Merge pull request #1379

Code cleanup: relocate functions & remove unused #defines

src/libespeak-ng/common.c

@@ -28,13 +28,15 @@
 #include <stdlib.h>
 #include <string.h>
 #include <sys/stat.h>
+#include <wctype.h>
 
 #include <espeak-ng/espeak_ng.h>
 #include <espeak-ng/speak_lib.h>
 #include <espeak-ng/encoding.h>
-
+#include <ucd/ucd.h>
 
 #include "common.h"
+#include "translate.h"
 
 #pragma GCC visibility push(default)
 
@@ -57,4 +59,254 @@ void strncpy0(char *to, const char *from, int size)
 	strncpy(to, from, size);
 	to[size-1] = 0;
 }
+
+int utf8_in(int *c, const char *buf)
+{
+	/* Read a unicode characater from a UTF8 string
+	 * Returns the number of UTF8 bytes used.
+	 * buf: position of buffer is moved, if character is read
+	 * c: holds UTF-16 representation of multibyte character by
+	 * skipping UTF-8 header bits of bytes in following way:
+	 * 2-byte character "ā":
+	 * hex            binary
+	 * c481           1100010010000001
+	 *    |           11000100  000001
+	 *    V              \    \ |    |
+	 * 0101           0000000100000001
+	 * 3-byte character "ꙅ":
+	 * ea9985 111010101001100110000101
+	 *            1010  011001  000101
+	 *    |       +  +--.\   \  |    |
+	 *    V        `--.  \`.  `.|    |
+	 *   A645         1010011001000101
+	 * 4-byte character "𠜎":
+	 * f0a09c8e 11110000101000001001110010001110
+	 *    V          000  100000  011100  001110
+	 *   02070e         000000100000011100001110
+	 */
+	return utf8_in2(c, buf, 0);
+}
 #pragma GCC visibility pop
+
+int utf8_out(unsigned int c, char *buf)
+{
+	// write a UTF-16 character into a buffer as UTF-8
+	// returns the number of bytes written
+
+	int n_bytes;
+	int j;
+	int shift;
+	static const char unsigned code[4] = { 0, 0xc0, 0xe0, 0xf0 };
+
+	if (c < 0x80) {
+		buf[0] = c;
+		return 1;
+	}
+	if (c >= 0x110000) {
+		buf[0] = ' '; // out of range character code
+		return 1;
+	}
+	if (c < 0x0800)
+		n_bytes = 1;
+	else if (c < 0x10000)
+		n_bytes = 2;
+	else
+		n_bytes = 3;
+
+	shift = 6*n_bytes;
+	buf[0] = code[n_bytes] | (c >> shift);
+	for (j = 0; j < n_bytes; j++) {
+		shift -= 6;
+		buf[j+1] = 0x80 + ((c >> shift) & 0x3f);
+	}
+	return n_bytes+1;
+}
+
+int utf8_in2(int *c, const char *buf, int backwards)
+{
+	// Reads a unicode characater from a UTF8 string
+	// Returns the number of UTF8 bytes used.
+	// c: holds integer representation of multibyte character
+	// buf: position of buffer is moved, if character is read
+	// backwards: set if we are moving backwards through the UTF8 string
+
+	int c1;
+	int n_bytes;
+	int ix;
+	static const unsigned char mask[4] = { 0xff, 0x1f, 0x0f, 0x07 };
+
+	// find the start of the next/previous character
+	while ((*buf & 0xc0) == 0x80) {
+		// skip over non-initial bytes of a multi-byte utf8 character
+		if (backwards)
+			buf--;
+		else
+			buf++;
+	}
+
+	n_bytes = 0;
+
+	if ((c1 = *buf++) & 0x80) {
+		if ((c1 & 0xe0) == 0xc0)
+			n_bytes = 1;
+		else if ((c1 & 0xf0) == 0xe0)
+			n_bytes = 2;
+		else if ((c1 & 0xf8) == 0xf0)
+			n_bytes = 3;
+
+		c1 &= mask[n_bytes];
+		for (ix = 0; ix < n_bytes; ix++)
+		{
+			if (!*buf)
+				/* Oops, truncated */
+				break;
+			c1 = (c1 << 6) + (*buf++ & 0x3f);
+		}
+		n_bytes = ix;
+	}
+	*c = c1;
+	return n_bytes+1;
+}
+
+
+int IsAlpha(unsigned int c)
+{
+	// Replacement for iswalph() which also checks for some in-word symbols
+
+	static const unsigned short extra_indic_alphas[] = {
+		0xa70, 0xa71, // Gurmukhi: tippi, addak
+		0
+	};
+
+	if (iswalpha(c))
+		return 1;
+
+	if (c < 0x300)
+		return 0;
+
+	if ((c >= 0x901) && (c <= 0xdf7)) {
+		// Indic scripts: Devanagari, Tamil, etc
+		if ((c & 0x7f) < 0x64)
+			return 1;
+		if (lookupwchar(extra_indic_alphas, c) != 0)
+			return 1;
+		if ((c >= 0xd7a) && (c <= 0xd7f))
+			return 1; // malaytalam chillu characters
+
+		return 0;
+	}
+
+	if ((c >= 0x5b0) && (c <= 0x5c2))
+		return 1; // Hebrew vowel marks
+
+	if (c == 0x0605)
+		return 1;
+
+	if ((c == 0x670) || ((c >= 0x64b) && (c <= 0x65e)))
+		return 1; // arabic vowel marks
+
+	if ((c >= 0x300) && (c <= 0x36f))
+		return 1; // combining accents
+
+	if ((c >= 0xf40) && (c <= 0xfbc))
+		return 1; // tibetan
+
+	if ((c >= 0x1100) && (c <= 0x11ff))
+		return 1; // Korean jamo
+
+	if ((c >= 0x2800) && (c <= 0x28ff))
+		return 1; // braille
+
+	if ((c > 0x3040) && (c <= 0xa700))
+		return 1; // Chinese/Japanese.  Should never get here, but Mac OS 10.4's iswalpha seems to be broken, so just make sure
+
+	return 0;
+}
+
+// brackets, also 0x2014 to 0x021f which don't need to be in this list
+static const unsigned short brackets[] = {
+	'(', ')', '[', ']', '{', '}', '<', '>', '"', '\'', '`',
+	0xab,   0xbb,   // double angle brackets
+	0x300a, 0x300b, // double angle brackets (ideograph)
+	0xe000+'<',     // private usage area
+	0
+};
+
+int IsBracket(int c)
+{
+	if ((c >= 0x2014) && (c <= 0x201f))
+		return 1;
+	return lookupwchar(brackets, c);
+}
+
+int IsDigit09(unsigned int c)
+{
+	if ((c >= '0') && (c <= '9'))
+		return 1;
+	return 0;
+}
+
+int IsDigit(unsigned int c)
+{
+	if (iswdigit(c))
+		return 1;
+
+	if ((c >= 0x966) && (c <= 0x96f))
+		return 1;
+
+	return 0;
+}
+
+int IsSpace(unsigned int c)
+{
+	if (c == 0)
+		return 0;
+	if ((c >= 0x2500) && (c < 0x25a0))
+		return 1; // box drawing characters
+	if ((c >= 0xfff9) && (c <= 0xffff))
+		return 1; // unicode specials
+	return iswspace(c);
+}
+
+int isspace2(unsigned int c)
+{
+	// can't use isspace() because on Windows, isspace(0xe1) gives TRUE !
+	int c2;
+
+	if (((c2 = (c & 0xff)) == 0) || (c > ' '))
+		return 0;
+	return 1;
+}
+
+int is_str_totally_null(const char* str, int size) {
+	// Tests if all bytes of str are null up to size
+	// This should never be reimplemented with integers, because
+	// this function has to work with unaligned char*
+	// (casting to int when unaligned may result in ungaranteed behaviors)
+	return (*str == 0 && memcmp(str, str+1, size-1) == 0);
+}
+
+int Read4Bytes(FILE *f)
+{
+	// Read 4 bytes (least significant first) into a word
+	int ix;
+	unsigned char c;
+	int acc = 0;
+
+	for (ix = 0; ix < 4; ix++) {
+		c = fgetc(f) & 0xff;
+		acc += (c << (ix*8));
+	}
+	return acc;
+}
+
+int towlower2(unsigned int c, Translator *translator)
+{
+	// check for non-standard upper to lower case conversions
+	if (c == 'I' && translator->langopts.dotless_i)
+		return 0x131; // I -> ı
+
+	return ucd_tolower(c);
+}
+
+

src/libespeak-ng/common.h

@@ -21,10 +21,25 @@
 #define ESPEAK_NG_COMMON_H
 
 #include "espeak-ng/espeak_ng.h"
+#include "translate.h"
 
 extern ESPEAK_NG_API int GetFileLength(const char *filename);
 extern ESPEAK_NG_API void strncpy0(char *to, const char *from, int size);
 
+int IsAlpha(unsigned int c);
+int IsBracket(int c);
+int IsDigit(unsigned int c);
+int IsDigit09(unsigned int c);
+int IsSpace(unsigned int c);
+int isspace2(unsigned int c);
+int is_str_totally_null(const char* str, int size); // Tests if all bytes of str up to size are null
+int Read4Bytes(FILE *f);
+int towlower2(unsigned int c, Translator *translator); // Supports Turkish I
+
+ESPEAK_NG_API int utf8_in(int *c, const char *buf);
+int utf8_in2(int *c, const char *buf, int backwards);
+int utf8_out(unsigned int c, char *buf);
+
 #ifdef __cplusplus
 }
 #endif

src/libespeak-ng/compiledata.c

@@ -35,11 +35,10 @@
 #include <espeak-ng/speak_lib.h>
 #include <espeak-ng/encoding.h>
 
-#include "common.h"                    // for GetFileLength, strncpy0
+#include "common.h"                    // for GetFileLength, strncpy0, ...
 #include "error.h"                    // for create_file_error_context
 #include "mnemonics.h"               // for LookupMnemName, MNEM_TAB
 #include "phoneme.h"                  // for PHONEME_TAB, PHONEME_TAB_LIST
-#include "readclause.h"               // for Read4Bytes
 #include "spect.h"                    // for SpectFrame, peak_t, SpectSeq
 #include "speech.h"			// for path_home, GetFileLength
 #include "synthdata.h"                // for LoadPhData

src/libespeak-ng/compiledict.c

@@ -39,10 +39,8 @@
 #include "error.h"                // for create_file_error_context
 #include "mnemonics.h"               // for LookupMnemName, MNEM_TAB
 #include "phoneme.h"              // for PHONEME_TAB_LIST, phonSWITCH, phone...
-#include "readclause.h"           // for towlower2
 #include "speech.h"		// for path_home
 #include "synthesize.h"           // for Write4Bytes
-#include "translate.h"            // for isspace2, IsDigit09, utf8_in, utf8_out
 
 static FILE *f_log = NULL;
 

src/libespeak-ng/dictionary.c

@@ -36,13 +36,15 @@
 #include "dictionary.h"
 #include "numbers.h"                       // for LookupAccentedLetter, Look...
 #include "phoneme.h"                       // for PHONEME_TAB, phVOWEL, phon...
-#include "readclause.h"                    // for WordToString2, is_str_tota...
+#include "readclause.h"                    // for WordToString2
 #include "speech.h"                        // for path_home
 #include "compiledict.h"                   // for DecodeRule
 #include "synthdata.h"                     // for PhonemeCode, InterpretPhoneme
 #include "synthesize.h"                    // for STRESS_IS_PRIMARY, phoneme...
 #include "translate.h"                     // for Translator, utf8_in, LANGU...
 
+static int LookupFlags(Translator *tr, const char *word, unsigned int **flags_out);
+
 typedef struct {
 	int points;
 	const char *phonemes;
@@ -761,96 +763,7 @@ int IsVowel(Translator *tr, int letter)
 	return IsLetter(tr, letter, LETTERGP_VOWEL2);
 }
 
-static int Unpronouncable2(Translator *tr, char *word)
-{
-	int c;
-	int end_flags;
-	char ph_buf[N_WORD_PHONEMES];
-
-	ph_buf[0] = 0;
-	c = word[-1];
-	word[-1] = ' '; // ensure there is a space before the "word"
-	end_flags = TranslateRules(tr, word, ph_buf, sizeof(ph_buf), NULL, FLAG_UNPRON_TEST, NULL);
-	word[-1] = c;
-	if ((end_flags == 0) || (end_flags & SUFX_UNPRON))
-		return 1;
-	return 0;
-}
-
-int Unpronouncable(Translator *tr, char *word, int posn)
-{
-	/* Determines whether a word in 'unpronouncable', i.e. whether it should
-	    be spoken as individual letters.
-
-	    This function may be language specific. This is a generic version.
-	 */
-
-	int c;
-	int c1 = 0;
-	int vowel_posn = 9;
-	int index;
-	int count;
-	ALPHABET *alphabet;
-
-	utf8_in(&c, word);
-	if ((tr->letter_bits_offset > 0) && (c < 0x241)) {
-		// Latin characters for a language with a non-latin alphabet
-		return 0;  // so we can re-translate the word as English
-	}
-
-	if (((alphabet = AlphabetFromChar(c)) != NULL)  && (alphabet->offset != tr->letter_bits_offset)) {
-		// Character is not in our alphabet
-		return 0;
-	}
-
-	if (tr->langopts.param[LOPT_UNPRONOUNCABLE] == 1)
-		return 0;
-
-	if (((c = *word) == ' ') || (c == 0) || (c == '\''))
-		return 0;
-
-	index = 0;
-	count = 0;
-	for (;;) {
-		index += utf8_in(&c, &word[index]);
-		if ((c == 0) || (c == ' '))
-			break;
-
-		if ((c == '\'') && ((count > 1) || (posn > 0)))
-			break; // "tv'" but not "l'"
-
-		if (count == 0)
-			c1 = c;
-
-		if ((c == '\'') && (tr->langopts.param[LOPT_UNPRONOUNCABLE] == 3)) {
-			// don't count apostrophe
-		} else
-			count++;
-
-		if (IsVowel(tr, c)) {
-			vowel_posn = count; // position of the first vowel
-			break;
-		}
-
-		if ((c != '\'') && !iswalpha(c))
-			return 0;
-	}
-
-	if ((vowel_posn > 2) && (tr->langopts.param[LOPT_UNPRONOUNCABLE] == 2)) {
-		// Lookup unpronounable rules in *_rules
-		return Unpronouncable2(tr, word);
-	}
-
-	if (c1 == tr->langopts.param[LOPT_UNPRONOUNCABLE])
-		vowel_posn--; // disregard this as the initial letter when counting
-
-	if (vowel_posn > (tr->langopts.max_initial_consonants+1))
-		return 1; // no vowel, or no vowel in first few letters
-
-	return 0;
-}
-
-static int GetVowelStress(Translator *tr, unsigned char *phonemes, signed char *vowel_stress, int *vowel_count, int *stressed_syllable, int control)
+int GetVowelStress(Translator *tr, unsigned char *phonemes, signed char *vowel_stress, int *vowel_count, int *stressed_syllable, int control)
 {
 	// control = 1, set stress to 1 for forced unstressed vowels
 	unsigned char phcode;
@@ -962,55 +875,11 @@ static int GetVowelStress(Translator *tr, unsigned char *phonemes, signed char *
 	return max_stress;
 }
 
-static char stress_phonemes[] = {
+const char stress_phonemes[] = {
 	phonSTRESS_D, phonSTRESS_U, phonSTRESS_2, phonSTRESS_3,
 	phonSTRESS_P, phonSTRESS_P2, phonSTRESS_TONIC
 };
 
-void ChangeWordStress(Translator *tr, char *word, int new_stress)
-{
-	int ix;
-	unsigned char *p;
-	int max_stress;
-	int vowel_count; // num of vowels + 1
-	int stressed_syllable = 0; // position of stressed syllable
-	unsigned char phonetic[N_WORD_PHONEMES];
-	signed char vowel_stress[N_WORD_PHONEMES/2];
-
-	strcpy((char *)phonetic, word);
-	max_stress = GetVowelStress(tr, phonetic, vowel_stress, &vowel_count, &stressed_syllable, 0);
-
-	if (new_stress >= STRESS_IS_PRIMARY) {
-		// promote to primary stress
-		for (ix = 1; ix < vowel_count; ix++) {
-			if (vowel_stress[ix] >= max_stress) {
-				vowel_stress[ix] = new_stress;
-				break;
-			}
-		}
-	} else {
-		// remove primary stress
-		for (ix = 1; ix < vowel_count; ix++) {
-			if (vowel_stress[ix] > new_stress) // >= allows for diminished stress (=1)
-				vowel_stress[ix] = new_stress;
-		}
-	}
-
-	// write out phonemes
-	ix = 1;
-	p = phonetic;
-	while (*p != 0) {
-		if ((phoneme_tab[*p]->type == phVOWEL) && !(phoneme_tab[*p]->phflags & phNONSYLLABIC)) {
-			if ((vowel_stress[ix] == STRESS_IS_DIMINISHED) || (vowel_stress[ix] > STRESS_IS_UNSTRESSED))
-				*word++ = stress_phonemes[(unsigned char)vowel_stress[ix]];
-
-			ix++;
-		}
-		*word++ = *p++;
-	}
-	*word = 0;
-}
-
 void SetWordStress(Translator *tr, char *output, unsigned int *dictionary_flags, int tonic, int control)
 {
 	/* Guess stress pattern of word.  This is language specific
@@ -2345,7 +2214,7 @@ int TranslateRules(Translator *tr, char *p_start, char *phonemes, int ph_size, c
 						if (letter == 0xe000+'(') {
 							if (pre_pause < tr->langopts.param[LOPT_BRACKET_PAUSE_ANNOUNCED])
 								pre_pause = tr->langopts.param[LOPT_BRACKET_PAUSE_ANNOUNCED]; // a bracket, already spoken by AnnouncePunctuation()
-						} 
+						}
 						if (IsBracket(letter)) {
 							if (pre_pause < tr->langopts.param[LOPT_BRACKET_PAUSE])
 								pre_pause = tr->langopts.param[LOPT_BRACKET_PAUSE];
@@ -2456,37 +2325,6 @@ int TranslateRules(Translator *tr, char *p_start, char *phonemes, int ph_size, c
 	return 0;
 }
 
-void ApplySpecialAttribute2(Translator *tr, char *phonemes, int dict_flags)
-{
-	// apply after the translation is complete
-
-	int ix;
-	int len;
-	char *p;
-
-	len = strlen(phonemes);
-
-	if (tr->langopts.param[LOPT_ALT] & 2) {
-		for (ix = 0; ix < (len-1); ix++) {
-			if (phonemes[ix] == phonSTRESS_P) {
-				p = &phonemes[ix+1];
-				if ((dict_flags & FLAG_ALT2_TRANS) != 0) {
-					if (*p == PhonemeCode('E'))
-						*p = PhonemeCode('e');
-					if (*p == PhonemeCode('O'))
-						*p = PhonemeCode('o');
-				} else {
-					if (*p == PhonemeCode('e'))
-						*p = PhonemeCode('E');
-					if (*p == PhonemeCode('o'))
-						*p = PhonemeCode('O');
-				}
-				break;
-			}
-		}
-	}
-}
-
 int TransposeAlphabet(Translator *tr, char *text)
 {
 	// transpose cyrillic alphabet (for example) into ascii (single byte) character codes
@@ -2856,6 +2694,21 @@ static const char *LookupDict2(Translator *tr, const char *word, const char *wor
 	return 0;
 }
 
+
+    static int utf8_nbytes(const char *buf)
+{
+	// Returns the number of bytes for the first UTF-8 character in buf
+
+	unsigned char c = (unsigned char)buf[0];
+	if (c < 0x80)
+		return 1;
+	if (c < 0xe0)
+		return 2;
+	if (c < 0xf0)
+		return 3;
+	return 4;
+}
+
 /* Lookup a specified word in the word dictionary.
    Returns phonetic data in 'phonetic' and bits in 'flags'
 
@@ -3022,7 +2875,7 @@ int Lookup(Translator *tr, const char *word, char *ph_out)
 	return flags0;
 }
 
-int LookupFlags(Translator *tr, const char *word, unsigned int **flags_out)
+static int LookupFlags(Translator *tr, const char *word, unsigned int **flags_out)
 {
 	char buf[100];
 	static unsigned int flags[2];

src/libespeak-ng/dictionary.h

@@ -31,22 +31,22 @@ extern "C"
 {
 #endif
 
+extern const char stress_phonemes[];
+
 int LoadDictionary(Translator *tr, const char *name, int no_error);
 int HashDictionary(const char *string);
 const char *EncodePhonemes(const char *p, char *outptr, int *bad_phoneme);
 void DecodePhonemes(const char *inptr, char *outptr);
 char *WritePhMnemonic(char *phon_out, PHONEME_TAB *ph, PHONEME_LIST *plist, int use_ipa, int *flags);
 const char *GetTranslatedPhonemeString(int phoneme_mode);
+int GetVowelStress(Translator *tr, unsigned char *phonemes, signed char *vowel_stress, int *vowel_count, int *stressed_syllable, int control);
 int IsVowel(Translator *tr, int letter);
-int Unpronouncable(Translator *tr, char *word, int posn);
-void ChangeWordStress(Translator *tr, char *word, int new_stress);
 void SetWordStress(Translator *tr, char *output, unsigned int *dictionary_flags, int tonic, int control);
 void AppendPhonemes(Translator *tr, char *string, int size, const char *ph);
 int TranslateRules(Translator *tr, char *p_start, char *phonemes, int ph_size, char *end_phonemes, int word_flags, unsigned int *dict_flags);
 int TransposeAlphabet(Translator *tr, char *text);
 int Lookup(Translator *tr, const char *word, char *ph_out);
 int LookupDictList(Translator *tr, char **wordptr, char *ph_out, unsigned int *flags, int end_flags, WORD_TAB *wtab);
-int LookupFlags(Translator *tr, const char *word, unsigned int **flags_out);
 int RemoveEnding(Translator *tr, char *word, int end_type, char *word_copy);
 
 #ifdef __cplusplus

src/libespeak-ng/klatt.h

@@ -112,9 +112,6 @@ typedef struct {
 #define Rnpp 10
 #define R1p  11
 #define R2p  12
-#define R3p  13
-#define R4p  14
-#define R5p  15
 #define R6p  16
 
 #define RGL  17

src/libespeak-ng/numbers.c

@@ -32,12 +32,13 @@
 #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, towlower2
+#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 "translate.h"   // for Translator, LANGUAGE_OPTIONS, WOR...
 #include "voice.h"       // for voice, voice_t
 
 #define M_LIGATURE  0x8000

src/libespeak-ng/readclause.c

@@ -131,23 +131,6 @@ int clause_type_from_codepoint(uint32_t c)
 	return CLAUSE_NONE;
 }
 
-int is_str_totally_null(const char* str, int size) {
-	// Tests if all bytes of str are null up to size
-	// This should never be reimplemented with integers, because
-	// this function has to work with unaligned char*
-	// (casting to int when unaligned may result in ungaranteed behaviors)
-	return (*str == 0 && memcmp(str, str+1, size-1) == 0);
-}
-
-int towlower2(unsigned int c, Translator *translator)
-{
-	// check for non-standard upper to lower case conversions
-	if (c == 'I' && translator->langopts.dotless_i)
-		return 0x131; // I -> ı
-
-	return ucd_tolower(c);
-}
-
 static int IsRomanU(unsigned int c)
 {
 	if ((c == 'I') || (c == 'V') || (c == 'X') || (c == 'L'))
@@ -288,20 +271,6 @@ static const char *LookupCharName(Translator *tr, int c, int only)
 	return buf;
 }
 
-int Read4Bytes(FILE *f)
-{
-	// Read 4 bytes (least significant first) into a word
-	int ix;
-	unsigned char c;
-	int acc = 0;
-
-	for (ix = 0; ix < 4; ix++) {
-		c = fgetc(f) & 0xff;
-		acc += (c << (ix*8));
-	}
-	return acc;
-}
-
 static int AnnouncePunctuation(Translator *tr, int c1, int *c2_ptr, char *output, int *bufix, int end_clause)
 {
 	// announce punctuation names

src/libespeak-ng/readclause.h

@@ -34,14 +34,9 @@ typedef struct {
 
 extern PARAM_STACK param_stack[];
 
-// Tests if all bytes of str up to size are null
-int is_str_totally_null(const char* str, int size);
-
 int clause_type_from_codepoint(uint32_t c);
-int towlower2(unsigned int c, Translator *translator); // Supports Turkish I
 int Eof(void);
 const char *WordToString2(unsigned int word);
-int Read4Bytes(FILE *f);
 int AddNameData(const char *name,
                 int wide);
 int ReadClause(Translator *tr,

src/libespeak-ng/soundicon.c

@@ -38,7 +38,6 @@
 #include "soundicon.h" 
 #include "common.h"                // for GetFileLength
 #include "error.h"                // for create_file_error_context
-#include "readclause.h"               // for Read4Bytes
 #include "speech.h"                   // for path_home, PATHSEP
 #include "synthesize.h"                   // for samplerate
 

src/libespeak-ng/spect.h

@@ -33,7 +33,6 @@ float polint(float xa[], float ya[], int n, float x);
 
 #define FRAME_WIDTH  1000  // max width for 8000kHz frame
 #define MAX_DISPLAY_FREQ 9500
-#define FRAME_HEIGHT 240
 
 #define T_AMPLITUDE 308
 #define T_AV      312

src/libespeak-ng/ssml.c

@@ -45,7 +45,7 @@
 #include "readclause.h"           // for PARAM_STACK, param_stack, AddNameData
 #include "soundicon.h"               // for LoadSoundFile2
 #include "synthesize.h"           // for SPEED_FACTORS, speed
-#include "translate.h"            // for CTRL_EMBEDDED, IsDigit09, utf8_out
+#include "translate.h"            // for CTRL_EMBEDDED
 #include "voice.h"                // for SelectVoice, SelectVoiceByName
 #include "speech.h"               // for MAKE_MEM_UNDEFINED
 

src/libespeak-ng/synth_mbrola.c

@@ -34,7 +34,6 @@
 
 #include "dictionary.h"
 #include "mbrola.h"
-#include "readclause.h"
 #include "setlengths.h"
 #include "synthdata.h"
 #include "wavegen.h"

src/libespeak-ng/tr_languages.c

@@ -31,6 +31,7 @@
 #include <espeak-ng/speak_lib.h>
 #include <espeak-ng/encoding.h>
 
+#include "common.h"
 #include "setlengths.h"          // for SetLengthMods
 #include "translate.h"           // for Translator, LANGUAGE_OPTIONS, L, NUM...
 

src/libespeak-ng/translate.c

@@ -33,6 +33,7 @@
 #include <espeak-ng/encoding.h>
 
 #include "translate.h"
+#include "common.h"
 #include "dictionary.h"           // for TranslateRules, LookupDictList, Cha...
 #include "phoneme.h"              // for phonSWITCH, PHONEME_TAB, phonPAUSE_...
 #include "phonemelist.h"          // for MakePhonemeList
@@ -104,111 +105,9 @@ static char source[N_TR_SOURCE+40]; // extra space for embedded command & voice
 int n_replace_phonemes;
 REPLACE_PHONEMES replace_phonemes[N_REPLACE_PHONEMES];
 
-// brackets, also 0x2014 to 0x021f which don't need to be in this list
-static const unsigned short brackets[] = {
-	'(', ')', '[', ']', '{', '}', '<', '>', '"', '\'', '`',
-	0xab,   0xbb,   // double angle brackets
-	0x300a, 0x300b, // double angle brackets (ideograph)
-	0xe000+'<',     // private usage area
-	0
-};
-
 // other characters which break a word, but don't produce a pause
 static const unsigned short breaks[] = { '_', 0 };
 
-int IsAlpha(unsigned int c)
-{
-	// Replacement for iswalph() which also checks for some in-word symbols
-
-	static const unsigned short extra_indic_alphas[] = {
-		0xa70, 0xa71, // Gurmukhi: tippi, addak
-		0
-	};
-
-	if (iswalpha(c))
-		return 1;
-
-	if (c < 0x300)
-		return 0;
-
-	if ((c >= 0x901) && (c <= 0xdf7)) {
-		// Indic scripts: Devanagari, Tamil, etc
-		if ((c & 0x7f) < 0x64)
-			return 1;
-		if (lookupwchar(extra_indic_alphas, c) != 0)
-			return 1;
-		if ((c >= 0xd7a) && (c <= 0xd7f))
-			return 1; // malaytalam chillu characters
-
-		return 0;
-	}
-
-	if ((c >= 0x5b0) && (c <= 0x5c2))
-		return 1; // Hebrew vowel marks
-
-	if (c == 0x0605)
-		return 1;
-
-	if ((c == 0x670) || ((c >= 0x64b) && (c <= 0x65e)))
-		return 1; // arabic vowel marks
-
-	if ((c >= 0x300) && (c <= 0x36f))
-		return 1; // combining accents
-
-	if ((c >= 0xf40) && (c <= 0xfbc))
-		return 1; // tibetan
-
-	if ((c >= 0x1100) && (c <= 0x11ff))
-		return 1; // Korean jamo
-
-	if ((c >= 0x2800) && (c <= 0x28ff))
-		return 1; // braille
-
-	if ((c > 0x3040) && (c <= 0xa700))
-		return 1; // Chinese/Japanese.  Should never get here, but Mac OS 10.4's iswalpha seems to be broken, so just make sure
-
-	return 0;
-}
-
-int IsDigit09(unsigned int c)
-{
-	if ((c >= '0') && (c <= '9'))
-		return 1;
-	return 0;
-}
-
-int IsDigit(unsigned int c)
-{
-	if (iswdigit(c))
-		return 1;
-
-	if ((c >= 0x966) && (c <= 0x96f))
-		return 1;
-
-	return 0;
-}
-
-static int IsSpace(unsigned int c)
-{
-	if (c == 0)
-		return 0;
-	if ((c >= 0x2500) && (c < 0x25a0))
-		return 1; // box drawing characters
-	if ((c >= 0xfff9) && (c <= 0xffff))
-		return 1; // unicode specials
-	return iswspace(c);
-}
-
-int isspace2(unsigned int c)
-{
-	// can't use isspace() because on Windows, isspace(0xe1) gives TRUE !
-	int c2;
-
-	if (((c2 = (c & 0xff)) == 0) || (c > ' '))
-		return 0;
-	return 1;
-}
-
 void DeleteTranslator(Translator *tr)
 {
 	if (!tr) return;
@@ -230,136 +129,6 @@ int lookupwchar(const unsigned short *list, int c)
 	return 0;
 }
 
-int IsBracket(int c)
-{
-	if ((c >= 0x2014) && (c <= 0x201f))
-		return 1;
-	return lookupwchar(brackets, c);
-}
-
-int utf8_nbytes(const char *buf)
-{
-	// Returns the number of bytes for the first UTF-8 character in buf
-
-	unsigned char c = (unsigned char)buf[0];
-	if (c < 0x80)
-		return 1;
-	if (c < 0xe0)
-		return 2;
-	if (c < 0xf0)
-		return 3;
-	return 4;
-}
-
-int utf8_in2(int *c, const char *buf, int backwards)
-{
-	// Reads a unicode characater from a UTF8 string
-	// Returns the number of UTF8 bytes used.
-	// c: holds integer representation of multibyte character
-	// buf: position of buffer is moved, if character is read
-	// backwards: set if we are moving backwards through the UTF8 string
-
-	int c1;
-	int n_bytes;
-	int ix;
-	static const unsigned char mask[4] = { 0xff, 0x1f, 0x0f, 0x07 };
-
-	// find the start of the next/previous character
-	while ((*buf & 0xc0) == 0x80) {
-		// skip over non-initial bytes of a multi-byte utf8 character
-		if (backwards)
-			buf--;
-		else
-			buf++;
-	}
-
-	n_bytes = 0;
-
-	if ((c1 = *buf++) & 0x80) {
-		if ((c1 & 0xe0) == 0xc0)
-			n_bytes = 1;
-		else if ((c1 & 0xf0) == 0xe0)
-			n_bytes = 2;
-		else if ((c1 & 0xf8) == 0xf0)
-			n_bytes = 3;
-
-		c1 &= mask[n_bytes];
-		for (ix = 0; ix < n_bytes; ix++)
-		{
-			if (!*buf)
-				/* Oops, truncated */
-				break;
-			c1 = (c1 << 6) + (*buf++ & 0x3f);
-		}
-		n_bytes = ix;
-	}
-	*c = c1;
-	return n_bytes+1;
-}
-
-#pragma GCC visibility push(default)
-int utf8_in(int *c, const char *buf)
-{
-	/* Read a unicode characater from a UTF8 string
-	 * Returns the number of UTF8 bytes used.
-	 * buf: position of buffer is moved, if character is read
-	 * c: holds UTF-16 representation of multibyte character by
-	 * skipping UTF-8 header bits of bytes in following way:
-	 * 2-byte character "ā":
-	 * hex            binary
-	 * c481           1100010010000001
-	 *    |           11000100  000001
-	 *    V              \    \ |    |
-	 * 0101           0000000100000001
-	 * 3-byte character "ꙅ":
-	 * ea9985 111010101001100110000101
-	 *            1010  011001  000101
-	 *    |       +  +--.\   \  |    |
-	 *    V        `--.  \`.  `.|    |
-	 *   A645         1010011001000101
-	 * 4-byte character "𠜎":
-	 * f0a09c8e 11110000101000001001110010001110
-	 *    V          000  100000  011100  001110
-	 *   02070e         000000100000011100001110
-	 */
-	return utf8_in2(c, buf, 0);
-}
-#pragma GCC visibility pop
-
-int utf8_out(unsigned int c, char *buf)
-{
-	// write a UTF-16 character into a buffer as UTF-8
-	// returns the number of bytes written
-
-	int n_bytes;
-	int j;
-	int shift;
-	static const char unsigned code[4] = { 0, 0xc0, 0xe0, 0xf0 };
-
-	if (c < 0x80) {
-		buf[0] = c;
-		return 1;
-	}
-	if (c >= 0x110000) {
-		buf[0] = ' '; // out of range character code
-		return 1;
-	}
-	if (c < 0x0800)
-		n_bytes = 1;
-	else if (c < 0x10000)
-		n_bytes = 2;
-	else
-		n_bytes = 3;
-
-	shift = 6*n_bytes;
-	buf[0] = code[n_bytes] | (c >> shift);
-	for (j = 0; j < n_bytes; j++) {
-		shift -= 6;
-		buf[j+1] = 0x80 + ((c >> shift) & 0x3f);
-	}
-	return n_bytes+1;
-}
-
 char *strchr_w(const char *s, int c)
 {
 	// return NULL for any non-ascii character

src/libespeak-ng/translate.h

@@ -64,9 +64,6 @@ extern "C"
 #define FLAG_ALT_TRANS      0x8000 // language specific
 #define FLAG_ALT2_TRANS    0x10000 // language specific
 #define FLAG_ALT3_TRANS    0x20000 // language specific
-#define FLAG_ALT4_TRANS    0x40000 // language specific
-#define FLAG_ALT5_TRANS    0x80000 // language specific
-#define FLAG_ALT6_TRANS   0x100000 // language specific
 #define FLAG_ALT7_TRANS   0x200000 // language specific
 
 #define FLAG_COMBINE      0x800000 // combine with the next word
@@ -661,21 +658,11 @@ extern int (*phoneme_callback)(const char *);
 #define LEADING_2_BITS 0xC0 // 0b11000000
 #define UTF8_TAIL_BITS 0x80 // 0b10000000
 
-ESPEAK_NG_API int utf8_in(int *c, const char *buf);
-int utf8_in2(int *c, const char *buf, int backwards);
-int utf8_out(unsigned int c, char *buf);
-int utf8_nbytes(const char *buf);
-
 int lookupwchar(const unsigned short *list, int c);
 char *strchr_w(const char *s, int c);
-int IsBracket(int c);
 void InitNamedata(void);
 void InitText(int flags);
 void InitText2(void);
-int IsDigit(unsigned int c);
-int IsDigit09(unsigned int c);
-int IsAlpha(unsigned int c);
-int isspace2(unsigned int c);
 ALPHABET *AlphabetFromChar(int c);
 
 Translator *SelectTranslator(const char *name);
@@ -686,8 +673,6 @@ void ProcessLanguageOptions(LANGUAGE_OPTIONS *langopts);
 
 void print_dictionary_flags(unsigned int *flags, char *buf, int buf_len);
 
-void ApplySpecialAttribute2(Translator *tr, char *phonemes, int dict_flags);
-
 int TranslateWord(Translator *tr, char *word1, WORD_TAB *wtab, char *word_out);
 void TranslateClause(Translator *tr, int *tone, char **voice_change);
 

src/libespeak-ng/translateword.c

@@ -36,7 +36,7 @@
 #include "translate.h"
 #include "translateword.h"
 #include "common.h"               // for strncpy0
-#include "dictionary.h"           // for TranslateRules, LookupDictList, Cha...
+#include "dictionary.h"           // for TranslateRules, LookupDictList
 #include "numbers.h"              // for SetSpellingStress, ...
 #include "phoneme.h"              // for phonSWITCH, PHONEME_TAB, phonPAUSE_...
 #include "readclause.h"           // for towlower2
@@ -46,10 +46,14 @@
 
 
 static void addPluralSuffixes(int flags, Translator *tr, char last_char, char *word_phonemes);
+static void ApplySpecialAttribute2(Translator *tr, char *phonemes, int dict_flags);
+static void ChangeWordStress(Translator *tr, char *word, int new_stress);
 static int CheckDottedAbbrev(char *word1);
 static int NonAsciiNumber(int letter);
 static char *SpeakIndividualLetters(Translator *tr, char *word, char *phonemes, int spell_word, ALPHABET *current_alphabet, char word_phonemes[]);
 static int TranslateLetter(Translator *tr, char *word, char *phonemes, int control, ALPHABET *current_alphabet);
+static int Unpronouncable(Translator *tr, char *word, int posn);
+static int Unpronouncable2(Translator *tr, char *word);
 
 int TranslateWord3(Translator *tr, char *word_start, WORD_TAB *wtab, char *word_out, bool *any_stressed_words, ALPHABET *current_alphabet, char word_phonemes[], size_t size_word_phonemes)
 {
@@ -667,6 +671,82 @@ int TranslateWord3(Translator *tr, char *word_start, WORD_TAB *wtab, char *word_
 }
 
 
+void ApplySpecialAttribute2(Translator *tr, char *phonemes, int dict_flags)
+{
+	// apply after the translation is complete
+
+	int ix;
+	int len;
+	char *p;
+
+	len = strlen(phonemes);
+
+	if (tr->langopts.param[LOPT_ALT] & 2) {
+		for (ix = 0; ix < (len-1); ix++) {
+			if (phonemes[ix] == phonSTRESS_P) {
+				p = &phonemes[ix+1];
+				if ((dict_flags & FLAG_ALT2_TRANS) != 0) {
+					if (*p == PhonemeCode('E'))
+						*p = PhonemeCode('e');
+					if (*p == PhonemeCode('O'))
+						*p = PhonemeCode('o');
+				} else {
+					if (*p == PhonemeCode('e'))
+						*p = PhonemeCode('E');
+					if (*p == PhonemeCode('o'))
+						*p = PhonemeCode('O');
+				}
+				break;
+			}
+		}
+	}
+}
+
+
+static void ChangeWordStress(Translator *tr, char *word, int new_stress)
+{
+	int ix;
+	unsigned char *p;
+	int max_stress;
+	int vowel_count; // num of vowels + 1
+	int stressed_syllable = 0; // position of stressed syllable
+	unsigned char phonetic[N_WORD_PHONEMES];
+	signed char vowel_stress[N_WORD_PHONEMES/2];
+
+	strcpy((char *)phonetic, word);
+	max_stress = GetVowelStress(tr, phonetic, vowel_stress, &vowel_count, &stressed_syllable, 0);
+
+	if (new_stress >= STRESS_IS_PRIMARY) {
+		// promote to primary stress
+		for (ix = 1; ix < vowel_count; ix++) {
+			if (vowel_stress[ix] >= max_stress) {
+				vowel_stress[ix] = new_stress;
+				break;
+			}
+		}
+	} else {
+		// remove primary stress
+		for (ix = 1; ix < vowel_count; ix++) {
+			if (vowel_stress[ix] > new_stress) // >= allows for diminished stress (=1)
+				vowel_stress[ix] = new_stress;
+		}
+	}
+
+	// write out phonemes
+	ix = 1;
+	p = phonetic;
+	while (*p != 0) {
+		if ((phoneme_tab[*p]->type == phVOWEL) && !(phoneme_tab[*p]->phflags & phNONSYLLABIC)) {
+			if ((vowel_stress[ix] == STRESS_IS_DIMINISHED) || (vowel_stress[ix] > STRESS_IS_UNSTRESSED))
+				*word++ = stress_phonemes[(unsigned char)vowel_stress[ix]];
+
+			ix++;
+		}
+		*word++ = *p++;
+	}
+	*word = 0;
+}
+
 static char *SpeakIndividualLetters(Translator *tr, char *word, char *phonemes, int spell_word, ALPHABET *current_alphabet, char word_phonemes[])
 {
 	int posn = 0;
@@ -1034,3 +1114,92 @@ static int NonAsciiNumber(int letter)
 	}
 	return -1;
 }
+
+static int Unpronouncable(Translator *tr, char *word, int posn)
+{
+	/* Determines whether a word in 'unpronouncable', i.e. whether it should
+	    be spoken as individual letters.
+
+	    This function may be language specific. This is a generic version.
+	 */
+
+	int c;
+	int c1 = 0;
+	int vowel_posn = 9;
+	int index;
+	int count;
+	ALPHABET *alphabet;
+
+	utf8_in(&c, word);
+	if ((tr->letter_bits_offset > 0) && (c < 0x241)) {
+		// Latin characters for a language with a non-latin alphabet
+		return 0;  // so we can re-translate the word as English
+	}
+
+	if (((alphabet = AlphabetFromChar(c)) != NULL)  && (alphabet->offset != tr->letter_bits_offset)) {
+		// Character is not in our alphabet
+		return 0;
+	}
+
+	if (tr->langopts.param[LOPT_UNPRONOUNCABLE] == 1)
+		return 0;
+
+	if (((c = *word) == ' ') || (c == 0) || (c == '\''))
+		return 0;
+
+	index = 0;
+	count = 0;
+	for (;;) {
+		index += utf8_in(&c, &word[index]);
+		if ((c == 0) || (c == ' '))
+			break;
+
+		if ((c == '\'') && ((count > 1) || (posn > 0)))
+			break; // "tv'" but not "l'"
+
+		if (count == 0)
+			c1 = c;
+
+		if ((c == '\'') && (tr->langopts.param[LOPT_UNPRONOUNCABLE] == 3)) {
+			// don't count apostrophe
+		} else
+			count++;
+
+		if (IsVowel(tr, c)) {
+			vowel_posn = count; // position of the first vowel
+			break;
+		}
+
+		if ((c != '\'') && !iswalpha(c))
+			return 0;
+	}
+
+	if ((vowel_posn > 2) && (tr->langopts.param[LOPT_UNPRONOUNCABLE] == 2)) {
+		// Lookup unpronounable rules in *_rules
+		return Unpronouncable2(tr, word);
+	}
+
+	if (c1 == tr->langopts.param[LOPT_UNPRONOUNCABLE])
+		vowel_posn--; // disregard this as the initial letter when counting
+
+	if (vowel_posn > (tr->langopts.max_initial_consonants+1))
+		return 1; // no vowel, or no vowel in first few letters
+
+	return 0;
+}
+
+static int Unpronouncable2(Translator *tr, char *word)
+{
+	int c;
+	int end_flags;
+	char ph_buf[N_WORD_PHONEMES];
+
+	ph_buf[0] = 0;
+	c = word[-1];
+	word[-1] = ' '; // ensure there is a space before the "word"
+	end_flags = TranslateRules(tr, word, ph_buf, sizeof(ph_buf), NULL, FLAG_UNPRON_TEST, NULL);
+	word[-1] = c;
+	if ((end_flags == 0) || (end_flags & SUFX_UNPRON))
+		return 1;
+	return 0;
+}