source/transliteration/transliteration_body.cxx

/**************************************************************
 *
 * Licensed to the Apache Software Foundation (ASF) under one
 * or more contributor license agreements.  See the NOTICE file
 * distributed with this work for additional information
 * regarding copyright ownership.  The ASF licenses this file
 * to you under the Apache License, Version 2.0 (the
 * "License"); you may not use this file except in compliance
 * with the License.  You may obtain a copy of the License at
 *
 *   http://www.apache.org/licenses/LICENSE-2.0
 *
 * Unless required by applicable law or agreed to in writing,
 * software distributed under the License is distributed on an
 * "AS IS" BASIS, WITHOUT WARRANTIES OR CONDITIONS OF ANY
 * KIND, either express or implied.  See the License for the
 * specific language governing permissions and limitations
 * under the License.
 *
 *************************************************************/


// MARKER(update_precomp.py): autogen include statement, do not remove
#include "precompiled_i18npool.hxx"

#include <rtl/ustrbuf.hxx>
#include <i18nutil/casefolding.hxx>
#include <i18nutil/unicode.hxx>

#include <comphelper/processfactory.hxx>
#include <osl/diagnose.h>

#include <string.h>

#include "characterclassificationImpl.hxx"
#include "breakiteratorImpl.hxx"

#define TRANSLITERATION_ALL
#include "transliteration_body.hxx"

using namespace ::com::sun::star::uno;
using namespace ::com::sun::star::lang;
using namespace ::rtl;

#define A2OU(x) OUString::createFromAscii(x)

namespace com { namespace sun { namespace star { namespace i18n {


Transliteration_body::Transliteration_body()
{
	nMappingType = 0;
	transliterationName = "Transliteration_body";
	implementationName = "com.sun.star.i18n.Transliteration.Transliteration_body";
}

sal_Int16 SAL_CALL Transliteration_body::getType() throw(RuntimeException)
{
	return TransliterationType::ONE_TO_ONE;
}

sal_Bool SAL_CALL Transliteration_body::equals(
	const OUString& /*str1*/, sal_Int32 /*pos1*/, sal_Int32 /*nCount1*/, sal_Int32& /*nMatch1*/,
	const OUString& /*str2*/, sal_Int32 /*pos2*/, sal_Int32 /*nCount2*/, sal_Int32& /*nMatch2*/)
	throw(RuntimeException)
{
	throw RuntimeException();
}

Sequence< OUString > SAL_CALL
Transliteration_body::transliterateRange( const OUString& str1, const OUString& str2 )
	throw( RuntimeException)
{
	Sequence< OUString > ostr(2);
	ostr[0] = str1;
	ostr[1] = str2;
	return ostr;
}


static sal_uInt8 lcl_getMappingTypeForToggleCase( sal_uInt8 nMappingType, sal_Unicode cChar )
{
    sal_uInt8 nRes = nMappingType;

    // take care of TOGGLE_CASE transliteration:
    // nMappingType should not be a combination of flags, thuse we decide now
    // which one to use.
    if (nMappingType == (MappingTypeLowerToUpper | MappingTypeUpperToLower))
    {
        const sal_Int16 nType = unicode::getUnicodeType( cChar );
        if (nType & 0x02 /* lower case*/)
            nRes = MappingTypeLowerToUpper;
        else
        {
            // should also work properly for non-upper characters like white spacs, numbers, ...
            nRes = MappingTypeUpperToLower;
        }
    }

    return nRes;
}


OUString SAL_CALL
Transliteration_body::transliterate(
    const OUString& inStr, sal_Int32 startPos, sal_Int32 nCount,
	Sequence< sal_Int32 >& offset)
    throw(RuntimeException)
{
#if 0
/* Performance optimization:
 * The two realloc() consume 48% (32% grow, 16% shrink) runtime of this method!
 * getValue() needs about 15%, so there is equal balance if we trade the second
 * (shrinking) realloc() for a getValue(). But if the caller initializes the
 * sequence to nCount elements there isn't any change in size necessary in most
 * cases (one-to-one mapping) and we gain 33%.
 *
 * Of that constellation the getValue() method takes 20% upon each call, so 40%
 * for both. By remembering the first calls' results we could gain some extra
 * percentage again, but unfortunately getValue() may return a reference to a
 * static buffer, so we can't store the pointer directly but would have to
 * copy-construct an array, which doesn't give us any advantage.
 *
 * Much more is accomplished by working directly on the sequence buffer
 * returned by getArray() instead of using operator[] for each and every
 * access.
 *
 * And while we're at it: now that we know the size in advance we don't need to
 * copy the buffer anymore, just create the real string buffer and let the
 * return value take ownership.
 *
 * All together these changes result in the new implementation needing only 62%
 * of the time of the old implementation (in other words: that one was 1.61
 * times slower ...)
 */

    // Allocate the max possible buffer. Try to use stack instead of heap which
    // would have to be reallocated most times anyway.
    const sal_Int32 nLocalBuf = 512 * NMAPPINGMAX;
    sal_Unicode aLocalBuf[nLocalBuf], *out = aLocalBuf, *aHeapBuf = NULL;

    const sal_Unicode *in = inStr.getStr() + startPos;

    if (nCount > 512)
        out = aHeapBuf =  (sal_Unicode*) malloc((nCount * NMAPPINGMAX) * sizeof(sal_Unicode));

        if (useOffset)
            offset.realloc(nCount * NMAPPINGMAX);
	sal_Int32 j = 0;
	for (sal_Int32 i = 0; i < nCount; i++) {
	    Mapping &map = casefolding::getValue(in, i, nCount, aLocale, nMappingType);
	    for (sal_Int32 k = 0; k < map.nmap; k++) {
                if (useOffset)
                    offset[j] = i + startPos;
		out[j++] = map.map[k];
	    }
	}
        if (useOffset)
            offset.realloc(j);

	OUString r(out, j);

	if (aHeapBuf)
	    free(aHeapBuf);

	return r;
#else
    const sal_Unicode *in = inStr.getStr() + startPos;

    // Two different blocks to eliminate the if(useOffset) condition inside the
    // inner k loop. Yes, on massive use even such small things do count.
    if ( useOffset )
    {
        sal_Int32 nOffCount = 0, i;
        for (i = 0; i < nCount; i++)
        {
            // take care of TOGGLE_CASE transliteration:
            sal_uInt8 nTmpMappingType = nMappingType;
            if (nMappingType == (MappingTypeLowerToUpper | MappingTypeUpperToLower))
                nTmpMappingType = lcl_getMappingTypeForToggleCase( nMappingType, in[i] );

            const Mapping &map = casefolding::getValue( in, i, nCount, aLocale, nTmpMappingType );
            nOffCount += map.nmap;
        }
        rtl_uString* pStr = x_rtl_uString_new_WithLength( nOffCount );  // our x_rtl_ustring.h
        sal_Unicode* out = pStr->buffer;

        if ( nOffCount != offset.getLength() )
            offset.realloc( nOffCount );

        sal_Int32 j = 0;
        sal_Int32 * pArr = offset.getArray();
        for (i = 0; i < nCount; i++)
        {
            // take care of TOGGLE_CASE transliteration:
            sal_uInt8 nTmpMappingType = nMappingType;
            if (nMappingType == (MappingTypeLowerToUpper | MappingTypeUpperToLower))
                nTmpMappingType = lcl_getMappingTypeForToggleCase( nMappingType, in[i] );

            const Mapping &map = casefolding::getValue( in, i, nCount, aLocale, nTmpMappingType );
            for (sal_Int32 k = 0; k < map.nmap; k++)
            {
                pArr[j] = i + startPos;
                out[j++] = map.map[k];
            }
        }
        out[j] = 0;

        return OUString( pStr, SAL_NO_ACQUIRE ); // take over ownership of <pStr>
    }
    else
    {
        // In the simple case of no offset sequence used we can eliminate the
        // first getValue() loop. We could also assume that most calls result
        // in identical string lengths, thus using a preallocated
        // OUStringBuffer could be an easy way to assemble the return string
        // without too much hassle. However, for single characters the
        // OUStringBuffer::append() method is quite expensive compared to a
        // simple array operation, so it pays here to copy the final result
        // instead.

        // Allocate the max possible buffer. Try to use stack instead of heap,
        // which would have to be reallocated most times anyways.
        const sal_Int32 nLocalBuf = 2048;
        sal_Unicode aLocalBuf[ nLocalBuf * NMAPPINGMAX ], *out = aLocalBuf, *pHeapBuf = NULL;
        if ( nCount > nLocalBuf )
            out = pHeapBuf = new sal_Unicode[ nCount * NMAPPINGMAX ];

        sal_Int32 j = 0;
        for ( sal_Int32 i = 0; i < nCount; i++)
        {
            // take care of TOGGLE_CASE transliteration:
            sal_uInt8 nTmpMappingType = nMappingType;
            if (nMappingType == (MappingTypeLowerToUpper | MappingTypeUpperToLower))
                nTmpMappingType = lcl_getMappingTypeForToggleCase( nMappingType, in[i] );

            const Mapping &map = casefolding::getValue( in, i, nCount, aLocale, nTmpMappingType );
            for (sal_Int32 k = 0; k < map.nmap; k++)
            {
                out[j++] = map.map[k];
            }
        }

        OUString aRet( out, j );
        if ( pHeapBuf )
            delete [] pHeapBuf;
        return aRet;
    }
#endif
}

OUString SAL_CALL
Transliteration_body::transliterateChar2String( sal_Unicode inChar ) throw(RuntimeException)
{
        const Mapping &map = casefolding::getValue(&inChar, 0, 1, aLocale, nMappingType);
        rtl_uString* pStr = x_rtl_uString_new_WithLength( map.nmap );  // our x_rtl_ustring.h
        sal_Unicode* out = pStr->buffer;
        sal_Int32 i;

        for (i = 0; i < map.nmap; i++)
            out[i] = map.map[i];
        out[i] = 0;

        return OUString( pStr, SAL_NO_ACQUIRE ); // take over ownership of <pStr>
}

sal_Unicode SAL_CALL
Transliteration_body::transliterateChar2Char( sal_Unicode inChar ) throw(MultipleCharsOutputException, RuntimeException)
{
        const Mapping &map = casefolding::getValue(&inChar, 0, 1, aLocale, nMappingType);
        if (map.nmap > 1)
            throw MultipleCharsOutputException();
        return map.map[0];
}

OUString SAL_CALL
Transliteration_body::folding( const OUString& inStr, sal_Int32 startPos, sal_Int32 nCount,
	Sequence< sal_Int32 >& offset) throw(RuntimeException)
{
	return this->transliterate(inStr, startPos, nCount, offset);
}

Transliteration_casemapping::Transliteration_casemapping()
{
	nMappingType = 0;
	transliterationName = "casemapping(generic)";
	implementationName = "com.sun.star.i18n.Transliteration.Transliteration_casemapping";
}

void SAL_CALL
Transliteration_casemapping::setMappingType( const sal_uInt8 rMappingType, const Locale& rLocale )
{
	nMappingType = rMappingType;
	aLocale = rLocale;
}

Transliteration_u2l::Transliteration_u2l()
{
	nMappingType = MappingTypeUpperToLower;
	transliterationName = "upper_to_lower(generic)";
	implementationName = "com.sun.star.i18n.Transliteration.Transliteration_u2l";
}

Transliteration_l2u::Transliteration_l2u()
{
	nMappingType = MappingTypeLowerToUpper;
	transliterationName = "lower_to_upper(generic)";
	implementationName = "com.sun.star.i18n.Transliteration.Transliteration_l2u";
}

Transliteration_togglecase::Transliteration_togglecase()
{
    // usually nMappingType must NOT be a combiantion of different flages here,
    // but we take care of that problem in Transliteration_body::transliterate above
    // before that value is used. There we will decide which of both is to be used on
    // a per character basis.
    nMappingType = MappingTypeLowerToUpper | MappingTypeUpperToLower;
    transliterationName = "toggle(generic)";
    implementationName = "com.sun.star.i18n.Transliteration.Transliteration_togglecase";
}

Transliteration_titlecase::Transliteration_titlecase()
{
    nMappingType = MappingTypeToTitle;
    transliterationName = "title(generic)";
    implementationName = "com.sun.star.i18n.Transliteration.Transliteration_titlecase";
}

#if 0
struct LigatureData
{
    sal_uInt32  cChar;
    sal_Char *  pUtf8Text;
};

// available Unicode ligatures:
// http://www.unicode.org/charts
// http://www.unicode.org/charts/PDF/UFB00.pdf
static LigatureData aLigatures[] =
{
    { 0x0FB00,     "ff" },
    { 0x0FB01,     "fi" },
    { 0x0FB02,     "fl" },
    { 0x0FB03,     "ffi" },
    { 0x0FB04,     "ffl" },
    { 0x0FB05,     "ft" },
    { 0x0FB06,     "st" },

    { 0x0FB13,     "\xD5\xB4\xD5\xB6" },     // Armenian small men now
    { 0x0FB14,     "\xD5\xB4\xD5\xA5" },     // Armenian small men ech
    { 0x0FB15,     "\xD5\xB4\xD5\xAB" },     // Armenian small men ini
    { 0x0FB16,     "\xD5\xBE\xD5\xB6" },     // Armenian small vew now
    { 0x0FB17,     "\xD5\xB4\xD5\xAD" },     // Armenian small men xeh
    { 0x00000,     "" }
};

static inline bool lcl_IsLigature( sal_uInt32 cChar )
{
    return (0x0FB00 <= cChar && cChar <= 0x0FB06) || (0x0FB13 <= cChar && cChar <= 0x0FB17);
}

static rtl::OUString lcl_ResolveLigature( sal_uInt32 cChar )
{
    rtl::OUString aRes;
    if (lcl_IsLigature( cChar ))
    {
        LigatureData *pFound = NULL;
        LigatureData *pData = aLigatures;
        while (!pFound && pData->cChar != 0)
        {
            if (pData->cChar == cChar)
                pFound = pData;
            ++pData;
        }
        if (pFound)
            aRes = rtl::OUString( pFound->pUtf8Text, strlen( pFound->pUtf8Text ), RTL_TEXTENCODING_UTF8 );
    }
    else
        aRes = rtl::OUString( &cChar, 1 );
    return aRes;
}
#endif // if 0

static rtl::OUString transliterate_titlecase_Impl(
    const OUString& inStr, sal_Int32 startPos, sal_Int32 nCount,
    const Locale &rLocale,
	Sequence< sal_Int32 >& offset )
    throw(RuntimeException)
{
    const OUString aText( inStr.copy( startPos, nCount ) );

    OUString aRes;
    if (aText.getLength() > 0)
    {
        Reference< XMultiServiceFactory > xMSF = ::comphelper::getProcessServiceFactory();
        CharacterClassificationImpl aCharClassImpl( xMSF );

        // because aCharClassImpl.toTitle does not handle ligatures or � but will raise
        // an exception we need to handle the first chara manually...

        // we don't want to change surrogates by accident, thuse we use proper code point iteration
        sal_Int32 nPos = 0;
        sal_uInt32 cFirstChar = aText.iterateCodePoints( &nPos );
        OUString aResolvedLigature( &cFirstChar, 1 ); //lcl_ResolveLigature( cFirstChar ) );
        // toUpper can be used to properly resolve ligatures and characters like �
        aResolvedLigature = aCharClassImpl.toUpper( aResolvedLigature, 0, aResolvedLigature.getLength(), rLocale );
        // since toTitle will leave all-uppercase text unchanged we first need to
        // use toLower to bring possible 2nd and following charas in lowercase
        aResolvedLigature = aCharClassImpl.toLower( aResolvedLigature, 0, aResolvedLigature.getLength(), rLocale );
        sal_Int32 nResolvedLen = aResolvedLigature.getLength();

        // now we can properly use toTitle to get the expected result for the resolved string.
        // The rest of the text should just become lowercase.
        aRes = aCharClassImpl.toTitle( aResolvedLigature, 0, nResolvedLen, rLocale );
        aRes += aCharClassImpl.toLower( aText, 1, aText.getLength() - 1, rLocale );
        offset.realloc( aRes.getLength() );

        sal_Int32 *pOffset = offset.getArray();
        sal_Int32 nLen = offset.getLength();
        for (sal_Int32 i = 0; i < nLen; ++i)
        {
            sal_Int32 nIdx = 0;
            if (i >= nResolvedLen)
                nIdx = i - nResolvedLen + 1;
            pOffset[i] = nIdx;
        }
    }
#if OSL_DEBUG_LEVEL > 1
    const sal_Int32 *pCOffset = offset.getConstArray();
    (void) pCOffset;
#endif

    return aRes;
}


// this function expects to be called on a word-by-word basis,
// namely that startPos points to the first char of the word
rtl::OUString SAL_CALL Transliteration_titlecase::transliterate(
    const OUString& inStr, sal_Int32 startPos, sal_Int32 nCount,
	Sequence< sal_Int32 >& offset )
    throw(RuntimeException)
{
    return transliterate_titlecase_Impl( inStr, startPos, nCount, aLocale, offset );
}


Transliteration_sentencecase::Transliteration_sentencecase()
{
    nMappingType = MappingTypeToTitle;  // though only to be applied to the first word...
    transliterationName = "sentence(generic)";
    implementationName = "com.sun.star.i18n.Transliteration.Transliteration_sentencecase";
}


// this function expects to be called on a sentence-by-sentence basis,
// namely that startPos points to the first word (NOT first char!) in the sentence
rtl::OUString SAL_CALL Transliteration_sentencecase::transliterate(
    const OUString& inStr, sal_Int32 startPos, sal_Int32 nCount,
    Sequence< sal_Int32 >& offset )
    throw(RuntimeException)
{
    return transliterate_titlecase_Impl( inStr, startPos, nCount, aLocale, offset );
}


} } } }