liblouis/compileTranslationTable.c

/* liblouis Braille Translation and Back-Translation Library

   Based on the Linux screenreader BRLTTY, copyright (C) 1999-2006 by The
   BRLTTY Team

   Copyright (C) 2004, 2005, 2006 ViewPlus Technologies, Inc. www.viewplus.com
   Copyright (C) 2004, 2005, 2006 JJB Software, Inc. www.jjb-software.com
   Copyright (C) 2016 Mike Gray, American Printing House for the Blind
   Copyright (C) 2016 Davy Kager, Dedicon

   This file is part of liblouis.

   liblouis is free software: you can redistribute it and/or modify it
   under the terms of the GNU Lesser General Public License as published
   by the Free Software Foundation, either version 2.1 of the License, or
   (at your option) any later version.

   liblouis is distributed in the hope that it will be useful, but
   WITHOUT ANY WARRANTY; without even the implied warranty of
   MERCHANTABILITY or FITNESS FOR A PARTICULAR PURPOSE. See the GNU
   Lesser General Public License for more details.

   You should have received a copy of the GNU Lesser General Public
   License along with liblouis. If not, see <http://www.gnu.org/licenses/>.
*/

/**
 * @file
 * @brief Read and compile translation tables
 */

#include <stddef.h>
#include <stdlib.h>
#include <stdio.h>
#include <stdarg.h>
#include <string.h>
#include <ctype.h>
#include <sys/stat.h>

#include "internal.h"
#include "config.h"

#define QUOTESUB 28 /* Stand-in for double quotes in strings */

/* needed to make debuggin easier */
#ifdef DEBUG
wchar_t wchar;
#endif

/* The following variables and functions make it possible to specify the
 * path on which all tables for liblouis and all files for liblouisutdml,
 * in their proper directories, will be found.
 */

static char *dataPathPtr;

char *EXPORT_CALL
lou_setDataPath(const char *path) {
	static char dataPath[MAXSTRING];
	dataPathPtr = NULL;
	if (path == NULL) return NULL;
	strcpy(dataPath, path);
	dataPathPtr = dataPath;
	return dataPathPtr;
}

char *EXPORT_CALL
lou_getDataPath(void) {
	return dataPathPtr;
}

/* End of dataPath code. */

static int
eqasc2uni(const unsigned char *a, const widechar *b, const int len) {
	int k;
	for (k = 0; k < len; k++)
		if ((widechar)a[k] != b[k]) return 0;
	return 1;
}

typedef struct CharsString {
	widechar length;
	widechar chars[MAXSTRING];
} CharsString;

static int errorCount;
static int warningCount;

typedef struct TranslationTableChainEntry {
	struct TranslationTableChainEntry *next;
	TranslationTableHeader *table;
	int tableListLength;
	char tableList[1];
} TranslationTableChainEntry;

static TranslationTableChainEntry *translationTableChain = NULL;

typedef struct DisplayTableChainEntry {
	struct DisplayTableChainEntry *next;
	DisplayTableHeader *table;
	int tableListLength;
	char tableList[1];
} DisplayTableChainEntry;

static DisplayTableChainEntry *displayTableChain = NULL;

/* predifined character classes */
static const char *characterClassNames[] = {
	"space",
	"letter",
	"digit",
	"punctuation",
	"uppercase",
	"lowercase",
	"math",
	"sign",
	"litdigit",
	NULL,
};

static const char *opcodeNames[CTO_None] = {
	"include",
	"locale",
	"undefined",
	"capsletter",
	"begcapsword",
	"endcapsword",
	"begcaps",
	"endcaps",
	"begcapsphrase",
	"endcapsphrase",
	"lencapsphrase",
	"letsign",
	"noletsignbefore",
	"noletsign",
	"noletsignafter",
	"numsign",
	"numericmodechars",
	"midendnumericmodechars",
	"numericnocontchars",
	"seqdelimiter",
	"seqbeforechars",
	"seqafterchars",
	"seqafterpattern",
	"seqafterexpression",
	"emphclass",
	"emphletter",
	"begemphword",
	"endemphword",
	"begemph",
	"endemph",
	"begemphphrase",
	"endemphphrase",
	"lenemphphrase",
	"capsmodechars",
	"emphmodechars",
	"begcomp",
	"compbegemph1",
	"compendemph1",
	"compbegemph2",
	"compendemph2",
	"compbegemph3",
	"compendemph3",
	"compcapsign",
	"compbegcaps",
	"compendcaps",
	"endcomp",
	"nocontractsign",
	"multind",
	"compdots",
	"comp6",
	"class",
	"after",
	"before",
	"noback",
	"nofor",
	"empmatchbefore",
	"empmatchafter",
	"swapcc",
	"swapcd",
	"swapdd",
	"space",
	"digit",
	"punctuation",
	"math",
	"sign",
	"letter",
	"uppercase",
	"lowercase",
	"grouping",
	"uplow",
	"litdigit",
	"display",
	"replace",
	"context",
	"correct",
	"pass2",
	"pass3",
	"pass4",
	"repeated",
	"repword",
	"capsnocont",
	"always",
	"exactdots",
	"nocross",
	"syllable",
	"nocont",
	"compbrl",
	"literal",
	"largesign",
	"word",
	"partword",
	"joinnum",
	"joinword",
	"lowword",
	"contraction",
	"sufword",
	"prfword",
	"begword",
	"begmidword",
	"midword",
	"midendword",
	"endword",
	"prepunc",
	"postpunc",
	"begnum",
	"midnum",
	"endnum",
	"decpoint",
	"hyphen",
	// "apostrophe",
	// "initial",
	"nobreak",
	"match",
	"backmatch",
	"attribute",
};

static short opcodeLengths[CTO_None] = { 0 };

static void
compileError(FileInfo *nested, const char *format, ...);

static void
free_tablefiles(char **tables);

static int
getAChar(FileInfo *nested) {
	/* Read a big endian, little endian or ASCII 8 file and convert it to
	 * 16- or 32-bit unsigned integers */
	int ch1 = 0, ch2 = 0;
	widechar character;
	if (nested->encoding == ascii8)
		if (nested->status == 2) {
			nested->status++;
			return nested->checkencoding[1];
		}
	while ((ch1 = fgetc(nested->in)) != EOF) {
		if (nested->status < 2) nested->checkencoding[nested->status] = ch1;
		nested->status++;
		if (nested->status == 2) {
			if (nested->checkencoding[0] == 0xfe && nested->checkencoding[1] == 0xff)
				nested->encoding = bigEndian;
			else if (nested->checkencoding[0] == 0xff && nested->checkencoding[1] == 0xfe)
				nested->encoding = littleEndian;
			else if (nested->checkencoding[0] < 128 && nested->checkencoding[1] < 128) {
				nested->encoding = ascii8;
				return nested->checkencoding[0];
			} else {
				compileError(nested,
						"encoding is neither big-endian, little-endian nor ASCII 8.");
				ch1 = EOF;
				break;
				;
			}
			continue;
		}
		switch (nested->encoding) {
		case noEncoding:
			break;
		case ascii8:
			return ch1;
			break;
		case bigEndian:
			ch2 = fgetc(nested->in);
			if (ch2 == EOF) break;
			character = (widechar)(ch1 << 8) | ch2;
			return (int)character;
			break;
		case littleEndian:
			ch2 = fgetc(nested->in);
			if (ch2 == EOF) break;
			character = (widechar)(ch2 << 8) | ch1;
			return (int)character;
			break;
		}
		if (ch1 == EOF || ch2 == EOF) break;
	}
	return EOF;
}

int EXPORT_CALL
_lou_getALine(FileInfo *nested) {
	/* Read a line of widechar's from an input file */
	int ch;
	int pch = 0;
	nested->linelen = 0;
	while ((ch = getAChar(nested)) != EOF) {
		if (ch == 13) continue;
		if (pch == '\\' && ch == 10) {
			nested->linelen--;
			pch = ch;
			continue;
		}
		if (ch == 10 || nested->linelen >= MAXSTRING - 1) break;
		nested->line[nested->linelen++] = (widechar)ch;
		pch = ch;
	}
	nested->line[nested->linelen] = 0;
	nested->linepos = 0;
	if (ch == EOF) return 0;
	nested->lineNumber++;
	return 1;
}

static inline int
atEndOfLine(FileInfo *nested) {
	return nested->linepos >= nested->linelen;
}

static inline int
atTokenDelimiter(FileInfo *nested) {
	return nested->line[nested->linepos] <= 32;
}

static int
getToken(FileInfo *nested, CharsString *result, const char *description, int *lastToken) {
	/* Find the next string of contiguous non-whitespace characters. If this
	 * is the last token on the line, return 2 instead of 1. */
	while (!atEndOfLine(nested) && atTokenDelimiter(nested)) nested->linepos++;
	result->length = 0;
	while (!atEndOfLine(nested) && !atTokenDelimiter(nested)) {
		int maxlen = MAXSTRING;
		if (result->length >= maxlen) {
			compileError(nested, "more than %d characters (bytes)", maxlen);
			return 0;
		} else
			result->chars[result->length++] = nested->line[nested->linepos++];
	}
	if (!result->length) {
		/* Not enough tokens */
		if (description) compileError(nested, "%s not specified.", description);
		return 0;
	}
	result->chars[result->length] = 0;
	while (!atEndOfLine(nested) && atTokenDelimiter(nested)) nested->linepos++;
	return (*lastToken = atEndOfLine(nested)) ? 2 : 1;
}

static void
compileError(FileInfo *nested, const char *format, ...) {
#ifndef __SYMBIAN32__
	char buffer[MAXSTRING];
	va_list arguments;
	va_start(arguments, format);
	vsnprintf(buffer, sizeof(buffer), format, arguments);
	va_end(arguments);
	if (nested)
		_lou_logMessage(LOU_LOG_ERROR, "%s:%d: error: %s", nested->fileName,
				nested->lineNumber, buffer);
	else
		_lou_logMessage(LOU_LOG_ERROR, "error: %s", buffer);
	errorCount++;
#endif
}

static void
compileWarning(FileInfo *nested, const char *format, ...) {
#ifndef __SYMBIAN32__
	char buffer[MAXSTRING];
	va_list arguments;
	va_start(arguments, format);
	vsnprintf(buffer, sizeof(buffer), format, arguments);
	va_end(arguments);
	if (nested)
		_lou_logMessage(LOU_LOG_WARN, "%s:%d: warning: %s", nested->fileName,
				nested->lineNumber, buffer);
	else
		_lou_logMessage(LOU_LOG_WARN, "warning: %s", buffer);
	warningCount++;
#endif
}

static int
allocateSpaceInTranslationTable(FileInfo *nested, TranslationTableOffset *offset,
		int count, TranslationTableHeader **table) {
	/* allocate memory for table and expand previously allocated memory if necessary */
	int spaceNeeded = ((count + OFFSETSIZE - 1) / OFFSETSIZE) * OFFSETSIZE;
	TranslationTableOffset newSize = (*table)->bytesUsed + spaceNeeded;
	TranslationTableOffset size = (*table)->tableSize;
	if (newSize > size) {
		TranslationTableHeader *newTable;
		newSize += (newSize / OFFSETSIZE);
		newTable = realloc(*table, newSize);
		if (!newTable) {
			compileError(nested, "Not enough memory for translation table.");
			_lou_outOfMemory();
		}
		memset(((unsigned char *)newTable) + size, 0, newSize - size);
		/* update references to the old table */
		{
			TranslationTableChainEntry *entry;
			for (entry = translationTableChain; entry != NULL; entry = entry->next)
				if (entry->table == *table)
					entry->table = (TranslationTableHeader *)newTable;
		}
		*table = (TranslationTableHeader *)newTable;
		(*table)->tableSize = newSize;
	}
	if (offset != NULL) {
		*offset = ((*table)->bytesUsed - sizeof(**table)) / OFFSETSIZE;
		(*table)->bytesUsed += spaceNeeded;
	}
	return 1;
}

static int
allocateSpaceInDisplayTable(FileInfo *nested, TranslationTableOffset *offset, int count,
		DisplayTableHeader **table) {
	/* allocate memory for table and expand previously allocated memory if necessary */
	int spaceNeeded = ((count + OFFSETSIZE - 1) / OFFSETSIZE) * OFFSETSIZE;
	TranslationTableOffset newSize = (*table)->bytesUsed + spaceNeeded;
	TranslationTableOffset size = (*table)->tableSize;
	if (newSize > size) {
		DisplayTableHeader *newTable;
		newSize += (newSize / OFFSETSIZE);
		newTable = realloc(*table, newSize);
		if (!newTable) {
			compileError(nested, "Not enough memory for display table.");
			_lou_outOfMemory();
		}
		memset(((unsigned char *)newTable) + size, 0, newSize - size);
		/* update references to the old table */
		{
			DisplayTableChainEntry *entry;
			for (entry = displayTableChain; entry != NULL; entry = entry->next)
				if (entry->table == *table) entry->table = (DisplayTableHeader *)newTable;
		}
		*table = (DisplayTableHeader *)newTable;
		(*table)->tableSize = newSize;
	}
	if (offset != NULL) {
		*offset = ((*table)->bytesUsed - sizeof(**table)) / OFFSETSIZE;
		(*table)->bytesUsed += spaceNeeded;
	}
	return 1;
}

static int
allocateTranslationTable(FileInfo *nested, TranslationTableHeader **table) {
	/* Allocate memory for the table and a guess on the number of rules */
	const TranslationTableOffset startSize = 2 * sizeof(**table);
	if (*table) return 1;
	TranslationTableOffset bytesUsed =
			sizeof(**table) + OFFSETSIZE; /* So no offset is ever zero */
	if (!(*table = malloc(startSize))) {
		compileError(nested, "Not enough memory");
		if (*table != NULL) free(*table);
		*table = NULL;
		_lou_outOfMemory();
	}
	memset(*table, 0, startSize);
	(*table)->tableSize = startSize;
	(*table)->bytesUsed = bytesUsed;
	return 1;
}

static int
allocateDisplayTable(FileInfo *nested, DisplayTableHeader **table) {
	/* Allocate memory for the table and a guess on the number of rules */
	const TranslationTableOffset startSize = 2 * sizeof(**table);
	if (*table) return 1;
	TranslationTableOffset bytesUsed =
			sizeof(**table) + OFFSETSIZE; /* So no offset is ever zero */
	if (!(*table = malloc(startSize))) {
		compileError(nested, "Not enough memory");
		if (*table != NULL) free(*table);
		*table = NULL;
		_lou_outOfMemory();
	}
	memset(*table, 0, startSize);
	(*table)->tableSize = startSize;
	(*table)->bytesUsed = bytesUsed;
	return 1;
}

static TranslationTableCharacter *
compile_findCharOrDots(widechar c, int m, TranslationTableHeader *table) {
	/* Look up a character or dot pattern. If m is 0 look up a character,
	 * otherwise look up a dot pattern. Although the algorithms are almost
	 * identical, different tables are needed for characters and dots because
	 * of the possibility of conflicts. */
	TranslationTableCharacter *character;
	TranslationTableOffset bucket;
	unsigned long int makeHash = _lou_charHash(c);
	if (m == 0)
		bucket = table->characters[makeHash];
	else
		bucket = table->dots[makeHash];
	while (bucket) {
		character = (TranslationTableCharacter *)&table->ruleArea[bucket];
		if (character->realchar == c) return character;
		bucket = character->next;
	}
	return NULL;
}

static TranslationTableCharacter *
addCharOrDots(FileInfo *nested, widechar c, int m, TranslationTableHeader **table) {
	/* See if a character or dot pattern is in the appropriate table. If not,
	 * insert it. In either
	 * case, return a pointer to it. */
	TranslationTableOffset bucket;
	TranslationTableCharacter *character;
	TranslationTableCharacter *oldchar;
	TranslationTableOffset offset;
	unsigned long int makeHash;
	if ((character = compile_findCharOrDots(c, m, *table))) return character;
	if (!allocateSpaceInTranslationTable(nested, &offset, sizeof(*character), table))
		return NULL;
	character = (TranslationTableCharacter *)&(*table)->ruleArea[offset];
	memset(character, 0, sizeof(*character));
	character->realchar = c;
	makeHash = _lou_charHash(c);
	if (m == 0)
		bucket = (*table)->characters[makeHash];
	else
		bucket = (*table)->dots[makeHash];
	if (!bucket) {
		if (m == 0)
			(*table)->characters[makeHash] = offset;
		else
			(*table)->dots[makeHash] = offset;
	} else {
		oldchar = (TranslationTableCharacter *)&(*table)->ruleArea[bucket];
		while (oldchar->next)
			oldchar = (TranslationTableCharacter *)&(*table)->ruleArea[oldchar->next];
		oldchar->next = offset;
	}
	return character;
}

static CharOrDots *
getCharOrDots(widechar c, int m, const DisplayTableHeader *table) {
	CharOrDots *cdPtr;
	TranslationTableOffset bucket;
	unsigned long int makeHash = _lou_charHash(c);
	if (m == 0)
		bucket = table->charToDots[makeHash];
	else
		bucket = table->dotsToChar[makeHash];
	while (bucket) {
		cdPtr = (CharOrDots *)&table->ruleArea[bucket];
		if (cdPtr->lookFor == c) return cdPtr;
		bucket = cdPtr->next;
	}
	return NULL;
}

widechar EXPORT_CALL
_lou_getDotsForChar(widechar c, const DisplayTableHeader *table) {
	CharOrDots *cdPtr = getCharOrDots(c, 0, table);
	if (cdPtr) return cdPtr->found;
	return LOU_DOTS;
}

widechar EXPORT_CALL
_lou_getCharFromDots(widechar d, const DisplayTableHeader *table) {
	CharOrDots *cdPtr = getCharOrDots(d, 1, table);
	if (cdPtr) return cdPtr->found;
	return '\0';
}

static int
putCharAndDots(FileInfo *nested, widechar c, widechar d, DisplayTableHeader **table) {
	TranslationTableOffset bucket;
	CharOrDots *cdPtr;
	CharOrDots *oldcdPtr = NULL;
	TranslationTableOffset offset;
	unsigned long int makeHash;
	if (!(cdPtr = getCharOrDots(c, 0, *table))) {
		if (!allocateSpaceInDisplayTable(nested, &offset, sizeof(*cdPtr), table))
			return 0;
		cdPtr = (CharOrDots *)&(*table)->ruleArea[offset];
		cdPtr->next = 0;
		cdPtr->lookFor = c;
		cdPtr->found = d;
		makeHash = _lou_charHash(c);
		bucket = (*table)->charToDots[makeHash];
		if (!bucket)
			(*table)->charToDots[makeHash] = offset;
		else {
			oldcdPtr = (CharOrDots *)&(*table)->ruleArea[bucket];
			while (oldcdPtr->next)
				oldcdPtr = (CharOrDots *)&(*table)->ruleArea[oldcdPtr->next];
			oldcdPtr->next = offset;
		}
	}
	if (!(cdPtr = getCharOrDots(d, 1, *table))) {
		if (!allocateSpaceInDisplayTable(nested, &offset, sizeof(*cdPtr), table))
			return 0;
		cdPtr = (CharOrDots *)&(*table)->ruleArea[offset];
		cdPtr->next = 0;
		cdPtr->lookFor = d;
		cdPtr->found = c;
		makeHash = _lou_charHash(d);
		bucket = (*table)->dotsToChar[makeHash];
		if (!bucket)
			(*table)->dotsToChar[makeHash] = offset;
		else {
			oldcdPtr = (CharOrDots *)&(*table)->ruleArea[bucket];
			while (oldcdPtr->next)
				oldcdPtr = (CharOrDots *)&(*table)->ruleArea[oldcdPtr->next];
			oldcdPtr->next = offset;
		}
	}
	return 1;
}

static inline const char *
getPartName(int actionPart) {
	return actionPart ? "action" : "test";
}

static int
passFindCharacters(FileInfo *nested, widechar *instructions, int end,
		widechar **characters, int *length) {
	int IC = 0;
	int lookback = 0;

	*characters = NULL;
	*length = 0;

	while (IC < end) {
		widechar instruction = instructions[IC];

		switch (instruction) {
		case pass_string:
		case pass_dots: {
			int count = instructions[IC + 1];
			IC += 2;
			if (count > lookback) {
				*characters = &instructions[IC + lookback];
				*length = count - lookback;
				return 1;
			} else {
				lookback -= count;
			}
			IC += count;
			continue;
		}

		case pass_attributes:
			IC += 5;
			if (instructions[IC - 2] == instructions[IC - 1] &&
					instructions[IC - 1] <= lookback) {
				lookback -= instructions[IC - 1];
				continue;
			}
			goto NO_CHARACTERS;

		case pass_swap:
			IC += 2;
			/* fall through */

		case pass_groupstart:
		case pass_groupend:
		case pass_groupreplace:
			IC += 3;

		NO_CHARACTERS : { return 1; }

		case pass_eq:
		case pass_lt:
		case pass_gt:
		case pass_lteq:
		case pass_gteq:
			IC += 3;
			continue;

		case pass_lookback:
			lookback += instructions[IC + 1];
			IC += 2;
			continue;

		case pass_not:
		case pass_startReplace:
		case pass_endReplace:
		case pass_first:
		case pass_last:
		case pass_copy:
		case pass_omit:
		case pass_plus:
		case pass_hyphen:
			IC += 1;
			continue;

		case pass_endTest:
			goto NO_CHARACTERS;

		default:
			compileError(nested, "unhandled test suboperand: \\x%02x", instruction);
			return 0;
		}
	}
	goto NO_CHARACTERS;
}

/* The following functions are called by addRule to handle various cases. */

static void
addForwardRuleWithSingleChar(FileInfo *nested, TranslationTableOffset newRuleOffset,
		TranslationTableRule *newRule, TranslationTableHeader **table) {
	/* direction = 0, newRule->charslen = 1 */
	TranslationTableRule *currentRule;
	TranslationTableOffset *currentOffsetPtr;
	TranslationTableCharacter *character;
	int m = 0;
	if (newRule->opcode == CTO_CompDots || newRule->opcode == CTO_Comp6) return;
	if (newRule->opcode >= CTO_Pass2 && newRule->opcode <= CTO_Pass4) m = 1;
	// get the character from the table, or if the character is not defined yet, define it
	// (without adding attributes)
	character = addCharOrDots(nested, newRule->charsdots[0], m, table);
	if (m != 1 && character->attributes & CTC_Letter &&
			(newRule->opcode == CTO_WholeWord || newRule->opcode == CTO_LargeSign)) {
		if ((*table)->noLetsignCount < LETSIGNSIZE)
			(*table)->noLetsign[(*table)->noLetsignCount++] = newRule->charsdots[0];
	}
	// if the new rule is a character definition rule, set the main definition rule of
	// this character to it
	// (possibly overwriting previous definition rules)
	// adding the attributes to the character has already been done elsewhere
	if (newRule->opcode >= CTO_Space && newRule->opcode < CTO_UpLow)
		character->definitionRule = newRuleOffset;
	// add the new rule to the list of rules associated with this character
	// if the new rule is a character definition rule, it is inserted at the end of the
	// list
	// otherwise it is inserted before the first character definition rule
	currentOffsetPtr = &character->otherRules;
	while (*currentOffsetPtr) {
		currentRule = (TranslationTableRule *)&(*table)->ruleArea[*currentOffsetPtr];
		if (currentRule->charslen == 0) break;
		if (currentRule->opcode >= CTO_Space && currentRule->opcode < CTO_UpLow)
			if (!(newRule->opcode >= CTO_Space && newRule->opcode < CTO_UpLow)) break;
		currentOffsetPtr = &currentRule->charsnext;
	}
	newRule->charsnext = *currentOffsetPtr;
	*currentOffsetPtr = newRuleOffset;
}

static void
addForwardRuleWithMultipleChars(TranslationTableOffset newRuleOffset,
		TranslationTableRule *newRule, TranslationTableHeader *table) {
	/* direction = 0 newRule->charslen > 1 */
	TranslationTableRule *currentRule = NULL;
	TranslationTableOffset *currentOffsetPtr =
			&table->forRules[_lou_stringHash(&newRule->charsdots[0], 0, NULL)];
	while (*currentOffsetPtr) {
		currentRule = (TranslationTableRule *)&table->ruleArea[*currentOffsetPtr];
		if (newRule->charslen > currentRule->charslen) break;
		if (newRule->charslen == currentRule->charslen)
			if ((currentRule->opcode == CTO_Always) && (newRule->opcode != CTO_Always))
				break;
		currentOffsetPtr = &currentRule->charsnext;
	}
	newRule->charsnext = *currentOffsetPtr;
	*currentOffsetPtr = newRuleOffset;
}

static void
addBackwardRuleWithSingleCell(FileInfo *nested, widechar cell,
		TranslationTableOffset newRuleOffset, TranslationTableRule *newRule,
		TranslationTableHeader **table) {
	/* direction = 1, newRule->dotslen = 1 */
	TranslationTableRule *currentRule;
	TranslationTableOffset *currentOffsetPtr;
	TranslationTableCharacter *dots;
	if (newRule->opcode == CTO_SwapCc || newRule->opcode == CTO_Repeated)
		return; /* too ambiguous */
	// get the cell from the table, or if the cell is not defined yet, define it (without
	// adding attributes)
	dots = addCharOrDots(nested, cell, 1, table);
	if (newRule->opcode >= CTO_Space && newRule->opcode < CTO_UpLow)
		dots->definitionRule = newRuleOffset;
	currentOffsetPtr = &dots->otherRules;
	while (*currentOffsetPtr) {
		currentRule = (TranslationTableRule *)&(*table)->ruleArea[*currentOffsetPtr];
		if (newRule->charslen > currentRule->charslen || currentRule->dotslen == 0) break;
		if (currentRule->opcode >= CTO_Space && currentRule->opcode < CTO_UpLow)
			if (!(newRule->opcode >= CTO_Space && newRule->opcode < CTO_UpLow)) break;
		currentOffsetPtr = &currentRule->dotsnext;
	}
	newRule->dotsnext = *currentOffsetPtr;
	*currentOffsetPtr = newRuleOffset;
}

static void
addBackwardRuleWithMultipleCells(widechar *cells, int count,
		TranslationTableOffset newRuleOffset, TranslationTableRule *newRule,
		TranslationTableHeader *table) {
	/* direction = 1, newRule->dotslen > 1 */
	TranslationTableRule *currentRule = NULL;
	TranslationTableOffset *currentOffsetPtr =
			&table->backRules[_lou_stringHash(cells, 0, NULL)];
	if (newRule->opcode == CTO_SwapCc) return;
	while (*currentOffsetPtr) {
		int currentLength;
		int newLength;
		currentRule = (TranslationTableRule *)&table->ruleArea[*currentOffsetPtr];
		currentLength = currentRule->dotslen + currentRule->charslen;
		newLength = count + newRule->charslen;
		if (newLength > currentLength) break;
		if (currentLength == newLength)
			if ((currentRule->opcode == CTO_Always) && (newRule->opcode != CTO_Always))
				break;
		currentOffsetPtr = &currentRule->dotsnext;
	}
	newRule->dotsnext = *currentOffsetPtr;
	*currentOffsetPtr = newRuleOffset;
}

static int
addForwardPassRule(TranslationTableOffset newRuleOffset, TranslationTableRule *newRule,
		TranslationTableHeader *table) {
	TranslationTableOffset *currentOffsetPtr;
	TranslationTableRule *currentRule;
	switch (newRule->opcode) {
	case CTO_Correct:
		currentOffsetPtr = &table->forPassRules[0];
		break;
	case CTO_Context:
		currentOffsetPtr = &table->forPassRules[1];
		break;
	case CTO_Pass2:
		currentOffsetPtr = &table->forPassRules[2];
		break;
	case CTO_Pass3:
		currentOffsetPtr = &table->forPassRules[3];
		break;
	case CTO_Pass4:
		currentOffsetPtr = &table->forPassRules[4];
		break;
	default:
		return 0;
	}
	while (*currentOffsetPtr) {
		currentRule = (TranslationTableRule *)&table->ruleArea[*currentOffsetPtr];
		if (newRule->charslen > currentRule->charslen) break;
		currentOffsetPtr = &currentRule->charsnext;
	}
	newRule->charsnext = *currentOffsetPtr;
	*currentOffsetPtr = newRuleOffset;
	return 1;
}

static int
addBackwardPassRule(TranslationTableOffset newRuleOffset, TranslationTableRule *newRule,
		TranslationTableHeader *table) {
	TranslationTableOffset *currentOffsetPtr;
	TranslationTableRule *currentRule;
	switch (newRule->opcode) {
	case CTO_Correct:
		currentOffsetPtr = &table->backPassRules[0];
		break;
	case CTO_Context:
		currentOffsetPtr = &table->backPassRules[1];
		break;
	case CTO_Pass2:
		currentOffsetPtr = &table->backPassRules[2];
		break;
	case CTO_Pass3:
		currentOffsetPtr = &table->backPassRules[3];
		break;
	case CTO_Pass4:
		currentOffsetPtr = &table->backPassRules[4];
		break;
	default:
		return 0;
	}
	while (*currentOffsetPtr) {
		currentRule = (TranslationTableRule *)&table->ruleArea[*currentOffsetPtr];
		if (newRule->charslen > currentRule->charslen) break;
		currentOffsetPtr = &currentRule->dotsnext;
	}
	newRule->dotsnext = *currentOffsetPtr;
	*currentOffsetPtr = newRuleOffset;
	return 1;
}

static int
addRule(FileInfo *nested, TranslationTableOpcode opcode, CharsString *ruleChars,
		CharsString *ruleDots, TranslationTableCharacterAttributes after,
		TranslationTableCharacterAttributes before, TranslationTableOffset *newRuleOffset,
		TranslationTableRule **newRule, int noback, int nofor,
		TranslationTableHeader **table) {
	/* Add a rule to the table, using the hash function to find the start of
	 * chains and chaining both the chars and dots strings */
	TranslationTableOffset ruleOffset;
	int ruleSize = sizeof(TranslationTableRule) - (DEFAULTRULESIZE * CHARSIZE);
	if (ruleChars) ruleSize += CHARSIZE * ruleChars->length;
	if (ruleDots) ruleSize += CHARSIZE * ruleDots->length;
	if (!allocateSpaceInTranslationTable(nested, &ruleOffset, ruleSize, table)) return 0;
	TranslationTableRule *rule = (TranslationTableRule *)&(*table)->ruleArea[ruleOffset];
	if (newRule) *newRule = rule;
	if (newRuleOffset) *newRuleOffset = ruleOffset;
	rule->opcode = opcode;
	rule->after = after;
	rule->before = before;
	if (ruleChars)
		memcpy(&rule->charsdots[0], &ruleChars->chars[0],
				CHARSIZE * (rule->charslen = ruleChars->length));
	else
		rule->charslen = 0;
	if (ruleDots)
		memcpy(&rule->charsdots[rule->charslen], &ruleDots->chars[0],
				CHARSIZE * (rule->dotslen = ruleDots->length));
	else
		rule->dotslen = 0;

	/* link new rule into table. */
	if (opcode == CTO_SwapCc || opcode == CTO_SwapCd || opcode == CTO_SwapDd) return 1;
	if (opcode >= CTO_Context && opcode <= CTO_Pass4)
		if (!(opcode == CTO_Context && rule->charslen > 0)) {
			if (!nofor)
				if (!addForwardPassRule(ruleOffset, rule, *table)) return 0;
			if (!noback)
				if (!addBackwardPassRule(ruleOffset, rule, *table)) return 0;
			return 1;
		}
	if (!nofor) {
		if (rule->charslen == 1)
			addForwardRuleWithSingleChar(nested, ruleOffset, rule, table);
		else if (rule->charslen > 1)
			addForwardRuleWithMultipleChars(ruleOffset, rule, *table);
	}
	if (!noback) {
		widechar *cells;
		int count;

		if (rule->opcode == CTO_Context) {
			cells = &rule->charsdots[0];
			count = rule->charslen;
		} else {
			cells = &rule->charsdots[rule->charslen];
			count = rule->dotslen;
		}

		if (count == 1)
			addBackwardRuleWithSingleCell(nested, *cells, ruleOffset, rule, table);
		else if (count > 1)
			addBackwardRuleWithMultipleCells(cells, count, ruleOffset, rule, *table);
	}
	return 1;
}

static const CharacterClass *
findCharacterClass(const CharsString *name, const TranslationTableHeader *table) {
	/* Find a character class, whether predefined or user-defined */
	const CharacterClass *class = table->characterClasses;
	while (class) {
		if ((name->length == class->length) &&
				(memcmp(&name->chars[0], class->name, CHARSIZE * name->length) == 0))
			return class;
		class = class->next;
	}
	return NULL;
}

static CharacterClass *
addCharacterClass(FileInfo *nested, const widechar *name, int length,
		TranslationTableHeader *table) {
	/* Define a character class, Whether predefined or user-defined */
	CharacterClass **classes = &table->characterClasses;
	;
	TranslationTableCharacterAttributes *nextAttribute =
			&table->nextCharacterClassAttribute;
	CharacterClass *class;
	if (*nextAttribute) {
		if (!(class = malloc(sizeof(*class) + CHARSIZE * (length - 1))))
			_lou_outOfMemory();
		else {
			memset(class, 0, sizeof(*class));
			memcpy(class->name, name, CHARSIZE * (class->length = length));
			class->attribute = *nextAttribute;
			if (*nextAttribute == CTC_Class4)
				*nextAttribute = CTC_UserDefined0;
			else if (*nextAttribute == CTC_UserDefined7)
				*nextAttribute = CTC_Class13;
			else
				*nextAttribute <<= 1;
			class->next = *classes;
			*classes = class;
			return class;
		}
	}
	compileError(nested, "character class table overflow.");
	return NULL;
}

static void
deallocateCharacterClasses(TranslationTableHeader *table) {
	CharacterClass **classes = &table->characterClasses;
	while (*classes) {
		CharacterClass *class = *classes;
		*classes = (*classes)->next;
		if (class) free(class);
	}
}

static int
allocateCharacterClasses(TranslationTableHeader *table) {
	/* Allocate memory for predifined character classes */
	int k = 0;
	table->characterClasses = NULL;
	table->nextCharacterClassAttribute = 1;
	while (characterClassNames[k]) {
		widechar wname[MAXSTRING];
		int length = (int)strlen(characterClassNames[k]);
		int kk;
		for (kk = 0; kk < length; kk++) wname[kk] = (widechar)characterClassNames[k][kk];
		if (!addCharacterClass(NULL, wname, length, table)) {
			deallocateCharacterClasses(table);
			return 0;
		}
		k++;
	}
	return 1;
}

static TranslationTableOpcode
getOpcode(FileInfo *nested, const CharsString *token) {
	static TranslationTableOpcode lastOpcode = 0;
	TranslationTableOpcode opcode = lastOpcode;

	do {
		if (token->length == opcodeLengths[opcode])
			if (eqasc2uni((unsigned char *)opcodeNames[opcode], &token->chars[0],
						token->length)) {
				lastOpcode = opcode;
				return opcode;
			}
		opcode++;
		if (opcode >= CTO_None) opcode = 0;
	} while (opcode != lastOpcode);
	compileError(nested, "opcode %s not defined.",
			_lou_showString(&token->chars[0], token->length, 0));
	return CTO_None;
}

TranslationTableOpcode EXPORT_CALL
_lou_findOpcodeNumber(const char *toFind) {
	/* Used by tools such as lou_debug */
	static TranslationTableOpcode lastOpcode = 0;
	TranslationTableOpcode opcode = lastOpcode;
	int length = (int)strlen(toFind);
	do {
		if (length == opcodeLengths[opcode] &&
				strcasecmp(toFind, opcodeNames[opcode]) == 0) {
			lastOpcode = opcode;
			return opcode;
		}
		opcode++;
		if (opcode >= CTO_None) opcode = 0;
	} while (opcode != lastOpcode);
	return CTO_None;
}

const char *EXPORT_CALL
_lou_findOpcodeName(TranslationTableOpcode opcode) {
	static char scratchBuf[MAXSTRING];
	/* Used by tools such as lou_debug */
	if (opcode < 0 || opcode >= CTO_None) {
		sprintf(scratchBuf, "%u", opcode);
		return scratchBuf;
	}
	return opcodeNames[opcode];
}

static widechar
hexValue(FileInfo *nested, const widechar *digits, int length) {
	int k;
	unsigned int binaryValue = 0;
	for (k = 0; k < length; k++) {
		unsigned int hexDigit = 0;
		if (digits[k] >= '0' && digits[k] <= '9')
			hexDigit = digits[k] - '0';
		else if (digits[k] >= 'a' && digits[k] <= 'f')
			hexDigit = digits[k] - 'a' + 10;
		else if (digits[k] >= 'A' && digits[k] <= 'F')
			hexDigit = digits[k] - 'A' + 10;
		else {
			compileError(nested, "invalid %d-digit hexadecimal number", length);
			return (widechar)0xffffffff;
		}
		binaryValue |= hexDigit << (4 * (length - 1 - k));
	}
	return (widechar)binaryValue;
}

#define MAXBYTES 7
static const unsigned int first0Bit[MAXBYTES] = { 0x80, 0xC0, 0xE0, 0xF0, 0xF8, 0xFC,
	0XFE };

static int
parseChars(FileInfo *nested, CharsString *result, CharsString *token) {
	int in = 0;
	int out = 0;
	int lastOutSize = 0;
	int lastIn;
	unsigned int ch = 0;
	int numBytes = 0;
	unsigned int utf32 = 0;
	int k;
	while (in < token->length) {
		ch = token->chars[in++] & 0xff;
		if (ch < 128) {
			if (ch == '\\') { /* escape sequence */
				switch (ch = token->chars[in]) {
				case '\\':
					break;
				case 'e':
					ch = 0x1b;
					break;
				case 'f':
					ch = 12;
					break;
				case 'n':
					ch = 10;
					break;
				case 'r':
					ch = 13;
					break;
				case 's':
					ch = ' ';
					break;
				case 't':
					ch = 9;
					break;
				case 'v':
					ch = 11;
					break;
				case 'w':
					ch = LOU_ENDSEGMENT;
					break;
				case 34:
					ch = QUOTESUB;
					break;
				case 'X':
					compileWarning(nested, "\\Xhhhh (with a capital 'X') is deprecated.");
				case 'x':
					if (token->length - in > 4) {
						ch = hexValue(nested, &token->chars[in + 1], 4);
						in += 4;
					}
					break;
				case 'Y':
					compileWarning(
							nested, "\\Yhhhhh (with a capital 'Y') is deprecated.");
				case 'y':
					if (CHARSIZE == 2) {
					not32:
						compileError(nested,
								"liblouis has not been compiled for 32-bit Unicode");
						break;
					}
					if (token->length - in > 5) {
						ch = hexValue(nested, &token->chars[in + 1], 5);
						in += 5;
					}
					break;
				case 'Z':
					compileWarning(
							nested, "\\Zhhhhhhhh (with a capital 'Z') is deprecated.");
				case 'z':
					if (CHARSIZE == 2) goto not32;
					if (token->length - in > 8) {
						ch = hexValue(nested, &token->chars[in + 1], 8);
						in += 8;
					}
					break;
				default:
					compileError(nested, "invalid escape sequence '\\%c'", ch);
					break;
				}
				in++;
			}
			if (out >= MAXSTRING - 1) {
				compileError(nested, "Token too long");
				result->length = MAXSTRING - 1;
				return 1;
			}
			result->chars[out++] = (widechar)ch;
			continue;
		}
		lastOutSize = out;
		lastIn = in;
		for (numBytes = MAXBYTES - 1; numBytes > 0; numBytes--)
			if (ch >= first0Bit[numBytes]) break;
		utf32 = ch & (0XFF - first0Bit[numBytes]);
		for (k = 0; k < numBytes; k++) {
			if (in >= MAXSTRING - 1) break;
			if (out >= MAXSTRING - 1) {
				compileError(nested, "Token too long");
				result->length = lastOutSize;
				return 1;
			}
			if (token->chars[in] < 128 || (token->chars[in] & 0x0040)) {
				compileWarning(nested, "invalid UTF-8. Assuming Latin-1.");
				result->chars[out++] = token->chars[lastIn];
				in = lastIn + 1;
				continue;
			}
			utf32 = (utf32 << 6) + (token->chars[in++] & 0x3f);
		}
		if (out >= MAXSTRING - 1) {
			compileError(nested, "Token too long");
			result->length = lastOutSize;
			return 1;
		}
		if (CHARSIZE == 2 && utf32 > 0xffff) utf32 = 0xffff;
		result->chars[out++] = (widechar)utf32;
	}
	result->length = out;
	return 1;
}

int EXPORT_CALL
_lou_extParseChars(const char *inString, widechar *outString) {
	/* Parse external character strings */
	CharsString wideIn;
	CharsString result;
	int k;
	for (k = 0; inString[k] && k < MAXSTRING - 1; k++) wideIn.chars[k] = inString[k];
	wideIn.chars[k] = 0;
	wideIn.length = k;
	parseChars(NULL, &result, &wideIn);
	if (errorCount) {
		errorCount = 0;
		return 0;
	}
	for (k = 0; k < result.length; k++) outString[k] = result.chars[k];
	return result.length;
}

static int
parseDots(FileInfo *nested, CharsString *cells, const CharsString *token) {
	/* get dot patterns */
	widechar cell = 0; /* assembly place for dots */
	int cellCount = 0;
	int index;
	int start = 0;

	for (index = 0; index < token->length; index++) {
		int started = index != start;
		widechar character = token->chars[index];
		switch (character) { /* or dots to make up Braille cell */
			{
				int dot;
			case '1':
				dot = LOU_DOT_1;
				goto haveDot;
			case '2':
				dot = LOU_DOT_2;
				goto haveDot;
			case '3':
				dot = LOU_DOT_3;
				goto haveDot;
			case '4':
				dot = LOU_DOT_4;
				goto haveDot;
			case '5':
				dot = LOU_DOT_5;
				goto haveDot;
			case '6':
				dot = LOU_DOT_6;
				goto haveDot;
			case '7':
				dot = LOU_DOT_7;
				goto haveDot;
			case '8':
				dot = LOU_DOT_8;
				goto haveDot;
			case '9':
				dot = LOU_DOT_9;
				goto haveDot;
			case 'a':
			case 'A':
				dot = LOU_DOT_10;
				goto haveDot;
			case 'b':
			case 'B':
				dot = LOU_DOT_11;
				goto haveDot;
			case 'c':
			case 'C':
				dot = LOU_DOT_12;
				goto haveDot;
			case 'd':
			case 'D':
				dot = LOU_DOT_13;
				goto haveDot;
			case 'e':
			case 'E':
				dot = LOU_DOT_14;
				goto haveDot;
			case 'f':
			case 'F':
				dot = LOU_DOT_15;
			haveDot:
				if (started && !cell) goto invalid;
				if (cell & dot) {
					compileError(nested, "dot specified more than once.");
					return 0;
				}
				cell |= dot;
				break;
			}
		case '0': /* blank */
			if (started) goto invalid;
			break;
		case '-': /* got all dots for this cell */
			if (!started) {
				compileError(nested, "missing cell specification.");
				return 0;
			}
			cells->chars[cellCount++] = cell | LOU_DOTS;
			cell = 0;
			start = index + 1;
			break;
		default:
		invalid:
			compileError(
					nested, "invalid dot number %s.", _lou_showString(&character, 1, 0));
			return 0;
		}
	}
	if (index == start) {
		compileError(nested, "missing cell specification.");
		return 0;
	}
	cells->chars[cellCount++] = cell | LOU_DOTS; /* last cell */
	cells->length = cellCount;
	return 1;
}

int EXPORT_CALL
_lou_extParseDots(const char *inString, widechar *outString) {
	/* Parse external dot patterns */
	CharsString wideIn;
	CharsString result;
	int k;
	for (k = 0; inString[k] && k < MAXSTRING - 1; k++) wideIn.chars[k] = inString[k];
	wideIn.chars[k] = 0;
	wideIn.length = k;
	parseDots(NULL, &result, &wideIn);
	if (errorCount) {
		errorCount = 0;
		return 0;
	}
	for (k = 0; k < result.length; k++) outString[k] = result.chars[k];
	outString[k] = 0;
	return result.length;
}

static int
getCharacters(FileInfo *nested, CharsString *characters, int *lastToken) {
	/* Get ruleChars string */
	CharsString token;
	if (getToken(nested, &token, "characters", lastToken))
		if (parseChars(nested, characters, &token)) return 1;
	return 0;
}

static int
getRuleCharsText(FileInfo *nested, CharsString *ruleChars, int *lastToken) {
	CharsString token;
	if (getToken(nested, &token, "Characters operand", lastToken))
		if (parseChars(nested, ruleChars, &token)) return 1;
	return 0;
}

static int
getRuleDotsText(FileInfo *nested, CharsString *ruleDots, int *lastToken) {
	CharsString token;
	if (getToken(nested, &token, "characters", lastToken))
		if (parseChars(nested, ruleDots, &token)) return 1;
	return 0;
}

static int
getRuleDotsPattern(FileInfo *nested, CharsString *ruleDots, int *lastToken) {
	/* Interpret the dets operand */
	CharsString token;
	if (getToken(nested, &token, "Dots operand", lastToken)) {
		if (token.length == 1 && token.chars[0] == '=') {
			ruleDots->length = 0;
			return 1;
		}
		if (parseDots(nested, ruleDots, &token)) return 1;
	}
	return 0;
}

static int
getCharacterClass(FileInfo *nested, const CharacterClass **class,
		const TranslationTableHeader *table, int *lastToken) {
	CharsString token;
	if (getToken(nested, &token, "character class name", lastToken)) {
		if ((*class = findCharacterClass(&token, table))) return 1;
		compileError(nested, "character class not defined.");
	}
	return 0;
}

static int
includeFile(FileInfo *nested, CharsString *includedFile, TranslationTableHeader **table,
		DisplayTableHeader **displayTable);

static TranslationTableOffset
findRuleName(const CharsString *name, const TranslationTableHeader *table) {
	const RuleName *nameRule = table->ruleNames;
	while (nameRule) {
		if ((name->length == nameRule->length) &&
				(memcmp(&name->chars[0], nameRule->name, CHARSIZE * name->length) == 0))
			return nameRule->ruleOffset;
		nameRule = nameRule->next;
	}
	return 0;
}

static int
addRuleName(FileInfo *nested, CharsString *name, TranslationTableOffset newRuleOffset,
		TranslationTableHeader *table) {
	int k;
	RuleName *nameRule;
	if (!(nameRule = malloc(sizeof(*nameRule) + CHARSIZE * (name->length - 1)))) {
		compileError(nested, "not enough memory");
		_lou_outOfMemory();
	}
	memset(nameRule, 0, sizeof(*nameRule));
	// a name is a sequence of characters in the ranges 'a'..'z' and 'A'..'Z'
	for (k = 0; k < name->length; k++) {
		widechar c = name->chars[k];
		if ((c >= 'a' && c <= 'z') || (c >= 'A' && c <= 'Z'))
			nameRule->name[k] = c;
		else {
			compileError(nested, "a name may contain only letters");
			return 0;
		}
	}
	nameRule->length = name->length;
	nameRule->ruleOffset = newRuleOffset;
	nameRule->next = table->ruleNames;
	table->ruleNames = nameRule;
	return 1;
}

static void
deallocateRuleNames(TranslationTableHeader *table) {
	RuleName **ruleNames = &table->ruleNames;
	while (*ruleNames) {
		RuleName *nameRule = *ruleNames;
		*ruleNames = nameRule->next;
		if (nameRule) free(nameRule);
	}
}

static int
compileSwapDots(FileInfo *nested, CharsString *source, CharsString *dest) {
	int k = 0;
	int kk = 0;
	CharsString dotsSource;
	CharsString dotsDest;
	dest->length = 0;
	dotsSource.length = 0;
	while (k <= source->length) {
		if (source->chars[k] != ',' && k != source->length)
			dotsSource.chars[dotsSource.length++] = source->chars[k];
		else {
			if (!parseDots(nested, &dotsDest, &dotsSource)) return 0;
			dest->chars[dest->length++] = dotsDest.length + 1;
			for (kk = 0; kk < dotsDest.length; kk++)
				dest->chars[dest->length++] = dotsDest.chars[kk];
			dotsSource.length = 0;
		}
		k++;
	}
	return 1;
}

static int
compileSwap(FileInfo *nested, TranslationTableOpcode opcode, int *lastToken,
		TranslationTableOffset *newRuleOffset, TranslationTableRule **newRule, int noback,
		int nofor, TranslationTableHeader **table) {
	CharsString ruleChars;
	CharsString ruleDots;
	CharsString name;
	CharsString matches;
	CharsString replacements;
	TranslationTableOffset ruleOffset;
	if (!getToken(nested, &name, "name operand", lastToken)) return 0;
	if (!getToken(nested, &matches, "matches operand", lastToken)) return 0;
	if (!getToken(nested, &replacements, "replacements operand", lastToken)) return 0;
	if (opcode == CTO_SwapCc || opcode == CTO_SwapCd) {
		if (!parseChars(nested, &ruleChars, &matches)) return 0;
	} else {
		if (!compileSwapDots(nested, &matches, &ruleChars)) return 0;
	}
	if (opcode == CTO_SwapCc) {
		if (!parseChars(nested, &ruleDots, &replacements)) return 0;
	} else {
		if (!compileSwapDots(nested, &replacements, &ruleDots)) return 0;
	}
	if (!addRule(nested, opcode, &ruleChars, &ruleDots, 0, 0, &ruleOffset, newRule,
				noback, nofor, table))
		return 0;
	if (!addRuleName(nested, &name, ruleOffset, *table)) return 0;
	if (newRuleOffset) *newRuleOffset = ruleOffset;
	return 1;
}

static int
getNumber(widechar *source, widechar *dest) {
	/* Convert a string of wide character digits to an integer */
	int k = 0;
	*dest = 0;
	while (source[k] >= '0' && source[k] <= '9') *dest = 10 * *dest + (source[k++] - '0');
	return k;
}

/* Start of multipass compiler */

static int
passGetAttributes(CharsString *passLine, int *passLinepos,
		TranslationTableCharacterAttributes *passAttributes, FileInfo *passNested) {
	int more = 1;
	*passAttributes = 0;
	while (more) {
		switch (passLine->chars[*passLinepos]) {
		case pass_any:
			*passAttributes = 0xffffffff;
			break;
		case pass_digit:
			*passAttributes |= CTC_Digit;
			break;
		case pass_litDigit:
			*passAttributes |= CTC_LitDigit;
			break;
		case pass_letter:
			*passAttributes |= CTC_Letter;
			break;
		case pass_math:
			*passAttributes |= CTC_Math;
			break;
		case pass_punctuation:
			*passAttributes |= CTC_Punctuation;
			break;
		case pass_sign:
			*passAttributes |= CTC_Sign;
			break;
		case pass_space:
			*passAttributes |= CTC_Space;
			break;
		case pass_uppercase:
			*passAttributes |= CTC_UpperCase;
			break;
		case pass_lowercase:
			*passAttributes |= CTC_LowerCase;
			break;
		case pass_class1:
			*passAttributes |= CTC_Class1;
			break;
		case pass_class2:
			*passAttributes |= CTC_Class2;
			break;
		case pass_class3:
			*passAttributes |= CTC_Class3;
			break;
		case pass_class4:
			*passAttributes |= CTC_Class4;
			break;
		default:
			more = 0;
			break;
		}
		if (more) (*passLinepos)++;
	}
	if (!*passAttributes) {
		compileError(passNested, "missing attribute");
		(*passLinepos)--;
		return 0;
	}
	return 1;
}

static int
passGetDots(CharsString *passLine, int *passLinepos, CharsString *passHoldString,
		FileInfo *passNested) {
	CharsString collectDots;
	collectDots.length = 0;
	while (*passLinepos < passLine->length &&
			(passLine->chars[*passLinepos] == '-' ||
					(passLine->chars[*passLinepos] >= '0' &&
							passLine->chars[*passLinepos] <= '9') ||
					((passLine->chars[*passLinepos] | 32) >= 'a' &&
							(passLine->chars[*passLinepos] | 32) <= 'f')))
		collectDots.chars[collectDots.length++] = passLine->chars[(*passLinepos)++];
	if (!parseDots(passNested, passHoldString, &collectDots)) return 0;
	return 1;
}

static int
passGetString(CharsString *passLine, int *passLinepos, CharsString *passHoldString,
		FileInfo *passNested) {
	passHoldString->length = 0;
	while (1) {
		if ((*passLinepos >= passLine->length) || !passLine->chars[*passLinepos]) {
			compileError(passNested, "unterminated string");
			return 0;
		}
		if (passLine->chars[*passLinepos] == 34) break;
		if (passLine->chars[*passLinepos] == QUOTESUB)
			passHoldString->chars[passHoldString->length++] = 34;
		else
			passHoldString->chars[passHoldString->length++] =
					passLine->chars[*passLinepos];
		(*passLinepos)++;
	}
	passHoldString->chars[passHoldString->length] = 0;
	(*passLinepos)++;
	return 1;
}

static int
passGetNumber(CharsString *passLine, int *passLinepos, widechar *passHoldNumber) {
	/* Convert a string of wide character digits to an integer */
	*passHoldNumber = 0;
	while ((*passLinepos < passLine->length) && (passLine->chars[*passLinepos] >= '0') &&
			(passLine->chars[*passLinepos] <= '9'))
		*passHoldNumber =
				10 * (*passHoldNumber) + (passLine->chars[(*passLinepos)++] - '0');
	return 1;
}

static int
passGetVariableNumber(FileInfo *nested, CharsString *passLine, int *passLinepos,
		widechar *passHoldNumber) {
	if (!passGetNumber(passLine, passLinepos, passHoldNumber)) return 0;
	if ((*passHoldNumber >= 0) && (*passHoldNumber < NUMVAR)) return 1;
	compileError(nested, "variable number out of range");
	return 0;
}

static int
passGetName(CharsString *passLine, int *passLinepos, CharsString *passHoldString) {
	passHoldString->length = 0;
	// a name is a sequence of characters in the ranges 'a'..'z' and 'A'..'Z'
	do {
		widechar c = passLine->chars[*passLinepos];
		if ((c >= 'a' && c <= 'z') || (c >= 'A' && c <= 'Z')) {
			passHoldString->chars[passHoldString->length++] = c;
			(*passLinepos)++;
		} else {
			break;
		}
	} while (*passLinepos < passLine->length);
	return 1;
}

static inline int
wantsString(TranslationTableOpcode opcode, int actionPart, int nofor) {
	if (opcode == CTO_Correct) return 1;
	if (opcode != CTO_Context) return 0;
	return !nofor == !actionPart;
}

static int
verifyStringOrDots(FileInfo *nested, TranslationTableOpcode opcode, int isString,
		int actionPart, int nofor) {
	if (!wantsString(opcode, actionPart, nofor) == !isString) return 1;

	compileError(nested, "%s are not allowed in the %s part of a %s translation %s rule.",
			isString ? "strings" : "dots", getPartName(actionPart),
			nofor ? "backward" : "forward", _lou_findOpcodeName(opcode));

	return 0;
}

static int
compilePassOpcode(FileInfo *nested, TranslationTableOpcode opcode,
		TranslationTableOffset *newRuleOffset, TranslationTableRule **newRule, int noback,
		int nofor, TranslationTableHeader **table) {
	static CharsString passRuleChars;
	static CharsString passRuleDots;
	/* Compile the operands of a pass opcode */
	widechar passSubOp;
	const CharacterClass *class;
	TranslationTableOffset ruleOffset = 0;
	TranslationTableRule *rule = NULL;
	int k;
	int kk = 0;
	int endTest = 0;
	widechar *passInstructions = passRuleDots.chars;
	int passIC = 0; /* Instruction counter */
	passRuleChars.length = 0;
	FileInfo *passNested = nested;
	CharsString passHoldString;
	widechar passHoldNumber;
	CharsString passLine;
	int passLinepos = 0;
	TranslationTableCharacterAttributes passAttributes;
	passHoldString.length = 0;
	for (k = nested->linepos; k < nested->linelen; k++)
		passHoldString.chars[passHoldString.length++] = nested->line[k];
#define SEPCHAR 0x0001
	for (k = 0; k < passHoldString.length && passHoldString.chars[k] > 32; k++)
		;
	if (k < passHoldString.length)
		passHoldString.chars[k] = SEPCHAR;
	else {
		compileError(passNested, "Invalid multipass operands");
		return 0;
	}
	parseChars(passNested, &passLine, &passHoldString);
	/* Compile test part */
	for (k = 0; k < passLine.length && passLine.chars[k] != SEPCHAR; k++)
		;
	endTest = k;
	passLine.chars[endTest] = pass_endTest;
	passLinepos = 0;
	while (passLinepos <= endTest) {
		if (passIC >= MAXSTRING) {
			compileError(passNested, "Test part in multipass operand too long");
			return 0;
		}
		switch ((passSubOp = passLine.chars[passLinepos])) {
		case pass_lookback:
			passInstructions[passIC++] = pass_lookback;
			passLinepos++;
			passGetNumber(&passLine, &passLinepos, &passHoldNumber);
			if (passHoldNumber == 0) passHoldNumber = 1;
			passInstructions[passIC++] = passHoldNumber;
			break;
		case pass_not:
			passInstructions[passIC++] = pass_not;
			passLinepos++;
			break;
		case pass_first:
			passInstructions[passIC++] = pass_first;
			passLinepos++;
			break;
		case pass_last:
			passInstructions[passIC++] = pass_last;
			passLinepos++;
			break;
		case pass_search:
			passInstructions[passIC++] = pass_search;
			passLinepos++;
			break;
		case pass_string:
			if (!verifyStringOrDots(nested, opcode, 1, 0, nofor)) {
				return 0;
			}
			passLinepos++;
			passInstructions[passIC++] = pass_string;
			passGetString(&passLine, &passLinepos, &passHoldString, passNested);
			goto testDoCharsDots;
		case pass_dots:
			if (!verifyStringOrDots(nested, opcode, 0, 0, nofor)) {
				return 0;
			}
			passLinepos++;
			passInstructions[passIC++] = pass_dots;
			passGetDots(&passLine, &passLinepos, &passHoldString, passNested);
		testDoCharsDots:
			if (passHoldString.length == 0) return 0;
			if (passIC >= MAXSTRING) {
				compileError(passNested,
						"@ operand in test part of multipass operand too long");
				return 0;
			}
			passInstructions[passIC++] = passHoldString.length;
			for (kk = 0; kk < passHoldString.length; kk++) {
				if (passIC >= MAXSTRING) {
					compileError(passNested,
							"@ operand in test part of multipass operand too long");
					return 0;
				}
				passInstructions[passIC++] = passHoldString.chars[kk];
			}
			break;
		case pass_startReplace:
			passInstructions[passIC++] = pass_startReplace;
			passLinepos++;
			break;
		case pass_endReplace:
			passInstructions[passIC++] = pass_endReplace;
			passLinepos++;
			break;
		case pass_variable:
			passLinepos++;
			if (!passGetVariableNumber(nested, &passLine, &passLinepos, &passHoldNumber))
				return 0;
			switch (passLine.chars[passLinepos]) {
			case pass_eq:
				passInstructions[passIC++] = pass_eq;
				goto doComp;
			case pass_lt:
				if (passLine.chars[passLinepos + 1] == pass_eq) {
					passLinepos++;
					passInstructions[passIC++] = pass_lteq;
				} else
					passInstructions[passIC++] = pass_lt;
				goto doComp;
			case pass_gt:
				if (passLine.chars[passLinepos + 1] == pass_eq) {
					passLinepos++;
					passInstructions[passIC++] = pass_gteq;
				} else
					passInstructions[passIC++] = pass_gt;
			doComp:
				passInstructions[passIC++] = passHoldNumber;
				passLinepos++;
				passGetNumber(&passLine, &passLinepos, &passHoldNumber);
				passInstructions[passIC++] = passHoldNumber;
				break;
			default:
				compileError(passNested, "incorrect comparison operator");
				return 0;
			}
			break;
		case pass_attributes:
			passLinepos++;
			if (!passGetAttributes(&passLine, &passLinepos, &passAttributes, passNested))
				return 0;
		insertAttributes:
			passInstructions[passIC++] = pass_attributes;
			passInstructions[passIC++] = passAttributes >> 16;
			passInstructions[passIC++] = passAttributes & 0xffff;
		getRange:
			if (passLine.chars[passLinepos] == pass_until) {
				passLinepos++;
				passInstructions[passIC++] = 1;
				passInstructions[passIC++] = 0xffff;
				break;
			}
			passGetNumber(&passLine, &passLinepos, &passHoldNumber);
			if (passHoldNumber == 0) {
				passHoldNumber = passInstructions[passIC++] = 1;
				passInstructions[passIC++] = 1; /* This is not an error */
				break;
			}
			passInstructions[passIC++] = passHoldNumber;
			if (passLine.chars[passLinepos] != pass_hyphen) {
				passInstructions[passIC++] = passHoldNumber;
				break;
			}
			passLinepos++;
			passGetNumber(&passLine, &passLinepos, &passHoldNumber);
			if (passHoldNumber == 0) {
				compileError(passNested, "invalid range");
				return 0;
			}
			passInstructions[passIC++] = passHoldNumber;
			break;
		case pass_groupstart:
		case pass_groupend:
			passLinepos++;
			passGetName(&passLine, &passLinepos, &passHoldString);
			ruleOffset = findRuleName(&passHoldString, *table);
			if (ruleOffset)
				rule = (TranslationTableRule *)&(*table)->ruleArea[ruleOffset];
			if (rule && rule->opcode == CTO_Grouping) {
				passInstructions[passIC++] = passSubOp;
				passInstructions[passIC++] = ruleOffset >> 16;
				passInstructions[passIC++] = ruleOffset & 0xffff;
				break;
			} else {
				compileError(passNested, "%s is not a grouping name",
						_lou_showString(
								&passHoldString.chars[0], passHoldString.length, 0));
				return 0;
			}
			break;
		case pass_swap:
			passLinepos++;
			passGetName(&passLine, &passLinepos, &passHoldString);
			if ((class = findCharacterClass(&passHoldString, *table))) {
				passAttributes = class->attribute;
				goto insertAttributes;
			}
			ruleOffset = findRuleName(&passHoldString, *table);
			if (ruleOffset)
				rule = (TranslationTableRule *)&(*table)->ruleArea[ruleOffset];
			if (rule &&
					(rule->opcode == CTO_SwapCc || rule->opcode == CTO_SwapCd ||
							rule->opcode == CTO_SwapDd)) {
				passInstructions[passIC++] = pass_swap;
				passInstructions[passIC++] = ruleOffset >> 16;
				passInstructions[passIC++] = ruleOffset & 0xffff;
				goto getRange;
			}
			compileError(passNested, "%s is neither a class name nor a swap name.",
					_lou_showString(&passHoldString.chars[0], passHoldString.length, 0));
			return 0;
		case pass_endTest:
			passInstructions[passIC++] = pass_endTest;
			passLinepos++;
			break;
		default:
			compileError(passNested, "incorrect operator '%c ' in test part",
					passLine.chars[passLinepos]);
			return 0;
		}

	} /* Compile action part */

	/* Compile action part */
	while (passLinepos < passLine.length && passLine.chars[passLinepos] <= 32)
		passLinepos++;
	while (passLinepos < passLine.length && passLine.chars[passLinepos] > 32) {
		if (passIC >= MAXSTRING) {
			compileError(passNested, "Action part in multipass operand too long");
			return 0;
		}
		switch ((passSubOp = passLine.chars[passLinepos])) {
		case pass_string:
			if (!verifyStringOrDots(nested, opcode, 1, 1, nofor)) {
				return 0;
			}
			passLinepos++;
			passInstructions[passIC++] = pass_string;
			passGetString(&passLine, &passLinepos, &passHoldString, passNested);
			goto actionDoCharsDots;
		case pass_dots:
			if (!verifyStringOrDots(nested, opcode, 0, 1, nofor)) {
				return 0;
			}
			passLinepos++;
			passGetDots(&passLine, &passLinepos, &passHoldString, passNested);
			passInstructions[passIC++] = pass_dots;
		actionDoCharsDots:
			if (passHoldString.length == 0) return 0;
			if (passIC >= MAXSTRING) {
				compileError(passNested,
						"@ operand in action part of multipass operand too long");
				return 0;
			}
			passInstructions[passIC++] = passHoldString.length;
			for (kk = 0; kk < passHoldString.length; kk++) {
				if (passIC >= MAXSTRING) {
					compileError(passNested,
							"@ operand in action part of multipass operand too long");
					return 0;
				}
				passInstructions[passIC++] = passHoldString.chars[kk];
			}
			break;
		case pass_variable:
			passLinepos++;
			if (!passGetVariableNumber(nested, &passLine, &passLinepos, &passHoldNumber))
				return 0;
			switch (passLine.chars[passLinepos]) {
			case pass_eq:
				passInstructions[passIC++] = pass_eq;
				passInstructions[passIC++] = passHoldNumber;
				passLinepos++;
				passGetNumber(&passLine, &passLinepos, &passHoldNumber);
				passInstructions[passIC++] = passHoldNumber;
				break;
			case pass_plus:
			case pass_hyphen:
				passInstructions[passIC++] = passLine.chars[passLinepos++];
				passInstructions[passIC++] = passHoldNumber;
				break;
			default:
				compileError(passNested, "incorrect variable operator in action part");
				return 0;
			}
			break;
		case pass_copy:
			passInstructions[passIC++] = pass_copy;
			passLinepos++;
			break;
		case pass_omit:
			passInstructions[passIC++] = pass_omit;
			passLinepos++;
			break;
		case pass_groupreplace:
		case pass_groupstart:
		case pass_groupend:
			passLinepos++;
			passGetName(&passLine, &passLinepos, &passHoldString);
			ruleOffset = findRuleName(&passHoldString, *table);
			if (ruleOffset)
				rule = (TranslationTableRule *)&(*table)->ruleArea[ruleOffset];
			if (rule && rule->opcode == CTO_Grouping) {
				passInstructions[passIC++] = passSubOp;
				passInstructions[passIC++] = ruleOffset >> 16;
				passInstructions[passIC++] = ruleOffset & 0xffff;
				break;
			}
			compileError(passNested, "%s is not a grouping name",
					_lou_showString(&passHoldString.chars[0], passHoldString.length, 0));
			return 0;
		case pass_swap:
			passLinepos++;
			passGetName(&passLine, &passLinepos, &passHoldString);
			ruleOffset = findRuleName(&passHoldString, *table);
			if (ruleOffset)
				rule = (TranslationTableRule *)&(*table)->ruleArea[ruleOffset];
			if (rule &&
					(rule->opcode == CTO_SwapCc || rule->opcode == CTO_SwapCd ||
							rule->opcode == CTO_SwapDd)) {
				passInstructions[passIC++] = pass_swap;
				passInstructions[passIC++] = ruleOffset >> 16;
				passInstructions[passIC++] = ruleOffset & 0xffff;
				break;
			}
			compileError(passNested, "%s is not a swap name.",
					_lou_showString(&passHoldString.chars[0], passHoldString.length, 0));
			return 0;
			break;
		default:
			compileError(passNested, "incorrect operator in action part");
			return 0;
		}
	}

	/* Analyze and add rule */
	passRuleDots.length = passIC;

	{
		widechar *characters;
		int length;
		int found = passFindCharacters(
				passNested, passInstructions, passRuleDots.length, &characters, &length);

		if (!found) return 0;

		if (characters) {
			for (k = 0; k < length; k += 1) passRuleChars.chars[k] = characters[k];
			passRuleChars.length = k;
		}
	}

	if (!addRule(passNested, opcode, &passRuleChars, &passRuleDots, 0, 0, newRuleOffset,
				newRule, noback, nofor, table))
		return 0;
	return 1;
}

/* End of multipass compiler */

static int
compileBrailleIndicator(FileInfo *nested, const char *ermsg,
		TranslationTableOpcode opcode, int *lastToken,
		TranslationTableOffset *newRuleOffset, TranslationTableRule **newRule, int noback,
		int nofor, TranslationTableHeader **table) {
	CharsString token;
	CharsString cells;
	if (getToken(nested, &token, ermsg, lastToken))
		if (parseDots(nested, &cells, &token))
			if (!addRule(nested, opcode, NULL, &cells, 0, 0, newRuleOffset, newRule,
						noback, nofor, table))
				return 0;
	return 1;
}

static int
compileNumber(FileInfo *nested, int *lastToken) {
	CharsString token;
	widechar dest;
	if (!getToken(nested, &token, "number", lastToken)) return 0;
	getNumber(&token.chars[0], &dest);
	if (!(dest > 0)) {
		compileError(nested, "a nonzero positive number is required");
		return 0;
	}
	return dest;
}

static int
compileGrouping(FileInfo *nested, int *lastToken, TranslationTableOffset *newRuleOffset,
		TranslationTableRule **newRule, int noback, int nofor,
		TranslationTableHeader **table, DisplayTableHeader **displayTable) {
	int k;
	CharsString name;
	CharsString groupChars;
	CharsString groupDots;
	CharsString dotsParsed;
	if (!getToken(nested, &name, "name operand", lastToken)) return 0;
	if (!getRuleCharsText(nested, &groupChars, lastToken)) return 0;
	if (!getToken(nested, &groupDots, "dots operand", lastToken)) return 0;
	for (k = 0; k < groupDots.length && groupDots.chars[k] != ','; k++)
		;
	if (k == groupDots.length) {
		compileError(
				nested, "Dots operand must consist of two cells separated by a comma");
		return 0;
	}
	groupDots.chars[k] = '-';
	if (!parseDots(nested, &dotsParsed, &groupDots)) return 0;
	if (groupChars.length != 2 || dotsParsed.length != 2) {
		compileError(nested,
				"two Unicode characters and two cells separated by a comma are needed.");
		return 0;
	}
	if (table) {
		TranslationTableOffset ruleOffset;
		TranslationTableCharacter *charsDotsPtr;
		charsDotsPtr = addCharOrDots(nested, groupChars.chars[0], 0, table);
		charsDotsPtr->attributes |= CTC_Math;
		charsDotsPtr->uppercase = charsDotsPtr->realchar;
		charsDotsPtr->lowercase = charsDotsPtr->realchar;
		charsDotsPtr = addCharOrDots(nested, groupChars.chars[1], 0, table);
		charsDotsPtr->attributes |= CTC_Math;
		charsDotsPtr->uppercase = charsDotsPtr->realchar;
		charsDotsPtr->lowercase = charsDotsPtr->realchar;
		charsDotsPtr = addCharOrDots(nested, dotsParsed.chars[0], 1, table);
		charsDotsPtr->attributes |= CTC_Math;
		charsDotsPtr->uppercase = charsDotsPtr->realchar;
		charsDotsPtr->lowercase = charsDotsPtr->realchar;
		charsDotsPtr = addCharOrDots(nested, dotsParsed.chars[1], 1, table);
		charsDotsPtr->attributes |= CTC_Math;
		charsDotsPtr->uppercase = charsDotsPtr->realchar;
		charsDotsPtr->lowercase = charsDotsPtr->realchar;
		if (!addRule(nested, CTO_Grouping, &groupChars, &dotsParsed, 0, 0, &ruleOffset,
					newRule, noback, nofor, table))
			return 0;
		if (!addRuleName(nested, &name, ruleOffset, *table)) return 0;
		if (newRuleOffset) *newRuleOffset = ruleOffset;
	}
	if (displayTable) {
		putCharAndDots(nested, groupChars.chars[0], dotsParsed.chars[0], displayTable);
		putCharAndDots(nested, groupChars.chars[1], dotsParsed.chars[1], displayTable);
	}
	if (table) {
		widechar endChar;
		widechar endDots;
		endChar = groupChars.chars[1];
		endDots = dotsParsed.chars[1];
		groupChars.length = dotsParsed.length = 1;
		if (!addRule(nested, CTO_Math, &groupChars, &dotsParsed, 0, 0, newRuleOffset,
					newRule, noback, nofor, table))
			return 0;
		groupChars.chars[0] = endChar;
		dotsParsed.chars[0] = endDots;
		if (!addRule(nested, CTO_Math, &groupChars, &dotsParsed, 0, 0, newRuleOffset,
					newRule, noback, nofor, table))
			return 0;
	}
	return 1;
}

static int
compileUplow(FileInfo *nested, int *lastToken, TranslationTableOffset *newRuleOffset,
		TranslationTableRule **newRule, int noback, int nofor,
		TranslationTableHeader **table, DisplayTableHeader **displayTable) {
	int k;
	TranslationTableCharacter *upperChar;
	TranslationTableCharacter *lowerChar;
	TranslationTableCharacter *upperCell = NULL;
	TranslationTableCharacter *lowerCell = NULL;
	CharsString ruleChars;
	CharsString ruleDots;
	CharsString upperDots;
	CharsString lowerDots;
	int haveLowerDots = 0;
	TranslationTableCharacterAttributes attr;
	if (!getRuleCharsText(nested, &ruleChars, lastToken)) return 0;
	if (!getToken(nested, &ruleDots, "dots operand", lastToken)) return 0;
	for (k = 0; k < ruleDots.length && ruleDots.chars[k] != ','; k++)
		;
	if (k == ruleDots.length) {
		if (!parseDots(nested, &upperDots, &ruleDots)) return 0;
		lowerDots.length = upperDots.length;
		for (k = 0; k < upperDots.length; k++) lowerDots.chars[k] = upperDots.chars[k];
		lowerDots.chars[k] = 0;
	} else {
		haveLowerDots = ruleDots.length;
		ruleDots.length = k;
		if (!parseDots(nested, &upperDots, &ruleDots)) return 0;
		ruleDots.length = 0;
		k++;
		for (; k < haveLowerDots; k++)
			ruleDots.chars[ruleDots.length++] = ruleDots.chars[k];
		if (!parseDots(nested, &lowerDots, &ruleDots)) return 0;
	}
	if (ruleChars.length != 2 || upperDots.length < 1) {
		compileError(nested,
				"Exactly two Unicode characters and at least one cell are required.");
		return 0;
	}
	if (haveLowerDots && lowerDots.length < 1) {
		compileError(nested, "at least one cell is required after the comma.");
		return 0;
	}
	if (table) {
		upperChar = addCharOrDots(nested, ruleChars.chars[0], 0, table);
		upperChar->attributes |= CTC_Letter | CTC_UpperCase;
		upperChar->uppercase = ruleChars.chars[0];
		upperChar->lowercase = ruleChars.chars[1];
		lowerChar = addCharOrDots(nested, ruleChars.chars[1], 0, table);
		lowerChar->attributes |= CTC_Letter | CTC_LowerCase;
		lowerChar->uppercase = ruleChars.chars[0];
		lowerChar->lowercase = ruleChars.chars[1];
		for (k = 0; k < upperDots.length; k++)
			if (!compile_findCharOrDots(upperDots.chars[k], 1, *table)) {
				attr = CTC_Letter | CTC_UpperCase;
				upperCell = addCharOrDots(nested, upperDots.chars[k], 1, table);
				upperCell->attributes |= attr;
				upperCell->uppercase = upperCell->realchar;
			}
		if (haveLowerDots) {
			for (k = 0; k < lowerDots.length; k++)
				if (!compile_findCharOrDots(lowerDots.chars[k], 1, *table)) {
					attr = CTC_Letter | CTC_LowerCase;
					lowerCell = addCharOrDots(nested, lowerDots.chars[k], 1, table);
					if (lowerDots.length != 1) attr = CTC_Space;
					lowerCell->attributes |= attr;
					lowerCell->lowercase = lowerCell->realchar;
				}
		} else if (upperCell != NULL && upperDots.length == 1)
			upperCell->attributes |= CTC_LowerCase;
		if (upperCell != NULL) upperCell->lowercase = lowerDots.chars[0];
		if (lowerCell != NULL) lowerCell->uppercase = upperDots.chars[0];
	}
	if (displayTable) {
		if (lowerDots.length == 1)
			putCharAndDots(nested, ruleChars.chars[1], lowerDots.chars[0], displayTable);
		if (upperDots.length == 1)
			putCharAndDots(nested, ruleChars.chars[0], upperDots.chars[0], displayTable);
	}
	if (table) {
		ruleChars.length = 1;
		ruleChars.chars[2] = ruleChars.chars[0];
		ruleChars.chars[0] = ruleChars.chars[1];
		if (!addRule(nested, CTO_LowerCase, &ruleChars, &lowerDots, 0, 0, newRuleOffset,
					newRule, noback, nofor, table))
			return 0;
		ruleChars.chars[0] = ruleChars.chars[2];
		if (!addRule(nested, CTO_UpperCase, &ruleChars, &upperDots, 0, 0, newRuleOffset,
					newRule, noback, nofor, table))
			return 0;
	}
	return 1;
}

/* Functions for compiling hyphenation tables */

typedef struct HyphenDict { /* hyphenation dictionary: finite state machine */
	int numStates;
	HyphenationState *states;
} HyphenDict;

#define DEFAULTSTATE 0xffff
#define HYPHENHASHSIZE 8191

typedef struct HyphenHashEntry {
	struct HyphenHashEntry *next;
	CharsString *key;
	int val;
} HyphenHashEntry;

typedef struct HyphenHashTab {
	HyphenHashEntry *entries[HYPHENHASHSIZE];
} HyphenHashTab;

/* a hash function from ASU - adapted from Gtk+ */
static unsigned int
hyphenStringHash(const CharsString *s) {
	int k;
	unsigned int h = 0, g;
	for (k = 0; k < s->length; k++) {
		h = (h << 4) + s->chars[k];
		if ((g = h & 0xf0000000)) {
			h = h ^ (g >> 24);
			h = h ^ g;
		}
	}
	return h;
}

static HyphenHashTab *
hyphenHashNew(void) {
	HyphenHashTab *hashTab;
	if (!(hashTab = malloc(sizeof(HyphenHashTab)))) _lou_outOfMemory();
	memset(hashTab, 0, sizeof(HyphenHashTab));
	return hashTab;
}

static void
hyphenHashFree(HyphenHashTab *hashTab) {
	int i;
	HyphenHashEntry *e, *next;
	for (i = 0; i < HYPHENHASHSIZE; i++)
		for (e = hashTab->entries[i]; e; e = next) {
			next = e->next;
			free(e->key);
			free(e);
		}
	free(hashTab);
}

/* assumes that key is not already present! */
static void
hyphenHashInsert(HyphenHashTab *hashTab, const CharsString *key, int val) {
	int i, j;
	HyphenHashEntry *e;
	i = hyphenStringHash(key) % HYPHENHASHSIZE;
	if (!(e = malloc(sizeof(HyphenHashEntry)))) _lou_outOfMemory();
	e->next = hashTab->entries[i];
	e->key = malloc((key->length + 1) * CHARSIZE);
	if (!e->key) _lou_outOfMemory();
	e->key->length = key->length;
	for (j = 0; j < key->length; j++) e->key->chars[j] = key->chars[j];
	e->val = val;
	hashTab->entries[i] = e;
}

/* return val if found, otherwise DEFAULTSTATE */
static int
hyphenHashLookup(HyphenHashTab *hashTab, const CharsString *key) {
	int i, j;
	HyphenHashEntry *e;
	if (key->length == 0) return 0;
	i = hyphenStringHash(key) % HYPHENHASHSIZE;
	for (e = hashTab->entries[i]; e; e = e->next) {
		if (key->length != e->key->length) continue;
		for (j = 0; j < key->length; j++)
			if (key->chars[j] != e->key->chars[j]) break;
		if (j == key->length) return e->val;
	}
	return DEFAULTSTATE;
}

static int
hyphenGetNewState(HyphenDict *dict, HyphenHashTab *hashTab, const CharsString *string) {
	hyphenHashInsert(hashTab, string, dict->numStates);
	/* predicate is true if dict->numStates is a power of two */
	if (!(dict->numStates & (dict->numStates - 1)))
		dict->states =
				realloc(dict->states, (dict->numStates << 1) * sizeof(HyphenationState));
	if (!dict->states) _lou_outOfMemory();
	dict->states[dict->numStates].hyphenPattern = 0;
	dict->states[dict->numStates].fallbackState = DEFAULTSTATE;
	dict->states[dict->numStates].numTrans = 0;
	dict->states[dict->numStates].trans.pointer = NULL;
	return dict->numStates++;
}

/* add a transition from state1 to state2 through ch - assumes that the
 * transition does not already exist */
static void
hyphenAddTrans(HyphenDict *dict, int state1, int state2, widechar ch) {
	int numTrans;
	numTrans = dict->states[state1].numTrans;
	if (numTrans == 0)
		dict->states[state1].trans.pointer = malloc(sizeof(HyphenationTrans));
	else if (!(numTrans & (numTrans - 1)))
		dict->states[state1].trans.pointer = realloc(dict->states[state1].trans.pointer,
				(numTrans << 1) * sizeof(HyphenationTrans));
	dict->states[state1].trans.pointer[numTrans].ch = ch;
	dict->states[state1].trans.pointer[numTrans].newState = state2;
	dict->states[state1].numTrans++;
}

static int
compileHyphenation(FileInfo *nested, CharsString *encoding, int *lastToken,
		TranslationTableHeader **table) {
	CharsString hyph;
	HyphenationTrans *holdPointer;
	HyphenHashTab *hashTab;
	CharsString word;
	char pattern[MAXSTRING + 1];
	unsigned int stateNum = 0, lastState = 0;
	int i, j, k = encoding->length;
	widechar ch;
	int found;
	HyphenHashEntry *e;
	HyphenDict dict;
	TranslationTableOffset holdOffset;
	/* Set aside enough space for hyphenation states and transitions in
	 * translation table. Must be done before anything else */
	allocateSpaceInTranslationTable(nested, NULL, 250000, table);
	hashTab = hyphenHashNew();
	dict.numStates = 1;
	dict.states = malloc(sizeof(HyphenationState));
	if (!dict.states) _lou_outOfMemory();
	dict.states[0].hyphenPattern = 0;
	dict.states[0].fallbackState = DEFAULTSTATE;
	dict.states[0].numTrans = 0;
	dict.states[0].trans.pointer = NULL;
	do {
		if (encoding->chars[0] == 'I') {
			if (!getToken(nested, &hyph, NULL, lastToken)) continue;
		} else {
			/* UTF-8 */
			if (!getToken(nested, &word, NULL, lastToken)) continue;
			parseChars(nested, &hyph, &word);
		}
		if (hyph.length == 0 || hyph.chars[0] == '#' || hyph.chars[0] == '%' ||
				hyph.chars[0] == '<')
			continue; /* comment */
		j = 0;
		pattern[j] = '0';
		for (i = 0; i < hyph.length; i++) {
			if (hyph.chars[i] >= '0' && hyph.chars[i] <= '9')
				pattern[j] = (char)hyph.chars[i];
			else {
				word.chars[j] = hyph.chars[i];
				pattern[++j] = '0';
			}
		}
		word.chars[j] = 0;
		word.length = j;
		pattern[j + 1] = 0;
		for (i = 0; pattern[i] == '0'; i++)
			;
		found = hyphenHashLookup(hashTab, &word);
		if (found != DEFAULTSTATE)
			stateNum = found;
		else
			stateNum = hyphenGetNewState(&dict, hashTab, &word);
		k = j + 2 - i;
		if (k > 0) {
			allocateSpaceInTranslationTable(
					nested, &dict.states[stateNum].hyphenPattern, k, table);
			memcpy(&(*table)->ruleArea[dict.states[stateNum].hyphenPattern], &pattern[i],
					k);
		}
		/* now, put in the prefix transitions */
		while (found == DEFAULTSTATE) {
			lastState = stateNum;
			ch = word.chars[word.length-- - 1];
			found = hyphenHashLookup(hashTab, &word);
			if (found != DEFAULTSTATE)
				stateNum = found;
			else
				stateNum = hyphenGetNewState(&dict, hashTab, &word);
			hyphenAddTrans(&dict, stateNum, lastState, ch);
		}
	} while (_lou_getALine(nested));
	/* put in the fallback states */
	for (i = 0; i < HYPHENHASHSIZE; i++) {
		for (e = hashTab->entries[i]; e; e = e->next) {
			for (j = 1; j <= e->key->length; j++) {
				word.length = 0;
				for (k = j; k < e->key->length; k++)
					word.chars[word.length++] = e->key->chars[k];
				stateNum = hyphenHashLookup(hashTab, &word);
				if (stateNum != DEFAULTSTATE) break;
			}
			if (e->val) dict.states[e->val].fallbackState = stateNum;
		}
	}
	hyphenHashFree(hashTab);
	/* Transfer hyphenation information to table */
	for (i = 0; i < dict.numStates; i++) {
		if (dict.states[i].numTrans == 0)
			dict.states[i].trans.offset = 0;
		else {
			holdPointer = dict.states[i].trans.pointer;
			allocateSpaceInTranslationTable(nested, &dict.states[i].trans.offset,
					dict.states[i].numTrans * sizeof(HyphenationTrans), table);
			memcpy(&(*table)->ruleArea[dict.states[i].trans.offset], holdPointer,
					dict.states[i].numTrans * sizeof(HyphenationTrans));
			free(holdPointer);
		}
	}
	allocateSpaceInTranslationTable(
			nested, &holdOffset, dict.numStates * sizeof(HyphenationState), table);
	(*table)->hyphenStatesArray = holdOffset;
	/* Prevents segmentation fault if table is reallocated */
	memcpy(&(*table)->ruleArea[(*table)->hyphenStatesArray], &dict.states[0],
			dict.numStates * sizeof(HyphenationState));
	free(dict.states);
	return 1;
}

static int
compileCharDef(FileInfo *nested, TranslationTableOpcode opcode,
		TranslationTableCharacterAttributes attributes, int *lastToken,
		TranslationTableOffset *newRuleOffset, TranslationTableRule **newRule, int noback,
		int nofor, TranslationTableHeader **table, DisplayTableHeader **displayTable) {
	CharsString ruleChars;
	CharsString ruleDots;
	if (!getRuleCharsText(nested, &ruleChars, lastToken)) return 0;
	if (!getRuleDotsPattern(nested, &ruleDots, lastToken)) return 0;
	if (ruleChars.length != 1) {
		compileError(nested, "Exactly one character is required.");
		return 0;
	}
	if (ruleDots.length < 1) {
		compileError(nested, "At least one cell is required.");
		return 0;
	}
	if (table) {
		TranslationTableCharacter *character;
		TranslationTableCharacter *cell = NULL;
		int k;
		if (attributes & (CTC_UpperCase | CTC_LowerCase)) attributes |= CTC_Letter;
		character = addCharOrDots(nested, ruleChars.chars[0], 0, table);
		character->attributes |= attributes;
		character->uppercase = character->lowercase = character->realchar;
		for (k = ruleDots.length - 1; k >= 0; k -= 1) {
			cell = compile_findCharOrDots(ruleDots.chars[k], 1, *table);
			if (!cell) {
				cell = addCharOrDots(nested, ruleDots.chars[k], 1, table);
				cell->uppercase = cell->lowercase = cell->realchar;
			}
		}
		if (ruleDots.length == 1) cell->attributes |= attributes;
	}
	if (displayTable && ruleDots.length == 1)
		putCharAndDots(nested, ruleChars.chars[0], ruleDots.chars[0], displayTable);
	if (table)
		if (!addRule(nested, opcode, &ruleChars, &ruleDots, 0, 0, newRuleOffset, newRule,
					noback, nofor, table))
			return 0;
	return 1;
}

static int
compileBeforeAfter(FileInfo *nested, int *lastToken) {
	/* 1=before, 2=after, 0=error */
	CharsString token;
	CharsString tmp;
	if (getToken(nested, &token, "last word before or after", lastToken))
		if (parseChars(nested, &tmp, &token)) {
			if (eqasc2uni((unsigned char *)"before", tmp.chars, 6)) return 1;
			if (eqasc2uni((unsigned char *)"after", tmp.chars, 5)) return 2;
		}
	return 0;
}

static int
compileRule(FileInfo *nested, TranslationTableOffset *newRuleOffset,
		TranslationTableRule **newRule, TranslationTableHeader **table,
		DisplayTableHeader **displayTable) {
	int lastToken = 0;
	int ok = 1;
	CharsString token;
	TranslationTableOpcode opcode;
	CharsString ruleChars;
	CharsString ruleDots;
	CharsString cells;
	CharsString scratchPad;
	CharsString emphClass;
	TranslationTableCharacterAttributes after = 0;
	TranslationTableCharacterAttributes before = 0;
	TranslationTableCharacter *c = NULL;
	widechar *patterns = NULL;
	int k, i;
	int noback, nofor;
	noback = nofor = 0;
doOpcode:
	if (!getToken(nested, &token, NULL, &lastToken)) return 1;	/* blank line */
	if (token.chars[0] == '#' || token.chars[0] == '<') return 1; /* comment */
	if (nested->lineNumber == 1 &&
			(eqasc2uni((unsigned char *)"ISO", token.chars, 3) ||
					eqasc2uni((unsigned char *)"UTF-8", token.chars, 5))) {
		if (table)
			compileHyphenation(nested, &token, &lastToken, table);
		else
			/* ignore the whole file */
			while (_lou_getALine(nested))
				;
		return 1;
	}
	opcode = getOpcode(nested, &token);
	switch (opcode) {
	case CTO_IncludeFile: {
		CharsString includedFile;
		if (getToken(nested, &token, "include file name", &lastToken))
			if (parseChars(nested, &includedFile, &token))
				if (!includeFile(nested, &includedFile, table, displayTable)) ok = 0;
		break;
	}
	case CTO_NoBack:
		if (nofor) {
			compileError(nested, "%s already specified.", _lou_findOpcodeName(CTO_NoFor));
			ok = 0;
			break;
		}
		noback = 1;
		goto doOpcode;
	case CTO_NoFor:
		if (noback) {
			compileError(
					nested, "%s already specified.", _lou_findOpcodeName(CTO_NoBack));
			ok = 0;
			break;
		}
		nofor = 1;
		goto doOpcode;
	case CTO_Space:
		compileCharDef(nested, opcode, CTC_Space, &lastToken, newRuleOffset, newRule,
				noback, nofor, table, displayTable);
		break;
	case CTO_Digit:
		compileCharDef(nested, opcode, CTC_Digit, &lastToken, newRuleOffset, newRule,
				noback, nofor, table, displayTable);
		break;
	case CTO_LitDigit:
		compileCharDef(nested, opcode, CTC_LitDigit, &lastToken, newRuleOffset, newRule,
				noback, nofor, table, displayTable);
		break;
	case CTO_Punctuation:
		compileCharDef(nested, opcode, CTC_Punctuation, &lastToken, newRuleOffset,
				newRule, noback, nofor, table, displayTable);
		break;
	case CTO_Math:
		compileCharDef(nested, opcode, CTC_Math, &lastToken, newRuleOffset, newRule,
				noback, nofor, table, displayTable);
		break;
	case CTO_Sign:
		compileCharDef(nested, opcode, CTC_Sign, &lastToken, newRuleOffset, newRule,
				noback, nofor, table, displayTable);