firebird-mirror/src/intl/lc_narrow.cpp

/*
 *	PROGRAM:	InterBase International support
 *	MODULE:		lc_narrow.cpp
 *	DESCRIPTION:	Common base for Narrow language drivers
 *			(full International collation)
 *
 * The contents of this file are subject to the Interbase Public
 * License Version 1.0 (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.Inprise.com/IPL.html
 *
 * Software distributed under the License is distributed on an
 * "AS IS" basis, WITHOUT WARRANTY OF ANY KIND, either express
 * or implied. See the License for the specific language governing
 * rights and limitations under the License.
 *
 * The Original Code was created by Inprise Corporation
 * and its predecessors. Portions created by Inprise Corporation are
 * Copyright (C) Inprise Corporation.
 *
 * All Rights Reserved.
 * Contributor(s): ______________________________________.
 */

#include "firebird.h"
#include "../intl/ldcommon.h"
#include "../jrd/CharSet.h"
#include "../jrd/IntlUtil.h"
#include "lc_narrow.h"
#include "ld_proto.h"
#include <limits.h>
#include <math.h>

using namespace Firebird;


static ULONG fam2_str_to_upper(texttype* obj, ULONG iLen, const BYTE* pStr, ULONG iOutLen, BYTE *pOutStr);
static ULONG fam2_str_to_lower(texttype* obj, ULONG iLen, const BYTE* pStr, ULONG iOutLen, BYTE *pOutStr);


const USHORT LANGFAM2_MAX_KEY	= MAX_KEY;
const BYTE ASCII_SPACE			= 32;
const UINT16 NULL_WEIGHT		= 0;
const UINT16 NULL_SECONDARY		= 0;
const UINT16 NULL_TERTIARY		= 0;


/*
 * key_length (obj, inLen)
 *
 * For an input string of (inLen) bytes, return the maximum
 * key buffer length.
 *
 * This is used for index buffer allocation within the
 * Engine.
 *
 *  Note:
 *	Strings containing (m) expand chars need 2*(m)*3
 *	bytes for key values of the expansion.  This is
 *	offset by string values that don't have corresponding
 *	secondary or tertiary key values.
 *	Let:
 *		n = length of input string
 *		np = count of "simple" bytes in the string, alphabetic
 *		     no secondary and no tertiary.
 *		ns = Has secondary or tertiary, but not both
 *		nt = Has tertiary and secondary.
 *		nc = is a COMPRESSED value.
 *		ne = Has an EXPAND value.
 *		nsp = Is a special value.
 *
 *		n = np + ns + nt + nc + ne + nsp
 *
 *		Key_length(n) =
 *			np
 *			+ 2 * ns
 *			+ 3 * nt
 *			+ 3 * (nc/2)
 *			+ 3 * 2 * ne 
 *			+ 2 * nsp
 *			+ 1 (if nsp > 0, for separating keys from special keys)
 *
 *	Clearly this is maximized when the string consists solely of
 *	EXPAND characters.  This degenerate case doesn't occur in
 *	standard text usage, except for short strings (1-2 characters).
 *
 *	Therefore, we compute the keylength based on the "normal" case
 *	of the (nt) term.  It is likely we could choose a probabilistic value
 *	(such as 2.5 * n) for the length of the key.
 *
 *	The degenerate case of short strings is handled by a minimal key
 *	length.
 */
USHORT LC_NARROW_key_length(texttype* obj, USHORT inLen)
{
/* fb_assert (inLen <= LANGFAM2_MAX_KEY); *//* almost certainly an error */

	// is it the first time key_length is called?
	if (obj->texttype_impl->texttype_bytes_per_key == 0)
	{
		BYTE bytesPerChar = 3;

		// if collation is not multi-level, the weights used is already know
		if (obj->texttype_impl->texttype_flags & TEXTTYPE_non_multi_level)
		{
			if (obj->texttype_impl->texttype_flags & TEXTTYPE_secondary_insensitive)
				--bytesPerChar;

			if (obj->texttype_impl->texttype_flags & TEXTTYPE_tertiary_insensitive)
				--bytesPerChar;
		}
		else	// scan the table to identify what weights are used
		{
			bool useSecondary = false;
			bool useTertiary = false;

			for (int ch = 0; ch <= 255; ++ch)
			{
				const SortOrderTblEntry* coll =
					&((const SortOrderTblEntry*)obj->texttype_impl->texttype_collation_table)[ch];

				if (coll->Secondary != NULL_SECONDARY)
					useSecondary = true;

				if (coll->Tertiary != NULL_TERTIARY)
					useTertiary = true;
			}

			if (!useSecondary)
				--bytesPerChar;

			if (!useTertiary)
				--bytesPerChar;
		}

		obj->texttype_impl->texttype_bytes_per_key = bytesPerChar;
	}

	USHORT len = obj->texttype_impl->texttype_bytes_per_key * MAX(inLen, 2);

	if (obj->texttype_impl->texttype_expand_table &&
		((const ExpandChar*) obj->texttype_impl->texttype_expand_table)[0].Ch)
	{
		len += (USHORT) log10(inLen + 1.0) * 4 * obj->texttype_impl->texttype_bytes_per_key;
	}

	return (MIN(len, LANGFAM2_MAX_KEY));
}



#ifdef DEBUG
#include <stdio.h>
static ULONG do_debug = 0;
#endif

/*
 *
 *  Convert a user string to a sequence that will collate bytewise.
 * 
 * RETURN:
 *		Length, in bytes, of returned key
 */
USHORT LC_NARROW_string_to_key(texttype* obj, USHORT iInLen, const BYTE* pInChar, USHORT iOutLen, BYTE *pOutChar,
							   USHORT key_type)
{
	fb_assert(pOutChar != NULL);
	fb_assert(pInChar != NULL);
/* fb_assert (iInLen   <= LANGFAM2_MAX_KEY); */
	fb_assert(iOutLen <= LANGFAM2_MAX_KEY);
	fb_assert(iOutLen >= LC_NARROW_key_length(obj, iInLen));

#ifdef DEBUG
/* Dump out the input string */
	if (do_debug) {
		printf("string: (%02d) '%*s'\n", iInLen, iInLen, pInChar);
		fflush(stdout);
	}
#endif	/* DEBUG */

	BYTE* outbuff = pOutChar;
	USHORT lprimary = 0;
	USHORT lsecondary = 0;
	USHORT ltertiary = 0;
	USHORT lspecial = 0;

	BYTE secondary[LANGFAM2_MAX_KEY];
	BYTE tertiary[LANGFAM2_MAX_KEY];
	BYTE special[LANGFAM2_MAX_KEY * 2];

/* point inbuff at last character */
	const BYTE* inbuff = pInChar + iInLen - 1;

	if (obj->texttype_pad_option)
	{
		/* skip backwards over all spaces & reset input length */
		while ((inbuff >= pInChar) && (*inbuff == ASCII_SPACE))
			inbuff--;
	}

	iInLen = (inbuff - pInChar + 1);
	
	USHORT i;

	for (i = 0; i < iInLen; i++, pInChar++) {
		fb_assert(lprimary < iOutLen);
		fb_assert(lsecondary < sizeof(secondary));
		fb_assert(ltertiary < sizeof(tertiary));
		fb_assert(lspecial < sizeof(special));

		const SortOrderTblEntry* coll =
			&((const SortOrderTblEntry*) obj->texttype_impl->
			  texttype_collation_table)[*pInChar];
		if (!(coll->IsExpand || coll->IsCompress)) {
			if (coll->Primary != NULL_WEIGHT && lprimary < iOutLen)
				outbuff[lprimary++] = coll->Primary + obj->texttype_impl->primary_sum;
			if (coll->Secondary != NULL_SECONDARY && lsecondary < sizeof(secondary))
				secondary[lsecondary++] = coll->Secondary;
			if (coll->Tertiary != NULL_TERTIARY && ltertiary < sizeof(tertiary))
				tertiary[ltertiary++] = coll->Tertiary;
		}
		else if (coll->IsExpand && coll->IsCompress) {
			/* Both flags set indicate a special value */

			if (obj->texttype_impl->texttype_flags & TEXTTYPE_specials_first)
			{
				if (coll->Primary != NULL_WEIGHT && lprimary < iOutLen)
					outbuff[lprimary++] = coll->Primary + obj->texttype_impl->ignore_sum;
				if (coll->Secondary != NULL_SECONDARY && lsecondary < sizeof(secondary))
					secondary[lsecondary++] = coll->Secondary;
				if (coll->Tertiary != NULL_TERTIARY && ltertiary < sizeof(tertiary))
					tertiary[ltertiary++] = coll->Tertiary;
			}
			else
			{
				if ((coll->Primary != NULL_WEIGHT) &&
					!(obj->texttype_impl->texttype_flags & TEXTTYPE_ignore_specials) &&
					 lspecial + 1 < sizeof(special))
				{
					special[lspecial++] = (i + 1);	/* position */
					special[lspecial++] = coll->Primary;
				}
			}
		}
		else if (coll->IsExpand) {
			const ExpandChar* exp = &((const ExpandChar*) obj->texttype_impl->texttype_expand_table)[0];
			while (exp->Ch && exp->Ch != *pInChar)
				exp++;
			fb_assert(exp->Ch == *pInChar);
			for (int j = 0; j < 2; j++) {
				if (j)
					coll =
						&((const SortOrderTblEntry*) obj->texttype_impl->
						  texttype_collation_table)[exp->ExpCh2];
				if (coll->Primary != NULL_WEIGHT && lprimary < iOutLen)
					outbuff[lprimary++] = coll->Primary;
				if (coll->Secondary != NULL_SECONDARY && lsecondary < sizeof(secondary))
					secondary[lsecondary++] = coll->Secondary;
				if (coll->Tertiary != NULL_TERTIARY && ltertiary < sizeof(tertiary))
					tertiary[ltertiary++] = coll->Tertiary;
			}
		}
		else {					/* (col->IsCompress) */
			bool complete = (USHORT) (i + 1) < iInLen;

			if (complete) {
				const CompressPair* cmp =
					&((const CompressPair*) obj->texttype_impl->
					  texttype_compress_table)[0];
				while (cmp->CharPair[0]) {
					if ((cmp->CharPair[0] == *pInChar) &&
						(cmp->CharPair[1] == *(pInChar + 1)))
					{
						/* Gobble the two-to-1 entry */
						coll = &cmp->NoCaseWeight;
						pInChar++;
						i++;
						break;
					}
					cmp++;
				}
			}

			// ASF: If key_type == INTL_KEY_PARTIAL and the last CompressPair
			// isn't complete, don't put the byte in the sortkey. If we put,
			// incorrect results occur when using index.
			if (key_type != INTL_KEY_PARTIAL || complete)
			{
				if (coll->Primary != NULL_WEIGHT && lprimary < iOutLen)
					outbuff[lprimary++] = coll->Primary;
				if (coll->Secondary != NULL_SECONDARY && lsecondary < sizeof(secondary))
					secondary[lsecondary++] = coll->Secondary;
				if (coll->Tertiary != NULL_TERTIARY && ltertiary < sizeof(tertiary))
					tertiary[ltertiary++] = coll->Tertiary;
			}
		}
	}

/* primary keys are already in output key */

	outbuff += lprimary;
	iOutLen -= lprimary;

	if (key_type == INTL_KEY_PARTIAL)
		/* return length of key */
		return (outbuff - pOutChar);

	bool useLevel = !(obj->texttype_impl->texttype_flags & TEXTTYPE_secondary_insensitive);

	if (!(obj->texttype_impl->texttype_flags & TEXTTYPE_non_multi_level))	// multi-level
	{
		if (key_type == INTL_KEY_SORT)
			useLevel = true;
	}

	if (useLevel)
	{
		/* put secondary keys into output key */
		if (obj->texttype_impl->texttype_flags & TEXTTYPE_reverse_secondary) {
			for (i = 0; i < lsecondary && iOutLen; i++) {
				*outbuff++ = secondary[lsecondary - i - 1];
				iOutLen--;
			}
		}
		else {
			for (i = 0; i < lsecondary && iOutLen; i++) {
				*outbuff++ = secondary[i];
				iOutLen--;
			}
		}
	}

	useLevel = !(obj->texttype_impl->texttype_flags & TEXTTYPE_tertiary_insensitive);

	if (!(obj->texttype_impl->texttype_flags & TEXTTYPE_non_multi_level))	// multi-level
	{
		if (key_type == INTL_KEY_SORT)
			useLevel = true;
	}

	if (useLevel)
	{
		/* put tertiary keys into output key */
		for (i = 0; i < ltertiary && iOutLen; i++) {
			*outbuff++ = tertiary[i];
			iOutLen--;
		}
	}

	/* put special keys into output key */
	if ((lspecial && iOutLen) &&
		!(obj->texttype_impl->texttype_flags & TEXTTYPE_ignore_specials))
	{
		/* Insert the marker-byte */
		*outbuff++ = 0;
		iOutLen--;
		for (i = 0; i < lspecial && iOutLen; i++) {
			*outbuff++ = special[i];
			iOutLen--;
		}
	}

#ifdef DEBUG
/* Dump out the computed key */
	if (do_debug) {
		printf("   key: (%02d) ", (outbuff - pOutChar));
		for (const UCHAR* p = pOutChar; p < outbuff; p++)
			printf("%2x ", *p);
		printf("\n");
		fflush(stdout);
	}
#endif

/* return length of key */
	return (outbuff - pOutChar);
}



#define	LC_HAVE_WAITING	1
#define LC_HAVE_SPECIAL	2

/* expansion char go before the expansion. */
/* eg: S-set collates before ss */

struct coltab_status {
	USHORT stat_flags;
	const SortOrderTblEntry* stat_waiting;
};




static SSHORT special_scan(texttype* obj, ULONG l1, const BYTE* s1, ULONG l2, const BYTE* s2)
{
	const SortOrderTblEntry* col1 = 0;
	const SortOrderTblEntry* col2 = 0;

	ULONG index1 = 0;
	ULONG index2 = 0;

	while (true) {
		/* Scan to find ignore char from l1 */
		while (l1) {
			col1 =
				&((const SortOrderTblEntry*) obj->texttype_impl->
				  texttype_collation_table)[*s1];

			if (col1->IsExpand && col1->IsCompress &&
				!(obj->texttype_impl->texttype_flags & TEXTTYPE_specials_first))
			{
				break;
			}

			l1--;
			s1++;
			index1++;
		}

		/* Scan to find ignore char from l2 */
		while (l2) {
			col2 =
				&((const SortOrderTblEntry*) obj->texttype_impl->
				  texttype_collation_table)[*s2];
			if (col2->IsExpand && col2->IsCompress &&
				!(obj->texttype_impl->texttype_flags & TEXTTYPE_specials_first))
			{
				break;
			}

			l2--;
			s2++;
			index2++;
		}
		if (!l1 && !l2)			/* All out of ignore characters */
			return 0;
		if (l1 && !l2)			/* Out in l2 only */
			return 1000;
		if (!l1 && l2)			/* Out in l1 only */
			return -1000;
		if (index1 < index2)	/* l1 has ignore ch before l2 */
			return -2000;
		if (index1 > index2)	/* l2 has ignore ch before l1 */
			return 2000;
		if (col1->Primary != col2->Primary)
			return (col1->Primary - col2->Primary);
		l1--;
		l2--;
		s1++;
		s2++;
		index1++;
		index2++;
	}
}


static const SortOrderTblEntry* get_coltab_entry(texttype* obj, const UCHAR** p,
	ULONG* l, coltab_status* stat, int* sum)
{
	*sum = obj->texttype_impl->primary_sum;

	if (stat->stat_flags & LC_HAVE_WAITING) {
		(*l)--;
		(*p)++;
		stat->stat_flags &= ~LC_HAVE_WAITING;
		fb_assert(stat->stat_waiting);
		return stat->stat_waiting;
	}

	stat->stat_waiting = NULL;
	while (*l) {
		const SortOrderTblEntry* col =
			&((const SortOrderTblEntry*) obj->texttype_impl->
			  texttype_collation_table)[**p];
		if (!(col->IsExpand || col->IsCompress))
		{
			/* Have col */
			(*l)--;
			(*p)++;
			return col;
		}
		else if (col->IsExpand && col->IsCompress)
		{
			if (obj->texttype_impl->texttype_flags & TEXTTYPE_specials_first)
			{
				*sum = obj->texttype_impl->ignore_sum;

				/* Have col */
				(*l)--;
				(*p)++;
				return col;
			}
			else
			{
				/* Both flags set indicate a special value */
				/* Need a new col */
				(*l)--;
				(*p)++;
				stat->stat_flags |= LC_HAVE_SPECIAL;
				continue;
			}
		}
		else if (col->IsExpand) {
			const ExpandChar* exp = &((const ExpandChar*) obj->texttype_impl->texttype_expand_table)[0];
			while (exp->Ch && exp->Ch != **p)
				exp++;
			fb_assert(exp->Ch == **p);
			/* Have coll1 */
			/* Have waiting */

			stat->stat_waiting =
				&((const SortOrderTblEntry*) obj->texttype_impl->
				  texttype_collation_table)[exp->ExpCh2];
			stat->stat_flags |= LC_HAVE_WAITING;
			return col;
		}
		else {					/* (col->IsCompress) */
			if (*l > 1) {
				const CompressPair* cmp =
					&((const CompressPair*) obj->texttype_impl->
					  texttype_compress_table)[0];
				while (cmp->CharPair[0]) {
					if ((cmp->CharPair[0] == **p) &&
						(cmp->CharPair[1] == *(*p + 1)))
					{
						/* Have Col */
						col = &cmp->NoCaseWeight;
						(*l) -= 2;
						(*p) += 2;
						return col;
					}
					cmp++;
				}
			}
			/* Have col */
			(*l)--;
			(*p)++;
			return col;
		}
	}
	return NULL;
}



#define XOR	^					/* C bitwise XOR operator - defined for readability */

SSHORT LC_NARROW_compare(texttype* obj, ULONG l1, const BYTE* s1, ULONG l2, const BYTE* s2, 
	INTL_BOOL* error_flag)
{
	fb_assert(obj != NULL);
	fb_assert(s1 != NULL);
	fb_assert(s2 != NULL);
	fb_assert(error_flag != NULL);

	*error_flag = false;

	if (obj->texttype_pad_option)
	{
		/* Start at EOS, scan backwards to find non-space */
		const BYTE* p = s1 + l1 - 1;
		while ((p >= s1) && (*p == ASCII_SPACE))
			p--;
		l1 = (p - s1 + 1);

		p = s2 + l2 - 1;
		while ((p >= s2) && (*p == ASCII_SPACE))
			p--;
		l2 = (p - s2 + 1);
	}

	const ULONG save_l1 = l1;
	const ULONG save_l2 = l2;
	const BYTE* const save_s1 = s1;
	const BYTE* const save_s2 = s2;
	SSHORT save_secondary = 0;
	SSHORT save_tertiary = 0;
	SSHORT save_quandary = 0;

	coltab_status stat1, stat2;
	stat1.stat_flags = 0;
	stat2.stat_flags = 0;
	
	const SortOrderTblEntry* col1 = 0;
	const SortOrderTblEntry* col2 = 0;

	while (true) {
		int sum1, sum2;

		col1 = get_coltab_entry(obj, &s1, &l1, &stat1, &sum1);
		col2 = get_coltab_entry(obj, &s2, &l2, &stat2, &sum2);

		if (!col1 || !col2)
			break;
		if (col1->Primary + sum1 != col2->Primary + sum2)
			return ((col1->Primary + sum1) - (col2->Primary + sum2));
		if ((obj->texttype_impl->texttype_flags & TEXTTYPE_secondary_insensitive) == 0 &&
			col1->Secondary != col2->Secondary)
		{
			if ((obj->texttype_impl->texttype_flags & TEXTTYPE_reverse_secondary) ||
				!save_secondary)
			{
				save_secondary = (col1->Secondary - col2->Secondary);
			}
		}
		else if ((obj->texttype_impl->texttype_flags & TEXTTYPE_tertiary_insensitive) == 0 &&
				 col1->Tertiary != col2->Tertiary)
		{
			if (!save_tertiary)
				save_tertiary = (col1->Tertiary - col2->Tertiary);
		}
		else if (((stat1.stat_flags & LC_HAVE_WAITING) XOR
				  (stat2.stat_flags & LC_HAVE_WAITING)) && !save_quandary)
		{
			if (obj->texttype_impl->texttype_flags & TEXTTYPE_expand_before)
				save_quandary = (stat1.stat_flags & LC_HAVE_WAITING) ? -1 : 1;
			else
				save_quandary = (stat1.stat_flags & LC_HAVE_WAITING) ? 1 : -1;
		}
	}

/* One of the strings ended */
	fb_assert(l1 == 0 || l2 == 0);
	fb_assert(col1 == NULL || col2 == NULL);

	if (col1 && !col2)
		return 500;

	if (!col1 && col2)
		return -500;

	if (l1 == 0 && l2 == 0) {
		if (save_secondary)
			return save_secondary;
		if (save_tertiary)
			return save_tertiary;
		if (save_quandary)
			return save_quandary;
		if (
			((stat1.stat_flags & LC_HAVE_SPECIAL)
			 || (stat2.stat_flags & LC_HAVE_SPECIAL))
			&& !(obj->texttype_impl->texttype_flags & TEXTTYPE_ignore_specials)
			&& !(obj->texttype_impl->texttype_flags & TEXTTYPE_specials_first))
		{
			return special_scan(obj, save_l1, save_s1, save_l2, save_s2);
		}
		return 0;
	}

	if (l1)
		return 600;
	return -600;
}



#ifdef DEBUG_COMPARE

/*
 * Debugging only
 * Routine used for comparing results from comparison algorithm
 * to results from key creation algorithm
 */
static SSHORT old_fam2_compare(texttype* obj, ULONG l1, const BYTE* s1,
	ULONG l2, const BYTE* s2, INTL_BOOL* error_flag)
{
	BYTE key1[LANGFAM2_MAX_KEY];
	BYTE key2[LANGFAM2_MAX_KEY];

	fb_assert(obj != NULL);
	fb_assert(s1 != NULL);
	fb_assert(s2 != NULL);

	const ULONG len1 = LC_NARROW_string_to_key(obj, l1, s1, sizeof(key1), key1, INTL_KEY_SORT);
	const ULONG len2 = LC_NARROW_string_to_key(obj, l2, s2, sizeof(key2), key2, INTL_KEY_SORT);
	const ULONG len = MIN(len1, len2);
	for (ULONG i = 0; i < len; i++) {
		if (key1[i] == key2[i])
			continue;
		else if (key1[i] < key2[i])
			return (-1);
		else
			return (1);
	}
	if (len1 < len2)
		return (-1);
	else if (len1 > len2)
		return (1);
	return (0);
}
#endif	/* DEBUG_COMPARE */


#ifdef DEBUG_COMPARE

#define	SIGN(x)	(((x) < 0) ? -1 : (((x) == 0) ? 0 : 1))

/*
 * Debugging only
 * Routine used for comparing results from comparision algorithm
 * to results from key creation algorithm
 */
static SSHORT fam2_compare(texttype* obj, ULONG l1, const BYTE* s1,
	ULONG l2, const BYTE* s2, INTL_BOOL* error_flag)
{
	SSHORT res1 = old_fam2_compare(obj, l1, s1, l2, s2);
	SSHORT res2 = LC_NARROW_compare(obj, l1, s1, l2, s2);

	if (SIGN(res1) != SIGN(res2)) {
		printf("different compares:\n%d %s\n%d %s\nold = %d new = %d\n",
				  l1, s1, l2, s2, res1, res2);
		fflush(stdout);
		do_debug = 1;
		res1 = old_fam2_compare(obj, l1, s1, l2, s2);
		res2 = LC_NARROW_compare(obj, l1, s1, l2, s2);
		do_debug = 0;
	}
	return res2;
}

#endif	/* DEBUG_COMPARE */

ULONG LC_NARROW_canonical(texttype* obj, ULONG srcLen, const UCHAR* src, ULONG dstLen, UCHAR* dst)
{
	fb_assert(dst != NULL);
	fb_assert(src != NULL);
	fb_assert(dstLen >= obj->texttype_canonical_width * srcLen);

	const BYTE* const inbuff = src;

	for (ULONG i = 0; i < srcLen; i++, src++)
	{
		const SortOrderTblEntry* coll =
			&((const SortOrderTblEntry*)obj->texttype_impl->texttype_collation_table)[*src];

		if ((obj->texttype_impl->texttype_flags & (TEXTTYPE_secondary_insensitive | TEXTTYPE_tertiary_insensitive)) == 0)
		{
			*reinterpret_cast<USHORT*>(dst) = (coll->Primary << 8) | (coll->Secondary << 4) | coll->Tertiary;
			dst += sizeof(USHORT);
		}
		else if ((obj->texttype_impl->texttype_flags & TEXTTYPE_secondary_insensitive) == 0)
		{
			*reinterpret_cast<USHORT*>(dst) = (coll->Primary << 8) | coll->Secondary;
			dst += sizeof(USHORT);
		}
		else if ((obj->texttype_impl->texttype_flags & TEXTTYPE_tertiary_insensitive) == 0)
		{
			*reinterpret_cast<USHORT*>(dst) = (coll->Primary << 8) | coll->Tertiary;
			dst += sizeof(USHORT);
		}
		else
			*dst++ = coll->Primary;
	}

	return src - inbuff;
}


void LC_NARROW_destroy(texttype* obj)
{
	delete obj->texttype_impl;
}



bool LC_NARROW_family2(
	texttype* tt,
	charset* cs,
	SSHORT country,
	USHORT flags,
	const SortOrderTblEntry* noCaseOrderTbl,
	const BYTE* toUpperConversionTbl,
	const BYTE* toLowerConversionTbl,
	const CompressPair* compressTbl,
	const ExpandChar* expansionTbl,
	const ASCII* name,
	USHORT attributes,
	const UCHAR* specificAttributes,
	ULONG specificAttributesLength)
{
	if (attributes & ~TEXTTYPE_ATTR_PAD_SPACE)
		return false;

	tt->texttype_version			= TEXTTYPE_VERSION_1;
	tt->texttype_name				= name;
	tt->texttype_country			= country;
	tt->texttype_pad_option			= (attributes & TEXTTYPE_ATTR_PAD_SPACE) ? true : false;
	tt->texttype_fn_key_length		= LC_NARROW_key_length;
	tt->texttype_fn_string_to_key	= LC_NARROW_string_to_key;
	tt->texttype_fn_compare			= LC_NARROW_compare;
	tt->texttype_fn_str_to_upper	= fam2_str_to_upper;
	tt->texttype_fn_str_to_lower	= fam2_str_to_lower;
	tt->texttype_fn_destroy			= LC_NARROW_destroy;
	tt->texttype_impl				= new TextTypeImpl;
	tt->texttype_impl->texttype_collation_table	= (const BYTE*) noCaseOrderTbl;
	tt->texttype_impl->texttype_toupper_table	= toUpperConversionTbl;
	tt->texttype_impl->texttype_tolower_table	= toLowerConversionTbl;
	tt->texttype_impl->texttype_compress_table	= (const BYTE*) compressTbl;
	tt->texttype_impl->texttype_expand_table	= (const BYTE*) expansionTbl;
	tt->texttype_impl->texttype_flags			= ((flags) & REVERSE) ? TEXTTYPE_reverse_secondary : 0;
	tt->texttype_impl->texttype_bytes_per_key	= 0;

	IntlUtil::SpecificAttributesMap map;
	Jrd::CharSet* charSet = NULL;

	try
	{
		charSet = Jrd::CharSet::createInstance(*getDefaultMemoryPool(), 0, cs);

		if (!IntlUtil::parseSpecificAttributes(charSet, specificAttributesLength, specificAttributes, &map))
		{
			delete charSet;
			return false;
		}

		delete charSet;
	}
	catch (...)
	{
		delete charSet;
		return false;
	}

	int validAttributeCount = 0;
	string value;

	if (map.get("SPECIALS-FIRST", value) && (value == "0" || value == "1"))
	{
		int maxPrimary = 0;
		int minPrimary = INT_MAX;
		int maxIgnore = 0;

		while (compressTbl->CharPair[0])
		{
			if (compressTbl->NoCaseWeight.Primary > maxPrimary)
				maxPrimary = compressTbl->NoCaseWeight.Primary;

			if (compressTbl->NoCaseWeight.Primary < minPrimary)
				minPrimary = compressTbl->NoCaseWeight.Primary;

			++compressTbl;
		}

		for (int ch = 0; ch <= 255; ++ch)
		{
			const SortOrderTblEntry* coll =
				&((const SortOrderTblEntry*)tt->texttype_impl->texttype_collation_table)[ch];

			if (!(coll->IsExpand || coll->IsCompress))
			{
				if (coll->Primary > maxPrimary)
					maxPrimary = coll->Primary;

				if (coll->Primary < minPrimary)
					minPrimary = coll->Primary;
			}
			else if (coll->IsExpand && coll->IsCompress)
			{
				if (coll->Primary > maxIgnore)
					maxIgnore = coll->Primary;
			}
		}

		if (maxIgnore > 0 && maxPrimary + maxIgnore - 1 <= 255)
		{
			++validAttributeCount;

			if (value == "1")
			{
				tt->texttype_impl->texttype_flags |= TEXTTYPE_specials_first;
				tt->texttype_impl->ignore_sum = minPrimary - 1;
				tt->texttype_impl->primary_sum = maxIgnore - 1;
			}
		}
	}

	if (map.count() - validAttributeCount != 0)
		return false;

	return true;
}


bool LC_NARROW_family3(
	texttype* tt,
	charset* cs,
	SSHORT country,
	USHORT flags,
	const SortOrderTblEntry* noCaseOrderTbl,
	const BYTE* toUpperConversionTbl,
	const BYTE* toLowerConversionTbl,
	const CompressPair* compressTbl,
	const ExpandChar* expansionTbl,
	const ASCII* name,
	USHORT attributes,
	const UCHAR* specificAttributes,
	ULONG specificAttributesLength)
{
	bool multiLevel = false;

	IntlUtil::SpecificAttributesMap map;
	Jrd::CharSet* charSet = NULL;
	string newSpecificAttributes;

	try
	{
		charSet = Jrd::CharSet::createInstance(*getDefaultMemoryPool(), 0, cs);

		if (!IntlUtil::parseSpecificAttributes(charSet, specificAttributesLength, specificAttributes, &map))
		{
			delete charSet;
			return false;
		}

		string value;
		if (map.get("MULTI-LEVEL", value))
		{
			if (value == "0")
				multiLevel = false;
			else if (value == "1")
				multiLevel = true;
			else
				return false;

			map.remove("MULTI-LEVEL");
		}

		newSpecificAttributes = IntlUtil::generateSpecificAttributes(charSet, map);
		specificAttributes = (const UCHAR*)newSpecificAttributes.begin();
		specificAttributesLength = newSpecificAttributes.length();

		delete charSet;
	}
	catch (...)
	{
		delete charSet;
		return false;
	}

	if (LC_NARROW_family2(tt, cs, country, flags, noCaseOrderTbl,
				toUpperConversionTbl, toLowerConversionTbl, compressTbl, expansionTbl, name,
				attributes & ~(TEXTTYPE_ATTR_CASE_INSENSITIVE | TEXTTYPE_ATTR_ACCENT_INSENSITIVE),
				specificAttributes, specificAttributesLength))
	{
		if (!multiLevel)
			tt->texttype_impl->texttype_flags |= TEXTTYPE_non_multi_level;

		if (attributes & (TEXTTYPE_ATTR_CASE_INSENSITIVE | TEXTTYPE_ATTR_ACCENT_INSENSITIVE))
		{
			tt->texttype_impl->texttype_flags |= TEXTTYPE_ignore_specials;

			if (multiLevel)
			{
				tt->texttype_flags |= TEXTTYPE_SEPARATE_UNIQUE;

				if ((attributes & (TEXTTYPE_ATTR_CASE_INSENSITIVE | TEXTTYPE_ATTR_ACCENT_INSENSITIVE)) ==
					TEXTTYPE_ATTR_ACCENT_INSENSITIVE)
				{
					tt->texttype_flags |= TEXTTYPE_UNSORTED_UNIQUE;
				}
			}

			if ((attributes & (TEXTTYPE_ATTR_CASE_INSENSITIVE | TEXTTYPE_ATTR_ACCENT_INSENSITIVE)) ==
				(TEXTTYPE_ATTR_CASE_INSENSITIVE | TEXTTYPE_ATTR_ACCENT_INSENSITIVE))
			{
				tt->texttype_canonical_width = 1;
			}
			else
				tt->texttype_canonical_width = 2;

			tt->texttype_fn_canonical = LC_NARROW_canonical;

			if (attributes & TEXTTYPE_ATTR_ACCENT_INSENSITIVE)
				tt->texttype_impl->texttype_flags |= TEXTTYPE_secondary_insensitive;

			if (attributes & TEXTTYPE_ATTR_CASE_INSENSITIVE)
				tt->texttype_impl->texttype_flags |= TEXTTYPE_tertiary_insensitive;
		}

		return true;
	}
	else
		return false;
}


/*
 *	Returns INTL_BAD_STR_LENGTH if output buffer was too small
 */
static ULONG fam2_str_to_upper(texttype* obj, ULONG iLen, const BYTE* pStr, ULONG iOutLen, BYTE *pOutStr)
{
	fb_assert(pStr != NULL);
	fb_assert(pOutStr != NULL);
	fb_assert(iOutLen >= iLen);
	const BYTE* const p = pOutStr;
	while (iLen && iOutLen) {
		*pOutStr++ = (obj->texttype_impl->texttype_toupper_table[(unsigned) *pStr]);
		pStr++;
		iLen--;
		iOutLen--;
	}
	if (iLen != 0)
		return (INTL_BAD_STR_LENGTH);
	return (pOutStr - p);
}


/*
 *	Returns INTL_BAD_STR_LENGTH if output buffer was too small
 */
static ULONG fam2_str_to_lower(texttype* obj, ULONG iLen, const BYTE* pStr, ULONG iOutLen, BYTE *pOutStr)
{
	fb_assert(pStr != NULL);
	fb_assert(pOutStr != NULL);
	fb_assert(iOutLen >= iLen);
	const BYTE* const p = pOutStr;
	while (iLen && iOutLen) {
		*pOutStr++ = (obj->texttype_impl->texttype_tolower_table[(unsigned) *pStr]);
		pStr++;
		iLen--;
		iOutLen--;
	}
	if (iLen != 0)
		return (INTL_BAD_STR_LENGTH);
	return (pOutStr - p);
}