1 | // © 2016 and later: Unicode, Inc. and others. |
2 | // License & terms of use: http://www.unicode.org/copyright.html |
3 | /* |
4 | ********************************************************************** |
5 | * Copyright (c) 2002-2016, International Business Machines |
6 | * Corporation and others. All Rights Reserved. |
7 | ********************************************************************** |
8 | */ |
9 | // |
10 | // rbbitblb.cpp |
11 | // |
12 | |
13 | |
14 | #include "unicode/utypes.h" |
15 | |
16 | #if !UCONFIG_NO_BREAK_ITERATION |
17 | |
18 | #include "unicode/unistr.h" |
19 | #include "rbbitblb.h" |
20 | #include "rbbirb.h" |
21 | #include "rbbiscan.h" |
22 | #include "rbbisetb.h" |
23 | #include "rbbidata.h" |
24 | #include "cstring.h" |
25 | #include "uassert.h" |
26 | #include "uvectr32.h" |
27 | #include "cmemory.h" |
28 | |
29 | U_NAMESPACE_BEGIN |
30 | |
31 | RBBITableBuilder::RBBITableBuilder(RBBIRuleBuilder *rb, RBBINode **rootNode, UErrorCode &status) : |
32 | fRB(rb), |
33 | fTree(*rootNode), |
34 | fStatus(&status), |
35 | fDStates(nullptr), |
36 | fSafeTable(nullptr) { |
37 | if (U_FAILURE(status)) { |
38 | return; |
39 | } |
40 | // fDStates is UVector<RBBIStateDescriptor *> |
41 | fDStates = new UVector(status); |
42 | if (U_SUCCESS(status) && fDStates == nullptr ) { |
43 | status = U_MEMORY_ALLOCATION_ERROR; |
44 | } |
45 | } |
46 | |
47 | |
48 | |
49 | RBBITableBuilder::~RBBITableBuilder() { |
50 | int i; |
51 | for (i=0; i<fDStates->size(); i++) { |
52 | delete (RBBIStateDescriptor *)fDStates->elementAt(i); |
53 | } |
54 | delete fDStates; |
55 | delete fSafeTable; |
56 | delete fLookAheadRuleMap; |
57 | } |
58 | |
59 | |
60 | //----------------------------------------------------------------------------- |
61 | // |
62 | // RBBITableBuilder::buildForwardTable - This is the main function for building |
63 | // the DFA state transition table from the RBBI rules parse tree. |
64 | // |
65 | //----------------------------------------------------------------------------- |
66 | void RBBITableBuilder::buildForwardTable() { |
67 | |
68 | if (U_FAILURE(*fStatus)) { |
69 | return; |
70 | } |
71 | |
72 | // If there were no rules, just return. This situation can easily arise |
73 | // for the reverse rules. |
74 | if (fTree==NULL) { |
75 | return; |
76 | } |
77 | |
78 | // |
79 | // Walk through the tree, replacing any references to $variables with a copy of the |
80 | // parse tree for the substition expression. |
81 | // |
82 | fTree = fTree->flattenVariables(); |
83 | #ifdef RBBI_DEBUG |
84 | if (fRB->fDebugEnv && uprv_strstr(fRB->fDebugEnv, "ftree" )) { |
85 | RBBIDebugPuts("\nParse tree after flattening variable references." ); |
86 | RBBINode::printTree(fTree, TRUE); |
87 | } |
88 | #endif |
89 | |
90 | // |
91 | // If the rules contained any references to {bof} |
92 | // add a {bof} <cat> <former root of tree> to the |
93 | // tree. Means that all matches must start out with the |
94 | // {bof} fake character. |
95 | // |
96 | if (fRB->fSetBuilder->sawBOF()) { |
97 | RBBINode *bofTop = new RBBINode(RBBINode::opCat); |
98 | RBBINode *bofLeaf = new RBBINode(RBBINode::leafChar); |
99 | // Delete and exit if memory allocation failed. |
100 | if (bofTop == NULL || bofLeaf == NULL) { |
101 | *fStatus = U_MEMORY_ALLOCATION_ERROR; |
102 | delete bofTop; |
103 | delete bofLeaf; |
104 | return; |
105 | } |
106 | bofTop->fLeftChild = bofLeaf; |
107 | bofTop->fRightChild = fTree; |
108 | bofLeaf->fParent = bofTop; |
109 | bofLeaf->fVal = 2; // Reserved value for {bof}. |
110 | fTree = bofTop; |
111 | } |
112 | |
113 | // |
114 | // Add a unique right-end marker to the expression. |
115 | // Appears as a cat-node, left child being the original tree, |
116 | // right child being the end marker. |
117 | // |
118 | RBBINode *cn = new RBBINode(RBBINode::opCat); |
119 | // Exit if memory allocation failed. |
120 | if (cn == NULL) { |
121 | *fStatus = U_MEMORY_ALLOCATION_ERROR; |
122 | return; |
123 | } |
124 | cn->fLeftChild = fTree; |
125 | fTree->fParent = cn; |
126 | RBBINode *endMarkerNode = cn->fRightChild = new RBBINode(RBBINode::endMark); |
127 | // Delete and exit if memory allocation failed. |
128 | if (cn->fRightChild == NULL) { |
129 | *fStatus = U_MEMORY_ALLOCATION_ERROR; |
130 | delete cn; |
131 | return; |
132 | } |
133 | cn->fRightChild->fParent = cn; |
134 | fTree = cn; |
135 | |
136 | // |
137 | // Replace all references to UnicodeSets with the tree for the equivalent |
138 | // expression. |
139 | // |
140 | fTree->flattenSets(); |
141 | #ifdef RBBI_DEBUG |
142 | if (fRB->fDebugEnv && uprv_strstr(fRB->fDebugEnv, "stree" )) { |
143 | RBBIDebugPuts("\nParse tree after flattening Unicode Set references." ); |
144 | RBBINode::printTree(fTree, TRUE); |
145 | } |
146 | #endif |
147 | |
148 | |
149 | // |
150 | // calculate the functions nullable, firstpos, lastpos and followpos on |
151 | // nodes in the parse tree. |
152 | // See the alogrithm description in Aho. |
153 | // Understanding how this works by looking at the code alone will be |
154 | // nearly impossible. |
155 | // |
156 | calcNullable(fTree); |
157 | calcFirstPos(fTree); |
158 | calcLastPos(fTree); |
159 | calcFollowPos(fTree); |
160 | if (fRB->fDebugEnv && uprv_strstr(fRB->fDebugEnv, "pos" )) { |
161 | RBBIDebugPuts("\n" ); |
162 | printPosSets(fTree); |
163 | } |
164 | |
165 | // |
166 | // For "chained" rules, modify the followPos sets |
167 | // |
168 | if (fRB->fChainRules) { |
169 | calcChainedFollowPos(fTree, endMarkerNode); |
170 | } |
171 | |
172 | // |
173 | // BOF (start of input) test fixup. |
174 | // |
175 | if (fRB->fSetBuilder->sawBOF()) { |
176 | bofFixup(); |
177 | } |
178 | |
179 | // |
180 | // Build the DFA state transition tables. |
181 | // |
182 | buildStateTable(); |
183 | mapLookAheadRules(); |
184 | flagAcceptingStates(); |
185 | flagLookAheadStates(); |
186 | flagTaggedStates(); |
187 | |
188 | // |
189 | // Update the global table of rule status {tag} values |
190 | // The rule builder has a global vector of status values that are common |
191 | // for all tables. Merge the ones from this table into the global set. |
192 | // |
193 | mergeRuleStatusVals(); |
194 | } |
195 | |
196 | |
197 | |
198 | //----------------------------------------------------------------------------- |
199 | // |
200 | // calcNullable. Impossible to explain succinctly. See Aho, section 3.9 |
201 | // |
202 | //----------------------------------------------------------------------------- |
203 | void RBBITableBuilder::calcNullable(RBBINode *n) { |
204 | if (n == NULL) { |
205 | return; |
206 | } |
207 | if (n->fType == RBBINode::setRef || |
208 | n->fType == RBBINode::endMark ) { |
209 | // These are non-empty leaf node types. |
210 | n->fNullable = FALSE; |
211 | return; |
212 | } |
213 | |
214 | if (n->fType == RBBINode::lookAhead || n->fType == RBBINode::tag) { |
215 | // Lookahead marker node. It's a leaf, so no recursion on children. |
216 | // It's nullable because it does not match any literal text from the input stream. |
217 | n->fNullable = TRUE; |
218 | return; |
219 | } |
220 | |
221 | |
222 | // The node is not a leaf. |
223 | // Calculate nullable on its children. |
224 | calcNullable(n->fLeftChild); |
225 | calcNullable(n->fRightChild); |
226 | |
227 | // Apply functions from table 3.40 in Aho |
228 | if (n->fType == RBBINode::opOr) { |
229 | n->fNullable = n->fLeftChild->fNullable || n->fRightChild->fNullable; |
230 | } |
231 | else if (n->fType == RBBINode::opCat) { |
232 | n->fNullable = n->fLeftChild->fNullable && n->fRightChild->fNullable; |
233 | } |
234 | else if (n->fType == RBBINode::opStar || n->fType == RBBINode::opQuestion) { |
235 | n->fNullable = TRUE; |
236 | } |
237 | else { |
238 | n->fNullable = FALSE; |
239 | } |
240 | } |
241 | |
242 | |
243 | |
244 | |
245 | //----------------------------------------------------------------------------- |
246 | // |
247 | // calcFirstPos. Impossible to explain succinctly. See Aho, section 3.9 |
248 | // |
249 | //----------------------------------------------------------------------------- |
250 | void RBBITableBuilder::calcFirstPos(RBBINode *n) { |
251 | if (n == NULL) { |
252 | return; |
253 | } |
254 | if (n->fType == RBBINode::leafChar || |
255 | n->fType == RBBINode::endMark || |
256 | n->fType == RBBINode::lookAhead || |
257 | n->fType == RBBINode::tag) { |
258 | // These are non-empty leaf node types. |
259 | // Note: In order to maintain the sort invariant on the set, |
260 | // this function should only be called on a node whose set is |
261 | // empty to start with. |
262 | n->fFirstPosSet->addElement(n, *fStatus); |
263 | return; |
264 | } |
265 | |
266 | // The node is not a leaf. |
267 | // Calculate firstPos on its children. |
268 | calcFirstPos(n->fLeftChild); |
269 | calcFirstPos(n->fRightChild); |
270 | |
271 | // Apply functions from table 3.40 in Aho |
272 | if (n->fType == RBBINode::opOr) { |
273 | setAdd(n->fFirstPosSet, n->fLeftChild->fFirstPosSet); |
274 | setAdd(n->fFirstPosSet, n->fRightChild->fFirstPosSet); |
275 | } |
276 | else if (n->fType == RBBINode::opCat) { |
277 | setAdd(n->fFirstPosSet, n->fLeftChild->fFirstPosSet); |
278 | if (n->fLeftChild->fNullable) { |
279 | setAdd(n->fFirstPosSet, n->fRightChild->fFirstPosSet); |
280 | } |
281 | } |
282 | else if (n->fType == RBBINode::opStar || |
283 | n->fType == RBBINode::opQuestion || |
284 | n->fType == RBBINode::opPlus) { |
285 | setAdd(n->fFirstPosSet, n->fLeftChild->fFirstPosSet); |
286 | } |
287 | } |
288 | |
289 | |
290 | |
291 | //----------------------------------------------------------------------------- |
292 | // |
293 | // calcLastPos. Impossible to explain succinctly. See Aho, section 3.9 |
294 | // |
295 | //----------------------------------------------------------------------------- |
296 | void RBBITableBuilder::calcLastPos(RBBINode *n) { |
297 | if (n == NULL) { |
298 | return; |
299 | } |
300 | if (n->fType == RBBINode::leafChar || |
301 | n->fType == RBBINode::endMark || |
302 | n->fType == RBBINode::lookAhead || |
303 | n->fType == RBBINode::tag) { |
304 | // These are non-empty leaf node types. |
305 | // Note: In order to maintain the sort invariant on the set, |
306 | // this function should only be called on a node whose set is |
307 | // empty to start with. |
308 | n->fLastPosSet->addElement(n, *fStatus); |
309 | return; |
310 | } |
311 | |
312 | // The node is not a leaf. |
313 | // Calculate lastPos on its children. |
314 | calcLastPos(n->fLeftChild); |
315 | calcLastPos(n->fRightChild); |
316 | |
317 | // Apply functions from table 3.40 in Aho |
318 | if (n->fType == RBBINode::opOr) { |
319 | setAdd(n->fLastPosSet, n->fLeftChild->fLastPosSet); |
320 | setAdd(n->fLastPosSet, n->fRightChild->fLastPosSet); |
321 | } |
322 | else if (n->fType == RBBINode::opCat) { |
323 | setAdd(n->fLastPosSet, n->fRightChild->fLastPosSet); |
324 | if (n->fRightChild->fNullable) { |
325 | setAdd(n->fLastPosSet, n->fLeftChild->fLastPosSet); |
326 | } |
327 | } |
328 | else if (n->fType == RBBINode::opStar || |
329 | n->fType == RBBINode::opQuestion || |
330 | n->fType == RBBINode::opPlus) { |
331 | setAdd(n->fLastPosSet, n->fLeftChild->fLastPosSet); |
332 | } |
333 | } |
334 | |
335 | |
336 | |
337 | //----------------------------------------------------------------------------- |
338 | // |
339 | // calcFollowPos. Impossible to explain succinctly. See Aho, section 3.9 |
340 | // |
341 | //----------------------------------------------------------------------------- |
342 | void RBBITableBuilder::calcFollowPos(RBBINode *n) { |
343 | if (n == NULL || |
344 | n->fType == RBBINode::leafChar || |
345 | n->fType == RBBINode::endMark) { |
346 | return; |
347 | } |
348 | |
349 | calcFollowPos(n->fLeftChild); |
350 | calcFollowPos(n->fRightChild); |
351 | |
352 | // Aho rule #1 |
353 | if (n->fType == RBBINode::opCat) { |
354 | RBBINode *i; // is 'i' in Aho's description |
355 | uint32_t ix; |
356 | |
357 | UVector *LastPosOfLeftChild = n->fLeftChild->fLastPosSet; |
358 | |
359 | for (ix=0; ix<(uint32_t)LastPosOfLeftChild->size(); ix++) { |
360 | i = (RBBINode *)LastPosOfLeftChild->elementAt(ix); |
361 | setAdd(i->fFollowPos, n->fRightChild->fFirstPosSet); |
362 | } |
363 | } |
364 | |
365 | // Aho rule #2 |
366 | if (n->fType == RBBINode::opStar || |
367 | n->fType == RBBINode::opPlus) { |
368 | RBBINode *i; // again, n and i are the names from Aho's description. |
369 | uint32_t ix; |
370 | |
371 | for (ix=0; ix<(uint32_t)n->fLastPosSet->size(); ix++) { |
372 | i = (RBBINode *)n->fLastPosSet->elementAt(ix); |
373 | setAdd(i->fFollowPos, n->fFirstPosSet); |
374 | } |
375 | } |
376 | |
377 | |
378 | |
379 | } |
380 | |
381 | //----------------------------------------------------------------------------- |
382 | // |
383 | // addRuleRootNodes Recursively walk a parse tree, adding all nodes flagged |
384 | // as roots of a rule to a destination vector. |
385 | // |
386 | //----------------------------------------------------------------------------- |
387 | void RBBITableBuilder::addRuleRootNodes(UVector *dest, RBBINode *node) { |
388 | if (node == NULL || U_FAILURE(*fStatus)) { |
389 | return; |
390 | } |
391 | if (node->fRuleRoot) { |
392 | dest->addElement(node, *fStatus); |
393 | // Note: rules cannot nest. If we found a rule start node, |
394 | // no child node can also be a start node. |
395 | return; |
396 | } |
397 | addRuleRootNodes(dest, node->fLeftChild); |
398 | addRuleRootNodes(dest, node->fRightChild); |
399 | } |
400 | |
401 | //----------------------------------------------------------------------------- |
402 | // |
403 | // calcChainedFollowPos. Modify the previously calculated followPos sets |
404 | // to implement rule chaining. NOT described by Aho |
405 | // |
406 | //----------------------------------------------------------------------------- |
407 | void RBBITableBuilder::calcChainedFollowPos(RBBINode *tree, RBBINode *endMarkNode) { |
408 | |
409 | UVector leafNodes(*fStatus); |
410 | if (U_FAILURE(*fStatus)) { |
411 | return; |
412 | } |
413 | |
414 | // get a list all leaf nodes |
415 | tree->findNodes(&leafNodes, RBBINode::leafChar, *fStatus); |
416 | if (U_FAILURE(*fStatus)) { |
417 | return; |
418 | } |
419 | |
420 | // Collect all leaf nodes that can start matches for rules |
421 | // with inbound chaining enabled, which is the union of the |
422 | // firstPosition sets from each of the rule root nodes. |
423 | |
424 | UVector ruleRootNodes(*fStatus); |
425 | addRuleRootNodes(&ruleRootNodes, tree); |
426 | |
427 | UVector matchStartNodes(*fStatus); |
428 | for (int j=0; j<ruleRootNodes.size(); ++j) { |
429 | RBBINode *node = static_cast<RBBINode *>(ruleRootNodes.elementAt(j)); |
430 | if (node->fChainIn) { |
431 | setAdd(&matchStartNodes, node->fFirstPosSet); |
432 | } |
433 | } |
434 | if (U_FAILURE(*fStatus)) { |
435 | return; |
436 | } |
437 | |
438 | int32_t endNodeIx; |
439 | int32_t startNodeIx; |
440 | |
441 | for (endNodeIx=0; endNodeIx<leafNodes.size(); endNodeIx++) { |
442 | RBBINode *endNode = (RBBINode *)leafNodes.elementAt(endNodeIx); |
443 | |
444 | // Identify leaf nodes that correspond to overall rule match positions. |
445 | // These include the endMarkNode in their followPos sets. |
446 | // |
447 | // Note: do not consider other end marker nodes, those that are added to |
448 | // look-ahead rules. These can't chain; a match immediately stops |
449 | // further matching. This leaves exactly one end marker node, the one |
450 | // at the end of the complete tree. |
451 | |
452 | if (!endNode->fFollowPos->contains(endMarkNode)) { |
453 | continue; |
454 | } |
455 | |
456 | // We've got a node that can end a match. |
457 | |
458 | // !!LBCMNoChain implementation: If this node's val correspond to |
459 | // the Line Break $CM char class, don't chain from it. |
460 | // TODO: Remove this. !!LBCMNoChain is deprecated, and is not used |
461 | // by any of the standard ICU rules. |
462 | if (fRB->fLBCMNoChain) { |
463 | UChar32 c = this->fRB->fSetBuilder->getFirstChar(endNode->fVal); |
464 | if (c != -1) { |
465 | // c == -1 occurs with sets containing only the {eof} marker string. |
466 | ULineBreak cLBProp = (ULineBreak)u_getIntPropertyValue(c, UCHAR_LINE_BREAK); |
467 | if (cLBProp == U_LB_COMBINING_MARK) { |
468 | continue; |
469 | } |
470 | } |
471 | } |
472 | |
473 | // Now iterate over the nodes that can start a match, looking for ones |
474 | // with the same char class as our ending node. |
475 | RBBINode *startNode; |
476 | for (startNodeIx = 0; startNodeIx<matchStartNodes.size(); startNodeIx++) { |
477 | startNode = (RBBINode *)matchStartNodes.elementAt(startNodeIx); |
478 | if (startNode->fType != RBBINode::leafChar) { |
479 | continue; |
480 | } |
481 | |
482 | if (endNode->fVal == startNode->fVal) { |
483 | // The end val (character class) of one possible match is the |
484 | // same as the start of another. |
485 | |
486 | // Add all nodes from the followPos of the start node to the |
487 | // followPos set of the end node, which will have the effect of |
488 | // letting matches transition from a match state at endNode |
489 | // to the second char of a match starting with startNode. |
490 | setAdd(endNode->fFollowPos, startNode->fFollowPos); |
491 | } |
492 | } |
493 | } |
494 | } |
495 | |
496 | |
497 | //----------------------------------------------------------------------------- |
498 | // |
499 | // bofFixup. Fixup for state tables that include {bof} beginning of input testing. |
500 | // Do an swizzle similar to chaining, modifying the followPos set of |
501 | // the bofNode to include the followPos nodes from other {bot} nodes |
502 | // scattered through the tree. |
503 | // |
504 | // This function has much in common with calcChainedFollowPos(). |
505 | // |
506 | //----------------------------------------------------------------------------- |
507 | void RBBITableBuilder::bofFixup() { |
508 | |
509 | if (U_FAILURE(*fStatus)) { |
510 | return; |
511 | } |
512 | |
513 | // The parse tree looks like this ... |
514 | // fTree root ---> <cat> |
515 | // / \ . |
516 | // <cat> <#end node> |
517 | // / \ . |
518 | // <bofNode> rest |
519 | // of tree |
520 | // |
521 | // We will be adding things to the followPos set of the <bofNode> |
522 | // |
523 | RBBINode *bofNode = fTree->fLeftChild->fLeftChild; |
524 | U_ASSERT(bofNode->fType == RBBINode::leafChar); |
525 | U_ASSERT(bofNode->fVal == 2); |
526 | |
527 | // Get all nodes that can be the start a match of the user-written rules |
528 | // (excluding the fake bofNode) |
529 | // We want the nodes that can start a match in the |
530 | // part labeled "rest of tree" |
531 | // |
532 | UVector *matchStartNodes = fTree->fLeftChild->fRightChild->fFirstPosSet; |
533 | |
534 | RBBINode *startNode; |
535 | int startNodeIx; |
536 | for (startNodeIx = 0; startNodeIx<matchStartNodes->size(); startNodeIx++) { |
537 | startNode = (RBBINode *)matchStartNodes->elementAt(startNodeIx); |
538 | if (startNode->fType != RBBINode::leafChar) { |
539 | continue; |
540 | } |
541 | |
542 | if (startNode->fVal == bofNode->fVal) { |
543 | // We found a leaf node corresponding to a {bof} that was |
544 | // explicitly written into a rule. |
545 | // Add everything from the followPos set of this node to the |
546 | // followPos set of the fake bofNode at the start of the tree. |
547 | // |
548 | setAdd(bofNode->fFollowPos, startNode->fFollowPos); |
549 | } |
550 | } |
551 | } |
552 | |
553 | //----------------------------------------------------------------------------- |
554 | // |
555 | // buildStateTable() Determine the set of runtime DFA states and the |
556 | // transition tables for these states, by the algorithm |
557 | // of fig. 3.44 in Aho. |
558 | // |
559 | // Most of the comments are quotes of Aho's psuedo-code. |
560 | // |
561 | //----------------------------------------------------------------------------- |
562 | void RBBITableBuilder::buildStateTable() { |
563 | if (U_FAILURE(*fStatus)) { |
564 | return; |
565 | } |
566 | RBBIStateDescriptor *failState; |
567 | // Set it to NULL to avoid uninitialized warning |
568 | RBBIStateDescriptor *initialState = NULL; |
569 | // |
570 | // Add a dummy state 0 - the stop state. Not from Aho. |
571 | int lastInputSymbol = fRB->fSetBuilder->getNumCharCategories() - 1; |
572 | failState = new RBBIStateDescriptor(lastInputSymbol, fStatus); |
573 | if (failState == NULL) { |
574 | *fStatus = U_MEMORY_ALLOCATION_ERROR; |
575 | goto ExitBuildSTdeleteall; |
576 | } |
577 | failState->fPositions = new UVector(*fStatus); |
578 | if (failState->fPositions == NULL) { |
579 | *fStatus = U_MEMORY_ALLOCATION_ERROR; |
580 | } |
581 | if (failState->fPositions == NULL || U_FAILURE(*fStatus)) { |
582 | goto ExitBuildSTdeleteall; |
583 | } |
584 | fDStates->addElement(failState, *fStatus); |
585 | if (U_FAILURE(*fStatus)) { |
586 | goto ExitBuildSTdeleteall; |
587 | } |
588 | |
589 | // initially, the only unmarked state in Dstates is firstpos(root), |
590 | // where toot is the root of the syntax tree for (r)#; |
591 | initialState = new RBBIStateDescriptor(lastInputSymbol, fStatus); |
592 | if (initialState == NULL) { |
593 | *fStatus = U_MEMORY_ALLOCATION_ERROR; |
594 | } |
595 | if (U_FAILURE(*fStatus)) { |
596 | goto ExitBuildSTdeleteall; |
597 | } |
598 | initialState->fPositions = new UVector(*fStatus); |
599 | if (initialState->fPositions == NULL) { |
600 | *fStatus = U_MEMORY_ALLOCATION_ERROR; |
601 | } |
602 | if (U_FAILURE(*fStatus)) { |
603 | goto ExitBuildSTdeleteall; |
604 | } |
605 | setAdd(initialState->fPositions, fTree->fFirstPosSet); |
606 | fDStates->addElement(initialState, *fStatus); |
607 | if (U_FAILURE(*fStatus)) { |
608 | goto ExitBuildSTdeleteall; |
609 | } |
610 | |
611 | // while there is an unmarked state T in Dstates do begin |
612 | for (;;) { |
613 | RBBIStateDescriptor *T = NULL; |
614 | int32_t tx; |
615 | for (tx=1; tx<fDStates->size(); tx++) { |
616 | RBBIStateDescriptor *temp; |
617 | temp = (RBBIStateDescriptor *)fDStates->elementAt(tx); |
618 | if (temp->fMarked == FALSE) { |
619 | T = temp; |
620 | break; |
621 | } |
622 | } |
623 | if (T == NULL) { |
624 | break; |
625 | } |
626 | |
627 | // mark T; |
628 | T->fMarked = TRUE; |
629 | |
630 | // for each input symbol a do begin |
631 | int32_t a; |
632 | for (a = 1; a<=lastInputSymbol; a++) { |
633 | // let U be the set of positions that are in followpos(p) |
634 | // for some position p in T |
635 | // such that the symbol at position p is a; |
636 | UVector *U = NULL; |
637 | RBBINode *p; |
638 | int32_t px; |
639 | for (px=0; px<T->fPositions->size(); px++) { |
640 | p = (RBBINode *)T->fPositions->elementAt(px); |
641 | if ((p->fType == RBBINode::leafChar) && (p->fVal == a)) { |
642 | if (U == NULL) { |
643 | U = new UVector(*fStatus); |
644 | if (U == NULL) { |
645 | *fStatus = U_MEMORY_ALLOCATION_ERROR; |
646 | goto ExitBuildSTdeleteall; |
647 | } |
648 | } |
649 | setAdd(U, p->fFollowPos); |
650 | } |
651 | } |
652 | |
653 | // if U is not empty and not in DStates then |
654 | int32_t ux = 0; |
655 | UBool UinDstates = FALSE; |
656 | if (U != NULL) { |
657 | U_ASSERT(U->size() > 0); |
658 | int ix; |
659 | for (ix=0; ix<fDStates->size(); ix++) { |
660 | RBBIStateDescriptor *temp2; |
661 | temp2 = (RBBIStateDescriptor *)fDStates->elementAt(ix); |
662 | if (setEquals(U, temp2->fPositions)) { |
663 | delete U; |
664 | U = temp2->fPositions; |
665 | ux = ix; |
666 | UinDstates = TRUE; |
667 | break; |
668 | } |
669 | } |
670 | |
671 | // Add U as an unmarked state to Dstates |
672 | if (!UinDstates) |
673 | { |
674 | RBBIStateDescriptor *newState = new RBBIStateDescriptor(lastInputSymbol, fStatus); |
675 | if (newState == NULL) { |
676 | *fStatus = U_MEMORY_ALLOCATION_ERROR; |
677 | } |
678 | if (U_FAILURE(*fStatus)) { |
679 | goto ExitBuildSTdeleteall; |
680 | } |
681 | newState->fPositions = U; |
682 | fDStates->addElement(newState, *fStatus); |
683 | if (U_FAILURE(*fStatus)) { |
684 | return; |
685 | } |
686 | ux = fDStates->size()-1; |
687 | } |
688 | |
689 | // Dtran[T, a] := U; |
690 | T->fDtran->setElementAt(ux, a); |
691 | } |
692 | } |
693 | } |
694 | return; |
695 | // delete local pointers only if error occured. |
696 | ExitBuildSTdeleteall: |
697 | delete initialState; |
698 | delete failState; |
699 | } |
700 | |
701 | |
702 | /** |
703 | * mapLookAheadRules |
704 | * |
705 | */ |
706 | void RBBITableBuilder::mapLookAheadRules() { |
707 | fLookAheadRuleMap = new UVector32(fRB->fScanner->numRules() + 1, *fStatus); |
708 | if (fLookAheadRuleMap == nullptr) { |
709 | *fStatus = U_MEMORY_ALLOCATION_ERROR; |
710 | } |
711 | if (U_FAILURE(*fStatus)) { |
712 | return; |
713 | } |
714 | fLookAheadRuleMap->setSize(fRB->fScanner->numRules() + 1); |
715 | int32_t laSlotsInUse = 0; |
716 | |
717 | for (int32_t n=0; n<fDStates->size(); n++) { |
718 | RBBIStateDescriptor *sd = (RBBIStateDescriptor *)fDStates->elementAt(n); |
719 | int32_t laSlotForState = 0; |
720 | |
721 | // Establish the look-ahead slot for this state, if the state covers |
722 | // any look-ahead nodes - corresponding to the '/' in look-ahead rules. |
723 | |
724 | // If any of the look-ahead nodes already have a slot assigned, use it, |
725 | // otherwise assign a new one. |
726 | |
727 | bool sawLookAheadNode = false; |
728 | for (int32_t ipos=0; ipos<sd->fPositions->size(); ++ipos) { |
729 | RBBINode *node = static_cast<RBBINode *>(sd->fPositions->elementAt(ipos)); |
730 | if (node->fType != RBBINode::NodeType::lookAhead) { |
731 | continue; |
732 | } |
733 | sawLookAheadNode = true; |
734 | int32_t ruleNum = node->fVal; // Set when rule was originally parsed. |
735 | U_ASSERT(ruleNum < fLookAheadRuleMap->size()); |
736 | U_ASSERT(ruleNum > 0); |
737 | int32_t laSlot = fLookAheadRuleMap->elementAti(ruleNum); |
738 | if (laSlot != 0) { |
739 | if (laSlotForState == 0) { |
740 | laSlotForState = laSlot; |
741 | } else { |
742 | // TODO: figure out if this can fail, change to setting an error code if so. |
743 | U_ASSERT(laSlot == laSlotForState); |
744 | } |
745 | } |
746 | } |
747 | if (!sawLookAheadNode) { |
748 | continue; |
749 | } |
750 | |
751 | if (laSlotForState == 0) { |
752 | laSlotForState = ++laSlotsInUse; |
753 | } |
754 | |
755 | // For each look ahead node covered by this state, |
756 | // set the mapping from the node's rule number to the look ahead slot. |
757 | // There can be multiple nodes/rule numbers going to the same la slot. |
758 | |
759 | for (int32_t ipos=0; ipos<sd->fPositions->size(); ++ipos) { |
760 | RBBINode *node = static_cast<RBBINode *>(sd->fPositions->elementAt(ipos)); |
761 | if (node->fType != RBBINode::NodeType::lookAhead) { |
762 | continue; |
763 | } |
764 | int32_t ruleNum = node->fVal; // Set when rule was originally parsed. |
765 | int32_t existingVal = fLookAheadRuleMap->elementAti(ruleNum); |
766 | (void)existingVal; |
767 | U_ASSERT(existingVal == 0 || existingVal == laSlotForState); |
768 | fLookAheadRuleMap->setElementAt(laSlotForState, ruleNum); |
769 | } |
770 | } |
771 | |
772 | } |
773 | |
774 | //----------------------------------------------------------------------------- |
775 | // |
776 | // flagAcceptingStates Identify accepting states. |
777 | // First get a list of all of the end marker nodes. |
778 | // Then, for each state s, |
779 | // if s contains one of the end marker nodes in its list of tree positions then |
780 | // s is an accepting state. |
781 | // |
782 | //----------------------------------------------------------------------------- |
783 | void RBBITableBuilder::flagAcceptingStates() { |
784 | if (U_FAILURE(*fStatus)) { |
785 | return; |
786 | } |
787 | UVector endMarkerNodes(*fStatus); |
788 | RBBINode *endMarker; |
789 | int32_t i; |
790 | int32_t n; |
791 | |
792 | if (U_FAILURE(*fStatus)) { |
793 | return; |
794 | } |
795 | |
796 | fTree->findNodes(&endMarkerNodes, RBBINode::endMark, *fStatus); |
797 | if (U_FAILURE(*fStatus)) { |
798 | return; |
799 | } |
800 | |
801 | for (i=0; i<endMarkerNodes.size(); i++) { |
802 | endMarker = (RBBINode *)endMarkerNodes.elementAt(i); |
803 | for (n=0; n<fDStates->size(); n++) { |
804 | RBBIStateDescriptor *sd = (RBBIStateDescriptor *)fDStates->elementAt(n); |
805 | if (sd->fPositions->indexOf(endMarker) >= 0) { |
806 | // Any non-zero value for fAccepting means this is an accepting node. |
807 | // The value is what will be returned to the user as the break status. |
808 | // If no other value was specified, force it to -1. |
809 | |
810 | if (sd->fAccepting==0) { |
811 | // State hasn't been marked as accepting yet. Do it now. |
812 | sd->fAccepting = fLookAheadRuleMap->elementAti(endMarker->fVal); |
813 | if (sd->fAccepting == 0) { |
814 | sd->fAccepting = -1; |
815 | } |
816 | } |
817 | if (sd->fAccepting==-1 && endMarker->fVal != 0) { |
818 | // Both lookahead and non-lookahead accepting for this state. |
819 | // Favor the look-ahead, because a look-ahead match needs to |
820 | // immediately stop the run-time engine. First match, not longest. |
821 | sd->fAccepting = fLookAheadRuleMap->elementAti(endMarker->fVal); |
822 | } |
823 | // implicit else: |
824 | // if sd->fAccepting already had a value other than 0 or -1, leave it be. |
825 | } |
826 | } |
827 | } |
828 | } |
829 | |
830 | |
831 | //----------------------------------------------------------------------------- |
832 | // |
833 | // flagLookAheadStates Very similar to flagAcceptingStates, above. |
834 | // |
835 | //----------------------------------------------------------------------------- |
836 | void RBBITableBuilder::flagLookAheadStates() { |
837 | if (U_FAILURE(*fStatus)) { |
838 | return; |
839 | } |
840 | UVector lookAheadNodes(*fStatus); |
841 | RBBINode *lookAheadNode; |
842 | int32_t i; |
843 | int32_t n; |
844 | |
845 | fTree->findNodes(&lookAheadNodes, RBBINode::lookAhead, *fStatus); |
846 | if (U_FAILURE(*fStatus)) { |
847 | return; |
848 | } |
849 | for (i=0; i<lookAheadNodes.size(); i++) { |
850 | lookAheadNode = (RBBINode *)lookAheadNodes.elementAt(i); |
851 | U_ASSERT(lookAheadNode->fType == RBBINode::NodeType::lookAhead); |
852 | |
853 | for (n=0; n<fDStates->size(); n++) { |
854 | RBBIStateDescriptor *sd = (RBBIStateDescriptor *)fDStates->elementAt(n); |
855 | int32_t positionsIdx = sd->fPositions->indexOf(lookAheadNode); |
856 | if (positionsIdx >= 0) { |
857 | U_ASSERT(lookAheadNode == sd->fPositions->elementAt(positionsIdx)); |
858 | int32_t lookaheadSlot = fLookAheadRuleMap->elementAti(lookAheadNode->fVal); |
859 | U_ASSERT(sd->fLookAhead == 0 || sd->fLookAhead == lookaheadSlot); |
860 | // if (sd->fLookAhead != 0 && sd->fLookAhead != lookaheadSlot) { |
861 | // printf("%s:%d Bingo. sd->fLookAhead:%d lookaheadSlot:%d\n", |
862 | // __FILE__, __LINE__, sd->fLookAhead, lookaheadSlot); |
863 | // } |
864 | sd->fLookAhead = lookaheadSlot; |
865 | } |
866 | } |
867 | } |
868 | } |
869 | |
870 | |
871 | |
872 | |
873 | //----------------------------------------------------------------------------- |
874 | // |
875 | // flagTaggedStates |
876 | // |
877 | //----------------------------------------------------------------------------- |
878 | void RBBITableBuilder::flagTaggedStates() { |
879 | if (U_FAILURE(*fStatus)) { |
880 | return; |
881 | } |
882 | UVector tagNodes(*fStatus); |
883 | RBBINode *tagNode; |
884 | int32_t i; |
885 | int32_t n; |
886 | |
887 | if (U_FAILURE(*fStatus)) { |
888 | return; |
889 | } |
890 | fTree->findNodes(&tagNodes, RBBINode::tag, *fStatus); |
891 | if (U_FAILURE(*fStatus)) { |
892 | return; |
893 | } |
894 | for (i=0; i<tagNodes.size(); i++) { // For each tag node t (all of 'em) |
895 | tagNode = (RBBINode *)tagNodes.elementAt(i); |
896 | |
897 | for (n=0; n<fDStates->size(); n++) { // For each state s (row in the state table) |
898 | RBBIStateDescriptor *sd = (RBBIStateDescriptor *)fDStates->elementAt(n); |
899 | if (sd->fPositions->indexOf(tagNode) >= 0) { // if s include the tag node t |
900 | sortedAdd(&sd->fTagVals, tagNode->fVal); |
901 | } |
902 | } |
903 | } |
904 | } |
905 | |
906 | |
907 | |
908 | |
909 | //----------------------------------------------------------------------------- |
910 | // |
911 | // mergeRuleStatusVals |
912 | // |
913 | // Update the global table of rule status {tag} values |
914 | // The rule builder has a global vector of status values that are common |
915 | // for all tables. Merge the ones from this table into the global set. |
916 | // |
917 | //----------------------------------------------------------------------------- |
918 | void RBBITableBuilder::mergeRuleStatusVals() { |
919 | // |
920 | // The basic outline of what happens here is this... |
921 | // |
922 | // for each state in this state table |
923 | // if the status tag list for this state is in the global statuses list |
924 | // record where and |
925 | // continue with the next state |
926 | // else |
927 | // add the tag list for this state to the global list. |
928 | // |
929 | int i; |
930 | int n; |
931 | |
932 | // Pre-set a single tag of {0} into the table. |
933 | // We will need this as a default, for rule sets with no explicit tagging. |
934 | if (fRB->fRuleStatusVals->size() == 0) { |
935 | fRB->fRuleStatusVals->addElement(1, *fStatus); // Num of statuses in group |
936 | fRB->fRuleStatusVals->addElement((int32_t)0, *fStatus); // and our single status of zero |
937 | } |
938 | |
939 | // For each state |
940 | for (n=0; n<fDStates->size(); n++) { |
941 | RBBIStateDescriptor *sd = (RBBIStateDescriptor *)fDStates->elementAt(n); |
942 | UVector *thisStatesTagValues = sd->fTagVals; |
943 | if (thisStatesTagValues == NULL) { |
944 | // No tag values are explicitly associated with this state. |
945 | // Set the default tag value. |
946 | sd->fTagsIdx = 0; |
947 | continue; |
948 | } |
949 | |
950 | // There are tag(s) associated with this state. |
951 | // fTagsIdx will be the index into the global tag list for this state's tag values. |
952 | // Initial value of -1 flags that we haven't got it set yet. |
953 | sd->fTagsIdx = -1; |
954 | int32_t thisTagGroupStart = 0; // indexes into the global rule status vals list |
955 | int32_t nextTagGroupStart = 0; |
956 | |
957 | // Loop runs once per group of tags in the global list |
958 | while (nextTagGroupStart < fRB->fRuleStatusVals->size()) { |
959 | thisTagGroupStart = nextTagGroupStart; |
960 | nextTagGroupStart += fRB->fRuleStatusVals->elementAti(thisTagGroupStart) + 1; |
961 | if (thisStatesTagValues->size() != fRB->fRuleStatusVals->elementAti(thisTagGroupStart)) { |
962 | // The number of tags for this state is different from |
963 | // the number of tags in this group from the global list. |
964 | // Continue with the next group from the global list. |
965 | continue; |
966 | } |
967 | // The lengths match, go ahead and compare the actual tag values |
968 | // between this state and the group from the global list. |
969 | for (i=0; i<thisStatesTagValues->size(); i++) { |
970 | if (thisStatesTagValues->elementAti(i) != |
971 | fRB->fRuleStatusVals->elementAti(thisTagGroupStart + 1 + i) ) { |
972 | // Mismatch. |
973 | break; |
974 | } |
975 | } |
976 | |
977 | if (i == thisStatesTagValues->size()) { |
978 | // We found a set of tag values in the global list that match |
979 | // those for this state. Use them. |
980 | sd->fTagsIdx = thisTagGroupStart; |
981 | break; |
982 | } |
983 | } |
984 | |
985 | if (sd->fTagsIdx == -1) { |
986 | // No suitable entry in the global tag list already. Add one |
987 | sd->fTagsIdx = fRB->fRuleStatusVals->size(); |
988 | fRB->fRuleStatusVals->addElement(thisStatesTagValues->size(), *fStatus); |
989 | for (i=0; i<thisStatesTagValues->size(); i++) { |
990 | fRB->fRuleStatusVals->addElement(thisStatesTagValues->elementAti(i), *fStatus); |
991 | } |
992 | } |
993 | } |
994 | } |
995 | |
996 | |
997 | |
998 | |
999 | |
1000 | |
1001 | |
1002 | //----------------------------------------------------------------------------- |
1003 | // |
1004 | // sortedAdd Add a value to a vector of sorted values (ints). |
1005 | // Do not replicate entries; if the value is already there, do not |
1006 | // add a second one. |
1007 | // Lazily create the vector if it does not already exist. |
1008 | // |
1009 | //----------------------------------------------------------------------------- |
1010 | void RBBITableBuilder::sortedAdd(UVector **vector, int32_t val) { |
1011 | int32_t i; |
1012 | |
1013 | if (*vector == NULL) { |
1014 | *vector = new UVector(*fStatus); |
1015 | } |
1016 | if (*vector == NULL || U_FAILURE(*fStatus)) { |
1017 | return; |
1018 | } |
1019 | UVector *vec = *vector; |
1020 | int32_t vSize = vec->size(); |
1021 | for (i=0; i<vSize; i++) { |
1022 | int32_t valAtI = vec->elementAti(i); |
1023 | if (valAtI == val) { |
1024 | // The value is already in the vector. Don't add it again. |
1025 | return; |
1026 | } |
1027 | if (valAtI > val) { |
1028 | break; |
1029 | } |
1030 | } |
1031 | vec->insertElementAt(val, i, *fStatus); |
1032 | } |
1033 | |
1034 | |
1035 | |
1036 | //----------------------------------------------------------------------------- |
1037 | // |
1038 | // setAdd Set operation on UVector |
1039 | // dest = dest union source |
1040 | // Elements may only appear once and must be sorted. |
1041 | // |
1042 | //----------------------------------------------------------------------------- |
1043 | void RBBITableBuilder::setAdd(UVector *dest, UVector *source) { |
1044 | int32_t destOriginalSize = dest->size(); |
1045 | int32_t sourceSize = source->size(); |
1046 | int32_t di = 0; |
1047 | MaybeStackArray<void *, 16> destArray, sourceArray; // Handle small cases without malloc |
1048 | void **destPtr, **sourcePtr; |
1049 | void **destLim, **sourceLim; |
1050 | |
1051 | if (destOriginalSize > destArray.getCapacity()) { |
1052 | if (destArray.resize(destOriginalSize) == NULL) { |
1053 | return; |
1054 | } |
1055 | } |
1056 | destPtr = destArray.getAlias(); |
1057 | destLim = destPtr + destOriginalSize; // destArray.getArrayLimit()? |
1058 | |
1059 | if (sourceSize > sourceArray.getCapacity()) { |
1060 | if (sourceArray.resize(sourceSize) == NULL) { |
1061 | return; |
1062 | } |
1063 | } |
1064 | sourcePtr = sourceArray.getAlias(); |
1065 | sourceLim = sourcePtr + sourceSize; // sourceArray.getArrayLimit()? |
1066 | |
1067 | // Avoid multiple "get element" calls by getting the contents into arrays |
1068 | (void) dest->toArray(destPtr); |
1069 | (void) source->toArray(sourcePtr); |
1070 | |
1071 | dest->setSize(sourceSize+destOriginalSize, *fStatus); |
1072 | |
1073 | while (sourcePtr < sourceLim && destPtr < destLim) { |
1074 | if (*destPtr == *sourcePtr) { |
1075 | dest->setElementAt(*sourcePtr++, di++); |
1076 | destPtr++; |
1077 | } |
1078 | // This check is required for machines with segmented memory, like i5/OS. |
1079 | // Direct pointer comparison is not recommended. |
1080 | else if (uprv_memcmp(destPtr, sourcePtr, sizeof(void *)) < 0) { |
1081 | dest->setElementAt(*destPtr++, di++); |
1082 | } |
1083 | else { /* *sourcePtr < *destPtr */ |
1084 | dest->setElementAt(*sourcePtr++, di++); |
1085 | } |
1086 | } |
1087 | |
1088 | // At most one of these two cleanup loops will execute |
1089 | while (destPtr < destLim) { |
1090 | dest->setElementAt(*destPtr++, di++); |
1091 | } |
1092 | while (sourcePtr < sourceLim) { |
1093 | dest->setElementAt(*sourcePtr++, di++); |
1094 | } |
1095 | |
1096 | dest->setSize(di, *fStatus); |
1097 | } |
1098 | |
1099 | |
1100 | |
1101 | //----------------------------------------------------------------------------- |
1102 | // |
1103 | // setEqual Set operation on UVector. |
1104 | // Compare for equality. |
1105 | // Elements must be sorted. |
1106 | // |
1107 | //----------------------------------------------------------------------------- |
1108 | UBool RBBITableBuilder::setEquals(UVector *a, UVector *b) { |
1109 | return a->equals(*b); |
1110 | } |
1111 | |
1112 | |
1113 | //----------------------------------------------------------------------------- |
1114 | // |
1115 | // printPosSets Debug function. Dump Nullable, firstpos, lastpos and followpos |
1116 | // for each node in the tree. |
1117 | // |
1118 | //----------------------------------------------------------------------------- |
1119 | #ifdef RBBI_DEBUG |
1120 | void RBBITableBuilder::printPosSets(RBBINode *n) { |
1121 | if (n==NULL) { |
1122 | return; |
1123 | } |
1124 | printf("\n" ); |
1125 | RBBINode::printNodeHeader(); |
1126 | RBBINode::printNode(n); |
1127 | RBBIDebugPrintf(" Nullable: %s\n" , n->fNullable?"TRUE" :"FALSE" ); |
1128 | |
1129 | RBBIDebugPrintf(" firstpos: " ); |
1130 | printSet(n->fFirstPosSet); |
1131 | |
1132 | RBBIDebugPrintf(" lastpos: " ); |
1133 | printSet(n->fLastPosSet); |
1134 | |
1135 | RBBIDebugPrintf(" followpos: " ); |
1136 | printSet(n->fFollowPos); |
1137 | |
1138 | printPosSets(n->fLeftChild); |
1139 | printPosSets(n->fRightChild); |
1140 | } |
1141 | #endif |
1142 | |
1143 | // |
1144 | // findDuplCharClassFrom() |
1145 | // |
1146 | bool RBBITableBuilder::findDuplCharClassFrom(IntPair *categories) { |
1147 | int32_t numStates = fDStates->size(); |
1148 | int32_t numCols = fRB->fSetBuilder->getNumCharCategories(); |
1149 | |
1150 | for (; categories->first < numCols-1; categories->first++) { |
1151 | for (categories->second=categories->first+1; categories->second < numCols; categories->second++) { |
1152 | // Initialized to different values to prevent returning true if numStates = 0 (implies no duplicates). |
1153 | uint16_t table_base = 0; |
1154 | uint16_t table_dupl = 1; |
1155 | for (int32_t state=0; state<numStates; state++) { |
1156 | RBBIStateDescriptor *sd = (RBBIStateDescriptor *)fDStates->elementAt(state); |
1157 | table_base = (uint16_t)sd->fDtran->elementAti(categories->first); |
1158 | table_dupl = (uint16_t)sd->fDtran->elementAti(categories->second); |
1159 | if (table_base != table_dupl) { |
1160 | break; |
1161 | } |
1162 | } |
1163 | if (table_base == table_dupl) { |
1164 | return true; |
1165 | } |
1166 | } |
1167 | } |
1168 | return false; |
1169 | } |
1170 | |
1171 | |
1172 | // |
1173 | // removeColumn() |
1174 | // |
1175 | void RBBITableBuilder::removeColumn(int32_t column) { |
1176 | int32_t numStates = fDStates->size(); |
1177 | for (int32_t state=0; state<numStates; state++) { |
1178 | RBBIStateDescriptor *sd = (RBBIStateDescriptor *)fDStates->elementAt(state); |
1179 | U_ASSERT(column < sd->fDtran->size()); |
1180 | sd->fDtran->removeElementAt(column); |
1181 | } |
1182 | } |
1183 | |
1184 | /* |
1185 | * findDuplicateState |
1186 | */ |
1187 | bool RBBITableBuilder::findDuplicateState(IntPair *states) { |
1188 | int32_t numStates = fDStates->size(); |
1189 | int32_t numCols = fRB->fSetBuilder->getNumCharCategories(); |
1190 | |
1191 | for (; states->first<numStates-1; states->first++) { |
1192 | RBBIStateDescriptor *firstSD = (RBBIStateDescriptor *)fDStates->elementAt(states->first); |
1193 | for (states->second=states->first+1; states->second<numStates; states->second++) { |
1194 | RBBIStateDescriptor *duplSD = (RBBIStateDescriptor *)fDStates->elementAt(states->second); |
1195 | if (firstSD->fAccepting != duplSD->fAccepting || |
1196 | firstSD->fLookAhead != duplSD->fLookAhead || |
1197 | firstSD->fTagsIdx != duplSD->fTagsIdx) { |
1198 | continue; |
1199 | } |
1200 | bool rowsMatch = true; |
1201 | for (int32_t col=0; col < numCols; ++col) { |
1202 | int32_t firstVal = firstSD->fDtran->elementAti(col); |
1203 | int32_t duplVal = duplSD->fDtran->elementAti(col); |
1204 | if (!((firstVal == duplVal) || |
1205 | ((firstVal == states->first || firstVal == states->second) && |
1206 | (duplVal == states->first || duplVal == states->second)))) { |
1207 | rowsMatch = false; |
1208 | break; |
1209 | } |
1210 | } |
1211 | if (rowsMatch) { |
1212 | return true; |
1213 | } |
1214 | } |
1215 | } |
1216 | return false; |
1217 | } |
1218 | |
1219 | |
1220 | bool RBBITableBuilder::findDuplicateSafeState(IntPair *states) { |
1221 | int32_t numStates = fSafeTable->size(); |
1222 | |
1223 | for (; states->first<numStates-1; states->first++) { |
1224 | UnicodeString *firstRow = static_cast<UnicodeString *>(fSafeTable->elementAt(states->first)); |
1225 | for (states->second=states->first+1; states->second<numStates; states->second++) { |
1226 | UnicodeString *duplRow = static_cast<UnicodeString *>(fSafeTable->elementAt(states->second)); |
1227 | bool rowsMatch = true; |
1228 | int32_t numCols = firstRow->length(); |
1229 | for (int32_t col=0; col < numCols; ++col) { |
1230 | int32_t firstVal = firstRow->charAt(col); |
1231 | int32_t duplVal = duplRow->charAt(col); |
1232 | if (!((firstVal == duplVal) || |
1233 | ((firstVal == states->first || firstVal == states->second) && |
1234 | (duplVal == states->first || duplVal == states->second)))) { |
1235 | rowsMatch = false; |
1236 | break; |
1237 | } |
1238 | } |
1239 | if (rowsMatch) { |
1240 | return true; |
1241 | } |
1242 | } |
1243 | } |
1244 | return false; |
1245 | } |
1246 | |
1247 | |
1248 | void RBBITableBuilder::removeState(IntPair duplStates) { |
1249 | const int32_t keepState = duplStates.first; |
1250 | const int32_t duplState = duplStates.second; |
1251 | U_ASSERT(keepState < duplState); |
1252 | U_ASSERT(duplState < fDStates->size()); |
1253 | |
1254 | RBBIStateDescriptor *duplSD = (RBBIStateDescriptor *)fDStates->elementAt(duplState); |
1255 | fDStates->removeElementAt(duplState); |
1256 | delete duplSD; |
1257 | |
1258 | int32_t numStates = fDStates->size(); |
1259 | int32_t numCols = fRB->fSetBuilder->getNumCharCategories(); |
1260 | for (int32_t state=0; state<numStates; ++state) { |
1261 | RBBIStateDescriptor *sd = (RBBIStateDescriptor *)fDStates->elementAt(state); |
1262 | for (int32_t col=0; col<numCols; col++) { |
1263 | int32_t existingVal = sd->fDtran->elementAti(col); |
1264 | int32_t newVal = existingVal; |
1265 | if (existingVal == duplState) { |
1266 | newVal = keepState; |
1267 | } else if (existingVal > duplState) { |
1268 | newVal = existingVal - 1; |
1269 | } |
1270 | sd->fDtran->setElementAt(newVal, col); |
1271 | } |
1272 | } |
1273 | } |
1274 | |
1275 | void RBBITableBuilder::removeSafeState(IntPair duplStates) { |
1276 | const int32_t keepState = duplStates.first; |
1277 | const int32_t duplState = duplStates.second; |
1278 | U_ASSERT(keepState < duplState); |
1279 | U_ASSERT(duplState < fSafeTable->size()); |
1280 | |
1281 | fSafeTable->removeElementAt(duplState); // Note that fSafeTable has a deleter function |
1282 | // and will auto-delete the removed element. |
1283 | int32_t numStates = fSafeTable->size(); |
1284 | for (int32_t state=0; state<numStates; ++state) { |
1285 | UnicodeString *sd = (UnicodeString *)fSafeTable->elementAt(state); |
1286 | int32_t numCols = sd->length(); |
1287 | for (int32_t col=0; col<numCols; col++) { |
1288 | int32_t existingVal = sd->charAt(col); |
1289 | int32_t newVal = existingVal; |
1290 | if (existingVal == duplState) { |
1291 | newVal = keepState; |
1292 | } else if (existingVal > duplState) { |
1293 | newVal = existingVal - 1; |
1294 | } |
1295 | sd->setCharAt(col, static_cast<char16_t>(newVal)); |
1296 | } |
1297 | } |
1298 | } |
1299 | |
1300 | |
1301 | /* |
1302 | * RemoveDuplicateStates |
1303 | */ |
1304 | int32_t RBBITableBuilder::removeDuplicateStates() { |
1305 | IntPair dupls = {3, 0}; |
1306 | int32_t numStatesRemoved = 0; |
1307 | |
1308 | while (findDuplicateState(&dupls)) { |
1309 | // printf("Removing duplicate states (%d, %d)\n", dupls.first, dupls.second); |
1310 | removeState(dupls); |
1311 | ++numStatesRemoved; |
1312 | } |
1313 | return numStatesRemoved; |
1314 | } |
1315 | |
1316 | |
1317 | //----------------------------------------------------------------------------- |
1318 | // |
1319 | // getTableSize() Calculate the size of the runtime form of this |
1320 | // state transition table. |
1321 | // |
1322 | //----------------------------------------------------------------------------- |
1323 | int32_t RBBITableBuilder::getTableSize() const { |
1324 | int32_t size = 0; |
1325 | int32_t numRows; |
1326 | int32_t numCols; |
1327 | int32_t rowSize; |
1328 | |
1329 | if (fTree == NULL) { |
1330 | return 0; |
1331 | } |
1332 | |
1333 | size = offsetof(RBBIStateTable, fTableData); // The header, with no rows to the table. |
1334 | |
1335 | numRows = fDStates->size(); |
1336 | numCols = fRB->fSetBuilder->getNumCharCategories(); |
1337 | |
1338 | rowSize = offsetof(RBBIStateTableRow, fNextState) + sizeof(uint16_t)*numCols; |
1339 | size += numRows * rowSize; |
1340 | return size; |
1341 | } |
1342 | |
1343 | |
1344 | //----------------------------------------------------------------------------- |
1345 | // |
1346 | // exportTable() export the state transition table in the format required |
1347 | // by the runtime engine. getTableSize() bytes of memory |
1348 | // must be available at the output address "where". |
1349 | // |
1350 | //----------------------------------------------------------------------------- |
1351 | void RBBITableBuilder::exportTable(void *where) { |
1352 | RBBIStateTable *table = (RBBIStateTable *)where; |
1353 | uint32_t state; |
1354 | int col; |
1355 | |
1356 | if (U_FAILURE(*fStatus) || fTree == NULL) { |
1357 | return; |
1358 | } |
1359 | |
1360 | int32_t catCount = fRB->fSetBuilder->getNumCharCategories(); |
1361 | if (catCount > 0x7fff || |
1362 | fDStates->size() > 0x7fff) { |
1363 | *fStatus = U_BRK_INTERNAL_ERROR; |
1364 | return; |
1365 | } |
1366 | |
1367 | table->fRowLen = offsetof(RBBIStateTableRow, fNextState) + sizeof(uint16_t) * catCount; |
1368 | table->fNumStates = fDStates->size(); |
1369 | table->fFlags = 0; |
1370 | if (fRB->fLookAheadHardBreak) { |
1371 | table->fFlags |= RBBI_LOOKAHEAD_HARD_BREAK; |
1372 | } |
1373 | if (fRB->fSetBuilder->sawBOF()) { |
1374 | table->fFlags |= RBBI_BOF_REQUIRED; |
1375 | } |
1376 | table->fReserved = 0; |
1377 | |
1378 | for (state=0; state<table->fNumStates; state++) { |
1379 | RBBIStateDescriptor *sd = (RBBIStateDescriptor *)fDStates->elementAt(state); |
1380 | RBBIStateTableRow *row = (RBBIStateTableRow *)(table->fTableData + state*table->fRowLen); |
1381 | U_ASSERT (-32768 < sd->fAccepting && sd->fAccepting <= 32767); |
1382 | U_ASSERT (-32768 < sd->fLookAhead && sd->fLookAhead <= 32767); |
1383 | row->fAccepting = (int16_t)sd->fAccepting; |
1384 | row->fLookAhead = (int16_t)sd->fLookAhead; |
1385 | row->fTagIdx = (int16_t)sd->fTagsIdx; |
1386 | for (col=0; col<catCount; col++) { |
1387 | row->fNextState[col] = (uint16_t)sd->fDtran->elementAti(col); |
1388 | } |
1389 | } |
1390 | } |
1391 | |
1392 | |
1393 | /** |
1394 | * Synthesize a safe state table from the main state table. |
1395 | */ |
1396 | void RBBITableBuilder::buildSafeReverseTable(UErrorCode &status) { |
1397 | // The safe table creation has three steps: |
1398 | |
1399 | // 1. Identifiy pairs of character classes that are "safe." Safe means that boundaries |
1400 | // following the pair do not depend on context or state before the pair. To test |
1401 | // whether a pair is safe, run it through the main forward state table, starting |
1402 | // from each state. If the the final state is the same, no matter what the starting state, |
1403 | // the pair is safe. |
1404 | // |
1405 | // 2. Build a state table that recognizes the safe pairs. It's similar to their |
1406 | // forward table, with a column for each input character [class], and a row for |
1407 | // each state. Row 1 is the start state, and row 0 is the stop state. Initially |
1408 | // create an additional state for each input character category; being in |
1409 | // one of these states means that the character has been seen, and is potentially |
1410 | // the first of a pair. In each of these rows, the entry for the second character |
1411 | // of a safe pair is set to the stop state (0), indicating that a match was found. |
1412 | // All other table entries are set to the state corresponding the current input |
1413 | // character, allowing that charcter to be the of a start following pair. |
1414 | // |
1415 | // Because the safe rules are to be run in reverse, moving backwards in the text, |
1416 | // the first and second pair categories are swapped when building the table. |
1417 | // |
1418 | // 3. Compress the table. There are typically many rows (states) that are |
1419 | // equivalent - that have zeroes (match completed) in the same columns - |
1420 | // and can be folded together. |
1421 | |
1422 | // Each safe pair is stored as two UChars in the safePair string. |
1423 | UnicodeString safePairs; |
1424 | |
1425 | int32_t numCharClasses = fRB->fSetBuilder->getNumCharCategories(); |
1426 | int32_t numStates = fDStates->size(); |
1427 | |
1428 | for (int32_t c1=0; c1<numCharClasses; ++c1) { |
1429 | for (int32_t c2=0; c2 < numCharClasses; ++c2) { |
1430 | int32_t wantedEndState = -1; |
1431 | int32_t endState = 0; |
1432 | for (int32_t startState = 1; startState < numStates; ++startState) { |
1433 | RBBIStateDescriptor *startStateD = static_cast<RBBIStateDescriptor *>(fDStates->elementAt(startState)); |
1434 | int32_t s2 = startStateD->fDtran->elementAti(c1); |
1435 | RBBIStateDescriptor *s2StateD = static_cast<RBBIStateDescriptor *>(fDStates->elementAt(s2)); |
1436 | endState = s2StateD->fDtran->elementAti(c2); |
1437 | if (wantedEndState < 0) { |
1438 | wantedEndState = endState; |
1439 | } else { |
1440 | if (wantedEndState != endState) { |
1441 | break; |
1442 | } |
1443 | } |
1444 | } |
1445 | if (wantedEndState == endState) { |
1446 | safePairs.append((char16_t)c1); |
1447 | safePairs.append((char16_t)c2); |
1448 | // printf("(%d, %d) ", c1, c2); |
1449 | } |
1450 | } |
1451 | // printf("\n"); |
1452 | } |
1453 | |
1454 | // Populate the initial safe table. |
1455 | // The table as a whole is UVector<UnicodeString> |
1456 | // Each row is represented by a UnicodeString, being used as a Vector<int16>. |
1457 | // Row 0 is the stop state. |
1458 | // Row 1 is the start sate. |
1459 | // Row 2 and beyond are other states, initially one per char class, but |
1460 | // after initial construction, many of the states will be combined, compacting the table. |
1461 | // The String holds the nextState data only. The four leading fields of a row, fAccepting, |
1462 | // fLookAhead, etc. are not needed for the safe table, and are omitted at this stage of building. |
1463 | |
1464 | U_ASSERT(fSafeTable == nullptr); |
1465 | fSafeTable = new UVector(uprv_deleteUObject, uhash_compareUnicodeString, numCharClasses + 2, status); |
1466 | for (int32_t row=0; row<numCharClasses + 2; ++row) { |
1467 | fSafeTable->addElement(new UnicodeString(numCharClasses, 0, numCharClasses+4), status); |
1468 | } |
1469 | |
1470 | // From the start state, each input char class transitions to the state for that input. |
1471 | UnicodeString &startState = *static_cast<UnicodeString *>(fSafeTable->elementAt(1)); |
1472 | for (int32_t charClass=0; charClass < numCharClasses; ++charClass) { |
1473 | // Note: +2 for the start & stop state. |
1474 | startState.setCharAt(charClass, static_cast<char16_t>(charClass+2)); |
1475 | } |
1476 | |
1477 | // Initially make every other state table row look like the start state row, |
1478 | for (int32_t row=2; row<numCharClasses+2; ++row) { |
1479 | UnicodeString &rowState = *static_cast<UnicodeString *>(fSafeTable->elementAt(row)); |
1480 | rowState = startState; // UnicodeString assignment, copies contents. |
1481 | } |
1482 | |
1483 | // Run through the safe pairs, set the next state to zero when pair has been seen. |
1484 | // Zero being the stop state, meaning we found a safe point. |
1485 | for (int32_t pairIdx=0; pairIdx<safePairs.length(); pairIdx+=2) { |
1486 | int32_t c1 = safePairs.charAt(pairIdx); |
1487 | int32_t c2 = safePairs.charAt(pairIdx + 1); |
1488 | |
1489 | UnicodeString &rowState = *static_cast<UnicodeString *>(fSafeTable->elementAt(c2 + 2)); |
1490 | rowState.setCharAt(c1, 0); |
1491 | } |
1492 | |
1493 | // Remove duplicate or redundant rows from the table. |
1494 | IntPair states = {1, 0}; |
1495 | while (findDuplicateSafeState(&states)) { |
1496 | // printf("Removing duplicate safe states (%d, %d)\n", states.first, states.second); |
1497 | removeSafeState(states); |
1498 | } |
1499 | } |
1500 | |
1501 | |
1502 | //----------------------------------------------------------------------------- |
1503 | // |
1504 | // getSafeTableSize() Calculate the size of the runtime form of this |
1505 | // safe state table. |
1506 | // |
1507 | //----------------------------------------------------------------------------- |
1508 | int32_t RBBITableBuilder::getSafeTableSize() const { |
1509 | int32_t size = 0; |
1510 | int32_t numRows; |
1511 | int32_t numCols; |
1512 | int32_t rowSize; |
1513 | |
1514 | if (fSafeTable == nullptr) { |
1515 | return 0; |
1516 | } |
1517 | |
1518 | size = offsetof(RBBIStateTable, fTableData); // The header, with no rows to the table. |
1519 | |
1520 | numRows = fSafeTable->size(); |
1521 | numCols = fRB->fSetBuilder->getNumCharCategories(); |
1522 | |
1523 | rowSize = offsetof(RBBIStateTableRow, fNextState) + sizeof(uint16_t)*numCols; |
1524 | size += numRows * rowSize; |
1525 | return size; |
1526 | } |
1527 | |
1528 | |
1529 | //----------------------------------------------------------------------------- |
1530 | // |
1531 | // exportSafeTable() export the state transition table in the format required |
1532 | // by the runtime engine. getTableSize() bytes of memory |
1533 | // must be available at the output address "where". |
1534 | // |
1535 | //----------------------------------------------------------------------------- |
1536 | void RBBITableBuilder::exportSafeTable(void *where) { |
1537 | RBBIStateTable *table = (RBBIStateTable *)where; |
1538 | uint32_t state; |
1539 | int col; |
1540 | |
1541 | if (U_FAILURE(*fStatus) || fSafeTable == nullptr) { |
1542 | return; |
1543 | } |
1544 | |
1545 | int32_t catCount = fRB->fSetBuilder->getNumCharCategories(); |
1546 | if (catCount > 0x7fff || |
1547 | fSafeTable->size() > 0x7fff) { |
1548 | *fStatus = U_BRK_INTERNAL_ERROR; |
1549 | return; |
1550 | } |
1551 | |
1552 | table->fRowLen = offsetof(RBBIStateTableRow, fNextState) + sizeof(uint16_t) * catCount; |
1553 | table->fNumStates = fSafeTable->size(); |
1554 | table->fFlags = 0; |
1555 | table->fReserved = 0; |
1556 | |
1557 | for (state=0; state<table->fNumStates; state++) { |
1558 | UnicodeString *rowString = (UnicodeString *)fSafeTable->elementAt(state); |
1559 | RBBIStateTableRow *row = (RBBIStateTableRow *)(table->fTableData + state*table->fRowLen); |
1560 | row->fAccepting = 0; |
1561 | row->fLookAhead = 0; |
1562 | row->fTagIdx = 0; |
1563 | row->fReserved = 0; |
1564 | for (col=0; col<catCount; col++) { |
1565 | row->fNextState[col] = rowString->charAt(col); |
1566 | } |
1567 | } |
1568 | } |
1569 | |
1570 | |
1571 | |
1572 | |
1573 | //----------------------------------------------------------------------------- |
1574 | // |
1575 | // printSet Debug function. Print the contents of a UVector |
1576 | // |
1577 | //----------------------------------------------------------------------------- |
1578 | #ifdef RBBI_DEBUG |
1579 | void RBBITableBuilder::printSet(UVector *s) { |
1580 | int32_t i; |
1581 | for (i=0; i<s->size(); i++) { |
1582 | const RBBINode *v = static_cast<const RBBINode *>(s->elementAt(i)); |
1583 | RBBIDebugPrintf("%5d" , v==NULL? -1 : v->fSerialNum); |
1584 | } |
1585 | RBBIDebugPrintf("\n" ); |
1586 | } |
1587 | #endif |
1588 | |
1589 | |
1590 | //----------------------------------------------------------------------------- |
1591 | // |
1592 | // printStates Debug Function. Dump the fully constructed state transition table. |
1593 | // |
1594 | //----------------------------------------------------------------------------- |
1595 | #ifdef RBBI_DEBUG |
1596 | void RBBITableBuilder::printStates() { |
1597 | int c; // input "character" |
1598 | int n; // state number |
1599 | |
1600 | RBBIDebugPrintf("state | i n p u t s y m b o l s \n" ); |
1601 | RBBIDebugPrintf(" | Acc LA Tag" ); |
1602 | for (c=0; c<fRB->fSetBuilder->getNumCharCategories(); c++) { |
1603 | RBBIDebugPrintf(" %2d" , c); |
1604 | } |
1605 | RBBIDebugPrintf("\n" ); |
1606 | RBBIDebugPrintf(" |---------------" ); |
1607 | for (c=0; c<fRB->fSetBuilder->getNumCharCategories(); c++) { |
1608 | RBBIDebugPrintf("---" ); |
1609 | } |
1610 | RBBIDebugPrintf("\n" ); |
1611 | |
1612 | for (n=0; n<fDStates->size(); n++) { |
1613 | RBBIStateDescriptor *sd = (RBBIStateDescriptor *)fDStates->elementAt(n); |
1614 | RBBIDebugPrintf(" %3d | " , n); |
1615 | RBBIDebugPrintf("%3d %3d %5d " , sd->fAccepting, sd->fLookAhead, sd->fTagsIdx); |
1616 | for (c=0; c<fRB->fSetBuilder->getNumCharCategories(); c++) { |
1617 | RBBIDebugPrintf(" %2d" , sd->fDtran->elementAti(c)); |
1618 | } |
1619 | RBBIDebugPrintf("\n" ); |
1620 | } |
1621 | RBBIDebugPrintf("\n\n" ); |
1622 | } |
1623 | #endif |
1624 | |
1625 | |
1626 | //----------------------------------------------------------------------------- |
1627 | // |
1628 | // printSafeTable Debug Function. Dump the fully constructed safe table. |
1629 | // |
1630 | //----------------------------------------------------------------------------- |
1631 | #ifdef RBBI_DEBUG |
1632 | void RBBITableBuilder::printReverseTable() { |
1633 | int c; // input "character" |
1634 | int n; // state number |
1635 | |
1636 | RBBIDebugPrintf(" Safe Reverse Table \n" ); |
1637 | if (fSafeTable == nullptr) { |
1638 | RBBIDebugPrintf(" --- nullptr ---\n" ); |
1639 | return; |
1640 | } |
1641 | RBBIDebugPrintf("state | i n p u t s y m b o l s \n" ); |
1642 | RBBIDebugPrintf(" | Acc LA Tag" ); |
1643 | for (c=0; c<fRB->fSetBuilder->getNumCharCategories(); c++) { |
1644 | RBBIDebugPrintf(" %2d" , c); |
1645 | } |
1646 | RBBIDebugPrintf("\n" ); |
1647 | RBBIDebugPrintf(" |---------------" ); |
1648 | for (c=0; c<fRB->fSetBuilder->getNumCharCategories(); c++) { |
1649 | RBBIDebugPrintf("---" ); |
1650 | } |
1651 | RBBIDebugPrintf("\n" ); |
1652 | |
1653 | for (n=0; n<fSafeTable->size(); n++) { |
1654 | UnicodeString *rowString = (UnicodeString *)fSafeTable->elementAt(n); |
1655 | RBBIDebugPrintf(" %3d | " , n); |
1656 | RBBIDebugPrintf("%3d %3d %5d " , 0, 0, 0); // Accepting, LookAhead, Tags |
1657 | for (c=0; c<fRB->fSetBuilder->getNumCharCategories(); c++) { |
1658 | RBBIDebugPrintf(" %2d" , rowString->charAt(c)); |
1659 | } |
1660 | RBBIDebugPrintf("\n" ); |
1661 | } |
1662 | RBBIDebugPrintf("\n\n" ); |
1663 | } |
1664 | #endif |
1665 | |
1666 | |
1667 | |
1668 | //----------------------------------------------------------------------------- |
1669 | // |
1670 | // printRuleStatusTable Debug Function. Dump the common rule status table |
1671 | // |
1672 | //----------------------------------------------------------------------------- |
1673 | #ifdef RBBI_DEBUG |
1674 | void RBBITableBuilder::printRuleStatusTable() { |
1675 | int32_t thisRecord = 0; |
1676 | int32_t nextRecord = 0; |
1677 | int i; |
1678 | UVector *tbl = fRB->fRuleStatusVals; |
1679 | |
1680 | RBBIDebugPrintf("index | tags \n" ); |
1681 | RBBIDebugPrintf("-------------------\n" ); |
1682 | |
1683 | while (nextRecord < tbl->size()) { |
1684 | thisRecord = nextRecord; |
1685 | nextRecord = thisRecord + tbl->elementAti(thisRecord) + 1; |
1686 | RBBIDebugPrintf("%4d " , thisRecord); |
1687 | for (i=thisRecord+1; i<nextRecord; i++) { |
1688 | RBBIDebugPrintf(" %5d" , tbl->elementAti(i)); |
1689 | } |
1690 | RBBIDebugPrintf("\n" ); |
1691 | } |
1692 | RBBIDebugPrintf("\n\n" ); |
1693 | } |
1694 | #endif |
1695 | |
1696 | |
1697 | //----------------------------------------------------------------------------- |
1698 | // |
1699 | // RBBIStateDescriptor Methods. This is a very struct-like class |
1700 | // Most access is directly to the fields. |
1701 | // |
1702 | //----------------------------------------------------------------------------- |
1703 | |
1704 | RBBIStateDescriptor::RBBIStateDescriptor(int lastInputSymbol, UErrorCode *fStatus) { |
1705 | fMarked = FALSE; |
1706 | fAccepting = 0; |
1707 | fLookAhead = 0; |
1708 | fTagsIdx = 0; |
1709 | fTagVals = NULL; |
1710 | fPositions = NULL; |
1711 | fDtran = NULL; |
1712 | |
1713 | fDtran = new UVector32(lastInputSymbol+1, *fStatus); |
1714 | if (U_FAILURE(*fStatus)) { |
1715 | return; |
1716 | } |
1717 | if (fDtran == NULL) { |
1718 | *fStatus = U_MEMORY_ALLOCATION_ERROR; |
1719 | return; |
1720 | } |
1721 | fDtran->setSize(lastInputSymbol+1); // fDtran needs to be pre-sized. |
1722 | // It is indexed by input symbols, and will |
1723 | // hold the next state number for each |
1724 | // symbol. |
1725 | } |
1726 | |
1727 | |
1728 | RBBIStateDescriptor::~RBBIStateDescriptor() { |
1729 | delete fPositions; |
1730 | delete fDtran; |
1731 | delete fTagVals; |
1732 | fPositions = NULL; |
1733 | fDtran = NULL; |
1734 | fTagVals = NULL; |
1735 | } |
1736 | |
1737 | U_NAMESPACE_END |
1738 | |
1739 | #endif /* #if !UCONFIG_NO_BREAK_ITERATION */ |
1740 | |