[apple/icu.git] / icuSources / i18n / anytrans.cpp

// © 2016 and later: Unicode, Inc. and others.
// License & terms of use: http://www.unicode.org/copyright.html
/*
*****************************************************************
* Copyright (c) 2002-2016 International Business Machines Corporation
* and others.  All Rights Reserved.
*****************************************************************
* Date        Name        Description
* 06/06/2002  aliu        Creation.
*****************************************************************
*/

#include "unicode/utypes.h"

#if !UCONFIG_NO_TRANSLITERATION

#include "unicode/uobject.h"
#include "unicode/uscript.h"

#include "anytrans.h"
#include "hash.h"
#include "mutex.h"
#include "nultrans.h"
#include "putilimp.h"
#include "tridpars.h"
#include "uinvchar.h"
#include "uvector.h"

//------------------------------------------------------------
// Constants

static const UChar TARGET_SEP = 45; // '-'
static const UChar VARIANT_SEP = 47; // '/'
static const UChar ANY[] = {0x41,0x6E,0x79,0}; // "Any"
static const UChar NULL_ID[] = {78,117,108,108,0}; // "Null"
static const UChar LATIN_PIVOT[] = {0x2D,0x4C,0x61,0x74,0x6E,0x3B,0x4C,0x61,0x74,0x6E,0x2D,0}; // "-Latn;Latn-"

// initial size for an Any-XXXX transform's cache of script-XXXX transforms
// (will grow as necessary, but we don't expect to have source text wit more than 7 scripts)
#define ANY_TRANS_CACHE_INIT_SIZE 7

//------------------------------------------------------------

U_CDECL_BEGIN
/**
 * Deleter function for Transliterator*.
 */
static void U_CALLCONV
_deleteTransliterator(void *obj) {
    delete (icu::Transliterator*) obj;
}
U_CDECL_END

//------------------------------------------------------------

U_NAMESPACE_BEGIN

//------------------------------------------------------------
// ScriptRunIterator

/**
 * Returns a series of ranges corresponding to scripts. They will be
 * of the form:
 *
 * ccccSScSSccccTTcTcccc   - c = common, S = first script, T = second
 * |            |          - first run (start, limit)
 *          |           |  - second run (start, limit)
 *
 * That is, the runs will overlap. The reason for this is so that a
 * transliterator can consider common characters both before and after
 * the scripts.
 */
class ScriptRunIterator : public UMemory {
private:
    const Replaceable& text;
    int32_t textStart;
    int32_t textLimit;

public:
    /**
     * The code of the current run, valid after next() returns.  May
     * be USCRIPT_INVALID_CODE if and only if the entire text is
     * COMMON/INHERITED.
     */
    UScriptCode scriptCode;

    /**
     * The start of the run, inclusive, valid after next() returns.
     */
    int32_t start;

    /**
     * The end of the run, exclusive, valid after next() returns.
     */
    int32_t limit;

    /**
     * Constructs a run iterator over the given text from start
     * (inclusive) to limit (exclusive).
     */
    ScriptRunIterator(const Replaceable& text, int32_t start, int32_t limit);

    /**
     * Returns TRUE if there are any more runs.  TRUE is always
     * returned at least once.  Upon return, the caller should
     * examine scriptCode, start, and limit.
     */
    UBool next();

    /**
     * Adjusts internal indices for a change in the limit index of the
     * given delta.  A positive delta means the limit has increased.
     */
    void adjustLimit(int32_t delta);

private:
    ScriptRunIterator(const ScriptRunIterator &other); // forbid copying of this class
    ScriptRunIterator &operator=(const ScriptRunIterator &other); // forbid copying of this class
};

ScriptRunIterator::ScriptRunIterator(const Replaceable& theText,
                                     int32_t myStart, int32_t myLimit) :
    text(theText)
{
    textStart = myStart;
    textLimit = myLimit;
    limit = myStart;
}

UBool ScriptRunIterator::next() {
    UChar32 ch;
    UScriptCode s;
    UErrorCode ec = U_ZERO_ERROR;

    scriptCode = USCRIPT_INVALID_CODE; // don't know script yet
    start = limit;

    // Are we done?
    if (start == textLimit) {
        return FALSE;
    }

    // Move start back to include adjacent COMMON or INHERITED
    // characters
    while (start > textStart) {
        ch = text.char32At(start - 1); // look back
        s = uscript_getScript(ch, &ec);
        if (s == USCRIPT_COMMON || s == USCRIPT_INHERITED) {
            --start;
        } else {
            break;
        }
    }

    // Move limit ahead to include COMMON, INHERITED, and characters
    // of the current script.
    while (limit < textLimit) {
        ch = text.char32At(limit); // look ahead
        s = uscript_getScript(ch, &ec);
        if (s != USCRIPT_COMMON && s != USCRIPT_INHERITED) {
            if (scriptCode == USCRIPT_INVALID_CODE) {
                scriptCode = s;
            } else if (s != scriptCode) {
                break;
            }
        }
        ++limit;
    }

    // Return TRUE even if the entire text is COMMON / INHERITED, in
    // which case scriptCode will be USCRIPT_INVALID_CODE.
    return TRUE;
}

void ScriptRunIterator::adjustLimit(int32_t delta) {
    limit += delta;
    textLimit += delta;
}

//------------------------------------------------------------
// AnyTransliterator

UOBJECT_DEFINE_RTTI_IMPLEMENTATION(AnyTransliterator)

AnyTransliterator::AnyTransliterator(const UnicodeString& id,
                                     const UnicodeString& theTarget,
                                     const UnicodeString& theVariant,
                                     UScriptCode theTargetScript,
                                     UErrorCode& ec) :
    Transliterator(id, NULL),
    targetScript(theTargetScript)
{
    cache = uhash_openSize(uhash_hashLong, uhash_compareLong, NULL, ANY_TRANS_CACHE_INIT_SIZE, &ec);
    if (U_FAILURE(ec)) {
        return;
    }
    uhash_setValueDeleter(cache, _deleteTransliterator);

    target = theTarget;
    if (theVariant.length() > 0) {
        target.append(VARIANT_SEP).append(theVariant);
    }
}

AnyTransliterator::~AnyTransliterator() {
    uhash_close(cache);
}

/**
 * Copy constructor.
 */
AnyTransliterator::AnyTransliterator(const AnyTransliterator& o) :
    Transliterator(o),
    target(o.target),
    targetScript(o.targetScript)
{
    // Don't copy the cache contents
    UErrorCode ec = U_ZERO_ERROR;
    cache = uhash_openSize(uhash_hashLong, uhash_compareLong, NULL, ANY_TRANS_CACHE_INIT_SIZE, &ec);
    if (U_FAILURE(ec)) {
        return;
    }
    uhash_setValueDeleter(cache, _deleteTransliterator);
}

/**
 * Transliterator API.
 */
Transliterator* AnyTransliterator::clone() const {
    return new AnyTransliterator(*this);
}

/**
 * Implements {@link Transliterator#handleTransliterate}.
 */
void AnyTransliterator::handleTransliterate(Replaceable& text, UTransPosition& pos,
                                            UBool isIncremental) const {
    int32_t allStart = pos.start;
    int32_t allLimit = pos.limit;

    ScriptRunIterator it(text, pos.contextStart, pos.contextLimit);

    while (it.next()) {
        // Ignore runs in the ante context
        if (it.limit <= allStart) continue;

        // Try to instantiate transliterator from it.scriptCode to
        // our target or target/variant
        Transliterator* t = getTransliterator(it.scriptCode);

        if (t == NULL) {
            // We have no transliterator.  Do nothing, but keep
            // pos.start up to date.
            pos.start = it.limit;
            continue;
        }

        // If the run end is before the transliteration limit, do
        // a non-incremental transliteration.  Otherwise do an
        // incremental one.
        UBool incremental = isIncremental && (it.limit >= allLimit);

        pos.start = uprv_max(allStart, it.start);
        pos.limit = uprv_min(allLimit, it.limit);
        int32_t limit = pos.limit;
        t->filteredTransliterate(text, pos, incremental);
        int32_t delta = pos.limit - limit;
        allLimit += delta;
        it.adjustLimit(delta);

        // We're done if we enter the post context
        if (it.limit >= allLimit) break;
    }

    // Restore limit.  pos.start is fine where the last transliterator
    // left it, or at the end of the last run.
    pos.limit = allLimit;
}

Transliterator* AnyTransliterator::getTransliterator(UScriptCode source) const {

    if (source == targetScript || source == USCRIPT_INVALID_CODE) {
        return NULL;
    }

    Transliterator* t = NULL;
    {
        Mutex m(NULL);
        t = (Transliterator*) uhash_iget(cache, (int32_t) source);
    }
    if (t == NULL) {
        UErrorCode ec = U_ZERO_ERROR;
        UnicodeString sourceName(uscript_getShortName(source), -1, US_INV);
        UnicodeString id(sourceName);
        id.append(TARGET_SEP).append(target);

        t = Transliterator::createInstance(id, UTRANS_FORWARD, ec);
        if (U_FAILURE(ec) || t == NULL) {
            delete t;

            // Try to pivot around Latin, our most common script
            id = sourceName;
            id.append(LATIN_PIVOT, -1).append(target);
            t = Transliterator::createInstance(id, UTRANS_FORWARD, ec);
            if (U_FAILURE(ec) || t == NULL) {
                delete t;
                t = NULL;
            }
        }

        if (t != NULL) {
            Transliterator *rt = NULL;
            {
                Mutex m(NULL);
                rt = static_cast<Transliterator *> (uhash_iget(cache, (int32_t) source));
                if (rt == NULL) {
                    // Common case, no race to cache this new transliterator.
                    uhash_iput(cache, (int32_t) source, t, &ec);
                } else {
                    // Race case, some other thread beat us to caching this transliterator.
                    Transliterator *temp = rt;
                    rt = t;    // Our newly created transliterator that lost the race & now needs deleting.
                    t  = temp; // The transliterator from the cache that we will return.
                }
            }
            delete rt;    // will be non-null only in case of races.
        }
    }
    return t;
}

/**
 * Return the script code for a given name, or -1 if not found.
 */
static UScriptCode scriptNameToCode(const UnicodeString& name) {
    char buf[128];
    UScriptCode code;
    UErrorCode ec = U_ZERO_ERROR;
    int32_t nameLen = name.length();
    UBool isInvariant = uprv_isInvariantUString(name.getBuffer(), nameLen);

    if (isInvariant) {
        name.extract(0, nameLen, buf, (int32_t)sizeof(buf), US_INV);
        buf[127] = 0;   // Make sure that we NULL terminate the string.
    }
    if (!isInvariant || uscript_getCode(buf, &code, 1, &ec) != 1 || U_FAILURE(ec))
    {
        code = USCRIPT_INVALID_CODE;
    }
    return code;
}

/**
 * Registers standard transliterators with the system.  Called by
 * Transliterator during initialization.  Scan all current targets and
 * register those that are scripts T as Any-T/V.
 */
void AnyTransliterator::registerIDs() {

    UErrorCode ec = U_ZERO_ERROR;
    Hashtable seen(TRUE, ec);

    int32_t sourceCount = Transliterator::_countAvailableSources();
    for (int32_t s=0; s<sourceCount; ++s) {
        UnicodeString source;
        Transliterator::_getAvailableSource(s, source);

        // Ignore the "Any" source
        if (source.caseCompare(ANY, 3, 0 /*U_FOLD_CASE_DEFAULT*/) == 0) continue;

        int32_t targetCount = Transliterator::_countAvailableTargets(source);
        for (int32_t t=0; t<targetCount; ++t) {
            UnicodeString target;
            Transliterator::_getAvailableTarget(t, source, target);

            // Only process each target once
            if (seen.geti(target) != 0) continue;
            ec = U_ZERO_ERROR;
            seen.puti(target, 1, ec);

            // Get the script code for the target.  If not a script, ignore.
            UScriptCode targetScript = scriptNameToCode(target);
            if (targetScript == USCRIPT_INVALID_CODE) continue;

            int32_t variantCount = Transliterator::_countAvailableVariants(source, target);
            // assert(variantCount >= 1);
            for (int32_t v=0; v<variantCount; ++v) {
                UnicodeString variant;
                Transliterator::_getAvailableVariant(v, source, target, variant);

                UnicodeString id;
                TransliteratorIDParser::STVtoID(UnicodeString(TRUE, ANY, 3), target, variant, id);
                ec = U_ZERO_ERROR;
                AnyTransliterator* t = new AnyTransliterator(id, target, variant,
                                                             targetScript, ec);
                if (U_FAILURE(ec)) {
                    delete t;
                } else {
                    Transliterator::_registerInstance(t);
                    Transliterator::_registerSpecialInverse(target, UnicodeString(TRUE, NULL_ID, 4), FALSE);
                }
            }
        }
    }
}

U_NAMESPACE_END

#endif /* #if !UCONFIG_NO_TRANSLITERATION */

//eof
Commit	Line	Data
f3c0d7a5 A	1	// © 2016 and later: Unicode, Inc. and others.
f3c0d7a5 A	2	// License & terms of use: http://www.unicode.org/copyright.html
b75a7d8f A	3	/*
b75a7d8f A	4	*****************************************************************
2ca993e8	5	* Copyright (c) 2002-2016 International Business Machines Corporation
b75a7d8f A	6	* and others. All Rights Reserved.
	7	*****************************************************************
	8	* Date Name Description
	9	* 06/06/2002 aliu Creation.
	10	*****************************************************************
	11	*/
	12
	13	#include "unicode/utypes.h"
	14
	15	#if !UCONFIG_NO_TRANSLITERATION
	16
	17	#include "unicode/uobject.h"
	18	#include "unicode/uscript.h"
57a6839d	19
b75a7d8f	20	#include "anytrans.h"
b75a7d8f	21	#include "hash.h"
57a6839d A	22	#include "mutex.h"
57a6839d A	23	#include "nultrans.h"
374ca955	24	#include "putilimp.h"
57a6839d	25	#include "tridpars.h"
73c04bcf	26	#include "uinvchar.h"
57a6839d	27	#include "uvector.h"
b75a7d8f A	28
	29	//------------------------------------------------------------
	30	// Constants
	31
	32	static const UChar TARGET_SEP = 45; // '-'
	33	static const UChar VARIANT_SEP = 47; // '/'
2ca993e8	34	static const UChar ANY[] = {0x41,0x6E,0x79,0}; // "Any"
b75a7d8f	35	static const UChar NULL_ID[] = {78,117,108,108,0}; // "Null"
2ca993e8 A	36	static const UChar LATIN_PIVOT[] = {0x2D,0x4C,0x61,0x74,0x6E,0x3B,0x4C,0x61,0x74,0x6E,0x2D,0}; // "-Latn;Latn-"
	37
	38	// initial size for an Any-XXXX transform's cache of script-XXXX transforms
	39	// (will grow as necessary, but we don't expect to have source text wit more than 7 scripts)
	40	#define ANY_TRANS_CACHE_INIT_SIZE 7
b75a7d8f A	41
	42	//------------------------------------------------------------
	43
	44	U_CDECL_BEGIN
	45	/**
	46	* Deleter function for Transliterator*.
	47	*/
	48	static void U_CALLCONV
	49	_deleteTransliterator(void *obj) {
57a6839d	50	delete (icu::Transliterator*) obj;
b75a7d8f A	51	}
	52	U_CDECL_END
	53
	54	//------------------------------------------------------------
	55
	56	U_NAMESPACE_BEGIN
	57
	58	//------------------------------------------------------------
	59	// ScriptRunIterator
	60
	61	/**
	62	* Returns a series of ranges corresponding to scripts. They will be
	63	* of the form:
	64	*
	65	* ccccSScSSccccTTcTcccc - c = common, S = first script, T = second
	66	* \| \| - first run (start, limit)
	67	* \| \| - second run (start, limit)
	68	*
	69	* That is, the runs will overlap. The reason for this is so that a
	70	* transliterator can consider common characters both before and after
	71	* the scripts.
	72	*/
	73	class ScriptRunIterator : public UMemory {
	74	private:
	75	const Replaceable& text;
	76	int32_t textStart;
	77	int32_t textLimit;
	78
	79	public:
	80	/**
	81	* The code of the current run, valid after next() returns. May
	82	* be USCRIPT_INVALID_CODE if and only if the entire text is
	83	* COMMON/INHERITED.
	84	*/
	85	UScriptCode scriptCode;
	86
	87	/**
	88	* The start of the run, inclusive, valid after next() returns.
	89	*/
	90	int32_t start;
	91
	92	/**
	93	* The end of the run, exclusive, valid after next() returns.
	94	*/
	95	int32_t limit;
57a6839d	96
b75a7d8f A	97	/**
	98	* Constructs a run iterator over the given text from start
	99	* (inclusive) to limit (exclusive).
	100	*/
	101	ScriptRunIterator(const Replaceable& text, int32_t start, int32_t limit);
	102
	103	/**
	104	* Returns TRUE if there are any more runs. TRUE is always
	105	* returned at least once. Upon return, the caller should
	106	* examine scriptCode, start, and limit.
	107	*/
	108	UBool next();
	109
	110	/**
	111	* Adjusts internal indices for a change in the limit index of the
	112	* given delta. A positive delta means the limit has increased.
	113	*/
	114	void adjustLimit(int32_t delta);
	115
	116	private:
	117	ScriptRunIterator(const ScriptRunIterator &other); // forbid copying of this class
	118	ScriptRunIterator &operator=(const ScriptRunIterator &other); // forbid copying of this class
	119	};
	120
	121	ScriptRunIterator::ScriptRunIterator(const Replaceable& theText,
	122	int32_t myStart, int32_t myLimit) :
	123	text(theText)
	124	{
	125	textStart = myStart;
	126	textLimit = myLimit;
	127	limit = myStart;
	128	}
	129
	130	UBool ScriptRunIterator::next() {
	131	UChar32 ch;
	132	UScriptCode s;
	133	UErrorCode ec = U_ZERO_ERROR;
	134
	135	scriptCode = USCRIPT_INVALID_CODE; // don't know script yet
	136	start = limit;
	137
	138	// Are we done?
	139	if (start == textLimit) {
	140	return FALSE;
	141	}
	142
	143	// Move start back to include adjacent COMMON or INHERITED
	144	// characters
	145	while (start > textStart) {
	146	ch = text.char32At(start - 1); // look back
	147	s = uscript_getScript(ch, &ec);
	148	if (s == USCRIPT_COMMON \|\| s == USCRIPT_INHERITED) {
	149	--start;
	150	} else {
	151	break;
	152	}
	153	}
	154
	155	// Move limit ahead to include COMMON, INHERITED, and characters
	156	// of the current script.
	157	while (limit < textLimit) {
	158	ch = text.char32At(limit); // look ahead
	159	s = uscript_getScript(ch, &ec);
	160	if (s != USCRIPT_COMMON && s != USCRIPT_INHERITED) {
161	if (scriptCode == USCRIPT_INVALID_CODE) {
162	scriptCode = s;
163	} else if (s != scriptCode) {
164	break;
165	}
166	}
167	++limit;
168	}
169
170	// Return TRUE even if the entire text is COMMON / INHERITED, in
171	// which case scriptCode will be USCRIPT_INVALID_CODE.
172	return TRUE;
173	}
174
175	void ScriptRunIterator::adjustLimit(int32_t delta) {
176	limit += delta;
177	textLimit += delta;
178	}
179
180	//------------------------------------------------------------
181	// AnyTransliterator
182
374ca955	183	UOBJECT_DEFINE_RTTI_IMPLEMENTATION(AnyTransliterator)
b75a7d8f A	184
	185	AnyTransliterator::AnyTransliterator(const UnicodeString& id,
	186	const UnicodeString& theTarget,
	187	const UnicodeString& theVariant,
	188	UScriptCode theTargetScript,
	189	UErrorCode& ec) :
	190	Transliterator(id, NULL),
57a6839d	191	targetScript(theTargetScript)
b75a7d8f	192	{
2ca993e8	193	cache = uhash_openSize(uhash_hashLong, uhash_compareLong, NULL, ANY_TRANS_CACHE_INIT_SIZE, &ec);
46f4442e A	194	if (U_FAILURE(ec)) {
	195	return;
	196	}
b75a7d8f A	197	uhash_setValueDeleter(cache, _deleteTransliterator);
	198
	199	target = theTarget;
	200	if (theVariant.length() > 0) {
	201	target.append(VARIANT_SEP).append(theVariant);
	202	}
	203	}
	204
	205	AnyTransliterator::~AnyTransliterator() {
	206	uhash_close(cache);
	207	}
	208
	209	/**
	210	* Copy constructor.
	211	*/
	212	AnyTransliterator::AnyTransliterator(const AnyTransliterator& o) :
	213	Transliterator(o),
	214	target(o.target),
	215	targetScript(o.targetScript)
	216	{
	217	// Don't copy the cache contents
	218	UErrorCode ec = U_ZERO_ERROR;
2ca993e8	219	cache = uhash_openSize(uhash_hashLong, uhash_compareLong, NULL, ANY_TRANS_CACHE_INIT_SIZE, &ec);
46f4442e A	220	if (U_FAILURE(ec)) {
	221	return;
	222	}
b75a7d8f A	223	uhash_setValueDeleter(cache, _deleteTransliterator);
	224	}
	225
	226	/**
	227	* Transliterator API.
	228	*/
	229	Transliterator* AnyTransliterator::clone() const {
	230	return new AnyTransliterator(*this);
	231	}
	232
	233	/**
	234	* Implements {@link Transliterator#handleTransliterate}.
	235	*/
	236	void AnyTransliterator::handleTransliterate(Replaceable& text, UTransPosition& pos,
	237	UBool isIncremental) const {
	238	int32_t allStart = pos.start;
	239	int32_t allLimit = pos.limit;
	240
	241	ScriptRunIterator it(text, pos.contextStart, pos.contextLimit);
	242
	243	while (it.next()) {
	244	// Ignore runs in the ante context
	245	if (it.limit <= allStart) continue;
	246
	247	// Try to instantiate transliterator from it.scriptCode to
	248	// our target or target/variant
	249	Transliterator* t = getTransliterator(it.scriptCode);
57a6839d	250
b75a7d8f A	251	if (t == NULL) {
	252	// We have no transliterator. Do nothing, but keep
	253	// pos.start up to date.
	254	pos.start = it.limit;
	255	continue;
	256	}
	257
	258	// If the run end is before the transliteration limit, do
	259	// a non-incremental transliteration. Otherwise do an
	260	// incremental one.
	261	UBool incremental = isIncremental && (it.limit >= allLimit);
57a6839d	262
b75a7d8f A	263	pos.start = uprv_max(allStart, it.start);
	264	pos.limit = uprv_min(allLimit, it.limit);
	265	int32_t limit = pos.limit;
	266	t->filteredTransliterate(text, pos, incremental);
	267	int32_t delta = pos.limit - limit;
	268	allLimit += delta;
	269	it.adjustLimit(delta);
	270
	271	// We're done if we enter the post context
	272	if (it.limit >= allLimit) break;
	273	}
	274
	275	// Restore limit. pos.start is fine where the last transliterator
	276	// left it, or at the end of the last run.
	277	pos.limit = allLimit;
	278	}
	279
	280	Transliterator* AnyTransliterator::getTransliterator(UScriptCode source) const {
	281
	282	if (source == targetScript \|\| source == USCRIPT_INVALID_CODE) {
	283	return NULL;
	284	}
	285
57a6839d A	286	Transliterator* t = NULL;
	287	{
	288	Mutex m(NULL);
	289	t = (Transliterator*) uhash_iget(cache, (int32_t) source);
	290	}
b75a7d8f A	291	if (t == NULL) {
b75a7d8f A	292	UErrorCode ec = U_ZERO_ERROR;
2ca993e8	293	UnicodeString sourceName(uscript_getShortName(source), -1, US_INV);
b75a7d8f A	294	UnicodeString id(sourceName);
b75a7d8f A	295	id.append(TARGET_SEP).append(target);
57a6839d	296
b75a7d8f A	297	t = Transliterator::createInstance(id, UTRANS_FORWARD, ec);
	298	if (U_FAILURE(ec) \|\| t == NULL) {
	299	delete t;
57a6839d	300
b75a7d8f A	301	// Try to pivot around Latin, our most common script
b75a7d8f A	302	id = sourceName;
4388f060	303	id.append(LATIN_PIVOT, -1).append(target);
b75a7d8f A	304	t = Transliterator::createInstance(id, UTRANS_FORWARD, ec);
	305	if (U_FAILURE(ec) \|\| t == NULL) {
	306	delete t;
	307	t = NULL;
	308	}
	309	}
	310
	311	if (t != NULL) {
57a6839d A	312	Transliterator *rt = NULL;
	313	{
	314	Mutex m(NULL);
	315	rt = static_cast<Transliterator *> (uhash_iget(cache, (int32_t) source));
	316	if (rt == NULL) {
	317	// Common case, no race to cache this new transliterator.
	318	uhash_iput(cache, (int32_t) source, t, &ec);
	319	} else {
	320	// Race case, some other thread beat us to caching this transliterator.
	321	Transliterator *temp = rt;
	322	rt = t; // Our newly created transliterator that lost the race & now needs deleting.
	323	t = temp; // The transliterator from the cache that we will return.
	324	}
	325	}
	326	delete rt; // will be non-null only in case of races.
b75a7d8f A	327	}
b75a7d8f A	328	}
b75a7d8f A	329	return t;
	330	}
	331
	332	/**
	333	* Return the script code for a given name, or -1 if not found.
	334	*/
73c04bcf	335	static UScriptCode scriptNameToCode(const UnicodeString& name) {
b75a7d8f A	336	char buf[128];
	337	UScriptCode code;
	338	UErrorCode ec = U_ZERO_ERROR;
73c04bcf A	339	int32_t nameLen = name.length();
73c04bcf A	340	UBool isInvariant = uprv_isInvariantUString(name.getBuffer(), nameLen);
57a6839d	341
73c04bcf A	342	if (isInvariant) {
	343	name.extract(0, nameLen, buf, (int32_t)sizeof(buf), US_INV);
	344	buf[127] = 0; // Make sure that we NULL terminate the string.
	345	}
	346	if (!isInvariant \|\| uscript_getCode(buf, &code, 1, &ec) != 1 \|\| U_FAILURE(ec))
	347	{
b75a7d8f A	348	code = USCRIPT_INVALID_CODE;
	349	}
	350	return code;
	351	}
	352
	353	/**
	354	* Registers standard transliterators with the system. Called by
	355	* Transliterator during initialization. Scan all current targets and
	356	* register those that are scripts T as Any-T/V.
	357	*/
	358	void AnyTransliterator::registerIDs() {
	359
374ca955 A	360	UErrorCode ec = U_ZERO_ERROR;
374ca955 A	361	Hashtable seen(TRUE, ec);
b75a7d8f A	362
	363	int32_t sourceCount = Transliterator::_countAvailableSources();
	364	for (int32_t s=0; s<sourceCount; ++s) {
	365	UnicodeString source;
	366	Transliterator::_getAvailableSource(s, source);
	367
	368	// Ignore the "Any" source
4388f060	369	if (source.caseCompare(ANY, 3, 0 /U_FOLD_CASE_DEFAULT/) == 0) continue;
b75a7d8f A	370
	371	int32_t targetCount = Transliterator::_countAvailableTargets(source);
	372	for (int32_t t=0; t<targetCount; ++t) {
	373	UnicodeString target;
	374	Transliterator::_getAvailableTarget(t, source, target);
	375
	376	// Only process each target once
	377	if (seen.geti(target) != 0) continue;
	378	ec = U_ZERO_ERROR;
	379	seen.puti(target, 1, ec);
57a6839d	380
b75a7d8f A	381	// Get the script code for the target. If not a script, ignore.
	382	UScriptCode targetScript = scriptNameToCode(target);
	383	if (targetScript == USCRIPT_INVALID_CODE) continue;
	384
	385	int32_t variantCount = Transliterator::_countAvailableVariants(source, target);
	386	// assert(variantCount >= 1);
	387	for (int32_t v=0; v<variantCount; ++v) {
	388	UnicodeString variant;
	389	Transliterator::_getAvailableVariant(v, source, target, variant);
57a6839d	390
b75a7d8f	391	UnicodeString id;
4388f060	392	TransliteratorIDParser::STVtoID(UnicodeString(TRUE, ANY, 3), target, variant, id);
b75a7d8f A	393	ec = U_ZERO_ERROR;
	394	AnyTransliterator* t = new AnyTransliterator(id, target, variant,
	395	targetScript, ec);
	396	if (U_FAILURE(ec)) {
	397	delete t;
	398	} else {
	399	Transliterator::_registerInstance(t);
4388f060	400	Transliterator::_registerSpecialInverse(target, UnicodeString(TRUE, NULL_ID, 4), FALSE);
b75a7d8f A	401	}
	402	}
	403	}
	404	}
	405	}
	406
	407	U_NAMESPACE_END
	408
	409	#endif /* #if !UCONFIG_NO_TRANSLITERATION */
	410
	411	//eof