ruby-sfst 0.1.0
Sign up to get free protection for your applications and to get access to all the features.
- data/CHANGELOG +1 -0
- data/Manifest +31 -0
- data/README.rdoc +25 -0
- data/Rakefile +22 -0
- data/ext/sfst_machine/alphabet.C +807 -0
- data/ext/sfst_machine/alphabet.h +281 -0
- data/ext/sfst_machine/basic.C +84 -0
- data/ext/sfst_machine/basic.h +24 -0
- data/ext/sfst_machine/compact.C +616 -0
- data/ext/sfst_machine/compact.h +98 -0
- data/ext/sfst_machine/determinise.C +304 -0
- data/ext/sfst_machine/extconf.rb +4 -0
- data/ext/sfst_machine/fst-compiler.C +2375 -0
- data/ext/sfst_machine/fst-compiler.h +113 -0
- data/ext/sfst_machine/fst-compiler.yy +213 -0
- data/ext/sfst_machine/fst.C +966 -0
- data/ext/sfst_machine/fst.h +365 -0
- data/ext/sfst_machine/interface.C +1838 -0
- data/ext/sfst_machine/interface.h +94 -0
- data/ext/sfst_machine/make-compact.C +328 -0
- data/ext/sfst_machine/make-compact.h +34 -0
- data/ext/sfst_machine/mem.h +74 -0
- data/ext/sfst_machine/operators.C +1131 -0
- data/ext/sfst_machine/sfst_machine.cc +411 -0
- data/ext/sfst_machine/utf8-scanner.C +2197 -0
- data/ext/sfst_machine/utf8-scanner.ll +179 -0
- data/ext/sfst_machine/utf8.C +146 -0
- data/ext/sfst_machine/utf8.h +19 -0
- data/lib/sfst.rb +99 -0
- data/ruby-sfst.gemspec +34 -0
- data/test/test_sfst.fst +3 -0
- data/test/test_sfst.rb +119 -0
- metadata +100 -0
@@ -0,0 +1,966 @@
|
|
1
|
+
|
2
|
+
/*******************************************************************/
|
3
|
+
/* */
|
4
|
+
/* FILE fst.C */
|
5
|
+
/* MODULE fst */
|
6
|
+
/* PROGRAM SFST */
|
7
|
+
/* AUTHOR Helmut Schmid, IMS, University of Stuttgart */
|
8
|
+
/* */
|
9
|
+
/* PURPOSE basic FST functions */
|
10
|
+
/* */
|
11
|
+
/*******************************************************************/
|
12
|
+
|
13
|
+
#include "fst.h"
|
14
|
+
|
15
|
+
using std::vector;
|
16
|
+
using std::istream;
|
17
|
+
using std::ostream;
|
18
|
+
using std::cerr;
|
19
|
+
|
20
|
+
const int BUFFER_SIZE=100000;
|
21
|
+
|
22
|
+
|
23
|
+
/*******************************************************************/
|
24
|
+
/* */
|
25
|
+
/* Arcs::size */
|
26
|
+
/* */
|
27
|
+
/*******************************************************************/
|
28
|
+
|
29
|
+
int Arcs::size() const
|
30
|
+
|
31
|
+
{
|
32
|
+
int n=0;
|
33
|
+
for( Arc *p=first_arcp; p; p=p->next ) n++;
|
34
|
+
for( Arc *p=first_epsilon_arcp; p; p=p->next ) n++;
|
35
|
+
return n;
|
36
|
+
}
|
37
|
+
|
38
|
+
|
39
|
+
/*******************************************************************/
|
40
|
+
/* */
|
41
|
+
/* Arcs::target_node */
|
42
|
+
/* */
|
43
|
+
/*******************************************************************/
|
44
|
+
|
45
|
+
Node *Arcs::target_node( Label l )
|
46
|
+
|
47
|
+
{
|
48
|
+
Arc *arc;
|
49
|
+
|
50
|
+
for( arc=first_arcp; arc; arc=arc->next)
|
51
|
+
if (arc->label() == l)
|
52
|
+
return arc->target_node();
|
53
|
+
|
54
|
+
return NULL;
|
55
|
+
}
|
56
|
+
|
57
|
+
const Node *Arcs::target_node( Label l ) const
|
58
|
+
|
59
|
+
{
|
60
|
+
const Arc *arc;
|
61
|
+
|
62
|
+
for( arc=first_arcp; arc; arc=arc->next)
|
63
|
+
if (arc->label() == l)
|
64
|
+
return arc->target_node();
|
65
|
+
|
66
|
+
return NULL;
|
67
|
+
}
|
68
|
+
|
69
|
+
|
70
|
+
/*******************************************************************/
|
71
|
+
/* */
|
72
|
+
/* Arcs::add_arc */
|
73
|
+
/* */
|
74
|
+
/*******************************************************************/
|
75
|
+
|
76
|
+
void Arcs::add_arc( Label l, Node *node, Transducer *a )
|
77
|
+
|
78
|
+
{
|
79
|
+
Arc *arc=a->new_arc( l, node );
|
80
|
+
|
81
|
+
if (l.is_epsilon()) {
|
82
|
+
arc->next = first_epsilon_arcp;
|
83
|
+
first_epsilon_arcp = arc;
|
84
|
+
}
|
85
|
+
else {
|
86
|
+
arc->next = first_arcp;
|
87
|
+
first_arcp = arc;
|
88
|
+
}
|
89
|
+
}
|
90
|
+
|
91
|
+
|
92
|
+
/*******************************************************************/
|
93
|
+
/* */
|
94
|
+
/* Arcs::remove_arc */
|
95
|
+
/* */
|
96
|
+
/*******************************************************************/
|
97
|
+
|
98
|
+
int Arcs::remove_arc( Arc *arc )
|
99
|
+
|
100
|
+
{
|
101
|
+
Arc **p = (arc->label().is_epsilon()) ? &first_epsilon_arcp : &first_arcp;
|
102
|
+
for( ; *p; p=&(*p)->next )
|
103
|
+
if (*p == arc) {
|
104
|
+
*p = arc->next;
|
105
|
+
return 1;
|
106
|
+
}
|
107
|
+
return 0;
|
108
|
+
}
|
109
|
+
|
110
|
+
|
111
|
+
/*******************************************************************/
|
112
|
+
/* */
|
113
|
+
/* Node::init */
|
114
|
+
/* */
|
115
|
+
/*******************************************************************/
|
116
|
+
|
117
|
+
void Node::init()
|
118
|
+
|
119
|
+
{
|
120
|
+
final = false;
|
121
|
+
visited = 0;
|
122
|
+
arcsp.init();
|
123
|
+
forwardp = NULL;
|
124
|
+
}
|
125
|
+
|
126
|
+
|
127
|
+
/*******************************************************************/
|
128
|
+
/* */
|
129
|
+
/* NodeNumbering::number_node */
|
130
|
+
/* */
|
131
|
+
/*******************************************************************/
|
132
|
+
|
133
|
+
void NodeNumbering::number_node( Node *node, Transducer &a )
|
134
|
+
|
135
|
+
{
|
136
|
+
if (!node->was_visited( a.vmark )) {
|
137
|
+
nummap[node] = nodes.size();
|
138
|
+
nodes.push_back(node);
|
139
|
+
for( ArcsIter p(node->arcs()); p; p++ ) {
|
140
|
+
Arc *arc=p;
|
141
|
+
number_node( arc->target_node(), a );
|
142
|
+
}
|
143
|
+
}
|
144
|
+
}
|
145
|
+
|
146
|
+
|
147
|
+
/*******************************************************************/
|
148
|
+
/* */
|
149
|
+
/* NodeNumbering::NodeNumbering */
|
150
|
+
/* */
|
151
|
+
/*******************************************************************/
|
152
|
+
|
153
|
+
NodeNumbering::NodeNumbering( Transducer &a )
|
154
|
+
|
155
|
+
{
|
156
|
+
a.incr_vmark();
|
157
|
+
number_node( a.root_node(), a );
|
158
|
+
}
|
159
|
+
|
160
|
+
|
161
|
+
/*******************************************************************/
|
162
|
+
/* */
|
163
|
+
/* Transducer::new_node */
|
164
|
+
/* */
|
165
|
+
/*******************************************************************/
|
166
|
+
|
167
|
+
Node *Transducer::new_node()
|
168
|
+
|
169
|
+
{
|
170
|
+
Node *node=(Node*)mem.alloc( sizeof(Node) );
|
171
|
+
|
172
|
+
node->init();
|
173
|
+
return node;
|
174
|
+
}
|
175
|
+
|
176
|
+
|
177
|
+
/*******************************************************************/
|
178
|
+
/* */
|
179
|
+
/* Transducer::new_arc */
|
180
|
+
/* */
|
181
|
+
/*******************************************************************/
|
182
|
+
|
183
|
+
Arc *Transducer::new_arc( Label l, Node *target )
|
184
|
+
|
185
|
+
{
|
186
|
+
Arc *arc=(Arc*)mem.alloc( sizeof(Arc) );
|
187
|
+
|
188
|
+
arc->init( l, target);
|
189
|
+
return arc;
|
190
|
+
}
|
191
|
+
|
192
|
+
|
193
|
+
/*******************************************************************/
|
194
|
+
/* */
|
195
|
+
/* Transducer::add_string */
|
196
|
+
/* */
|
197
|
+
/*******************************************************************/
|
198
|
+
|
199
|
+
void Transducer::add_string( char *s, bool extended )
|
200
|
+
|
201
|
+
{
|
202
|
+
Node *node=root_node();
|
203
|
+
|
204
|
+
Label l;
|
205
|
+
while (!(l = alphabet.next_label(s, extended)).is_epsilon()) {
|
206
|
+
alphabet.insert(l);
|
207
|
+
Arcs *arcs=node->arcs();
|
208
|
+
node = arcs->target_node( l );
|
209
|
+
if (node == NULL) {
|
210
|
+
node = new_node();
|
211
|
+
arcs->add_arc( l, node, this );
|
212
|
+
}
|
213
|
+
}
|
214
|
+
node->set_final(1);
|
215
|
+
}
|
216
|
+
|
217
|
+
|
218
|
+
/*******************************************************************/
|
219
|
+
/* */
|
220
|
+
/* Transducer::Transducer */
|
221
|
+
/* */
|
222
|
+
/*******************************************************************/
|
223
|
+
|
224
|
+
Transducer::Transducer( vector<Label> &path )
|
225
|
+
: root(), mem()
|
226
|
+
{
|
227
|
+
Node *node=root_node();
|
228
|
+
|
229
|
+
vmark = 0;
|
230
|
+
deterministic = minimised = true;
|
231
|
+
for( size_t i=0; i<path.size(); i++ ) {
|
232
|
+
Arcs *arcs=node->arcs();
|
233
|
+
node = new_node();
|
234
|
+
arcs->add_arc( path[i], node, this );
|
235
|
+
}
|
236
|
+
node->set_final(1);
|
237
|
+
}
|
238
|
+
|
239
|
+
|
240
|
+
/*******************************************************************/
|
241
|
+
/* */
|
242
|
+
/* Transducer::Transducer */
|
243
|
+
/* */
|
244
|
+
/*******************************************************************/
|
245
|
+
|
246
|
+
Transducer::Transducer( istream &is, const Alphabet *a, bool verbose )
|
247
|
+
: root(), mem()
|
248
|
+
{
|
249
|
+
bool extended=false;
|
250
|
+
int n=0;
|
251
|
+
char buffer[10000];
|
252
|
+
|
253
|
+
vmark = 0;
|
254
|
+
deterministic = true;
|
255
|
+
minimised = false;
|
256
|
+
if (a) {
|
257
|
+
alphabet.copy(*a);
|
258
|
+
extended = true;
|
259
|
+
}
|
260
|
+
while (is.getline(buffer, 10000)) {
|
261
|
+
if (verbose && ++n % 10000 == 0) {
|
262
|
+
if (n == 10000)
|
263
|
+
cerr << "\n";
|
264
|
+
cerr << "\r" << n << " words";
|
265
|
+
}
|
266
|
+
// delete final whitespace characters
|
267
|
+
int l;
|
268
|
+
for( l=strlen(buffer)-1; l>=0; l-- )
|
269
|
+
if ((buffer[l] != ' ' && buffer[l] != '\t' && buffer[l] != '\r') ||
|
270
|
+
(l > 0 && buffer[l-1] == '\\'))
|
271
|
+
break;
|
272
|
+
buffer[l+1] = 0;
|
273
|
+
|
274
|
+
add_string(buffer,extended);
|
275
|
+
}
|
276
|
+
if (verbose && n >= 10000)
|
277
|
+
cerr << "\n";
|
278
|
+
}
|
279
|
+
|
280
|
+
|
281
|
+
/*******************************************************************/
|
282
|
+
/* */
|
283
|
+
/* Transducer::Transducer */
|
284
|
+
/* */
|
285
|
+
/*******************************************************************/
|
286
|
+
|
287
|
+
Transducer::Transducer( char *s, const Alphabet *a, bool extended )
|
288
|
+
: root(), mem()
|
289
|
+
{
|
290
|
+
vmark = 0;
|
291
|
+
deterministic = minimised = true;
|
292
|
+
if (a)
|
293
|
+
alphabet.copy(*a);
|
294
|
+
add_string(s, extended);
|
295
|
+
}
|
296
|
+
|
297
|
+
|
298
|
+
/*******************************************************************/
|
299
|
+
/* */
|
300
|
+
/* Transducer::clear */
|
301
|
+
/* */
|
302
|
+
/*******************************************************************/
|
303
|
+
|
304
|
+
void Transducer::clear()
|
305
|
+
|
306
|
+
{
|
307
|
+
vmark = 0;
|
308
|
+
deterministic = minimised = false;
|
309
|
+
root.init();
|
310
|
+
mem.clear();
|
311
|
+
alphabet.clear();
|
312
|
+
}
|
313
|
+
|
314
|
+
|
315
|
+
/*******************************************************************/
|
316
|
+
/* */
|
317
|
+
/* Transducer::store_symbols */
|
318
|
+
/* */
|
319
|
+
/*******************************************************************/
|
320
|
+
|
321
|
+
void Transducer::store_symbols(Node *node, SymbolMap &symbol, LabelSet &labels)
|
322
|
+
|
323
|
+
{
|
324
|
+
if (!node->was_visited( vmark )) {
|
325
|
+
Arcs *arcs=node->arcs();
|
326
|
+
for( ArcsIter p(arcs); p; p++ ) {
|
327
|
+
Arc *arc=p;
|
328
|
+
Label l=arc->label();
|
329
|
+
|
330
|
+
labels.insert(l);
|
331
|
+
|
332
|
+
Character c = l.upper_char();
|
333
|
+
if (symbol.find(c) == symbol.end()) {
|
334
|
+
const char *s = alphabet.code2symbol(c);
|
335
|
+
if (s)
|
336
|
+
symbol[c] = fst_strdup(s);
|
337
|
+
}
|
338
|
+
|
339
|
+
c = l.lower_char();
|
340
|
+
if (symbol.find(c) == symbol.end()) {
|
341
|
+
const char *s = alphabet.code2symbol(c);
|
342
|
+
if (s)
|
343
|
+
symbol[c] = fst_strdup(s);
|
344
|
+
}
|
345
|
+
|
346
|
+
store_symbols( arc->target_node(), symbol, labels );
|
347
|
+
}
|
348
|
+
}
|
349
|
+
}
|
350
|
+
|
351
|
+
|
352
|
+
/*******************************************************************/
|
353
|
+
/* */
|
354
|
+
/* Transducer::minimise_alphabet */
|
355
|
+
/* */
|
356
|
+
/*******************************************************************/
|
357
|
+
|
358
|
+
void Transducer::minimise_alphabet()
|
359
|
+
|
360
|
+
{
|
361
|
+
SymbolMap symbols;
|
362
|
+
LabelSet labels;
|
363
|
+
incr_vmark();
|
364
|
+
store_symbols(root_node(), symbols, labels);
|
365
|
+
alphabet.clear();
|
366
|
+
for( SymbolMap::iterator it=symbols.begin(); it!=symbols.end(); it++ ) {
|
367
|
+
alphabet.add_symbol( it->second, it->first );
|
368
|
+
free(it->second);
|
369
|
+
}
|
370
|
+
for( LabelSet::iterator it=labels.begin(); it!=labels.end(); it++ )
|
371
|
+
alphabet.insert(*it);
|
372
|
+
}
|
373
|
+
|
374
|
+
|
375
|
+
/*******************************************************************/
|
376
|
+
/* */
|
377
|
+
/* Transducer::minimise */
|
378
|
+
/* */
|
379
|
+
/*******************************************************************/
|
380
|
+
|
381
|
+
Transducer &Transducer::minimise( bool verbose )
|
382
|
+
|
383
|
+
{
|
384
|
+
if (minimised)
|
385
|
+
return copy();
|
386
|
+
|
387
|
+
Transducer *a1, *a2;
|
388
|
+
|
389
|
+
a1 = &reverse();
|
390
|
+
a2 = &a1->determinise();
|
391
|
+
delete a1;
|
392
|
+
|
393
|
+
a1 = &a2->reverse();
|
394
|
+
delete a2;
|
395
|
+
|
396
|
+
a2 = &a1->determinise();
|
397
|
+
delete a1;
|
398
|
+
|
399
|
+
a2->minimised = true;
|
400
|
+
a2->minimise_alphabet();
|
401
|
+
|
402
|
+
return *a2;
|
403
|
+
}
|
404
|
+
|
405
|
+
|
406
|
+
/*******************************************************************/
|
407
|
+
/* */
|
408
|
+
/* Transducer::enumerate_paths_node */
|
409
|
+
/* */
|
410
|
+
/*******************************************************************/
|
411
|
+
|
412
|
+
void Transducer::enumerate_paths_node( Node *node, vector<Label> &path,
|
413
|
+
NodeHashSet &previous,
|
414
|
+
vector<Transducer*> &result )
|
415
|
+
{
|
416
|
+
if (node->is_final())
|
417
|
+
result.push_back(new Transducer(path));
|
418
|
+
|
419
|
+
for( ArcsIter it(node->arcs()); it; it++ ) {
|
420
|
+
Arc *arc=it;
|
421
|
+
|
422
|
+
NodeHashSet::iterator it=previous.insert(node).first;
|
423
|
+
path.push_back(arc->label());
|
424
|
+
enumerate_paths_node( arc->target_node(), path, previous, result );
|
425
|
+
path.pop_back();
|
426
|
+
previous.erase(it);
|
427
|
+
}
|
428
|
+
}
|
429
|
+
|
430
|
+
|
431
|
+
/*******************************************************************/
|
432
|
+
/* */
|
433
|
+
/* Transducer::enumerate_paths */
|
434
|
+
/* */
|
435
|
+
/*******************************************************************/
|
436
|
+
|
437
|
+
bool Transducer::enumerate_paths( vector<Transducer*> &result )
|
438
|
+
|
439
|
+
{
|
440
|
+
if (is_infinitely_ambiguous())
|
441
|
+
return true;
|
442
|
+
for( size_t i=0; i<result.size(); i++ )
|
443
|
+
delete result[i];
|
444
|
+
result.clear();
|
445
|
+
|
446
|
+
vector<Label> path;
|
447
|
+
NodeHashSet previous;
|
448
|
+
enumerate_paths_node( root_node(), path, previous, result );
|
449
|
+
return false;
|
450
|
+
}
|
451
|
+
|
452
|
+
|
453
|
+
|
454
|
+
|
455
|
+
/*******************************************************************/
|
456
|
+
/* */
|
457
|
+
/* Transducer::print_strings_node */
|
458
|
+
/* */
|
459
|
+
/*******************************************************************/
|
460
|
+
|
461
|
+
int Transducer::print_strings_node(Node *node, char *buffer, int pos,
|
462
|
+
FILE *file, bool with_brackets )
|
463
|
+
{
|
464
|
+
int result = 0;
|
465
|
+
|
466
|
+
if (node->was_visited( vmark )) {
|
467
|
+
if (node->forward() != NULL) { // cycle detected
|
468
|
+
cerr << "Warning: cyclic analyses (cycle aborted)\n";
|
469
|
+
return 0;
|
470
|
+
}
|
471
|
+
node->set_forward(node); // used like a flag for loop detection
|
472
|
+
}
|
473
|
+
if (pos == BUFFER_SIZE)
|
474
|
+
throw "Output string in function print_strings_node is too long";
|
475
|
+
if (node->is_final()) {
|
476
|
+
buffer[pos] = '\0';
|
477
|
+
fprintf(file,"%s\n", buffer);
|
478
|
+
result = 1;
|
479
|
+
}
|
480
|
+
for( ArcsIter i(node->arcs()); i; i++ ) {
|
481
|
+
int p=pos;
|
482
|
+
Arc *arc=i;
|
483
|
+
Label l=arc->label();
|
484
|
+
alphabet.write_label(l, buffer, &p, with_brackets);
|
485
|
+
result |= print_strings_node(arc->target_node(), buffer, p,
|
486
|
+
file, with_brackets );
|
487
|
+
}
|
488
|
+
node->set_forward(NULL);
|
489
|
+
|
490
|
+
return result;
|
491
|
+
}
|
492
|
+
|
493
|
+
|
494
|
+
/*******************************************************************/
|
495
|
+
/* */
|
496
|
+
/* Transducer::print_strings */
|
497
|
+
/* */
|
498
|
+
/*******************************************************************/
|
499
|
+
|
500
|
+
int Transducer::print_strings( FILE *file, bool with_brackets )
|
501
|
+
|
502
|
+
{
|
503
|
+
char buffer[BUFFER_SIZE];
|
504
|
+
incr_vmark();
|
505
|
+
return print_strings_node( root_node(), buffer, 0, file, with_brackets );
|
506
|
+
}
|
507
|
+
|
508
|
+
|
509
|
+
/*******************************************************************/
|
510
|
+
/* */
|
511
|
+
/* Transducer::analyze_string */
|
512
|
+
/* */
|
513
|
+
/*******************************************************************/
|
514
|
+
|
515
|
+
bool Transducer::analyze_string( char *string, FILE *file, bool with_brackets )
|
516
|
+
|
517
|
+
{
|
518
|
+
Transducer a1(string, &alphabet, false);
|
519
|
+
Transducer *a2=&(*this || a1);
|
520
|
+
Transducer *a3=&(a2->lower_level());
|
521
|
+
delete a2;
|
522
|
+
a2 = &a3->minimise();
|
523
|
+
delete a3;
|
524
|
+
|
525
|
+
a2->alphabet.copy(alphabet);
|
526
|
+
bool result = a2->print_strings( file, with_brackets );
|
527
|
+
delete a2;
|
528
|
+
return result;
|
529
|
+
}
|
530
|
+
|
531
|
+
|
532
|
+
/*******************************************************************/
|
533
|
+
/* */
|
534
|
+
/* Transducer::generate_string */
|
535
|
+
/* */
|
536
|
+
/*******************************************************************/
|
537
|
+
|
538
|
+
bool Transducer::generate_string( char *string, FILE *file, bool with_brackets)
|
539
|
+
|
540
|
+
{
|
541
|
+
Transducer a1(string, &alphabet, false);
|
542
|
+
Transducer *a2=&(a1 || *this);
|
543
|
+
Transducer *a3=&(a2->upper_level());
|
544
|
+
delete a2;
|
545
|
+
a2 = &a3->minimise();
|
546
|
+
delete a3;
|
547
|
+
|
548
|
+
a2->alphabet.copy(alphabet);
|
549
|
+
bool result = a2->print_strings( file, with_brackets );
|
550
|
+
delete a2;
|
551
|
+
return result;
|
552
|
+
}
|
553
|
+
|
554
|
+
|
555
|
+
/*******************************************************************/
|
556
|
+
/* */
|
557
|
+
/* complete */
|
558
|
+
/* */
|
559
|
+
/*******************************************************************/
|
560
|
+
|
561
|
+
static void complete( Node *node, Alphabet &alphabet, int vmark)
|
562
|
+
|
563
|
+
{
|
564
|
+
if (node->was_visited( vmark ))
|
565
|
+
return;
|
566
|
+
for( ArcsIter p(node->arcs()); p; p++ ) {
|
567
|
+
Arc *arc=p;
|
568
|
+
if (!arc->label().is_epsilon())
|
569
|
+
alphabet.insert(arc->label());
|
570
|
+
complete(arc->target_node(), alphabet, vmark);
|
571
|
+
}
|
572
|
+
}
|
573
|
+
|
574
|
+
|
575
|
+
/*******************************************************************/
|
576
|
+
/* */
|
577
|
+
/* Transducer::complete_alphabet */
|
578
|
+
/* */
|
579
|
+
/*******************************************************************/
|
580
|
+
|
581
|
+
void Transducer::complete_alphabet()
|
582
|
+
|
583
|
+
{
|
584
|
+
incr_vmark();
|
585
|
+
complete(root_node(), alphabet, vmark);
|
586
|
+
}
|
587
|
+
|
588
|
+
|
589
|
+
/*******************************************************************/
|
590
|
+
/* */
|
591
|
+
/* print_node */
|
592
|
+
/* */
|
593
|
+
/*******************************************************************/
|
594
|
+
|
595
|
+
static void print_node( ostream &s, Node *node, NodeNumbering &index,
|
596
|
+
long vmark, Alphabet &abc )
|
597
|
+
|
598
|
+
{
|
599
|
+
if (!node->was_visited( vmark )) {
|
600
|
+
Arcs *arcs=node->arcs();
|
601
|
+
if (node->is_final())
|
602
|
+
s << "final\t" << index[node] << "\n";
|
603
|
+
for( ArcsIter p(arcs); p; p++ ) {
|
604
|
+
Arc *arc=p;
|
605
|
+
s << index[node] << "\t";
|
606
|
+
s << abc.write_label(arc->label()) << "\t";
|
607
|
+
s << index[arc->target_node()] << "\n";
|
608
|
+
}
|
609
|
+
for( ArcsIter p(arcs); p; p++ ) {
|
610
|
+
Arc *arc=p;
|
611
|
+
print_node( s, arc->target_node(), index, vmark, abc );
|
612
|
+
}
|
613
|
+
}
|
614
|
+
}
|
615
|
+
|
616
|
+
|
617
|
+
/*******************************************************************/
|
618
|
+
/* */
|
619
|
+
/* operator<< */
|
620
|
+
/* */
|
621
|
+
/*******************************************************************/
|
622
|
+
|
623
|
+
ostream &operator<<( ostream &s, Transducer &a )
|
624
|
+
|
625
|
+
{
|
626
|
+
NodeNumbering index(a);
|
627
|
+
a.incr_vmark();
|
628
|
+
print_node( s, a.root_node(), index, a.vmark, a.alphabet );
|
629
|
+
return s;
|
630
|
+
}
|
631
|
+
|
632
|
+
|
633
|
+
/*******************************************************************/
|
634
|
+
/* */
|
635
|
+
/* store_node_info */
|
636
|
+
/* */
|
637
|
+
/*******************************************************************/
|
638
|
+
|
639
|
+
static void store_node_info( FILE *file, Node *node )
|
640
|
+
|
641
|
+
{
|
642
|
+
// write final flag
|
643
|
+
char c=node->is_final();
|
644
|
+
fwrite(&c,sizeof(c),1,file);
|
645
|
+
|
646
|
+
// write the number of arcs
|
647
|
+
int nn = node->arcs()->size();
|
648
|
+
if (nn > 65535)
|
649
|
+
throw "Error: in function store_node\n";
|
650
|
+
unsigned short n=(unsigned short)nn;
|
651
|
+
fwrite(&n,sizeof(n),1,file);
|
652
|
+
}
|
653
|
+
|
654
|
+
|
655
|
+
/*******************************************************************/
|
656
|
+
/* */
|
657
|
+
/* store_arc_label */
|
658
|
+
/* */
|
659
|
+
/*******************************************************************/
|
660
|
+
|
661
|
+
static void store_arc_label( FILE *file, Arc *arc )
|
662
|
+
|
663
|
+
{
|
664
|
+
Label l=arc->label();
|
665
|
+
Character lc=l.lower_char();
|
666
|
+
Character uc=l.upper_char();
|
667
|
+
fwrite(&lc,sizeof(lc),1,file);
|
668
|
+
fwrite(&uc,sizeof(uc),1,file);
|
669
|
+
}
|
670
|
+
|
671
|
+
|
672
|
+
/*******************************************************************/
|
673
|
+
/* */
|
674
|
+
/* store_node */
|
675
|
+
/* */
|
676
|
+
/*******************************************************************/
|
677
|
+
|
678
|
+
static void store_node( FILE *file, Node *node, NodeNumbering &index,
|
679
|
+
long vmark )
|
680
|
+
{
|
681
|
+
if (!node->was_visited( vmark )) {
|
682
|
+
|
683
|
+
store_node_info( file, node );
|
684
|
+
|
685
|
+
// write the arcs
|
686
|
+
for( ArcsIter p(node->arcs()); p; p++ ) {
|
687
|
+
Arc *arc=p;
|
688
|
+
store_arc_label( file, arc );
|
689
|
+
unsigned int t=index[arc->target_node()];
|
690
|
+
fwrite(&t,sizeof(t),1,file);
|
691
|
+
store_node(file, arc->target_node(), index, vmark );
|
692
|
+
}
|
693
|
+
}
|
694
|
+
}
|
695
|
+
|
696
|
+
|
697
|
+
/*******************************************************************/
|
698
|
+
/* */
|
699
|
+
/* store_lowmem_node */
|
700
|
+
/* */
|
701
|
+
/*******************************************************************/
|
702
|
+
|
703
|
+
static void store_lowmem_node( FILE *file, Node *node, NodeNumbering &index,
|
704
|
+
vector<unsigned int> &startpos)
|
705
|
+
{
|
706
|
+
store_node_info( file, node );
|
707
|
+
|
708
|
+
// write the arcs
|
709
|
+
for( ArcsIter p(node->arcs()); p; p++ ) {
|
710
|
+
Arc *arc=p;
|
711
|
+
store_arc_label( file, arc );
|
712
|
+
unsigned int t=startpos[index[arc->target_node()]];
|
713
|
+
fwrite(&t,sizeof(t),1,file);
|
714
|
+
}
|
715
|
+
}
|
716
|
+
|
717
|
+
|
718
|
+
/*******************************************************************/
|
719
|
+
/* */
|
720
|
+
/* Transducer::store_lowmem */
|
721
|
+
/* */
|
722
|
+
/*******************************************************************/
|
723
|
+
|
724
|
+
void Transducer::store_lowmem( FILE *file )
|
725
|
+
|
726
|
+
{
|
727
|
+
fputc('l',file);
|
728
|
+
alphabet.store(file);
|
729
|
+
|
730
|
+
// storing size of index table
|
731
|
+
NodeNumbering index(*this);
|
732
|
+
|
733
|
+
// compute the start position of the first node
|
734
|
+
unsigned int pos=(unsigned int)ftell(file);
|
735
|
+
vector<unsigned int> startpos;
|
736
|
+
for( size_t i=0; i<index.number_of_nodes(); i++ ) {
|
737
|
+
startpos.push_back(pos);
|
738
|
+
Node *node=index.get_node(i);
|
739
|
+
Arcs *arcs=node->arcs();
|
740
|
+
pos += sizeof(char) // size of final flag
|
741
|
+
+ sizeof(unsigned short) // size of number of arcs
|
742
|
+
+ arcs->size() * (sizeof(Character) * 2 + sizeof(unsigned int)); // size of n arcs
|
743
|
+
}
|
744
|
+
|
745
|
+
// storing nodes
|
746
|
+
for( size_t i=0; i<index.number_of_nodes(); i++ )
|
747
|
+
store_lowmem_node( file, index.get_node(i), index, startpos );
|
748
|
+
}
|
749
|
+
|
750
|
+
|
751
|
+
/*******************************************************************/
|
752
|
+
/* */
|
753
|
+
/* Transducer::store */
|
754
|
+
/* */
|
755
|
+
/*******************************************************************/
|
756
|
+
|
757
|
+
void Transducer::store( FILE *file )
|
758
|
+
|
759
|
+
{
|
760
|
+
fputc('a',file);
|
761
|
+
|
762
|
+
NodeNumbering index(*this);
|
763
|
+
incr_vmark();
|
764
|
+
unsigned int n=index.number_of_nodes();
|
765
|
+
fwrite(&n,sizeof(n),1,file);
|
766
|
+
store_node( file, root_node(), index, vmark );
|
767
|
+
|
768
|
+
alphabet.store(file);
|
769
|
+
}
|
770
|
+
|
771
|
+
|
772
|
+
/*******************************************************************/
|
773
|
+
/* */
|
774
|
+
/* read_node */
|
775
|
+
/* */
|
776
|
+
/*******************************************************************/
|
777
|
+
|
778
|
+
static void read_node( FILE *file, Node *node, Node **p, Transducer *a )
|
779
|
+
{
|
780
|
+
char c;
|
781
|
+
fread(&c,sizeof(c),1,file);
|
782
|
+
node->set_final(c);
|
783
|
+
|
784
|
+
unsigned short n;
|
785
|
+
fread( &n, sizeof(n), 1, file);
|
786
|
+
|
787
|
+
for( int i=0; i<n; i++ ) {
|
788
|
+
Character lc,uc;
|
789
|
+
unsigned int t;
|
790
|
+
fread(&lc,sizeof(lc),1,file);
|
791
|
+
fread(&uc,sizeof(uc),1,file);
|
792
|
+
fread(&t,sizeof(t),1,file);
|
793
|
+
if (ferror(file))
|
794
|
+
throw "Error encountered while reading transducer from file";
|
795
|
+
if (p[t])
|
796
|
+
node->add_arc( Label(lc,uc), p[t], a );
|
797
|
+
else {
|
798
|
+
p[t] = a->new_node();
|
799
|
+
node->add_arc( Label(lc,uc), p[t], a );
|
800
|
+
read_node(file, p[t], p, a );
|
801
|
+
}
|
802
|
+
}
|
803
|
+
}
|
804
|
+
|
805
|
+
|
806
|
+
/*******************************************************************/
|
807
|
+
/* */
|
808
|
+
/* Transducer::read_transducer_binary */
|
809
|
+
/* */
|
810
|
+
/*******************************************************************/
|
811
|
+
|
812
|
+
void Transducer::read_transducer_binary( FILE *file )
|
813
|
+
|
814
|
+
{
|
815
|
+
if (fgetc(file) != 'a')
|
816
|
+
throw "Error: wrong file format (not a standard transducer)\n";
|
817
|
+
|
818
|
+
vmark = deterministic = 0;
|
819
|
+
unsigned int n;
|
820
|
+
fread(&n,sizeof(n),1,file); // number of nodes
|
821
|
+
if (ferror(file))
|
822
|
+
throw "Error encountered while reading transducer from file";
|
823
|
+
|
824
|
+
Node **p=new Node*[n]; // maps indices to nodes
|
825
|
+
p[0] = root_node();
|
826
|
+
for( unsigned int i=1; i<n; i++)
|
827
|
+
p[i] = NULL;
|
828
|
+
read_node( file, root_node(), p, this );
|
829
|
+
delete[] p;
|
830
|
+
|
831
|
+
alphabet.read(file);
|
832
|
+
|
833
|
+
vmark = 1;
|
834
|
+
deterministic = minimised = 1;
|
835
|
+
}
|
836
|
+
|
837
|
+
|
838
|
+
/*******************************************************************/
|
839
|
+
/* */
|
840
|
+
/* error_message */
|
841
|
+
/* */
|
842
|
+
/*******************************************************************/
|
843
|
+
|
844
|
+
static void error_message( size_t line )
|
845
|
+
|
846
|
+
{
|
847
|
+
static char message[1000];
|
848
|
+
sprintf(message, "Error: in line %u of text transducer file",
|
849
|
+
(unsigned int)line);
|
850
|
+
throw message;
|
851
|
+
}
|
852
|
+
|
853
|
+
|
854
|
+
/*******************************************************************/
|
855
|
+
/* */
|
856
|
+
/* Transducer::create_node */
|
857
|
+
/* */
|
858
|
+
/*******************************************************************/
|
859
|
+
|
860
|
+
Node *Transducer::create_node( vector<Node*> &node, char *s, size_t line )
|
861
|
+
|
862
|
+
{
|
863
|
+
char *p;
|
864
|
+
long n = strtol(s, &p, 10);
|
865
|
+
|
866
|
+
if (s == p || n < 0)
|
867
|
+
error_message( line );
|
868
|
+
if ((long)node.size() <= n)
|
869
|
+
node.resize(n+1, NULL);
|
870
|
+
if (node[n] == NULL)
|
871
|
+
node[n] = new Node;
|
872
|
+
|
873
|
+
return node[n];
|
874
|
+
}
|
875
|
+
|
876
|
+
|
877
|
+
/*******************************************************************/
|
878
|
+
/* */
|
879
|
+
/* next_string */
|
880
|
+
/* */
|
881
|
+
/*******************************************************************/
|
882
|
+
|
883
|
+
static char *next_string( char* &s, size_t line )
|
884
|
+
|
885
|
+
{
|
886
|
+
// scan the input up to the next tab or newline character
|
887
|
+
// and unquote symbols preceded by a backslash
|
888
|
+
char *p = s;
|
889
|
+
char *q = s;
|
890
|
+
while (*q!=0 && *q!='\t' && *q!='\n' && *q!='\r') {
|
891
|
+
if (*q == '\\')
|
892
|
+
q++;
|
893
|
+
*(p++) = *(q++);
|
894
|
+
}
|
895
|
+
if (p == s)
|
896
|
+
error_message(line); // no string found
|
897
|
+
|
898
|
+
char *result=s;
|
899
|
+
// skip over following whitespace
|
900
|
+
while (*q == ' ' || *q == '\t' || *q == '\n' || *q == '\r')
|
901
|
+
q++;
|
902
|
+
|
903
|
+
if (*q == 0)
|
904
|
+
s = NULL; // end of string was reached
|
905
|
+
else
|
906
|
+
s = q; // move the string pointer s
|
907
|
+
|
908
|
+
*p = 0; // mark the end of the result string
|
909
|
+
|
910
|
+
return result;
|
911
|
+
}
|
912
|
+
|
913
|
+
|
914
|
+
/*******************************************************************/
|
915
|
+
/* */
|
916
|
+
/* Transducer::read_transducer_text */
|
917
|
+
/* */
|
918
|
+
/*******************************************************************/
|
919
|
+
|
920
|
+
void Transducer::read_transducer_text( FILE *file )
|
921
|
+
|
922
|
+
{
|
923
|
+
vector<Node*> nodes;
|
924
|
+
nodes.push_back(root_node());
|
925
|
+
|
926
|
+
vmark = deterministic = 0;
|
927
|
+
char buffer[10000];
|
928
|
+
for( size_t line=0; fgets(buffer, 10000, file ); line++ ) {
|
929
|
+
char *p = buffer;
|
930
|
+
char *s = next_string(p, line);
|
931
|
+
if (strcmp(s, "final") == 0) {
|
932
|
+
s = next_string(p, line);
|
933
|
+
create_node( nodes, s, line )->set_final(true);
|
934
|
+
}
|
935
|
+
else {
|
936
|
+
Node *node = create_node( nodes, s, line );
|
937
|
+
s = next_string(p, line);
|
938
|
+
Label l = alphabet.next_label( s, 2 );
|
939
|
+
if (*s != 0 || l == Label::epsilon)
|
940
|
+
error_message( line );
|
941
|
+
s = next_string(p, line);
|
942
|
+
Node *target = create_node( nodes, s, line );
|
943
|
+
node->add_arc( l, target, this );
|
944
|
+
}
|
945
|
+
}
|
946
|
+
|
947
|
+
vmark = 1;
|
948
|
+
deterministic = minimised = 1;
|
949
|
+
}
|
950
|
+
|
951
|
+
|
952
|
+
/*******************************************************************/
|
953
|
+
/* */
|
954
|
+
/* Transducer::Transducer */
|
955
|
+
/* */
|
956
|
+
/*******************************************************************/
|
957
|
+
|
958
|
+
Transducer::Transducer( FILE *file, bool binary )
|
959
|
+
|
960
|
+
{
|
961
|
+
if (binary)
|
962
|
+
read_transducer_binary( file );
|
963
|
+
else
|
964
|
+
read_transducer_text( file );
|
965
|
+
}
|
966
|
+
|