HomoFast-eSpeak-Persian/src/libespeak-ng/numbers.c

/*
 * Copyright (C) 2005 to 2015 by Jonathan Duddington
 * email: [email protected]
 * Copyright (C) 2015-2016, 2020 Reece H. Dunn
 *
 * This program is free software; you can redistribute it and/or modify
 * it under the terms of the GNU General Public License as published by
 * the Free Software Foundation; either version 3 of the License, or
 * (at your option) any later version.
 *
 * This program 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 General Public License for more details.
 *
 * You should have received a copy of the GNU General Public License
 * along with this program; if not, see: <http://www.gnu.org/licenses/>.
 */

#include "config.h"

#include <ctype.h>
#include <stdbool.h>
#include <stdint.h>
#include <stdio.h>
#include <stdlib.h>
#include <string.h>
#include <wctype.h>

#include <espeak-ng/espeak_ng.h>
#include <espeak-ng/speak_lib.h>
#include <espeak-ng/encoding.h>

#include "numbers.h"
#include "common.h"
#include "dictionary.h"  // for Lookup, TranslateRules, EncodePhonemes, Look...
#include "phoneme.h"     // for phonSWITCH, PHONEME_TAB, phonEND_WORD, phonP...
#include "readclause.h"  // for WordToString2
#include "synthdata.h"   // for SelectPhonemeTable
#include "synthesize.h"  // for phoneme_tab
#include "translate.h"   // for Translator, LANGUAGE_OPTIONS, IsDigit09, WOR...
#include "voice.h"       // for voice, voice_t

#define M_LIGATURE  0x8000
#define M_NAME      0
#define M_SMALLCAP  1
#define M_TURNED    2
#define M_REVERSED  3
#define M_CURL      4

#define M_ACUTE     5
#define M_BREVE     6
#define M_CARON     7
#define M_CEDILLA   8
#define M_CIRCUMFLEX 9
#define M_DIAERESIS 10
#define M_DOUBLE_ACUTE 11
#define M_DOT_ABOVE 12
#define M_GRAVE     13
#define M_MACRON    14
#define M_OGONEK    15
#define M_RING      16
#define M_STROKE    17
#define M_TILDE     18

#define M_BAR       19
#define M_RETROFLEX 20
#define M_HOOK      21

#define M_MIDDLE_DOT  M_DOT_ABOVE // duplicate of M_DOT_ABOVE
#define M_IMPLOSIVE   M_HOOK

static int n_digit_lookup;
static char *digit_lookup;
static int speak_missing_thousands;
static int number_control;

typedef struct {
	const char *name;
	int accent_flags;    // bit 0, say before the letter name
} ACCENTS;

// these are tokens to look up in the *_list file.
static ACCENTS accents_tab[] = {
	{ "_lig", 1 },
	{ "_smc", 0 },  // smallcap
	{ "_tur", 0 },  // turned
	{ "_rev", 0 },  // reversed
	{ "_crl", 0 },  // curl

	{ "_acu", 0 },  // acute
	{ "_brv", 0 },  // breve
	{ "_hac", 0 },  // caron/hacek
	{ "_ced", 0 },  // cedilla
	{ "_cir", 0 },  // circumflex
	{ "_dia", 0 },  // diaeresis
	{ "_ac2", 0 },  // double acute
	{ "_dot", 0 },  // dot
	{ "_grv", 0 },  // grave
	{ "_mcn", 0 },  // macron
	{ "_ogo", 0 },  // ogonek
	{ "_rng", 0 },  // ring
	{ "_stk", 0 },  // stroke
	{ "_tld", 0 },  // tilde

	{ "_bar", 0 },  // bar
	{ "_rfx", 0 },  // retroflex
	{ "_hok", 0 },  // hook
};

#define CAPITAL  0
#define LETTER(ch, mod1, mod2) (ch-59)+(mod1 << 6)+(mod2 << 11)
#define LIGATURE(ch1, ch2, mod1) (ch1-59)+((ch2-59) << 6)+(mod1 << 12)+M_LIGATURE

#define L_ALPHA   60 // U+3B1
#define L_SCHWA   61 // U+259
#define L_OPEN_E  62 // U+25B
#define L_GAMMA   63 // U+3B3
#define L_IOTA    64 // U+3B9

#define L_PHI     67 // U+3C6
#define L_ESH     68 // U+283
#define L_UPSILON 69 // U+3C5
#define L_EZH     70 // U+292
#define L_GLOTTAL 71 // U+294
#define L_RTAP    72 // U+27E
#define L_RLONG   73 // U+27C

static const short non_ascii_tab[] = {
	0,
	0x3b1, 0x259, 0x25b, 0x3b3, 0x3b9, 0x153, 0x3c9,
	0x3c6, 0x283, 0x3c5, 0x292, 0x294, 0x27e, 0x27c
};

// characters U+00e0 to U+017f
static const unsigned short letter_accents_0e0[] = {
	LETTER('a', M_GRAVE, 0), // U+00e0
	LETTER('a', M_ACUTE, 0),
	LETTER('a', M_CIRCUMFLEX, 0),
	LETTER('a', M_TILDE, 0),
	LETTER('a', M_DIAERESIS, 0),
	LETTER('a', M_RING, 0),
	LIGATURE('a', 'e', 0),
	LETTER('c', M_CEDILLA, 0),
	LETTER('e', M_GRAVE, 0),
	LETTER('e', M_ACUTE, 0),
	LETTER('e', M_CIRCUMFLEX, 0),
	LETTER('e', M_DIAERESIS, 0),
	LETTER('i', M_GRAVE, 0),
	LETTER('i', M_ACUTE, 0),
	LETTER('i', M_CIRCUMFLEX, 0),
	LETTER('i', M_DIAERESIS, 0),
	LETTER('d', M_NAME, 0), // eth U+00f0
	LETTER('n', M_TILDE, 0),
	LETTER('o', M_GRAVE, 0),
	LETTER('o', M_ACUTE, 0),
	LETTER('o', M_CIRCUMFLEX, 0),
	LETTER('o', M_TILDE, 0),
	LETTER('o', M_DIAERESIS, 0),
	0, // division sign
	LETTER('o', M_STROKE, 0),
	LETTER('u', M_GRAVE, 0),
	LETTER('u', M_ACUTE, 0),
	LETTER('u', M_CIRCUMFLEX, 0),
	LETTER('u', M_DIAERESIS, 0),
	LETTER('y', M_ACUTE, 0),
	LETTER('t', M_NAME, 0), // thorn
	LETTER('y', M_DIAERESIS, 0),
	CAPITAL, // U+0100
	LETTER('a', M_MACRON, 0),
	CAPITAL,
	LETTER('a', M_BREVE, 0),
	CAPITAL,
	LETTER('a', M_OGONEK, 0),
	CAPITAL,
	LETTER('c', M_ACUTE, 0),
	CAPITAL,
	LETTER('c', M_CIRCUMFLEX, 0),
	CAPITAL,
	LETTER('c', M_DOT_ABOVE, 0),
	CAPITAL,
	LETTER('c', M_CARON, 0),
	CAPITAL,
	LETTER('d', M_CARON, 0),
	CAPITAL, // U+0110
	LETTER('d', M_STROKE, 0),
	CAPITAL,
	LETTER('e', M_MACRON, 0),
	CAPITAL,
	LETTER('e', M_BREVE, 0),
	CAPITAL,
	LETTER('e', M_DOT_ABOVE, 0),
	CAPITAL,
	LETTER('e', M_OGONEK, 0),
	CAPITAL,
	LETTER('e', M_CARON, 0),
	CAPITAL,
	LETTER('g', M_CIRCUMFLEX, 0),
	CAPITAL,
	LETTER('g', M_BREVE, 0),
	CAPITAL, // U+0120
	LETTER('g', M_DOT_ABOVE, 0),
	CAPITAL,
	LETTER('g', M_CEDILLA, 0),
	CAPITAL,
	LETTER('h', M_CIRCUMFLEX, 0),
	CAPITAL,
	LETTER('h', M_STROKE, 0),
	CAPITAL,
	LETTER('i', M_TILDE, 0),
	CAPITAL,
	LETTER('i', M_MACRON, 0),
	CAPITAL,
	LETTER('i', M_BREVE, 0),
	CAPITAL,
	LETTER('i', M_OGONEK, 0),
	CAPITAL, // U+0130
	LETTER('i', M_NAME, 0), // dotless i
	CAPITAL,
	LIGATURE('i', 'j', 0),
	CAPITAL,
	LETTER('j', M_CIRCUMFLEX, 0),
	CAPITAL,
	LETTER('k', M_CEDILLA, 0),
	LETTER('k', M_NAME, 0), // kra
	CAPITAL,
	LETTER('l', M_ACUTE, 0),
	CAPITAL,
	LETTER('l', M_CEDILLA, 0),
	CAPITAL,
	LETTER('l', M_CARON, 0),
	CAPITAL,
	LETTER('l', M_MIDDLE_DOT, 0), // U+0140
	CAPITAL,
	LETTER('l', M_STROKE, 0),
	CAPITAL,
	LETTER('n', M_ACUTE, 0),
	CAPITAL,
	LETTER('n', M_CEDILLA, 0),
	CAPITAL,
	LETTER('n', M_CARON, 0),
	LETTER('n', M_NAME, 0), // apostrophe n
	CAPITAL,
	LETTER('n', M_NAME, 0), // eng
	CAPITAL,
	LETTER('o', M_MACRON, 0),
	CAPITAL,
	LETTER('o', M_BREVE, 0),
	CAPITAL, // U+0150
	LETTER('o', M_DOUBLE_ACUTE, 0),
	CAPITAL,
	LIGATURE('o', 'e', 0),
	CAPITAL,
	LETTER('r', M_ACUTE, 0),
	CAPITAL,
	LETTER('r', M_CEDILLA, 0),
	CAPITAL,
	LETTER('r', M_CARON, 0),
	CAPITAL,
	LETTER('s', M_ACUTE, 0),
	CAPITAL,
	LETTER('s', M_CIRCUMFLEX, 0),
	CAPITAL,
	LETTER('s', M_CEDILLA, 0),
	CAPITAL, // U+0160
	LETTER('s', M_CARON, 0),
	CAPITAL,
	LETTER('t', M_CEDILLA, 0),
	CAPITAL,
	LETTER('t', M_CARON, 0),
	CAPITAL,
	LETTER('t', M_STROKE, 0),
	CAPITAL,
	LETTER('u', M_TILDE, 0),
	CAPITAL,
	LETTER('u', M_MACRON, 0),
	CAPITAL,
	LETTER('u', M_BREVE, 0),
	CAPITAL,
	LETTER('u', M_RING, 0),
	CAPITAL, // U+0170
	LETTER('u', M_DOUBLE_ACUTE, 0),
	CAPITAL,
	LETTER('u', M_OGONEK, 0),
	CAPITAL,
	LETTER('w', M_CIRCUMFLEX, 0),
	CAPITAL,
	LETTER('y', M_CIRCUMFLEX, 0),
	CAPITAL, // Y-DIAERESIS
	CAPITAL,
	LETTER('z', M_ACUTE, 0),
	CAPITAL,
	LETTER('z', M_DOT_ABOVE, 0),
	CAPITAL,
	LETTER('z', M_CARON, 0),
	LETTER('s', M_NAME, 0), // long-s U+17f
};

// characters U+0250 to U+029F
static const unsigned short letter_accents_250[] = {
	LETTER('a', M_TURNED, 0), // U+250
	LETTER(L_ALPHA, 0, 0),
	LETTER(L_ALPHA, M_TURNED, 0),
	LETTER('b', M_IMPLOSIVE, 0),
	0, // open-o
	LETTER('c', M_CURL, 0),
	LETTER('d', M_RETROFLEX, 0),
	LETTER('d', M_IMPLOSIVE, 0),
	LETTER('e', M_REVERSED, 0), // U+258
	0, // schwa
	LETTER(L_SCHWA, M_HOOK, 0),
	0, // open-e
	LETTER(L_OPEN_E, M_REVERSED, 0),
	LETTER(L_OPEN_E, M_HOOK, M_REVERSED),
	0,
	LETTER('j', M_BAR, 0),
	LETTER('g', M_IMPLOSIVE, 0), // U+260
	LETTER('g', 0, 0),
	LETTER('g', M_SMALLCAP, 0),
	LETTER(L_GAMMA, 0, 0),
	0, // ramshorn
	LETTER('h', M_TURNED, 0),
	LETTER('h', M_HOOK, 0),
	0,
	LETTER('i', M_BAR, 0), // U+268
	LETTER(L_IOTA, 0, 0),
	LETTER('i', M_SMALLCAP, 0),
	LETTER('l', M_TILDE, 0),
	LETTER('l', M_BAR, 0),
	LETTER('l', M_RETROFLEX, 0),
	LIGATURE('l', 'z', 0),
	LETTER('m', M_TURNED, 0),
	0,
	LETTER('m', M_HOOK, 0),
	0,
	LETTER('n', M_RETROFLEX, 0),
	LETTER('n', M_SMALLCAP, 0),
	LETTER('o', M_BAR, 0),
	LIGATURE('o', 'e', M_SMALLCAP),
	0,
	LETTER(L_PHI, 0, 0), // U+278
	LETTER('r', M_TURNED, 0),
	LETTER(L_RLONG, M_TURNED, 0),
	LETTER('r', M_RETROFLEX, M_TURNED),
	0,
	LETTER('r', M_RETROFLEX, 0),
	0, // r-tap
	LETTER(L_RTAP, M_REVERSED, 0),
	LETTER('r', M_SMALLCAP, 0), // U+280
	LETTER('r', M_TURNED, M_SMALLCAP),
	LETTER('s', M_RETROFLEX, 0),
	0, // esh
	LETTER('j', M_HOOK, 0),
	LETTER(L_ESH, M_REVERSED, 0),
	LETTER(L_ESH, M_CURL, 0),
	LETTER('t', M_TURNED, 0),
	LETTER('t', M_RETROFLEX, 0), // U+288
	LETTER('u', M_BAR, 0),
	LETTER(L_UPSILON, 0, 0),
	LETTER('v', M_HOOK, 0),
	LETTER('v', M_TURNED, 0),
	LETTER('w', M_TURNED, 0),
	LETTER('y', M_TURNED, 0),
	LETTER('y', M_SMALLCAP, 0),
	LETTER('z', M_RETROFLEX, 0), // U+290
	LETTER('z', M_CURL, 0),
	0, // ezh
	LETTER(L_EZH, M_CURL, 0),
	0, // glottal stop
	LETTER(L_GLOTTAL, M_REVERSED, 0),
	LETTER(L_GLOTTAL, M_TURNED, 0),
	0,
	0, // bilabial click U+298
	LETTER('b', M_SMALLCAP, 0),
	0,
	LETTER('g', M_IMPLOSIVE, M_SMALLCAP),
	LETTER('h', M_SMALLCAP, 0),
	LETTER('j', M_CURL, 0),
	LETTER('k', M_TURNED, 0),
	LETTER('l', M_SMALLCAP, 0),
	LETTER('q', M_HOOK, 0), // U+2a0
	LETTER(L_GLOTTAL, M_STROKE, 0),
	LETTER(L_GLOTTAL, M_STROKE, M_REVERSED),
	LIGATURE('d', 'z', 0),
	0, // dezh
	LIGATURE('d', 'z', M_CURL),
	LIGATURE('t', 's', 0),
	0, // tesh
	LIGATURE('t', 's', M_CURL),
};

static int LookupLetter2(Translator *tr, unsigned int letter, char *ph_buf)
{
	int len;
	char single_letter[10];

	single_letter[0] = 0;
	single_letter[1] = '_';
	len = utf8_out(letter, &single_letter[2]);
	single_letter[len+2] = ' ';
	single_letter[len+3] = 0;

	if (Lookup(tr, &single_letter[1], ph_buf) == 0) {
		single_letter[1] = ' ';
		if (Lookup(tr, &single_letter[2], ph_buf) == 0)
			TranslateRules(tr, &single_letter[2], ph_buf, 20, NULL, 0, NULL);
	}
	return ph_buf[0];
}

void LookupAccentedLetter(Translator *tr, unsigned int letter, char *ph_buf)
{
	// lookup the character in the accents table
	int accent_data = 0;
	int accent1 = 0;
	int accent2 = 0;
	int flags1, flags2;
	int basic_letter;
	int letter2 = 0;
	char ph_letter1[30];
	char ph_letter2[30];
	char ph_accent1[30];
	char ph_accent2[30];

	ph_accent2[0] = 0;

	if ((letter >= 0xe0) && (letter < 0x17f))
		accent_data = letter_accents_0e0[letter - 0xe0];
	else if ((letter >= 0x250) && (letter <= 0x2a8))
		accent_data = letter_accents_250[letter - 0x250];

	if (accent_data != 0) {
		basic_letter = (accent_data & 0x3f) + 59;
		if (basic_letter < 'a')
			basic_letter = non_ascii_tab[basic_letter-59];

		if (accent_data & M_LIGATURE) {
			letter2 = (accent_data >> 6) & 0x3f;
			letter2 += 59;
			accent2 = (accent_data >> 12) & 0x7;
		} else {
			accent1 = (accent_data >> 6) & 0x1f;
			accent2 = (accent_data >> 11) & 0xf;
		}

		if ((accent1 == 0) && !(accent_data & M_LIGATURE)) {
			// just a letter name, not an accented character or ligature
			return;
		}

		if ((flags1 = Lookup(tr, accents_tab[accent1].name, ph_accent1)) != 0) {
			if (LookupLetter2(tr, basic_letter, ph_letter1) != 0) {
				if (accent2 != 0) {
					flags2 = Lookup(tr, accents_tab[accent2].name, ph_accent2);
					if (flags2 & FLAG_ACCENT_BEFORE) {
						strcpy(ph_buf, ph_accent2);
						ph_buf += strlen(ph_buf);
						ph_accent2[0] = 0;
					}
				}
				if (letter2 != 0) {
					// ligature
					LookupLetter2(tr, letter2, ph_letter2);
					sprintf(ph_buf, "%s%c%s%c%s%s", ph_accent1, phonPAUSE_VSHORT, ph_letter1, phonSTRESS_P, ph_letter2, ph_accent2);
				} else {
					if (accent1 == 0)
						strcpy(ph_buf, ph_letter1);
					else if ((tr->langopts.accents & 1) || (flags1 & FLAG_ACCENT_BEFORE) || (accents_tab[accent1].accent_flags & 1))
						sprintf(ph_buf, "%s%c%c%s", ph_accent1, phonPAUSE_VSHORT, phonSTRESS_P, ph_letter1);
					else
						sprintf(ph_buf, "%c%s%c%s%c", phonSTRESS_2, ph_letter1, phonPAUSE_VSHORT, ph_accent1, phonPAUSE_VSHORT);
				}
			}
		}
	}
}

void LookupLetter(Translator *tr, unsigned int letter, int next_byte, char *ph_buf1, int control)
{
	// control, bit 0:  not the first letter of a word

	int len;
	char single_letter[10] = { 0, 0 };
	unsigned int dict_flags[2];
	char ph_buf3[40];

	ph_buf1[0] = 0;
	len = utf8_out(letter, &single_letter[2]);
	single_letter[len+2] = ' ';

	if (next_byte == -1) {
		// speaking normal text, not individual characters
		if (Lookup(tr, &single_letter[2], ph_buf1) != 0)
			return;

		single_letter[1] = '_';
		if (Lookup(tr, &single_letter[1], ph_buf3) != 0)
			return; // the character is specified as _* so ignore it when speaking normal text

		// check whether this character is specified for English
		if (tr->translator_name == L('e', 'n'))
			return; // we are already using English

		SetTranslator3(ESPEAKNG_DEFAULT_VOICE);
		if (Lookup(translator3, &single_letter[2], ph_buf3) != 0) {
			// yes, switch to English and re-translate the word
			sprintf(ph_buf1, "%c", phonSWITCH);
		}
		SelectPhonemeTable(voice->phoneme_tab_ix); // revert to original phoneme table
		return;
	}

	if ((letter <= 32) || iswspace(letter)) {
		// lookup space as _&32 etc.
		sprintf(&single_letter[1], "_#%d ", letter);
		Lookup(tr, &single_letter[1], ph_buf1);
		return;
	}

	if (next_byte != ' ')
		next_byte = RULE_SPELLING;
	single_letter[3+len] = next_byte; // follow by space-space if the end of the word, or space-31

	single_letter[1] = '_';

	// if the $accent flag is set for this letter, use the accents table (below)
	dict_flags[1] = 0;

	if (Lookup(tr, &single_letter[1], ph_buf3) == 0) {
		single_letter[1] = ' ';
		if (Lookup(tr, &single_letter[2], ph_buf3) == 0)
			TranslateRules(tr, &single_letter[2], ph_buf3, sizeof(ph_buf3), NULL, FLAG_NO_TRACE, NULL);
	}

	if (ph_buf3[0] == 0)
		LookupAccentedLetter(tr, letter, ph_buf3);

	strcpy(ph_buf1, ph_buf3);
	if ((ph_buf1[0] == 0) || (ph_buf1[0] == phonSWITCH))
		return;

	dict_flags[0] = 0;
	dict_flags[1] = 0;
	SetWordStress(tr, ph_buf1, dict_flags, -1, control & 1);
}


#define L_SUB 0x4000 // subscript
#define L_SUP 0x8000 // superscript



// this list must be in ascending order
static unsigned short derived_letters[] = {
	0x00aa, 'a'+L_SUP,
	0x00b2, '2'+L_SUP,
	0x00b3, '3'+L_SUP,
	0x00b9, '1'+L_SUP,
	0x00ba, 'o'+L_SUP,
	0x02b0, 'h'+L_SUP,
	0x02b1, 0x266+L_SUP,
	0x02b2, 'j'+L_SUP,
	0x02b3, 'r'+L_SUP,
	0x02b4, 0x279+L_SUP,
	0x02b5, 0x27b+L_SUP,
	0x02b6, 0x281+L_SUP,
	0x02b7, 'w'+L_SUP,
	0x02b8, 'y'+L_SUP,
	0x02c0, 0x294+L_SUP,
	0x02c1, 0x295+L_SUP,
	0x02e0, 0x263+L_SUP,
	0x02e1, 'l'+L_SUP,
	0x02e2, 's'+L_SUP,
	0x02e3, 'x'+L_SUP,
	0x2070, '0'+L_SUP,
	0x2071, 'i'+L_SUP,
	0x2074, '4'+L_SUP,
	0x2075, '5'+L_SUP,
	0x2076, '6'+L_SUP,
	0x2077, '7'+L_SUP,
	0x2078, '8'+L_SUP,
	0x2079, '9'+L_SUP,
	0x207a, '+'+L_SUP,
	0x207b, '-'+L_SUP,
	0x207c, '='+L_SUP,
	0x207d, '('+L_SUP,
	0x207e, ')'+L_SUP,
	0x207f, 'n'+L_SUP,
	0x2080, '0'+L_SUB,
	0x2081, '1'+L_SUB,
	0x2082, '2'+L_SUB,
	0x2083, '3'+L_SUB,
	0x2084, '4'+L_SUB,
	0x2085, '5'+L_SUB,
	0x2086, '6'+L_SUB,
	0x2087, '7'+L_SUB,
	0x2088, '8'+L_SUB,
	0x2089, '9'+L_SUB,
	0x208a, '+'+L_SUB,
	0x208b, '-'+L_SUB,
	0x208c, '='+L_SUB,
	0x208d, '('+L_SUB,
	0x208e, ')'+L_SUB,
	0x2090, 'a'+L_SUB,
	0x2091, 'e'+L_SUB,
	0x2092, 'o'+L_SUB,
	0x2093, 'x'+L_SUB,
	0x2094, 0x259+L_SUB,
	0x2095, 'h'+L_SUB,
	0x2096, 'k'+L_SUB,
	0x2097, 'l'+L_SUB,
	0x2098, 'm'+L_SUB,
	0x2099, 'n'+L_SUB,
	0x209a, 'p'+L_SUB,
	0x209b, 's'+L_SUB,
	0x209c, 't'+L_SUB,
	0, 0
};



int IsSuperscript(int letter)
{
	// is this a subscript or superscript letter ?
	int ix;
	int c;

	for (ix = 0; (c = derived_letters[ix]) != 0; ix += 2) {
		if (c > letter)
			break;
		if (c == letter)
			return derived_letters[ix+1];
	}
	return 0;
}

void SetSpellingStress(Translator *tr, char *phonemes, int control, int n_chars)
{
	// Individual letter names, reduce the stress of some.
	int ix;
	unsigned int c;
	int n_stress = 0;
	int prev = 0;
	int count;
	unsigned char buf[N_WORD_PHONEMES];

	for (ix = 0; (c = phonemes[ix]) != 0; ix++) {
		if ((c == phonSTRESS_P) && (prev != phonSWITCH))
			n_stress++;
		buf[ix] = prev = c;
	}
	buf[ix] = 0;

	count = 0;
	prev = 0;
	for (ix = 0; (c = buf[ix]) != 0; ix++) {
		if ((c == phonSTRESS_P) && (n_chars > 1) && (prev != phonSWITCH)) {
			count++;

			if (tr->langopts.spelling_stress == 1) {
				// stress on initial letter when spelling
				if (count > 1)
					c = phonSTRESS_3;
			} else {
				if (count != n_stress) {
					if (((count % 3) != 0) || (count == n_stress-1))
						c = phonSTRESS_3; // reduce to secondary stress
				}
			}
		} else if (c == 0xff) {
			if ((control < 2) || (ix == 0))
				continue; // don't insert pauses

			if (((count % 3) == 0) || (control > 2))
				c = phonPAUSE_NOLINK; // pause following a primary stress
			else
				c = phonPAUSE_VSHORT;
		}
		*phonemes++ = prev = c;
	}
	if (control >= 2)
		*phonemes++ = phonPAUSE_NOLINK;
	*phonemes = 0;
}

// Numbers

static char ph_ordinal2[12];
static char ph_ordinal2x[12];

static int CheckDotOrdinal(Translator *tr, char *word, char *word_end, WORD_TAB *wtab, int roman)
{
	int ordinal = 0;
	int c2;
	int nextflags;

	if ((tr->langopts.numbers & NUM_ORDINAL_DOT) && ((word_end[0] == '.') || (wtab[0].flags & FLAG_HAS_DOT)) && !(wtab[1].flags & FLAG_NOSPACE)) {
		if (roman || !(wtab[1].flags & FLAG_FIRST_UPPER)) {
			if (word_end[0] == '.')
				utf8_in(&c2, &word_end[2]);
			else
				utf8_in(&c2, &word_end[0]);

			if ((word_end[0] != 0) && (word_end[1] != 0) && ((c2 == 0) || (wtab[0].flags & FLAG_COMMA_AFTER) || IsAlpha(c2))) {
				// ordinal number is indicated by dot after the number
				// but not if the next word starts with an upper-case letter
				// (c2 == 0) is for cases such as, "2.,"
				ordinal = 2;
				if (word_end[0] == '.')
					word_end[0] = ' ';

				if ((roman == 0) && (tr->translator_name == L('h', 'u'))) {
					// lang=hu don't treat dot as ordinal indicator if the next word is a month name ($alt). It may have a suffix.
					nextflags = 0;
					if (IsAlpha(c2))
						nextflags = TranslateWord(tr, &word_end[2], NULL, NULL);

					if ((tr->prev_dict_flags[0] & FLAG_ALT_TRANS) && ((c2 == 0) || (wtab[0].flags & FLAG_COMMA_AFTER) || iswdigit(c2)))
						ordinal = 0; // TEST  09.02.10

					if (nextflags & FLAG_ALT_TRANS)
						ordinal = 0;

					if (nextflags & FLAG_ALT3_TRANS) {
						if (word[-2] == '-')
							ordinal = 0; // e.g. december 2-5. között

						if (tr->prev_dict_flags[0] & (FLAG_ALT_TRANS | FLAG_ALT3_TRANS))
							ordinal = 0x22;
					}
				}
			}
		}
	}
	return ordinal;
}

static int hu_number_e(const char *word, int thousandplex, int value)
{
	// lang-hu: variant form of numbers when followed by hyphen and a suffix starting with 'a' or 'e' (but not a, e, az, ez, azt, ezt, att. ett

	if ((word[0] == 'a') || (word[0] == 'e')) {
		if ((word[1] == ' ') || (word[1] == 'z') || ((word[1] == 't') && (word[2] == 't')))
			return 0;

		if (((thousandplex == 1) || ((value % 1000) == 0)) && (word[1] == 'l'))
			return 0; // 1000-el

		return 1;
	}
	return 0;
}

int TranslateRoman(Translator *tr, char *word, char *ph_out, WORD_TAB *wtab)
{
	int c;
	char *p;
	const char *p2;
	int acc;
	int prev;
	int value;
	int subtract;
	int repeat = 0;
	int n_digits = 0;
	char *word_start;
	int num_control = 0;
	unsigned int flags[2];
	char ph_roman[30];
	char number_chars[N_WORD_BYTES];

	static const char *roman_numbers = "ixcmvld";
	static const int roman_values[] = { 1, 10, 100, 1000, 5, 50, 500 };

	acc = 0;
	prev = 0;
	subtract = 0x7fff;
	ph_out[0] = 0;
	flags[0] = 0;
	flags[1] = 0;

	if (((tr->langopts.numbers & NUM_ROMAN_CAPITALS) && !(wtab[0].flags & FLAG_ALL_UPPER)) || IsDigit09(word[-2]))
		return 0; // not '2xx'

	if (word[1] == ' ') {
		if ((tr->langopts.numbers & (NUM_ROMAN_CAPITALS | NUM_ROMAN_ORDINAL | NUM_ORDINAL_DOT)) && (wtab[0].flags & FLAG_HAS_DOT)) {
			// allow single letter Roman ordinal followed by dot.
		} else
			return 0; // only one letter, don't speak as a Roman Number
	}

	word_start = word;
	while ((c = *word++) != ' ') {
		if ((p2 = strchr(roman_numbers, c)) == NULL)
			return 0;

		value = roman_values[p2 - roman_numbers];
		if (value == prev) {
			repeat++;
			if (repeat >= 3)
				return 0;
		} else
			repeat = 0;

		if ((prev > 1) && (prev != 10) && (prev != 100)) {
			if (value >= prev)
				return 0;
		}
		if ((prev != 0) && (prev < value)) {
			if (((acc % 10) != 0) || ((prev*10) < value))
				return 0;
			subtract = prev;
			value -= subtract;
		} else if (value >= subtract)
			return 0;
		else
			acc += prev;
		prev = value;
		n_digits++;
	}

	if (IsDigit09(word[0]))
		return 0; // e.g. 'xx2'

	acc += prev;
	if (acc < tr->langopts.min_roman)
		return 0;

	if (acc > tr->langopts.max_roman)
		return 0;

	Lookup(tr, "_roman", ph_roman); // precede by "roman" if _rom is defined in *_list
	p = &ph_out[0];

	if ((tr->langopts.numbers & NUM_ROMAN_AFTER) == 0) {
		strcpy(ph_out, ph_roman);
		p = &ph_out[strlen(ph_roman)];
	}

	sprintf(number_chars, "  %d %s    ", acc, tr->langopts.roman_suffix);

	if (word[0] == '.') {
		// dot has not been removed.  This implies that there was no space after it
		return 0;
	}

	if (CheckDotOrdinal(tr, word_start, word, wtab, 1))
		wtab[0].flags |= FLAG_ORDINAL;

	if (tr->langopts.numbers & NUM_ROMAN_ORDINAL) {
		if (tr->translator_name == L('h', 'u')) {
			if (!(wtab[0].flags & FLAG_ORDINAL)) {
				if ((wtab[0].flags & FLAG_HYPHEN_AFTER) && hu_number_e(word, 0, acc)) {
					// should use the 'e' form of the number
					num_control |= 1;
				} else
					return 0;
			}
		} else
			wtab[0].flags |= FLAG_ORDINAL;
	}

	tr->prev_dict_flags[0] = 0;
	tr->prev_dict_flags[1] = 0;
	TranslateNumber(tr, &number_chars[2], p, flags, wtab, num_control);

	if (tr->langopts.numbers & NUM_ROMAN_AFTER)
		strcat(ph_out, ph_roman);

	return 1;
}

static const char *M_Variant(int value)
{
	// returns M, or perhaps MA or MB for some cases

	bool teens = false;

	if (((value % 100) > 10) && ((value % 100) < 20))
		teens = true;

	switch (translator->langopts.numbers2 & NUM2_THOUSANDS_VAR_BITS)
	{
	case NUM2_THOUSANDS_VAR1: // lang=ru  use singular for xx1 except for x11
		if ((teens == false) && ((value % 10) == 1))
			return "1M";
		break;
	case NUM2_THOUSANDS_VAR2: // lang=cs,sk
		if ((value >= 2) && (value <= 4))
			return "0MA";
		break;
	case NUM2_THOUSANDS_VAR3: // lang=pl
		if ((teens == false) && (((value % 10) >= 2) && ((value % 10) <= 4)))
			return "0MA";
		break;
	case NUM2_THOUSANDS_VAR4: // lang=lt
		if ((teens == true) || ((value % 10) == 0))
			return "0MB";
		if ((value % 10) == 1)
			return "0MA";
		break;
	case NUM2_THOUSANDS_VAR5: // lang=bs,hr,sr
		if (teens == false) {
			if ((value % 10) == 1)
				return "1M";
			if (((value % 10) >= 2) && ((value % 10) <= 4))
				return "0MA";
		}
		break;
	}
	return "0M";
}

static int LookupThousands(Translator *tr, int value, int thousandplex, int thousands_exact, char *ph_out)
{
	// thousands_exact:  bit 0  no hundreds,tens,or units,  bit 1  ordinal numberr
	int found;
	int found_value = 0;
	char string[14];
	char ph_of[12];
	char ph_thousands[40];
	char ph_buf[40];

	ph_of[0] = 0;

	// first look for a match with the exact value of thousands
	if (value > 0) {
		if (thousands_exact & 1) {
			if (thousands_exact & 2) {
				// ordinal number
				sprintf(string, "_%dM%do", value, thousandplex);
				found_value = Lookup(tr, string, ph_thousands);
			}
			if (!found_value && (number_control & 1)) {
				// look for the 'e' variant
				sprintf(string, "_%dM%de", value, thousandplex);
				found_value = Lookup(tr, string, ph_thousands);
			}
			if (!found_value) {
				// is there a different pronunciation if there are no hundreds,tens,or units ? (LANG=ta)
				sprintf(string, "_%dM%dx", value, thousandplex);
				found_value = Lookup(tr, string, ph_thousands);
			}
		}
		if (found_value == 0) {
			sprintf(string, "_%dM%d", value, thousandplex);
			found_value = Lookup(tr, string, ph_thousands);
		}
	}

	if (found_value == 0) {
		if ((value % 100) >= 20)
			Lookup(tr, "_0of", ph_of);

		found = 0;
		if (thousands_exact & 1) {
			if (thousands_exact & 2) {
				// ordinal number
				sprintf(string, "_%s%do", M_Variant(value), thousandplex);
				found = Lookup(tr, string, ph_thousands);
			}
			if (!found && (number_control & 1)) {
				// look for the 'e' variant
				sprintf(string, "_%s%de", M_Variant(value), thousandplex);
				found = Lookup(tr, string, ph_thousands);
			}
			if (!found) {
				// is there a different pronunciation if there are no hundreds,tens,or units ?
				sprintf(string, "_%s%dx", M_Variant(value), thousandplex);
				found = Lookup(tr, string, ph_thousands);
			}
		}
		if (found == 0) {
			sprintf(string, "_%s%d", M_Variant(value), thousandplex);

			if (Lookup(tr, string, ph_thousands) == 0) {
				if (thousandplex > 3) {
					sprintf(string, "_0M%d", thousandplex-1);
					if (Lookup(tr, string, ph_buf) == 0) {
						// say "millions" if this name is not available and neither is the next lower
						Lookup(tr, "_0M2", ph_thousands);
						speak_missing_thousands = 3;
					}
				}
				if (ph_thousands[0] == 0) {
					// repeat "thousand" if higher order names are not available
					sprintf(string, "_%dM1", value);
					if ((found_value = Lookup(tr, string, ph_thousands)) == 0)
						Lookup(tr, "_0M1", ph_thousands);
					speak_missing_thousands = 2;
				}
			}
		}
	}
	sprintf(ph_out, "%s%s", ph_of, ph_thousands);

	if ((value == 1) && (thousandplex == 1) && (tr->langopts.numbers & NUM_OMIT_1_THOUSAND))
		return 1;

	return found_value;
}

static int LookupNum2(Translator *tr, int value, int thousandplex, const int control, char *ph_out)
{
	// Lookup a 2 digit number
	// control bit 0: ordinal number
	// control bit 1: final tens and units (not number of thousands) (use special form of '1', LANG=de "eins")
	// control bit 2: tens and units only, no higher digits
	// control bit 3: use feminine form of '2' (for thousands
	// control bit 4: speak zero tens
	// control bit 5: variant of ordinal number (lang=hu)
	//         bit 8   followed by decimal fraction
	//         bit 9: use #f form for both tens and units (lang=ml)

	int found;
	int ix;
	int units;
	int tens;
	int is_ordinal;
	int used_and = 0;
	int found_ordinal = 0;
	int next_phtype;
	int ord_type = 'o';
	char string[12]; // for looking up entries in *_list
	char ph_ordinal[20];
	char ph_tens[50];
	char ph_digits[50];
	char ph_and[12];

	units = value % 10;
	tens = value / 10;

	found = 0;
	ph_ordinal[0] = 0;
	ph_tens[0] = 0;
	ph_digits[0] = 0;
	ph_and[0] = 0;

	if (control & 0x20)
		ord_type = 'q';

	is_ordinal = control & 1;

	if ((control & 2) && (n_digit_lookup == 2)) {
		// pronunciation of the final 2 digits has already been found
		strcpy(ph_out, digit_lookup);
	} else {
		if (digit_lookup[0] == 0) {
			// is there a special pronunciation for this 2-digit number
			if (control & 8) {
				// is there a feminine or thousands-variant form?
				sprintf(string, "_%dfx", value);
				if ((found = Lookup(tr, string, ph_digits)) == 0) {
					sprintf(string, "_%df", value);
					found = Lookup(tr, string, ph_digits);
				}
			} else if (is_ordinal) {
				strcpy(ph_ordinal, ph_ordinal2);

				if (control & 4) {
					sprintf(string, "_%d%cx", value, ord_type); // LANG=hu, special word for 1. 2. when there are no higher digits
					if ((found = Lookup(tr, string, ph_digits)) != 0) {
						if (ph_ordinal2x[0] != 0)
							strcpy(ph_ordinal, ph_ordinal2x); // alternate pronunciation (lang=an)
					}
				}
				if (found == 0) {
					sprintf(string, "_%d%c", value, ord_type);
					found = Lookup(tr, string, ph_digits);
				}
				found_ordinal = found;
			}

			if (found == 0) {
				if (control & 2) {
					// the final tens and units of a number
					if (number_control & 1) {
						// look for 'e' variant
						sprintf(string, "_%de", value);
						found = Lookup(tr, string, ph_digits);
					}
				} else {
					// followed by hundreds or thousands etc
					if ((tr->langopts.numbers2 & NUM2_ORDINAL_AND_THOUSANDS) && (thousandplex <= 1))
						sprintf(string, "_%do", value); // LANG=TA
					else
						sprintf(string, "_%da", value);
					found = Lookup(tr, string, ph_digits);
				}

				if (!found) {
					if ((is_ordinal) && (tr->langopts.numbers2 & NUM2_NO_TEEN_ORDINALS)) {
						// don't use numbers 10-99 to make ordinals, always use _1Xo etc (lang=pt)
					} else {
						sprintf(string, "_%d", value);
						found = Lookup(tr, string, ph_digits);
					}
				}
			}
		}

		// no, speak as tens+units

		if ((value < 10) && (control & 0x10)) {
			// speak leading zero
			Lookup(tr, "_0", ph_tens);
		} else {
			if (found)
				ph_tens[0] = 0;
			else {
				if (is_ordinal) {
					sprintf(string, "_%dX%c", tens, ord_type);
					if (Lookup(tr, string, ph_tens) != 0) {
						found_ordinal = 1;

						if ((units != 0) && (tr->langopts.numbers2 & NUM2_MULTIPLE_ORDINAL)) {
							// Use the ordinal form of tens as well as units. Add the ordinal ending
							strcat(ph_tens, ph_ordinal2);
						}
					}
				}
				if (found_ordinal == 0) {
					if (control & 0x200)
						sprintf(string, "_%dXf", tens);
					else
						sprintf(string, "_%dX", tens);
					Lookup(tr, string, ph_tens);
				}

				if ((ph_tens[0] == 0) && (tr->langopts.numbers & NUM_VIGESIMAL)) {
					// tens not found,  (for example) 73 is 60+13
					units = (value % 20);
					sprintf(string, "_%dX", tens & 0xfe);
					Lookup(tr, string, ph_tens);
				}

				ph_digits[0] = 0;
				if (units > 0) {
					found = 0;

					if ((control & 2) && (digit_lookup[0] != 0)) {
						// we have an entry for this digit (possibly together with the next word)
						strcpy(ph_digits, digit_lookup);
						found_ordinal = 1;
						ph_ordinal[0] = 0;
					} else {
						if (control & 8) {
							// is there a variant form of this number?
							sprintf(string, "_%df", units);
							found = Lookup(tr, string, ph_digits);
						}
						if ((is_ordinal) && ((tr->langopts.numbers & NUM_SWAP_TENS) == 0)) {
							// ordinal
							sprintf(string, "_%d%c", units, ord_type);
							if ((found = Lookup(tr, string, ph_digits)) != 0)
								found_ordinal = 1;
						}
						if (found == 0) {
							if ((number_control & 1) && (control & 2)) {
								// look for 'e' variant
								sprintf(string, "_%de", units);
								found = Lookup(tr, string, ph_digits);
							} else if (((control & 2) == 0) || ((tr->langopts.numbers & NUM_SWAP_TENS) != 0)) {
								// followed by hundreds or thousands (or tens)
								if ((tr->langopts.numbers2 & NUM2_ORDINAL_AND_THOUSANDS) && (thousandplex <= 1))
									sprintf(string, "_%do", units);  // LANG=TA,  only for 100s, 1000s
								else
									sprintf(string, "_%da", units);
								found = Lookup(tr, string, ph_digits);
							}
						}
						if (found == 0) {
							sprintf(string, "_%d", units);
							Lookup(tr, string, ph_digits);
						}
					}
				}
			}
		}

		if ((is_ordinal) && (found_ordinal == 0) && (ph_ordinal[0] == 0)) {
			if ((value >= 20) && (((value % 10) == 0) || (tr->langopts.numbers & NUM_SWAP_TENS)))
				Lookup(tr, "_ord20", ph_ordinal);
			if (ph_ordinal[0] == 0)
				Lookup(tr, "_ord", ph_ordinal);
		}

		if ((tr->langopts.numbers & (NUM_SWAP_TENS | NUM_AND_UNITS)) && (ph_tens[0] != 0) && (ph_digits[0] != 0)) {
			Lookup(tr, "_0and", ph_and);

			if ((is_ordinal) && (tr->langopts.numbers2 & NUM2_ORDINAL_NO_AND))
				ph_and[0] = 0;

			if (tr->langopts.numbers & NUM_SWAP_TENS)
				sprintf(ph_out, "%s%s%s%s", ph_digits, ph_and, ph_tens, ph_ordinal);
			else
				sprintf(ph_out, "%s%s%s%s", ph_tens, ph_and, ph_digits, ph_ordinal);
			used_and = 1;
		} else {
			if (tr->langopts.numbers & NUM_SINGLE_VOWEL) {
				// remove vowel from the end of tens if units starts with a vowel (LANG=Italian)
				if (((ix = strlen(ph_tens)-1) >= 0) && (ph_digits[0] != 0)) {
					if ((next_phtype = phoneme_tab[(unsigned int)(ph_digits[0])]->type) == phSTRESS)
						next_phtype = phoneme_tab[(unsigned int)(ph_digits[1])]->type;

					if ((phoneme_tab[(unsigned int)(ph_tens[ix])]->type == phVOWEL) && (next_phtype == phVOWEL))
						ph_tens[ix] = 0;
				}
			}

			if ((tr->langopts.numbers2 & NUM2_ORDINAL_DROP_VOWEL) && (ph_ordinal[0] != 0)) {
				ix = sprintf(ph_out, "%s%s", ph_tens, ph_digits);
				if ((ix > 0) && (phoneme_tab[(unsigned char)(ph_out[ix-1])]->type == phVOWEL))
					ix--;
				sprintf(&ph_out[ix], "%s", ph_ordinal);
			} else
				sprintf(ph_out, "%s%s%s", ph_tens, ph_digits, ph_ordinal);
		}
	}

	if (tr->langopts.numbers & NUM_SINGLE_STRESS_L) {
		// only one primary stress, on the first part (tens)
		found = 0;
		for (ix = 0; ix < (signed)strlen(ph_out); ix++) {
			if (ph_out[ix] == phonSTRESS_P) {
				if (found)
					ph_out[ix] = phonSTRESS_3;
				else
					found = 1;
			}
		}
	} else if (tr->langopts.numbers & NUM_SINGLE_STRESS) {
		// only one primary stress
		found = 0;
		for (ix = strlen(ph_out)-1; ix >= 0; ix--) {
			if (ph_out[ix] == phonSTRESS_P) {
				if (found)
					ph_out[ix] = phonSTRESS_3;
				else
					found = 1;
			}
		}
	}
	return used_and;
}

static int LookupNum3(Translator *tr, int value, char *ph_out, bool suppress_null, int thousandplex, int control)
{
	// Translate a 3 digit number
	//  control  bit 0,  previous thousands
	//           bit 1,  ordinal number
	//           bit 5   variant form of ordinal number
	//           bit 8   followed by decimal fraction

	int found;
	int hundreds;
	int tensunits;
	int x;
	int ix;
	int exact;
	int ordinal;
	int tplex;
	bool say_zero_hundred = false;
	bool say_one_hundred;
	char string[12]; // for looking up entries in **_list
	char buf1[100];
	char buf2[100];
	char ph_100[20];
	char ph_10T[20];
	char ph_digits[50];
	char ph_thousands[50];
	char ph_hundred_and[12];
	char ph_thousand_and[12];

	ordinal = control & 0x22;
	hundreds = value / 100;
	tensunits = value % 100;
	buf1[0] = 0;

	ph_thousands[0] = 0;
	ph_thousand_and[0] = 0;

	if ((tr->langopts.numbers & NUM_ZERO_HUNDRED) && ((control & 1) || (hundreds >= 10)))
		say_zero_hundred = true; // lang=vi

	if ((hundreds > 0) || say_zero_hundred) {
		found = 0;
		if (ordinal && (tensunits == 0)) {
			// ordinal number, with no tens or units
			found = Lookup(tr, "_0Co", ph_100);
		}
		if (found == 0) {
			if (tensunits == 0) {
				// special form for exact hundreds?
				found = Lookup(tr, "_0C0", ph_100);
			}
			if (!found)
				Lookup(tr, "_0C", ph_100);
		}

		if (((tr->langopts.numbers & NUM_1900) != 0) && (hundreds == 19)) {
			// speak numbers such as 1984 as years: nineteen-eighty-four
		} else if (hundreds >= 10) {
			ph_digits[0] = 0;

			exact = 0;
			if ((value % 1000) == 0)
				exact = 1;

			tplex = thousandplex+1;
			if (tr->langopts.numbers2 & NUM2_MYRIADS)
				tplex = 0;

			if (LookupThousands(tr, hundreds / 10, tplex, exact | ordinal, ph_10T) == 0) {
				x = 0;
				if (tr->langopts.numbers2 & (1 << tplex))
					x = 8; // use variant (feminine) for before thousands and millions
				if (tr->translator_name == L('m', 'l'))
					x = 0x208;
				LookupNum2(tr, hundreds/10, thousandplex, x, ph_digits);
			}

			if (tr->langopts.numbers2 & NUM2_SWAP_THOUSANDS)
				sprintf(ph_thousands, "%s%c%s%c", ph_10T, phonEND_WORD, ph_digits, phonEND_WORD);
			else
				sprintf(ph_thousands, "%s%c%s%c", ph_digits, phonEND_WORD, ph_10T, phonEND_WORD);

			hundreds %= 10;
			if ((hundreds == 0) && (say_zero_hundred == false))
				ph_100[0] = 0;
			suppress_null = true;
			control |= 1;
		}

		ph_digits[0] = 0;

		if ((hundreds > 0) || say_zero_hundred) {
			if ((tr->langopts.numbers & NUM_AND_HUNDRED) && ((control & 1) || (ph_thousands[0] != 0)))
				Lookup(tr, "_0and", ph_thousand_and);

			suppress_null = true;

			found = 0;
			if ((ordinal)
			    && ((tensunits == 0) || (tr->langopts.numbers2 & NUM2_MULTIPLE_ORDINAL))) {
				// ordinal number
				sprintf(string, "_%dCo", hundreds);
				found = Lookup(tr, string, ph_digits);

				if ((tr->langopts.numbers2 & NUM2_MULTIPLE_ORDINAL) && (tensunits > 0)) {
					// Use ordinal form of hundreds, as well as for tens and units
					// Add ordinal suffix to the hundreds
					strcat(ph_digits, ph_ordinal2);
				}
			}

			if ((hundreds == 0) && say_zero_hundred)
				Lookup(tr, "_0", ph_digits);
			else {
				if ((hundreds == 1) && (tr->langopts.numbers2 & NUM2_OMIT_1_HUNDRED_ONLY) && ((control & 1) == 0)) {
					// only look for special 100 if there are previous thousands
				} else {
					if ((!found) && (tensunits == 0)) {
						// is there a special pronunciation for exactly n00 ?
						sprintf(string, "_%dC0", hundreds);
						found = Lookup(tr, string, ph_digits);
					}

					if (!found) {
						sprintf(string, "_%dC", hundreds);
						found = Lookup(tr, string, ph_digits);  // is there a specific pronunciation for n-hundred ?
					}
				}

				if (found)
					ph_100[0] = 0;
				else {
					say_one_hundred = true;
					if (hundreds == 1) {
						if ((tr->langopts.numbers & NUM_OMIT_1_HUNDRED) != 0)
							say_one_hundred = false;
					}

					if (say_one_hundred == true)
						LookupNum2(tr, hundreds, thousandplex, 0, ph_digits);
				}
			}
		}

		sprintf(buf1, "%s%s%s%s", ph_thousands, ph_thousand_and, ph_digits, ph_100);
	}

	ph_hundred_and[0] = 0;
	if (tensunits > 0) {
		if ((control & 2) && (tr->langopts.numbers2 & NUM2_MULTIPLE_ORDINAL)) {
			// Don't use "and" if we apply ordinal to both hundreds and units
		} else {
			if ((value > 100) || ((control & 1) && (thousandplex == 0))) {
				if ((tr->langopts.numbers & NUM_HUNDRED_AND) || ((tr->langopts.numbers & NUM_HUNDRED_AND_DIGIT) && (tensunits < 10)))
					Lookup(tr, "_0and", ph_hundred_and);
			}
			if ((tr->langopts.numbers & NUM_THOUSAND_AND) && (hundreds == 0) && ((control & 1) || (ph_thousands[0] != 0)))
				Lookup(tr, "_0and", ph_hundred_and);
		}
	}

	buf2[0] = 0;

	if ((tensunits != 0) || (suppress_null == false)) {
		x = 0;
		if (thousandplex == 0) {
			x = 2; // allow "eins" for 1 rather than "ein"
			if (ordinal)
				x = 3; // ordinal number
			if ((value < 100) && !(control & 1))
				x |= 4; // tens and units only, no higher digits
			if (ordinal & 0x20)
				x |= 0x20; // variant form of ordinal number
		} else if (tr->langopts.numbers2 & (1 << thousandplex))
			x = 8; // use variant (feminine) for before thousands and millions

		if ((tr->translator_name == L('m', 'l')) && (thousandplex == 1))
			x |= 0x208; // use #f form for both tens and units

		if ((tr->langopts.numbers2 & NUM2_ZERO_TENS) && ((control & 1) || (hundreds > 0))) {
			// LANG=zh,
			x |= 0x10;
		}

		if (LookupNum2(tr, tensunits, thousandplex, x | (control & 0x100), buf2) != 0) {
			if (tr->langopts.numbers & NUM_SINGLE_AND)
				ph_hundred_and[0] = 0; // don't put 'and' after 'hundred' if there's 'and' between tens and units
		}
	} else {
		if (ph_ordinal2[0] != 0) {
			ix = strlen(buf1);
			if ((ix > 0) && (buf1[ix-1] == phonPAUSE_SHORT))
				buf1[ix-1] = 0; // remove pause before adding ordinal suffix
			strcpy(buf2, ph_ordinal2);
		}
	}

	sprintf(ph_out, "%s%s%c%s", buf1, ph_hundred_and, phonEND_WORD, buf2);

	return 0;
}

static bool CheckThousandsGroup(char *word, int group_len)
{
	// Is this a group of 3 digits which looks like a thousands group?
	int ix;

	for (ix = 0; ix < group_len; ix++) {
		if (!IsDigit09(word[ix]))
			return false;
	}

	if (IsDigit09(word[group_len]) || IsDigit09(word[-1]))
		return false;

	return true;
}

static int TranslateNumber_1(Translator *tr, char *word, char *ph_out, unsigned int *flags, WORD_TAB *wtab, int control)
{
	//  Number translation with various options
	// the "word" may be up to 4 digits
	// "words" of 3 digits may be preceded by another number "word" for thousands or millions

	int n_digits;
	int value;
	int ix;
	int digix;
	unsigned char c;
	bool suppress_null = false;
	int decimal_point = 0;
	int thousandplex = 0;
	int thousands_exact = 1;
	int thousands_inc = 0;
	int prev_thousands = 0;
	int ordinal = 0;
	int this_value;
	int decimal_count;
	int max_decimal_count;
	int decimal_mode;
	int suffix_ix;
	int skipwords = 0;
	int group_len;
	int len;
	char *p;
	char string[32]; // for looking up entries in **_list
	char buf1[100];
	char ph_append[50];
	char ph_buf[200];
	char ph_buf2[50];
	char ph_zeros[50];
	char suffix[30]; // string[] must be long enough for sizeof(suffix)+2
	char buf_digit_lookup[50];

	static const char str_pause[2] = { phonPAUSE_NOLINK, 0 };

	*flags = 0;
	n_digit_lookup = 0;
	buf_digit_lookup[0] = 0;
	digit_lookup = buf_digit_lookup;
	number_control = control;

	for (ix = 0; IsDigit09(word[ix]); ix++) ;
	n_digits = ix;
	value = this_value = atoi(word);

	group_len = 3;
	if (tr->langopts.numbers2 & NUM2_MYRIADS)
		group_len = 4;

	// is there a previous thousands part (as a previous "word") ?
	if ((n_digits == group_len) && (word[-2] == tr->langopts.thousands_sep) && IsDigit09(word[-3]))
		prev_thousands = 1;
	else if ((tr->langopts.thousands_sep == ' ') || (tr->langopts.numbers & NUM_ALLOW_SPACE)) {
		// thousands groups can be separated by spaces
		if ((n_digits == 3) && !(wtab->flags & FLAG_MULTIPLE_SPACES) && IsDigit09(word[-2]))
			prev_thousands = 1;
	}
	if (prev_thousands == 0)
		speak_missing_thousands = 0;

	ph_ordinal2[0] = 0;
	ph_zeros[0] = 0;

	if (prev_thousands || (word[0] != '0')) {
		// don't check for ordinal if the number has a leading zero
		ordinal = CheckDotOrdinal(tr, word, &word[ix], wtab, 0);
	}

	if ((word[ix] == '.') && !IsDigit09(word[ix+1]) && !IsDigit09(word[ix+2]) && !(wtab[1].flags & FLAG_NOSPACE)) {
		// remove dot unless followed by another number
		word[ix] = 0;
	}

	if ((ordinal == 0) || (tr->translator_name == L('h', 'u'))) {
		// NOTE lang=hu, allow both dot and ordinal suffix, eg. "december 21.-én"
		// look for an ordinal number suffix after the number
		ix++;
		p = suffix;
		if (wtab[0].flags & FLAG_HYPHEN_AFTER) {
			*p++ = '-';
			ix++;
		}
		while ((word[ix] != 0) && (word[ix] != ' ') && (ix < (int)(sizeof(suffix)-1)))
			*p++ = word[ix++];
		*p = 0;

		if (suffix[0] != 0) {
			if ((tr->langopts.ordinal_indicator != NULL) && (strcmp(suffix, tr->langopts.ordinal_indicator) == 0))
				ordinal = 2;
			else if (!IsDigit09(suffix[0])) { // not _#9 (tab)
				sprintf(string, "_#%s", suffix);
				if (Lookup(tr, string, ph_ordinal2)) {
					// this is an ordinal suffix
					ordinal = 2;
					flags[0] |= FLAG_SKIPWORDS;
					skipwords = 1;
					sprintf(string, "_x#%s", suffix);
					Lookup(tr, string, ph_ordinal2x); // is there an alternate pronunciation?
				}
			}
		}
	}

	if (wtab[0].flags & FLAG_ORDINAL)
		ordinal = 2;

	ph_append[0] = 0;
	ph_buf2[0] = 0;

	if ((word[0] == '0') && (prev_thousands == 0) && (word[1] != ' ') && (word[1] != tr->langopts.decimal_sep)) {
		if ((n_digits == 2) && (word[3] == ':') && IsDigit09(word[5]) && isspace(word[7])) {
			// looks like a time 02:30, omit the leading zero
		} else {
			if (n_digits > 3) {
				flags[0] &= ~FLAG_SKIPWORDS;
				return 0; // long number string with leading zero, speak as individual digits
			}

			// speak leading zeros
			for (ix = 0; (word[ix] == '0') && (ix < (n_digits-1)); ix++)
				Lookup(tr, "_0", &ph_zeros[strlen(ph_zeros)]);
		}
	}

	if ((tr->langopts.numbers & NUM_ALLOW_SPACE) && (word[n_digits] == ' '))
		thousands_inc = 1;
	else if (word[n_digits] == tr->langopts.thousands_sep)
		thousands_inc = 2;

	suffix_ix = n_digits+2;
	if (thousands_inc > 0) {
		// if the following "words" are three-digit groups, count them and add
		// a "thousand"/"million" suffix to this one
		digix = n_digits + thousands_inc;

		while (((wtab[thousandplex+1].flags & FLAG_MULTIPLE_SPACES) == 0) && CheckThousandsGroup(&word[digix], group_len)) {
			for (ix = 0; ix < group_len; ix++) {
				if (word[digix+ix] != '0') {
					thousands_exact = 0;
					break;
				}
			}

			thousandplex++;
			digix += group_len;
			if ((word[digix] == tr->langopts.thousands_sep) || ((tr->langopts.numbers & NUM_ALLOW_SPACE) && (word[digix] == ' '))) {
				suffix_ix = digix+2;
				digix += thousands_inc;
			} else
				break;
		}
	}

	if ((value == 0) && prev_thousands)
		suppress_null = true;

	if (tr->translator_name == L('h', 'u')) {
		// variant form of numbers when followed by hyphen and a suffix starting with 'a' or 'e' (but not a, e, az, ez, azt, ezt
		if ((wtab[thousandplex].flags & FLAG_HYPHEN_AFTER) && (thousands_exact == 1) && hu_number_e(&word[suffix_ix], thousandplex, value))
			number_control |= 1; // use _1e variant of number
	}

	if ((word[n_digits] == tr->langopts.decimal_sep) && IsDigit09(word[n_digits+1])) {
		// this "word" ends with a decimal point
		Lookup(tr, "_dpt", ph_append);
		decimal_point = 0x100;
	} else if (suppress_null == false) {
		if (thousands_inc > 0) {
			if (thousandplex > 0) {
				if ((suppress_null == false) && (LookupThousands(tr, value, thousandplex, thousands_exact, ph_append))) {
					// found an exact match for N thousand
					value = 0;
					suppress_null = true;
				}
			}
		}
	} else if (speak_missing_thousands == 1) {
		// speak this thousandplex if there was no word for the previous thousandplex
		sprintf(string, "_0M%d", thousandplex+1);
		if (Lookup(tr, string, buf1) == 0) {
			sprintf(string, "_0M%d", thousandplex);
			Lookup(tr, string, ph_append);
		}
	}

	if ((ph_append[0] == 0) && (word[n_digits] == '.') && (thousandplex == 0))
		Lookup(tr, "_.", ph_append);

	if (thousandplex == 0) {
		char *p2;
		// look for combinations of the number with the next word
		p = word;
		while (IsDigit09(p[1])) p++; // just use the last digit
		if (IsDigit09(p[-1])) {
			p2 = p - 1;
			if (LookupDictList(tr, &p2, buf_digit_lookup, flags, FLAG_SUFX, wtab)) // lookup 2 digits
				n_digit_lookup = 2;
		}

		if ((buf_digit_lookup[0] == 0) && (*p != '0')) {
			// LANG=hu ?
			// not found, lookup only the last digit (?? but not if dot-ordinal has been found)
			if (LookupDictList(tr, &p, buf_digit_lookup, flags, FLAG_SUFX, wtab)) // don't match '0', or entries with $only
				n_digit_lookup = 1;
		}

		if (prev_thousands == 0) {
			if ((decimal_point == 0) && (ordinal == 0)) {
				// Look for special pronunciation for this number in isolation (LANG=kl)
				sprintf(string, "_%dn", value);
				if (Lookup(tr, string, ph_out))
					return 1;
			}

			if (tr->langopts.numbers2 & NUM2_PERCENT_BEFORE) {
				// LANG=si, say "percent" before the number
				p2 = word;
				while ((*p2 != ' ') && (*p2 != 0))
					p2++;
				if (p2[1] == '%') {
					Lookup(tr, "%", ph_out);
					ph_out += strlen(ph_out);
					p2[1] = ' ';
				}
			}
		}

	}

	LookupNum3(tr, value, ph_buf, suppress_null, thousandplex, prev_thousands | ordinal | decimal_point);
	if ((thousandplex > 0) && (tr->langopts.numbers2 & NUM2_SWAP_THOUSANDS))
		sprintf(ph_out, "%s%s%c%s%s", ph_zeros, ph_append, phonEND_WORD, ph_buf2, ph_buf);
	else
		sprintf(ph_out, "%s%s%s%c%s", ph_zeros, ph_buf2, ph_buf, phonEND_WORD, ph_append);


	while (decimal_point) {
		n_digits++;

		decimal_count = 0;
		while (IsDigit09(word[n_digits+decimal_count]))
			decimal_count++;

		max_decimal_count = 2;
		switch (decimal_mode = (tr->langopts.numbers & NUM_DFRACTION_BITS))
		{
		case NUM_DFRACTION_4:
			max_decimal_count = 5;
			// fallthrough:
		case NUM_DFRACTION_2:
			// French/Polish decimal fraction
			while (word[n_digits] == '0') {
				Lookup(tr, "_0", buf1);
				strcat(ph_out, buf1);
				decimal_count--;
				n_digits++;
			}
			if ((decimal_count <= max_decimal_count) && IsDigit09(word[n_digits])) {
				LookupNum3(tr, atoi(&word[n_digits]), buf1, false, 0, 0);
				strcat(ph_out, buf1);
				n_digits += decimal_count;
			}
			break;
		case NUM_DFRACTION_1: // italian, say "hundredths" if leading zero
		case NUM_DFRACTION_5: // hungarian, always say "tenths" etc.
		case NUM_DFRACTION_6: // kazakh, always say "tenths" etc, before the decimal fraction
			LookupNum3(tr, atoi(&word[n_digits]), ph_buf, false, 0, 0);
			if ((word[n_digits] == '0') || (decimal_mode != NUM_DFRACTION_1)) {
				// decimal part has leading zeros, so add a "hundredths" or "thousandths" suffix
				sprintf(string, "_0Z%d", decimal_count);
				if (Lookup(tr, string, buf1) == 0)
					break; // revert to speaking single digits

				if (decimal_mode == NUM_DFRACTION_6)
					strcat(ph_out, buf1);
				else
					strcat(ph_buf, buf1);
			}
			strcat(ph_out, ph_buf);
			n_digits += decimal_count;
			break;
		case NUM_DFRACTION_3:
			// Romanian decimal fractions
			if ((decimal_count <= 4) && (word[n_digits] != '0')) {
				LookupNum3(tr, atoi(&word[n_digits]), buf1, false, 0, 0);
				strcat(ph_out, buf1);
				n_digits += decimal_count;
			}
			break;
		case NUM_DFRACTION_7:
			// alternative form of decimal fraction digits, except the final digit
			while (decimal_count-- > 1) {
				sprintf(string, "_%cd", word[n_digits]);
				if (Lookup(tr, string, buf1) == 0)
					break;
				n_digits++;
				strcat(ph_out, buf1);
			}
		}

		while (IsDigit09(c = word[n_digits]) && (strlen(ph_out) < (N_WORD_PHONEMES - 10))) {
			// speak any remaining decimal fraction digits individually
			value = word[n_digits++] - '0';
			LookupNum2(tr, value, 0, 2, buf1);
			len = strlen(ph_out);
			sprintf(&ph_out[len], "%c%s", phonEND_WORD, buf1);
		}

		// something after the decimal part ?
		if (Lookup(tr, "_dpt2", buf1))
			strcat(ph_out, buf1);

		if ((c == tr->langopts.decimal_sep) && IsDigit09(word[n_digits+1])) {
			Lookup(tr, "_dpt", buf1);
			strcat(ph_out, buf1);
		} else
			decimal_point = 0;
	}
	if ((ph_out[0] != 0) && (ph_out[0] != phonSWITCH)) {
		int next_char;
		char *p;
		p = &word[n_digits+1];

		p += utf8_in(&next_char, p);
		if ((tr->langopts.numbers & NUM_NOPAUSE) && (next_char == ' '))
			utf8_in(&next_char, p);

		if (!iswalpha(next_char) && (thousands_exact == 0))
			strcat(ph_out, str_pause); // don't add pause for 100s,  6th, etc.
	}

	*flags |= FLAG_FOUND;
	speak_missing_thousands--;

	if (skipwords)
		dictionary_skipwords = skipwords;
	return 1;
}

int TranslateNumber(Translator *tr, char *word1, char *ph_out, unsigned int *flags, WORD_TAB *wtab, int control)
{
	if ((option_sayas == SAYAS_DIGITS1) || (wtab[0].flags & FLAG_INDIVIDUAL_DIGITS))
		return 0; // speak digits individually

	if (tr->langopts.numbers != 0)
		return TranslateNumber_1(tr, word1, ph_out, flags, wtab, control);
	return 0;
}