[1.29.04] Allow up to 10 parts in multiple-word entries in *_list files.

lang=hr: Recognize Roman numbers only when all-capitals.
lang=en: Change to word-final [I] (phoneme [i]).


git-svn-id: https://espeak.svn.sourceforge.net/svnroot/espeak/trunk@84 d46cf337-b52f-0410-862d-fd96e6ae7743

src/compiledict.cpp

@@ -384,12 +384,19 @@ int compile_line(char *linebuf, char *dict_line, int *hash)
 	}
 	length += n_flag_codes;
 
-	if((multiple_string != NULL) && (multiple_words > 0) && (multiple_words <= 4))
+	if((multiple_string != NULL) && (multiple_words > 0))
 	{
-		dict_line[length++] = 40 + multiple_words;
-		ix = multiple_string_end - multiple_string;
-		memcpy(&dict_line[length],multiple_string,ix);
-		length += ix;
+		if(multiple_words > 10)
+		{
+			fprintf(f_log,"%5d: Two many parts in a multi-word entry: %d\n",linenum,multiple_words);
+		}
+		else
+		{
+			dict_line[length++] = 40 + multiple_words;
+			ix = multiple_string_end - multiple_string;
+			memcpy(&dict_line[length],multiple_string,ix);
+			length += ix;
+		}
 	}
 	dict_line[0] = length;
 

src/dictionary.cpp

@@ -36,7 +36,7 @@
 #include "translate.h"
 
 
-
+int dictionary_skipwords;
 char dictionary_name[40];
 
 extern MNEM_TAB mnem_flags[];
@@ -2716,12 +2716,13 @@ int Translator::LookupDict2(char *word, char *word2, char *phonetic, unsigned in
 			else
 			if(flag > 40)
 			{
-				// flags 41,42,or 43  match more than one word
+				// flags 41 to 50  match more than one word
 				// This comes after the other flags
 				n_chars = next - p;
 				if(memcmp(word2,p,n_chars)==0)
 				{
-					dictionary_flags |= ((flag-40) << 5);   // set (bits 5-7) to 1,2,or 3
+					dictionary_flags |= FLAG_SKIPWORDS;
+					dictionary_skipwords = (flag - 40);
 					p = next;
 					word_end = word2 + n_chars;
 				}

src/numbers.cpp

@@ -214,9 +214,6 @@ int Translator::TranslateRoman(char *word, char *ph_out)
 	static char *roman_numbers = "ixcmvld";
 	static int roman_values[] = {1,10,100,1000,5,50,500};
  
-	if((langopts.numbers & NUM_ROMAN) == 0)
-		return(0);
-
 	acc = 0;
 	prev = 0;
 	subtract = 0x7fff;

src/synthdata.cpp

@@ -35,7 +35,7 @@
 #include "translate.h"
 #include "wave.h"
 
-const char *version_string = "1.29.02  10.Sep.07";
+const char *version_string = "1.29.04  11.Sep.07";
 const int version_phdata  = 0x012901;
 
 int option_device_number = -1;
@@ -204,7 +204,7 @@ static unsigned int LookupSound2(int index, unsigned int other_phcode, int contr
 {//================================================================================
 // control=1  get formant transition data only
 
-	int code;
+	unsigned int code;
 	unsigned int value, value2;
 	
 	while((value = phoneme_index[index++]) != 0)
@@ -223,12 +223,17 @@ static unsigned int LookupSound2(int index, unsigned int other_phcode, int contr
 					}
 					break;
 				case 1:
-					if(control==0)
-						seq_len_adjust = value2 >> 8;
+					if(control==0)
+					{
+						seq_len_adjust = value2 >> 8;
+					}
 					break;
 				case 2:
-					if(control==0)
-						seq_len_adjust = -(value2 >> 8);
+					if(control==0)
+					{
+						seq_len_adjust = value2 >> 8;
+						seq_len_adjust = -seq_len_adjust;
+					}
 					break;
 				case 3:
 					if(control==0)

src/tr_languages.cpp

@@ -304,7 +304,7 @@ Translator *SelectTranslator(const char *name)
  			tr->langopts.max_initial_consonants = 5;
 			tr->langopts.spelling_stress = 1;
 
-			tr->langopts.numbers = 0x1c0d + 0x4000 + NUM_ROMAN;
+			tr->langopts.numbers = 0x1c0d + 0x4000 + NUM_ROMAN_UC;
 			tr->langopts.numbers2 = 0x4a;  // variant numbers before thousands,milliards
 			tr->langopts.replace_chars = replace_cyrillic;
 			tr->langopts.replacement_chars = replace_cyrillic_latin;

src/translate.cpp

@@ -671,8 +671,12 @@ if((wmark > 0) && (wmark < 8))
 		if(!found & ((word_flags & FLAG_UPPERS) != FLAG_FIRST_UPPER))
 		{
 			// either all upper or all lower case
-			if((found = TranslateRoman(word,phonemes)) != 0)
-				dictionary_flags |= FLAG_ABBREV;      // don't spell capital Roman numbers as individual letters
+
+			if((langopts.numbers & NUM_ROMAN) || ((langopts.numbers & NUM_ROMAN_UC) && (word_flags & FLAG_ALL_UPPER)))
+			{
+				if((found = TranslateRoman(word,phonemes)) != 0)
+					dictionary_flags |= FLAG_ABBREV;      // don't spell capital Roman numbers as individual letters
+			}
 		}
 
 		if((wflags & FLAG_ALL_UPPER) && (clause_upper_count <= clause_lower_count) &&
@@ -1233,7 +1237,8 @@ int Translator::TranslateWord2(char *word, WORD_TAB *wtab, int pre_pause, int ne
 				}
 				else
 				{
-					flags |= FLAG_SKIPWORDS_1;
+					flags |= FLAG_SKIPWORDS;
+					dictionary_skipwords = 1;
 				}
 			}
 		}
@@ -1562,7 +1567,7 @@ int Translator::TranslateChar(char *ptr, int prev_in, unsigned int c, unsigned i
 
 	int ix;
 	unsigned int word;
-	unsigned int new_c, c2;
+	unsigned int new_c, c2, c_lower;
 	int upper_case = 0;
 	static int ignore_next = 0;
 
@@ -1578,16 +1583,16 @@ int Translator::TranslateChar(char *ptr, int prev_in, unsigned int c, unsigned i
 
 	// there is a list of character codes to be substituted with alternative codes
 
-	if(iswupper(c))
+	if(iswupper(c_lower = c))
 	{
-		c = towlower(c);
+		c_lower = towlower(c);
 		upper_case = 1;
 	}
 
 	new_c = 0;
 	for(ix=0; (word = langopts.replace_chars[ix]) != 0; ix++)
 	{
-		if(c == (word & 0xffff))
+		if(c_lower == (word & 0xffff))
 		{
 			if((word >> 16) == 0)
 			{
@@ -2217,7 +2222,8 @@ if((c == '/') && (langopts.testing & 2) && isdigit(next_in) && IsAlpha(prev_out)
 		else
 		{
 			dict_flags = TranslateWord2(word, &words[ix], words[ix].pre_pause, words[ix+1].pre_pause);
-			ix += ((dict_flags >> 5) & 7);  // dictionary indicates skip next word(s)
+			if(dict_flags & FLAG_SKIPWORDS)
+				ix += dictionary_skipwords;  // dictionary indicates skip next word(s)
 
 			if((dict_flags & FLAG_DOT) && (ix == word_count-1) && (terminator == CLAUSE_PERIOD))
 			{

src/translate.h

@@ -34,8 +34,7 @@
 
 /* flags from word dictionary */
 // bits 0-3  stressed syllable,  7=unstressed
-#define FLAG_SKIPWORDS        0xe0  /* bits 5,6,7  number of words to skip */
-#define FLAG_SKIPWORDS_1      0x20
+#define FLAG_SKIPWORDS        0x80
 #define FLAG_PREPAUSE        0x100
 #define FLAG_ONLY            0x200
 #define BITNUM_FLAG_ONLY         9  // bit 9 is set
@@ -277,7 +276,8 @@ typedef struct {
 	unsigned char *length_mods;
 	unsigned char *length_mods0;
 
-#define NUM_ROMAN   0x20000
+#define NUM_ROMAN    0x20000
+#define NUM_ROMAN_UC 0x40000
 	// bits0-1=which numbers routine to use.
 	// bit2=  thousands separator must be space
 	// bit3=  , decimal separator, not .
@@ -293,6 +293,7 @@ typedef struct {
 	// bits13-15  post-decimal-digits 0=single digits, 1=(LANG=it) 2=(LANG=pl) 3=(LANG=ro)
 	// bit16=dot after number indicates ordinal
 	// bit17=recognize roman numbers
+	// bit18=Roman numbers only if upper case
 
 	int numbers;
 
@@ -501,6 +502,8 @@ extern char ctrl_embedded;    // to allow an alternative CTRL for embedded comma
 extern char *p_textinput;
 extern wchar_t *p_wchar_input;
 extern int ungot_char;
+extern int dictionary_skipwords;
+
 extern int (* uri_callback)(int, const char *, const char *);
 extern int (* phoneme_callback)(const char *);
 extern void SetLengthMods(Translator *tr, int value);