cst_val *en_exp_letters(const char *lets)
{
/* returns these as list of single char symbols */
char *aaa;
cst_val *r;
int i;
aaa = cst_alloc(char,2);
aaa[1] = '\0';
for (r=0,i=0; lets[i] != '\0'; i++)
{
aaa[0] = lets[i];
if (isupper((int)aaa[0]))
aaa[0] = tolower((int)aaa[0]);
if (strchr("0123456789",aaa[0]))
r = cons_val(string_val(digit2num[aaa[0]-'0']),r);
else if (cst_streq(aaa,"a"))
r = cons_val(string_val("_a"),r);
else
r = cons_val(string_val(aaa),r);
}
cst_free(aaa);
return val_reverse(r);
}
static cst_val *lex_lookup_addenda(const char *wp,const cst_lexicon *l,
int *found)
{
/* For those other words */
int i,j;
cst_val *phones;
phones = NULL;
for (i=0; l->addenda[i]; i++)
{
if (((wp[0] == '0') ||
(wp[0] == l->addenda[i][0][0]) ||
(l->addenda[i][0][0] == '0')) &&
(cst_streq(wp+1,l->addenda[i][0]+1)))
{
for (j=1; l->addenda[i][j]; j++)
phones = cons_val(string_val(l->addenda[i][j]),phones);
*found = TRUE;
return val_reverse(phones);
}
}
return NULL;
}
cst_val *get_wavelist(const char *wavelistfile)
{
cst_val *l = 0;
cst_tokenstream *ts;
const char *token;
int i=0;
ts = ts_open(wavelistfile);
if (!ts)
{
fprintf(stderr,"combine_waves: can't open \"%s\"\n",wavelistfile);
return 0;
}
while ((token=ts_get(ts)) != 0)
{
l = cons_val(string_val(token),l);
i++;
}
if (i%2 != 0)
{
fprintf(stderr,"combine_waves: doesn't have matched pairs \"%s\"\n",wavelistfile);
delete_val(l);
l = 0;
}
ts_close(ts);
return val_reverse(l);
}
cst_val *cst_args(char **argv, int argc,
const char *description,
cst_features *args)
{
/* parses the given arguments wrt the description */
cst_features *op_types = new_features();
cst_val *files = NULL;
int i;
const char *type;
parse_description(description,op_types);
for (i=1; i<argc; i++)
{
if (argv[i][0] == '-')
{
if ((!feat_present(op_types,argv[i])) ||
(cst_streq("-h",argv[i])) ||
(cst_streq("-?",argv[i])) ||
(cst_streq("--help",argv[i])) ||
(cst_streq("-help",argv[i])))
parse_usage(argv[0],"","",description);
else
{
type = feat_string(op_types,argv[i]);
if (cst_streq("<binary>",type))
feat_set_string(args,argv[i],"true");
else
{
if (i+1 == argc)
parse_usage(argv[0],
"missing argument for ",argv[i],
description);
if (cst_streq("<int>",type))
feat_set_int(args,argv[i],atoi(argv[i+1]));
else if (cst_streq("<float>",type))
feat_set_float(args,argv[i],atof(argv[i+1]));
else if (cst_streq("<string>",type))
feat_set_string(args,argv[i],argv[i+1]);
else
parse_usage(argv[0],
"unknown arg type ",type,
description);
i++;
}
}
}
else
files = cons_val(string_val(argv[i]),files);
}
delete_features(op_types);
return val_reverse(files);
}
cst_val *lex_lookup(const cst_lexicon *l, const char *word, const char *pos,
const cst_features *feats)
{
int index;
int p;
const char *q;
char *wp;
cst_val *phones = 0;
int found = FALSE;
wp = cst_alloc(char,cst_strlen(word)+2);
cst_sprintf(wp,"%c%s",(pos ? pos[0] : '0'),word);
if (l->addenda)
phones = lex_lookup_addenda(wp,l,&found);
if (!found)
{
index = lex_lookup_bsearch(l,wp);
if (index >= 0)
{
if (l->phone_hufftable)
{
for (p=index-2; l->data[p]; p--)
for (q=l->phone_hufftable[l->data[p]]; *q; q++)
phones = cons_val(string_val(l->phone_table[(unsigned char)*q]),
phones);
}
else /* no compression -- should we still support this ? */
{
for (p=index-2; l->data[p]; p--)
phones = cons_val(string_val(l->phone_table[l->data[p]]),
phones);
}
phones = val_reverse(phones);
}
else if (l->lts_function)
{
phones = (l->lts_function)(l,word,"",feats);
}
else if (l->lts_rule_set)
{
phones = lts_apply(word,
"", /* more features if we had them */
l->lts_rule_set);
}
}
cst_free(wp);
return phones;
}
static cst_val *add_lts_boundary_marks(const cst_val *l)
{
cst_val *l1;
const cst_val *v;
l1 = cons_val(string_val("#"),NULL);
for (v=l;v;v=val_cdr(v))
{
l1=cons_val(val_car(v),l1);
}
l1 = cons_val(string_val("#"),l1);
l1 = val_reverse(l1);
return l1;
}
cst_val *en_exp_ordinal(const char *rawnumstring)
{
/* return ordinal for digit string */
cst_val *card, *o;
const cst_val *t;
const char *l;
const char *ord;
char *numstring;
int i,j;
numstring = cst_strdup(rawnumstring);
for (j=i=0; i < cst_strlen(rawnumstring); i++)
if (rawnumstring[i] != ',')
{
numstring[j] = rawnumstring[i];
j++;
}
numstring[j] = '\0';
card = val_reverse(en_exp_number(numstring));
cst_free(numstring);
l = val_string(val_car(card));
ord = 0;
for (i=0; i<10; i++)
if (cst_streq(l,digit2num[i]))
ord = ord2num[i];
if (!ord)
for (i=0; i<10; i++)
if (cst_streq(l,digit2teen[i]))
ord = ord2teen[i];
if (!ord)
for (i=0; i<10; i++)
if (cst_streq(l,digit2enty[i]))
ord = ord2enty[i];
if (cst_streq(l,"hundred"))
ord = "hundredth";
if (cst_streq(l,"thousand"))
ord = "thousandth";
if (cst_streq(l,"billion"))
ord = "billtionth";
if (!ord) /* dunno, so don't convert anything */
return card;
o = cons_val(string_val(ord),0);
for (t=val_cdr(card); t; t=val_cdr(t))
o = cons_val(val_car(t),o);
delete_val(card);
return o;
}
cst_val *en_exp_digits(const char *numstring)
{
/* Expand given token to list of words pronouncing it as digits */
cst_val *d = 0;
const char *p;
for (p=numstring; *p; p++)
{
if ((*p >= '0') && (*p <= '9'))
d = cons_val(string_val(digit2num[*p-'0']),d);
else
d = cons_val(string_val("umpty"),d);
}
return val_reverse(d);
}
cst_val *cst_utf8_explode(const cst_string *utf8string)
{
/* return a list of utf-8 characters as strings */
const unsigned char *xxx = (const unsigned char *)utf8string;
cst_val *chars=NULL;
int i, l=0;
char utf8char[5];
#import "OpenEarsStaticAnalysisToggle.h"
#ifdef STATICANALYZEDEPENDENCIES
#define __clang_analyzer__ 1
#endif
#if !defined(__clang_analyzer__) || defined(STATICANALYZEDEPENDENCIES)
#undef __clang_analyzer__
for (i=0; xxx[i]; i++)
{
if (xxx[i] < 0x80) /* one byte */
{
sprintf(utf8char,"%c",xxx[i]);
l = 1;
}
else if (xxx[i] < 0xe0) /* two bytes */
{
sprintf(utf8char,"%c%c",xxx[i],xxx[i+1]);
i++;
l = 2;
}
else if (xxx[i] < 0xff) /* three bytes */
{
sprintf(utf8char,"%c%c%c",xxx[i],xxx[i+1],xxx[i+2]);
i++; i++;
l = 3;
}
else
{
sprintf(utf8char,"%c%c%c%c",xxx[i],xxx[i+1],xxx[i+2],xxx[i+3]);
i++; i++; i++;
l = 4;
}
chars = cons_val(string_val(utf8char),chars);
}
return val_reverse(chars);
#endif
}
cst_val *cmu_grapheme_lex_lts_function(const struct lexicon_struct *l,
const char *word, const char *pos,
const cst_features *feats)
{
cst_val *phones = 0;
cst_val *utflets = 0;
const cst_val *v;
char ord[10];
int i,phindex;
/* string to utf8 chars */
utflets = cst_utf8_explode(word);
for (v=utflets; v; v=val_cdr(v))
{
/* We will add the found phones in reverse order and reverse then */
/* afterwards */
cst_utf8_as_hex(val_string(val_car(v)),ord);
phindex = cst_find_u2sampa(ord);
if (phindex < 0)
printf("awb_debug no sampa %s %s\n",val_string(val_car(v)),ord);
for (i=4; (phindex>=0) && (i>0); i--)
{
if (unicode_sampa_mapping[phindex][i])
phones = cons_val(string_val(unicode_sampa_mapping[phindex][i]),
phones);
}
}
phones = val_reverse(phones);
#if 1
printf("cmu_grapheme_lex.c: word \"%s\" ",word);
val_print(stdout,phones);
printf("\n");
#endif
delete_val(utflets);
return phones;
}
cst_val *cst_lex_load_addenda(const cst_lexicon *lex, const char *lexfile)
{ /* Load an addend from given file, check its phones wrt lex */
cst_tokenstream *lf;
const cst_string *line;
cst_val *e = NULL;
cst_val *na = NULL;
int i;
lf = ts_open(lexfile,"\n","","","");
if (lf == NULL)
{
cst_errmsg("lex_add_addenda: cannot open lexicon file\n");
return NULL;;
}
while (!ts_eof(lf))
{
line = ts_get(lf);
if (line[0] == '#')
continue; /* a comment */
for (i=0; line[i]; i++)
{
if (line[i] != ' ')
break;
}
if (line[i])
{
e = cst_lex_make_entry(lex,line);
if (e)
na = cons_val(e,na);
}
else
continue; /* a blank line */
}
ts_close(lf);
return val_reverse(na);
}
cst_val *cst_utf8_explode(const cst_string *utf8string)
{
/* return a list of utf-8 characters as strings */
const unsigned char *xxx = (const unsigned char *)utf8string;
cst_val *chars=NULL;
int i, l=0;
char utf8char[5];
for (i=0; xxx[i]; i++)
{
if (xxx[i] < 0x80) /* one byte */
{
sprintf(utf8char,"%c",xxx[i]);
l = 1;
}
else if (xxx[i] < 0xe0) /* two bytes */
{
sprintf(utf8char,"%c%c",xxx[i],xxx[i+1]);
i++;
l = 2;
}
else if (xxx[i] < 0xff) /* three bytes */
{
sprintf(utf8char,"%c%c%c",xxx[i],xxx[i+1],xxx[i+2]);
i++; i++;
l = 3;
}
else
{
sprintf(utf8char,"%c%c%c%c",xxx[i],xxx[i+1],xxx[i+2],xxx[i+3]);
i++; i++; i++;
l = 4;
}
chars = cons_val(string_val(utf8char),chars);
}
return val_reverse(chars);
}
cst_val *lts_rewrites_word(const char *word, const cst_lts_rewrites *r)
{
cst_val *w, *p;
char x[2];
int i;
x[1] = '\0';
w = cons_val(string_val("#"),NULL);
for (i=0; word[i]; i++)
{
x[0] = word[i];
w = cons_val(string_val(x),w);
}
w = cons_val(string_val("#"),w);
w = val_reverse(w);
p = lts_rewrites(w,r);
delete_val(w);
return p;
}
cst_val *lts_rewrites(const cst_val *itape, const cst_lts_rewrites *r)
{
/* Returns list of rewritten "letters" to "phones" by r */
cst_val *LC;
const cst_val *RC, *i;
const cst_val *rule;
cst_val *otape;
LC = cons_val(val_car(itape),NULL);
RC = val_cdr(itape);
otape = NULL;
while (val_cdr(RC))
{
rule = find_rewrite_rule(LC,RC,r);
if (!rule)
break;
/* val_print(stdout,rule);
printf("\n"); */
/* Shift itape head */
for (i=val_car(val_cdr(rule)); i; i=val_cdr(i))
{
LC = cons_val(val_car(RC),LC);
RC = val_cdr(RC);
}
/* Output things to otape */
for (i=val_car(val_cdr(val_cdr(val_cdr(rule)))); i; i=val_cdr(i))
otape = cons_val(val_car(i),otape);
}
delete_val_list(LC);
return val_reverse(otape);
}
cst_val *cst_lex_make_entry(const cst_lexicon *lex, const cst_string *entry)
{ /* if replace then replace entry in addenda of lex with entry */
/* else append entry to addenda of lex */
cst_tokenstream *e;
cst_val *phones = NULL;
cst_val *ventry;
const cst_string *w, *p;
cst_string *word;
cst_string *pos;
int i;
e = ts_open_string(entry,
cst_ts_default_whitespacesymbols,
"","","");
w = ts_get(e);
if (w[0] == '"') /* it was a quoted entry */
{ /* so reparse it */
ts_close(e);
e = ts_open_string(entry,
cst_ts_default_whitespacesymbols,
"","","");
w = ts_get_quoted_token(e,'"','\\');
}
word = cst_strdup(w);
p = ts_get(e);
if (!cst_streq(":",p)) /* there is a real pos */
{
pos = cst_strdup(p);
p = ts_get(e);
if (!cst_streq(":",p)) /* there is a real pos */
{
cst_fprintf(stdout,"add_addenda: lex %s: expected \":\" in %s\n",
lex->name,
word);
cst_free(word);
cst_free(pos);
ts_close(e);
return NULL;
}
}
else
pos = cst_strdup("nil");
while (!ts_eof(e))
{
p = ts_get(e);
/* Check its a legal phone */
for (i=0; lex->phone_table[i]; i++)
{
if (cst_streq(p,lex->phone_table[i]))
break;
}
if (cst_streq("#",p)) /* comment to end of line */
break;
else if (cst_streq("",p)) /* trailing ws at eoln causes this */
break;
else if (lex->phone_table[i])
/* Only add it if its a valid phone */
phones = cons_val(string_val(p),phones);
else
{
cst_fprintf(stdout,"add_addenda: lex: %s word %s phone %s not in lexicon phoneset\n",
lex->name,
word,
p);
}
}
ventry = cons_val(string_val(word),cons_val(string_val(pos),
val_reverse(phones)));
cst_free(word);
cst_free(pos);
ts_close(e);
#if 0
printf("entry: ");
val_print(stdout,ventry);
printf("\n");
#endif
return ventry;
}
cst_val *lex_lookup_return_pos(const cst_lexicon *l, const char *word, const char *pos, int *return_pos)
{
int index,p;
char *wp, *buf;
cst_val *phones = 0;
int found = FALSE;
wp = cst_alloc(char,strlen(word)+2);
buf = cst_alloc(char,3);
sprintf(wp,"%c%s",(pos ? pos[0] : '0'),word);
#ifdef CECUM_DEBUG
printf ("lex_lookup_return_pos looks for %s \n", word);
#endif
if (l->addenda)
phones = CECUM_lex_lookup_addenda (wp,l,&found);
if (!found)
{
index = CECUM_lex_lookup_bsearch (l, l->entry_index,0,l->num_entries,wp);
#ifdef CECUM_DEBUG
printf ("lex_lookup_return_pos has found the word at pos %d \n", index);
#endif
if (index >= 0)
{
for (p=l->entry_index[index].phone_index; l->phones[p]; p++)
phones = cons_val(string_val(l->phone_table[l->phones[p]]), phones);
phones = val_reverse(phones);
if (!pos)
{
//aggiungi in testa il pos se non lo ho!
*return_pos = TRUE;
sprintf(buf,"%c",l->entry_index[index].word_pos[0]);
phones = cons_val(string_val(buf), phones);
}
//val_print(stdout,phones);
//printf("\n");
}
else
{
#ifdef CECUM_DEBUG
printf ("lex_lookup_return_pos: word not found \n", index);
#endif
/*
E' a partire da questo punto che si verifica in genere il problema su cui stavamo
lavorando. Se la parola non viene trovata nel dizionario, viene restituito un indice -1.
Ora, lts_rules_set (e lts_apply) non viene eseguita perchè in ifd_lex.c noi abbiamo
posto
ifd_lex.lts_rule_set = 0;
Del resto, nella lingua italiana non sembrano essere fornite rules compatibili con lst_apply,
a differenza di quello che accade con la lingua inglese
Il puntatore a funzione l->lst_function viene, al contrario, eseguito. Qui la funzione
richiamata è italian_lts_function() (dentro ifd_lex.c). Sarà quest'ultima a richiamare
più volte una serie di routine di rewrites (è qui che si verifica l'errore su cui stiamo
lavorando). Per esempio, è qui che verrà richiamata ifd_mid_lts_rewrites ()
*/
if (l->lts_rule_set)
phones = lts_apply(word,
"", /* more features if we had them */
l->lts_rule_set);
else if (l->lts_function)
phones = (l->lts_function)(l,word,"");
}
}
cst_free(wp);
cst_free(buf);
return phones;
}