
 *
 * <p>Example of use:
 * <pre>
 * \code
 *     UErrorCode success = U_ZERO_ERROR;
 *     Collator* myCollator = Collator::createInstance(success);
 *     CollationKey* keys = new CollationKey [3];
 *     myCollator->getCollationKey("Tom", keys[0], success );
 *     myCollator->getCollationKey("Dick", keys[1], success );
 *     myCollator->getCollationKey("Harry", keys[2], success );
 *
 *     // Inside body of sort routine, compare keys this way:
 *     CollationKey tmp;
 *     if(keys[0].compareTo( keys[1] ) > 0 ) {
 *         tmp = keys[0]; keys[0] = keys[1]; keys[1] = tmp;
 *     }
 *     //...
 * \endcode
 * </pre>
 * <p>Because Collator::compare()'s algorithm is complex, it is faster to sort
 * long lists of words by retrieving collation keys with Collator::getCollationKey().
 * You can then cache the collation keys and compare them using CollationKey::compareTo().
 * <p>
 * <strong>Note:</strong> <code>Collator</code>s with different Locale,
 * CollationStrength and DecompositionMode settings will return different
 * CollationKeys for the same set of strings. Locales have specific
 * collation rules, and the way in which secondary and tertiary differences
 * are taken into account, for example, will result in different CollationKeys
 * for same strings.
 * <p>

 * @see          Collator
 * @see          RuleBasedCollator
 * @version      1.3 12/18/96
 * @author       Helena Shih
 * @stable ICU 2.0
 */
class U_I18N_API CollationKey : public UObject {
public:
    /**
    * This creates an empty collation key based on the null string.  An empty
    * collation key contains no sorting information.  When comparing two empty
    * collation keys, the result is Collator::EQUAL.  Comparing empty collation key
    * with non-empty collation key is always Collator::LESS.
    * @stable ICU 2.0
    */
    CollationKey();


    /**
    * Creates a collation key based on the collation key values.
    * @param values the collation key values
    * @param count number of collation key values, including trailing nulls.
    * @stable ICU 2.0
    */
    CollationKey(const  uint8_t*    values,
                int32_t     count);

    /**
    * Copy constructor.
    * @param other    the object to be copied.
    * @stable ICU 2.0
    */
    CollationKey(const CollationKey& other);

    /**
    * Sort key destructor.
    * @stable ICU 2.0
    */
    virtual ~CollationKey();

    /**
    * Assignment operator
    * @param other    the object to be copied.
    * @stable ICU 2.0
    */
    const   CollationKey&   operator=(const CollationKey& other);

    /**
    * Compare if two collation keys are the same.
    * @param source the collation key to compare to.
    * @return Returns true if two collation keys are equal, false otherwise.
    * @stable ICU 2.0
    */
    UBool                   operator==(const CollationKey& source) const;

    /**
    * Compare if two collation keys are not the same.
    * @param source the collation key to compare to.
    * @return Returns TRUE if two collation keys are different, FALSE otherwise.
    * @stable ICU 2.0
    */
    UBool                   operator!=(const CollationKey& source) const;


    /**
    * Test to see if the key is in an invalid state. The key will be in an
    * invalid state if it couldn't allocate memory for some operation.
    * @return Returns TRUE if the key is in an invalid, FALSE otherwise.
    * @stable ICU 2.0
    */
    UBool                   isBogus(void) const;

    /**
    * Returns a pointer to the collation key values. The storage is owned
    * by the collation key and the pointer will become invalid if the key
    * is deleted.
    * @param count the output parameter of number of collation key values,
    * including any trailing nulls.
    * @return a pointer to the collation key values.
    * @stable ICU 2.0
    */
    const    uint8_t*       getByteArray(int32_t& count) const;

#ifdef U_USE_COLLATION_KEY_DEPRECATES
    /**
    * Extracts the collation key values into a new array. The caller owns
    * this storage and should free it.
    * @param count the output parameter of number of collation key values,
    * including any trailing nulls.
    * @obsolete ICU 2.6. Use getByteArray instead since this API will be removed in that release.
    */
    uint8_t*                toByteArray(int32_t& count) const;
#endif

#ifndef U_HIDE_DEPRECATED_API 
    /**
    * Convenience method which does a string(bit-wise) comparison of the
    * two collation keys.
    * @param target target collation key to be compared with
    * @return Returns Collator::LESS if sourceKey &lt; targetKey,
    * Collator::GREATER if sourceKey > targetKey and Collator::EQUAL
    * otherwise.
    * @deprecated ICU 2.6 use the overload with error code
    */
    Collator::EComparisonResult compareTo(const CollationKey& target) const;
#endif  /* U_HIDE_DEPRECATED_API */

    /**
    * Convenience method which does a string(bit-wise) comparison of the
    * two collation keys.
    * @param target target collation key to be compared with
    * @param status error code
    * @return Returns UCOL_LESS if sourceKey &lt; targetKey,
    * UCOL_GREATER if sourceKey > targetKey and UCOL_EQUAL
    * otherwise.
    * @stable ICU 2.6
    */
    UCollationResult compareTo(const CollationKey& target, UErrorCode &status) const;

    /**
    * Creates an integer that is unique to the collation key.  NOTE: this
    * is not the same as String.hashCode.
    * <p>Example of use:
    * <pre>
    * .    UErrorCode status = U_ZERO_ERROR;
    * .    Collator *myCollation = Collator::createInstance(Locale::US, status);
    * .    if (U_FAILURE(status)) return;
    * .    CollationKey key1, key2;
    * .    UErrorCode status1 = U_ZERO_ERROR, status2 = U_ZERO_ERROR;
    * .    myCollation->getCollationKey("abc", key1, status1);
    * .    if (U_FAILURE(status1)) { delete myCollation; return; }
    * .    myCollation->getCollationKey("ABC", key2, status2);
    * .    if (U_FAILURE(status2)) { delete myCollation; return; }
    * .    // key1.hashCode() != key2.hashCode()
    * </pre>
    * @return the hash value based on the string's collation order.
    * @see UnicodeString#hashCode
    * @stable ICU 2.0
    */
    int32_t                 hashCode(void) const;

    /**
     * ICU "poor man's RTTI", returns a UClassID for the actual class.
     * @stable ICU 2.2
     */
    virtual UClassID getDynamicClassID() const;

    /**
     * ICU "poor man's RTTI", returns a UClassID for this class.
     * @stable ICU 2.2
     */
    static UClassID U_EXPORT2 getStaticClassID();

private:
    /**
     * Replaces the current bytes buffer with a new one of newCapacity
     * and copies length bytes from the old buffer to the new one.
     * @return the new buffer, or NULL if the allocation failed
     */
    uint8_t *reallocate(int32_t newCapacity, int32_t length);
    /**
     * Set a new length for a new sort key in the existing fBytes.
     */
    void setLength(int32_t newLength);

    uint8_t *getBytes() {
        return (fFlagAndLength >= 0) ? fUnion.fStackBuffer : fUnion.fFields.fBytes;
    }
    const uint8_t *getBytes() const {
        return (fFlagAndLength >= 0) ? fUnion.fStackBuffer : fUnion.fFields.fBytes;
    }
    int32_t getCapacity() const {
        return (fFlagAndLength >= 0) ? (int32_t)sizeof(fUnion) : fUnion.fFields.fCapacity;
    }
    int32_t getLength() const { return fFlagAndLength & 0x7fffffff; }

    /**
    * Set the CollationKey to a "bogus" or invalid state
    * @return this CollationKey
    */
    CollationKey&           setToBogus(void);
    /**
    * Resets this CollationKey to an empty state
    * @return this CollationKey
    */
    CollationKey&           reset(void);

    /**
    * Allow private access to RuleBasedCollator
    */
    friend  class           RuleBasedCollator;
    friend  class           CollationKeyByteSink;

    // Class fields. sizeof(CollationKey) is intended to be 48 bytes
    // on a machine with 64-bit pointers.
    // We use a union to maximize the size of the internal buffer,
    // similar to UnicodeString but not as tight and complex.

    // (implicit) *vtable;
    /**
     * Sort key length and flag.
     * Bit 31 is set if the buffer is heap-allocated.
     * Bits 30..0 contain the sort key length.
     */
    int32_t fFlagAndLength;
    /**
    * Unique hash value of this CollationKey.
    * Special value 2 if the key is bogus.
    */
    mutable int32_t fHashCode;
    /**
     * fUnion provides 32 bytes for the internal buffer or for
     * pointer+capacity.
     */
    union StackBufferOrFields {
        /** fStackBuffer is used iff fFlagAndLength>=0, else fFields is used */
        uint8_t fStackBuffer[32];
        struct {
            uint8_t *fBytes;
            int32_t fCapacity;
        } fFields;
    } fUnion;
};

inline UBool
CollationKey::operator!=(const CollationKey& other) const
{
    return !(*this == other);
}

inline UBool
CollationKey::isBogus() const
{
    return fHashCode == 2;  // kBogusHashCode
}

inline const uint8_t*
CollationKey::getByteArray(int32_t &count) const
{
    count = getLength();
    return getBytes();
}

U_NAMESPACE_END

#endif /* #if !UCONFIG_NO_COLLATION */

#endif
Commit	Line	Data
b75a7d8f	1	/*
73c04bcf	2	*****************************************************************************
57a6839d	3	* Copyright (C) 1996-2014, International Business Machines Corporation and others.
374ca955	4	* All Rights Reserved.
73c04bcf	5	*****************************************************************************
374ca955 A	6	*
	7	* File sortkey.h
	8	*
	9	* Created by: Helena Shih
	10	*
	11	* Modification History:
	12	*
	13	* Date Name Description
	14	*
	15	* 6/20/97 helena Java class name change.
	16	* 8/18/97 helena Added internal API documentation.
	17	* 6/26/98 erm Changed to use byte arrays and memcmp.
73c04bcf	18	*****************************************************************************
374ca955	19	*/
b75a7d8f A	20
	21	#ifndef SORTKEY_H
	22	#define SORTKEY_H
	23
	24	#include "unicode/utypes.h"
	25
73c04bcf A	26	/**
	27	* \file
	28	* \brief C++ API: Keys for comparing strings multiple times.
	29	*/
	30
b75a7d8f A	31	#if !UCONFIG_NO_COLLATION
	32
	33	#include "unicode/uobject.h"
	34	#include "unicode/unistr.h"
	35	#include "unicode/coll.h"
	36
	37	U_NAMESPACE_BEGIN
	38
	39	/* forward declaration */
	40	class RuleBasedCollator;
57a6839d	41	class CollationKeyByteSink;
b75a7d8f A	42
b75a7d8f A	43	/**
73c04bcf	44	*
b75a7d8f A	45	* Collation keys are generated by the Collator class. Use the CollationKey objects
	46	* instead of Collator to compare strings multiple times. A CollationKey
	47	* preprocesses the comparison information from the Collator object to
	48	* make the comparison faster. If you are not going to comparing strings
	49	* multiple times, then using the Collator object is generally faster,
	50	* since it only processes as much of the string as needed to make a
	51	* comparison.
	52	* <p> For example (with strength == tertiary)
	53	* <p>When comparing "Abernathy" to "Baggins-Smythworthy", Collator
	54	* only needs to process a couple of characters, while a comparison
	55	* with CollationKeys will process all of the characters. On the other hand,
	56	* if you are doing a sort of a number of fields, it is much faster to use
	57	* CollationKeys, since you will be comparing strings multiple times.
	58	* <p>Typical use of CollationKeys are in databases, where you store a CollationKey
	59	* in a hidden field, and use it for sorting or indexing.
	60	*
	61	* <p>Example of use:
	62	* <pre>
	63	* \code
	64	* UErrorCode success = U_ZERO_ERROR;
	65	* Collator* myCollator = Collator::createInstance(success);
	66	* CollationKey* keys = new CollationKey [3];
	67	* myCollator->getCollationKey("Tom", keys[0], success );
	68	* myCollator->getCollationKey("Dick", keys[1], success );
	69	* myCollator->getCollationKey("Harry", keys[2], success );
	70	*
	71	* // Inside body of sort routine, compare keys this way:
	72	* CollationKey tmp;
	73	* if(keys[0].compareTo( keys[1] ) > 0 ) {
	74	* tmp = keys[0]; keys[0] = keys[1]; keys[1] = tmp;
	75	* }
	76	* //...
	77	* \endcode
	78	* </pre>
	79	* <p>Because Collator::compare()'s algorithm is complex, it is faster to sort
	80	* long lists of words by retrieving collation keys with Collator::getCollationKey().
	81	* You can then cache the collation keys and compare them using CollationKey::compareTo().
	82	* <p>
	83	* <strong>Note:</strong> <code>Collator</code>s with different Locale,
	84	* CollationStrength and DecompositionMode settings will return different
	85	* CollationKeys for the same set of strings. Locales have specific
	86	* collation rules, and the way in which secondary and tertiary differences
	87	* are taken into account, for example, will result in different CollationKeys
	88	* for same strings.
	89	* <p>
	90
	91	* @see Collator
	92	* @see RuleBasedCollator
	93	* @version 1.3 12/18/96
	94	* @author Helena Shih
51004dcb	95	* @stable ICU 2.0
b75a7d8f A	96	*/
	97	class U_I18N_API CollationKey : public UObject {
	98	public:
	99	/**
	100	* This creates an empty collation key based on the null string. An empty
	101	* collation key contains no sorting information. When comparing two empty
	102	* collation keys, the result is Collator::EQUAL. Comparing empty collation key
	103	* with non-empty collation key is always Collator::LESS.
51004dcb	104	* @stable ICU 2.0
b75a7d8f A	105	*/
	106	CollationKey();
	107
	108
	109	/**
	110	* Creates a collation key based on the collation key values.
	111	* @param values the collation key values
	112	* @param count number of collation key values, including trailing nulls.
51004dcb	113	* @stable ICU 2.0
b75a7d8f A	114	*/
	115	CollationKey(const uint8_t* values,
	116	int32_t count);
	117
	118	/**
	119	* Copy constructor.
	120	* @param other the object to be copied.
51004dcb	121	* @stable ICU 2.0
b75a7d8f A	122	*/
	123	CollationKey(const CollationKey& other);
	124
	125	/**
	126	* Sort key destructor.
51004dcb	127	* @stable ICU 2.0
b75a7d8f	128	*/
374ca955	129	virtual ~CollationKey();
b75a7d8f A	130
	131	/**
	132	* Assignment operator
	133	* @param other the object to be copied.
51004dcb	134	* @stable ICU 2.0
b75a7d8f A	135	*/
	136	const CollationKey& operator=(const CollationKey& other);
	137
	138	/**
	139	* Compare if two collation keys are the same.
	140	* @param source the collation key to compare to.
	141	* @return Returns true if two collation keys are equal, false otherwise.
51004dcb	142	* @stable ICU 2.0
b75a7d8f A	143	*/
	144	UBool operator==(const CollationKey& source) const;
	145
	146	/**
	147	* Compare if two collation keys are not the same.
	148	* @param source the collation key to compare to.
	149	* @return Returns TRUE if two collation keys are different, FALSE otherwise.
51004dcb	150	* @stable ICU 2.0
b75a7d8f A	151	*/
	152	UBool operator!=(const CollationKey& source) const;
	153
	154
	155	/**
	156	* Test to see if the key is in an invalid state. The key will be in an
	157	* invalid state if it couldn't allocate memory for some operation.
	158	* @return Returns TRUE if the key is in an invalid, FALSE otherwise.
51004dcb	159	* @stable ICU 2.0
b75a7d8f A	160	*/
	161	UBool isBogus(void) const;
	162
	163	/**
	164	* Returns a pointer to the collation key values. The storage is owned
	165	* by the collation key and the pointer will become invalid if the key
	166	* is deleted.
	167	* @param count the output parameter of number of collation key values,
	168	* including any trailing nulls.
	169	* @return a pointer to the collation key values.
51004dcb	170	* @stable ICU 2.0
b75a7d8f A	171	*/
	172	const uint8_t* getByteArray(int32_t& count) const;
	173
	174	#ifdef U_USE_COLLATION_KEY_DEPRECATES
	175	/**
	176	* Extracts the collation key values into a new array. The caller owns
	177	* this storage and should free it.
	178	* @param count the output parameter of number of collation key values,
	179	* including any trailing nulls.
	180	* @obsolete ICU 2.6. Use getByteArray instead since this API will be removed in that release.
	181	*/
	182	uint8_t* toByteArray(int32_t& count) const;
	183	#endif
	184
51004dcb	185	#ifndef U_HIDE_DEPRECATED_API
b75a7d8f A	186	/**
	187	* Convenience method which does a string(bit-wise) comparison of the
	188	* two collation keys.
	189	* @param target target collation key to be compared with
	190	* @return Returns Collator::LESS if sourceKey < targetKey,
	191	* Collator::GREATER if sourceKey > targetKey and Collator::EQUAL
	192	* otherwise.
	193	* @deprecated ICU 2.6 use the overload with error code
	194	*/
	195	Collator::EComparisonResult compareTo(const CollationKey& target) const;
51004dcb	196	#endif /* U_HIDE_DEPRECATED_API */
b75a7d8f A	197
	198	/**
	199	* Convenience method which does a string(bit-wise) comparison of the
	200	* two collation keys.
	201	* @param target target collation key to be compared with
	202	* @param status error code
	203	* @return Returns UCOL_LESS if sourceKey < targetKey,
	204	* UCOL_GREATER if sourceKey > targetKey and UCOL_EQUAL
	205	* otherwise.
51004dcb	206	* @stable ICU 2.6
b75a7d8f A	207	*/
	208	UCollationResult compareTo(const CollationKey& target, UErrorCode &status) const;
	209
	210	/**
	211	* Creates an integer that is unique to the collation key. NOTE: this
	212	* is not the same as String.hashCode.
	213	* <p>Example of use:
	214	* <pre>
	215	* . UErrorCode status = U_ZERO_ERROR;
	216	* . Collator *myCollation = Collator::createInstance(Locale::US, status);
	217	* . if (U_FAILURE(status)) return;
	218	* . CollationKey key1, key2;
	219	* . UErrorCode status1 = U_ZERO_ERROR, status2 = U_ZERO_ERROR;
	220	* . myCollation->getCollationKey("abc", key1, status1);
	221	* . if (U_FAILURE(status1)) { delete myCollation; return; }
	222	* . myCollation->getCollationKey("ABC", key2, status2);
	223	* . if (U_FAILURE(status2)) { delete myCollation; return; }
	224	* . // key1.hashCode() != key2.hashCode()
	225	* </pre>
	226	* @return the hash value based on the string's collation order.
	227	* @see UnicodeString#hashCode
51004dcb	228	* @stable ICU 2.0
b75a7d8f A	229	*/
	230	int32_t hashCode(void) const;
	231
	232	/**
	233	* ICU "poor man's RTTI", returns a UClassID for the actual class.
51004dcb	234	* @stable ICU 2.2
b75a7d8f	235	*/
374ca955	236	virtual UClassID getDynamicClassID() const;
b75a7d8f A	237
	238	/**
	239	* ICU "poor man's RTTI", returns a UClassID for this class.
51004dcb	240	* @stable ICU 2.2
b75a7d8f	241	*/
374ca955	242	static UClassID U_EXPORT2 getStaticClassID();
b75a7d8f A	243
	244	private:
	245	/**
51004dcb A	246	* Replaces the current bytes buffer with a new one of newCapacity
	247	* and copies length bytes from the old buffer to the new one.
	248	* @return the new buffer, or NULL if the allocation failed
	249	*/
	250	uint8_t *reallocate(int32_t newCapacity, int32_t length);
4388f060 A	251	/**
	252	* Set a new length for a new sort key in the existing fBytes.
	253	*/
	254	void setLength(int32_t newLength);
b75a7d8f	255
51004dcb A	256	uint8_t *getBytes() {
	257	return (fFlagAndLength >= 0) ? fUnion.fStackBuffer : fUnion.fFields.fBytes;
	258	}
	259	const uint8_t *getBytes() const {
	260	return (fFlagAndLength >= 0) ? fUnion.fStackBuffer : fUnion.fFields.fBytes;
	261	}
	262	int32_t getCapacity() const {
	263	return (fFlagAndLength >= 0) ? (int32_t)sizeof(fUnion) : fUnion.fFields.fCapacity;
	264	}
	265	int32_t getLength() const { return fFlagAndLength & 0x7fffffff; }
b75a7d8f	266
374ca955 A	267	/**
	268	* Set the CollationKey to a "bogus" or invalid state
	269	* @return this CollationKey
	270	*/
b75a7d8f	271	CollationKey& setToBogus(void);
374ca955 A	272	/**
	273	* Resets this CollationKey to an empty state
	274	* @return this CollationKey
	275	*/
b75a7d8f	276	CollationKey& reset(void);
51004dcb	277
374ca955 A	278	/**
	279	* Allow private access to RuleBasedCollator
	280	*/
b75a7d8f	281	friend class RuleBasedCollator;
51004dcb A	282	friend class CollationKeyByteSink;
	283
	284	// Class fields. sizeof(CollationKey) is intended to be 48 bytes
	285	// on a machine with 64-bit pointers.
	286	// We use a union to maximize the size of the internal buffer,
	287	// similar to UnicodeString but not as tight and complex.
	288
	289	// (implicit) *vtable;
374ca955	290	/**
51004dcb A	291	* Sort key length and flag.
	292	* Bit 31 is set if the buffer is heap-allocated.
	293	* Bits 30..0 contain the sort key length.
	294	*/
	295	int32_t fFlagAndLength;
374ca955	296	/**
51004dcb A	297	* Unique hash value of this CollationKey.
51004dcb A	298	* Special value 2 if the key is bogus.
374ca955	299	*/
51004dcb	300	mutable int32_t fHashCode;
374ca955	301	/**
51004dcb A	302	* fUnion provides 32 bytes for the internal buffer or for
	303	* pointer+capacity.
	304	*/
	305	union StackBufferOrFields {
	306	/** fStackBuffer is used iff fFlagAndLength>=0, else fFields is used */
	307	uint8_t fStackBuffer[32];
	308	struct {
	309	uint8_t *fBytes;
	310	int32_t fCapacity;
	311	} fFields;
	312	} fUnion;
b75a7d8f A	313	};
b75a7d8f A	314
b75a7d8f A	315	inline UBool
	316	CollationKey::operator!=(const CollationKey& other) const
	317	{
	318	return !(*this == other);
	319	}
	320
	321	inline UBool
	322	CollationKey::isBogus() const
	323	{
51004dcb	324	return fHashCode == 2; // kBogusHashCode
b75a7d8f A	325	}
	326
	327	inline const uint8_t*
	328	CollationKey::getByteArray(int32_t &count) const
	329	{
51004dcb A	330	count = getLength();
51004dcb A	331	return getBytes();
b75a7d8f A	332	}
	333
	334	U_NAMESPACE_END
	335
	336	#endif /* #if !UCONFIG_NO_COLLATION */
	337
	338	#endif