added an item.
[ragel.git] / redfsm / gendata.cpp
blobdac4f6b2fa0e12a02767c5e115a047bc1312c3ae
1 /*
2 * Copyright 2005-2007 Adrian Thurston <thurston@cs.queensu.ca>
3 */
5 /* This file is part of Ragel.
7 * Ragel is free software; you can redistribute it and/or modify
8 * it under the terms of the GNU General Public License as published by
9 * the Free Software Foundation; either version 2 of the License, or
10 * (at your option) any later version.
12 * Ragel is distributed in the hope that it will be useful,
13 * but WITHOUT ANY WARRANTY; without even the implied warranty of
14 * MERCHANTABILITY or FITNESS FOR A PARTICULAR PURPOSE. See the
15 * GNU General Public License for more details.
17 * You should have received a copy of the GNU General Public License
18 * along with Ragel; if not, write to the Free Software
19 * Foundation, Inc., 59 Temple Place, Suite 330, Boston, MA 02111-1307 USA
22 #include "gendata.h"
23 #include <iostream>
25 using std::cerr;
26 using std::endl;
28 /* Total error count. */
29 int gblErrorCount = 0;
31 CodeGenData::CodeGenData( ostream &out )
33 sourceFileName(0),
34 fsmName(0),
35 out(out),
36 redFsm(0),
37 allActions(0),
38 allActionTables(0),
39 allConditions(0),
40 allCondSpaces(0),
41 allStates(0),
42 nameIndex(0),
43 startState(-1),
44 errState(-1),
45 getKeyExpr(0),
46 accessExpr(0),
47 pExpr(0),
48 peExpr(0),
49 csExpr(0),
50 topExpr(0),
51 stackExpr(0),
52 actExpr(0),
53 tokstartExpr(0),
54 tokendExpr(0),
55 wantComplete(0),
56 hasLongestMatch(false),
57 hasEnd(true),
58 dataPrefix(true),
59 writeFirstFinal(true),
60 writeErr(true),
61 writeCS(true)
65 void CodeGenData::createMachine()
67 redFsm = new RedFsmAp();
70 void CodeGenData::initActionList( unsigned long length )
72 allActions = new Action[length];
73 for ( unsigned long a = 0; a < length; a++ )
74 actionList.append( allActions+a );
77 void CodeGenData::newAction( int anum, char *name, int line,
78 int col, InlineList *inlineList )
80 allActions[anum].actionId = anum;
81 allActions[anum].name = name;
82 allActions[anum].loc.line = line;
83 allActions[anum].loc.col = col;
84 allActions[anum].inlineList = inlineList;
87 void CodeGenData::initActionTableList( unsigned long length )
89 allActionTables = new RedAction[length];
92 void CodeGenData::initStateList( unsigned long length )
94 allStates = new RedStateAp[length];
95 for ( unsigned long s = 0; s < length; s++ )
96 redFsm->stateList.append( allStates+s );
98 /* We get the start state as an offset, set the pointer now. */
99 if ( startState >= 0 )
100 redFsm->startState = allStates + startState;
101 if ( errState >= 0 )
102 redFsm->errState = allStates + errState;
103 for ( EntryIdVect::Iter en = entryPointIds; en.lte(); en++ )
104 redFsm->entryPoints.insert( allStates + *en );
106 /* The nextStateId is no longer used to assign state ids (they come in set
107 * from the frontend now), however generation code still depends on it.
108 * Should eventually remove this variable. */
109 redFsm->nextStateId = redFsm->stateList.length();
112 void CodeGenData::setStartState( unsigned long startState )
114 this->startState = startState;
117 void CodeGenData::setErrorState( unsigned long errState )
119 this->errState = errState;
122 void CodeGenData::addEntryPoint( char *name, unsigned long entryState )
124 entryPointIds.append( entryState );
125 entryPointNames.append( name );
128 void CodeGenData::initTransList( int snum, unsigned long length )
130 /* Could preallocate the out range to save time growing it. For now do
131 * nothing. */
134 void CodeGenData::newTrans( int snum, int tnum, Key lowKey,
135 Key highKey, long targ, long action )
137 /* Get the current state and range. */
138 RedStateAp *curState = allStates + snum;
139 RedTransList &destRange = curState->outRange;
141 if ( curState == redFsm->errState )
142 return;
144 /* Make the new transitions. */
145 RedStateAp *targState = targ >= 0 ? (allStates + targ) :
146 wantComplete ? redFsm->getErrorState() : 0;
147 RedAction *actionTable = action >= 0 ? (allActionTables + action) : 0;
148 RedTransAp *trans = redFsm->allocateTrans( targState, actionTable );
149 RedTransEl transEl( lowKey, highKey, trans );
151 if ( wantComplete ) {
152 /* If the machine is to be complete then we need to fill any gaps with
153 * the error transitions. */
154 if ( destRange.length() == 0 ) {
155 /* Range is currently empty. */
156 if ( keyOps->minKey < lowKey ) {
157 /* The first range doesn't start at the low end. */
158 Key fillHighKey = lowKey;
159 fillHighKey.decrement();
161 /* Create the filler with the state's error transition. */
162 RedTransEl newTel( keyOps->minKey, fillHighKey, redFsm->getErrorTrans() );
163 destRange.append( newTel );
166 else {
167 /* The range list is not empty, get the the last range. */
168 RedTransEl *last = &destRange[destRange.length()-1];
169 Key nextKey = last->highKey;
170 nextKey.increment();
171 if ( nextKey < lowKey ) {
172 /* There is a gap to fill. Make the high key. */
173 Key fillHighKey = lowKey;
174 fillHighKey.decrement();
176 /* Create the filler with the state's error transtion. */
177 RedTransEl newTel( nextKey, fillHighKey, redFsm->getErrorTrans() );
178 destRange.append( newTel );
183 /* Filler taken care of. Append the range. */
184 destRange.append( RedTransEl( lowKey, highKey, trans ) );
187 void CodeGenData::finishTransList( int snum )
189 /* Get the current state and range. */
190 RedStateAp *curState = allStates + snum;
191 RedTransList &destRange = curState->outRange;
193 if ( curState == redFsm->errState )
194 return;
196 /* If building a complete machine we may need filler on the end. */
197 if ( wantComplete ) {
198 /* Check if there are any ranges already. */
199 if ( destRange.length() == 0 ) {
200 /* Fill with the whole alphabet. */
201 /* Add the range on the lower and upper bound. */
202 RedTransEl newTel( keyOps->minKey, keyOps->maxKey, redFsm->getErrorTrans() );
203 destRange.append( newTel );
205 else {
206 /* Get the last and check for a gap on the end. */
207 RedTransEl *last = &destRange[destRange.length()-1];
208 if ( last->highKey < keyOps->maxKey ) {
209 /* Make the high key. */
210 Key fillLowKey = last->highKey;
211 fillLowKey.increment();
213 /* Create the new range with the error trans and append it. */
214 RedTransEl newTel( fillLowKey, keyOps->maxKey, redFsm->getErrorTrans() );
215 destRange.append( newTel );
221 void CodeGenData::setId( int snum, int id )
223 RedStateAp *curState = allStates + snum;
224 curState->id = id;
227 void CodeGenData::setFinal( int snum )
229 RedStateAp *curState = allStates + snum;
230 curState->isFinal = true;
234 void CodeGenData::setStateActions( int snum, long toStateAction,
235 long fromStateAction, long eofAction )
237 RedStateAp *curState = allStates + snum;
238 if ( toStateAction >= 0 )
239 curState->toStateAction = allActionTables + toStateAction;
240 if ( fromStateAction >= 0 )
241 curState->fromStateAction = allActionTables + fromStateAction;
242 if ( eofAction >= 0 )
243 curState->eofAction = allActionTables + eofAction;
246 void CodeGenData::resolveTargetStates( InlineList *inlineList )
248 for ( InlineList::Iter item = *inlineList; item.lte(); item++ ) {
249 switch ( item->type ) {
250 case InlineItem::Goto: case InlineItem::Call:
251 case InlineItem::Next: case InlineItem::Entry:
252 item->targState = allStates + item->targId;
253 break;
254 default:
255 break;
258 if ( item->children != 0 )
259 resolveTargetStates( item->children );
263 void CodeGenData::closeMachine()
265 for ( ActionList::Iter a = actionList; a.lte(); a++ )
266 resolveTargetStates( a->inlineList );
268 /* Note that even if we want a complete graph we do not give the error
269 * state a default transition. All machines break out of the processing
270 * loop when in the error state. */
272 for ( RedStateList::Iter st = redFsm->stateList; st.lte(); st++ ) {
273 for ( StateCondList::Iter sci = st->stateCondList; sci.lte(); sci++ )
274 st->stateCondVect.append( sci );
279 bool CodeGenData::setAlphType( char *data )
281 HostType *alphType = findAlphTypeInternal( data );
282 if ( alphType == 0 )
283 return false;
285 thisKeyOps.setAlphType( alphType );
286 return true;
289 void CodeGenData::initCondSpaceList( ulong length )
291 allCondSpaces = new CondSpace[length];
292 for ( ulong c = 0; c < length; c++ )
293 condSpaceList.append( allCondSpaces + c );
296 void CodeGenData::newCondSpace( int cnum, int condSpaceId, Key baseKey )
298 CondSpace *cond = allCondSpaces + cnum;
299 cond->condSpaceId = condSpaceId;
300 cond->baseKey = baseKey;
303 void CodeGenData::condSpaceItem( int cnum, long condActionId )
305 CondSpace *cond = allCondSpaces + cnum;
306 cond->condSet.append( allActions + condActionId );
309 void CodeGenData::initStateCondList( int snum, ulong length )
311 /* Could preallocate these, as we could with transitions. */
314 void CodeGenData::addStateCond( int snum, Key lowKey, Key highKey, long condNum )
316 RedStateAp *curState = allStates + snum;
318 /* Create the new state condition. */
319 StateCond *stateCond = new StateCond;
320 stateCond->lowKey = lowKey;
321 stateCond->highKey = highKey;
323 /* Assign it a cond space. */
324 CondSpace *condSpace = allCondSpaces + condNum;
325 stateCond->condSpace = condSpace;
327 curState->stateCondList.append( stateCond );
331 CondSpace *CodeGenData::findCondSpace( Key lowKey, Key highKey )
333 for ( CondSpaceList::Iter cs = condSpaceList; cs.lte(); cs++ ) {
334 Key csHighKey = cs->baseKey;
335 csHighKey += keyOps->alphSize() * (1 << cs->condSet.length());
337 if ( lowKey >= cs->baseKey && highKey <= csHighKey )
338 return cs;
340 return 0;
343 Condition *CodeGenData::findCondition( Key key )
345 for ( ConditionList::Iter cond = conditionList; cond.lte(); cond++ ) {
346 Key upperKey = cond->baseKey + (1 << cond->condSet.length());
347 if ( cond->baseKey <= key && key <= upperKey )
348 return cond;
350 return 0;
353 Key CodeGenData::findMaxKey()
355 Key maxKey = keyOps->maxKey;
356 for ( RedStateList::Iter st = redFsm->stateList; st.lte(); st++ ) {
357 assert( st->outSingle.length() == 0 );
358 assert( st->defTrans == 0 );
360 long rangeLen = st->outRange.length();
361 if ( rangeLen > 0 ) {
362 Key highKey = st->outRange[rangeLen-1].highKey;
363 if ( highKey > maxKey )
364 maxKey = highKey;
367 return maxKey;
370 void CodeGenData::findFinalActionRefs()
372 for ( RedStateList::Iter st = redFsm->stateList; st.lte(); st++ ) {
373 /* Rerence count out of single transitions. */
374 for ( RedTransList::Iter rtel = st->outSingle; rtel.lte(); rtel++ ) {
375 if ( rtel->value->action != 0 ) {
376 rtel->value->action->numTransRefs += 1;
377 for ( ActionTable::Iter item = rtel->value->action->key; item.lte(); item++ )
378 item->value->numTransRefs += 1;
382 /* Reference count out of range transitions. */
383 for ( RedTransList::Iter rtel = st->outRange; rtel.lte(); rtel++ ) {
384 if ( rtel->value->action != 0 ) {
385 rtel->value->action->numTransRefs += 1;
386 for ( ActionTable::Iter item = rtel->value->action->key; item.lte(); item++ )
387 item->value->numTransRefs += 1;
391 /* Reference count default transition. */
392 if ( st->defTrans != 0 && st->defTrans->action != 0 ) {
393 st->defTrans->action->numTransRefs += 1;
394 for ( ActionTable::Iter item = st->defTrans->action->key; item.lte(); item++ )
395 item->value->numTransRefs += 1;
398 /* Reference count to state actions. */
399 if ( st->toStateAction != 0 ) {
400 st->toStateAction->numToStateRefs += 1;
401 for ( ActionTable::Iter item = st->toStateAction->key; item.lte(); item++ )
402 item->value->numToStateRefs += 1;
405 /* Reference count from state actions. */
406 if ( st->fromStateAction != 0 ) {
407 st->fromStateAction->numFromStateRefs += 1;
408 for ( ActionTable::Iter item = st->fromStateAction->key; item.lte(); item++ )
409 item->value->numFromStateRefs += 1;
412 /* Reference count EOF actions. */
413 if ( st->eofAction != 0 ) {
414 st->eofAction->numEofRefs += 1;
415 for ( ActionTable::Iter item = st->eofAction->key; item.lte(); item++ )
416 item->value->numEofRefs += 1;
421 void CodeGenData::analyzeAction( Action *act, InlineList *inlineList )
423 for ( InlineList::Iter item = *inlineList; item.lte(); item++ ) {
424 /* Only consider actions that are referenced. */
425 if ( act->numRefs() > 0 ) {
426 if ( item->type == InlineItem::Goto || item->type == InlineItem::GotoExpr )
427 redFsm->bAnyActionGotos = true;
428 else if ( item->type == InlineItem::Call || item->type == InlineItem::CallExpr )
429 redFsm->bAnyActionCalls = true;
430 else if ( item->type == InlineItem::Ret )
431 redFsm->bAnyActionRets = true;
434 /* Check for various things in regular actions. */
435 if ( act->numTransRefs > 0 || act->numToStateRefs > 0 || act->numFromStateRefs > 0 ) {
436 /* Any returns in regular actions? */
437 if ( item->type == InlineItem::Ret )
438 redFsm->bAnyRegActionRets = true;
440 /* Any next statements in the regular actions? */
441 if ( item->type == InlineItem::Next || item->type == InlineItem::NextExpr )
442 redFsm->bAnyRegNextStmt = true;
444 /* Any by value control in regular actions? */
445 if ( item->type == InlineItem::CallExpr || item->type == InlineItem::GotoExpr )
446 redFsm->bAnyRegActionByValControl = true;
448 /* Any references to the current state in regular actions? */
449 if ( item->type == InlineItem::Curs )
450 redFsm->bAnyRegCurStateRef = true;
452 if ( item->type == InlineItem::Break )
453 redFsm->bAnyRegBreak = true;
456 if ( item->children != 0 )
457 analyzeAction( act, item->children );
461 void CodeGenData::analyzeActionList( RedAction *redAct, InlineList *inlineList )
463 for ( InlineList::Iter item = *inlineList; item.lte(); item++ ) {
464 /* Any next statements in the action table? */
465 if ( item->type == InlineItem::Next || item->type == InlineItem::NextExpr )
466 redAct->bAnyNextStmt = true;
468 /* Any references to the current state. */
469 if ( item->type == InlineItem::Curs )
470 redAct->bAnyCurStateRef = true;
472 if ( item->type == InlineItem::Break )
473 redAct->bAnyBreakStmt = true;
475 if ( item->children != 0 )
476 analyzeActionList( redAct, item->children );
480 /* Assign ids to referenced actions. */
481 void CodeGenData::assignActionIds()
483 int nextActionId = 0;
484 for ( ActionList::Iter act = actionList; act.lte(); act++ ) {
485 /* Only ever interested in referenced actions. */
486 if ( act->numRefs() > 0 )
487 act->actionId = nextActionId++;
491 void CodeGenData::setValueLimits()
493 redFsm->maxSingleLen = 0;
494 redFsm->maxRangeLen = 0;
495 redFsm->maxKeyOffset = 0;
496 redFsm->maxIndexOffset = 0;
497 redFsm->maxActListId = 0;
498 redFsm->maxActionLoc = 0;
499 redFsm->maxActArrItem = 0;
500 redFsm->maxSpan = 0;
501 redFsm->maxCondSpan = 0;
502 redFsm->maxFlatIndexOffset = 0;
503 redFsm->maxCondOffset = 0;
504 redFsm->maxCondLen = 0;
505 redFsm->maxCondSpaceId = 0;
506 redFsm->maxCondIndexOffset = 0;
508 /* In both of these cases the 0 index is reserved for no value, so the max
509 * is one more than it would be if they started at 0. */
510 redFsm->maxIndex = redFsm->transSet.length();
511 redFsm->maxCond = condSpaceList.length();
513 /* The nextStateId - 1 is the last state id assigned. */
514 redFsm->maxState = redFsm->nextStateId - 1;
516 for ( CondSpaceList::Iter csi = condSpaceList; csi.lte(); csi++ ) {
517 if ( csi->condSpaceId > redFsm->maxCondSpaceId )
518 redFsm->maxCondSpaceId = csi->condSpaceId;
521 for ( RedStateList::Iter st = redFsm->stateList; st.lte(); st++ ) {
522 /* Maximum cond length. */
523 if ( st->stateCondList.length() > redFsm->maxCondLen )
524 redFsm->maxCondLen = st->stateCondList.length();
526 /* Maximum single length. */
527 if ( st->outSingle.length() > redFsm->maxSingleLen )
528 redFsm->maxSingleLen = st->outSingle.length();
530 /* Maximum range length. */
531 if ( st->outRange.length() > redFsm->maxRangeLen )
532 redFsm->maxRangeLen = st->outRange.length();
534 /* The key offset index offset for the state after last is not used, skip it.. */
535 if ( ! st.last() ) {
536 redFsm->maxCondOffset += st->stateCondList.length();
537 redFsm->maxKeyOffset += st->outSingle.length() + st->outRange.length()*2;
538 redFsm->maxIndexOffset += st->outSingle.length() + st->outRange.length() + 1;
541 /* Max cond span. */
542 if ( st->condList != 0 ) {
543 unsigned long long span = keyOps->span( st->condLowKey, st->condHighKey );
544 if ( span > redFsm->maxCondSpan )
545 redFsm->maxCondSpan = span;
548 /* Max key span. */
549 if ( st->transList != 0 ) {
550 unsigned long long span = keyOps->span( st->lowKey, st->highKey );
551 if ( span > redFsm->maxSpan )
552 redFsm->maxSpan = span;
555 /* Max cond index offset. */
556 if ( ! st.last() ) {
557 if ( st->condList != 0 )
558 redFsm->maxCondIndexOffset += keyOps->span( st->condLowKey, st->condHighKey );
561 /* Max flat index offset. */
562 if ( ! st.last() ) {
563 if ( st->transList != 0 )
564 redFsm->maxFlatIndexOffset += keyOps->span( st->lowKey, st->highKey );
565 redFsm->maxFlatIndexOffset += 1;
569 for ( ActionTableMap::Iter at = redFsm->actionMap; at.lte(); at++ ) {
570 /* Maximum id of action lists. */
571 if ( at->actListId+1 > redFsm->maxActListId )
572 redFsm->maxActListId = at->actListId+1;
574 /* Maximum location of items in action array. */
575 if ( at->location+1 > redFsm->maxActionLoc )
576 redFsm->maxActionLoc = at->location+1;
578 /* Maximum values going into the action array. */
579 if ( at->key.length() > redFsm->maxActArrItem )
580 redFsm->maxActArrItem = at->key.length();
581 for ( ActionTable::Iter item = at->key; item.lte(); item++ ) {
582 if ( item->value->actionId > redFsm->maxActArrItem )
583 redFsm->maxActArrItem = item->value->actionId;
590 /* Gather various info on the machine. */
591 void CodeGenData::analyzeMachine()
593 /* Find the true count of action references. */
594 findFinalActionRefs();
596 /* Check if there are any calls in action code. */
597 for ( ActionList::Iter act = actionList; act.lte(); act++ ) {
598 /* Record the occurrence of various kinds of actions. */
599 if ( act->numToStateRefs > 0 )
600 redFsm->bAnyToStateActions = true;
601 if ( act->numFromStateRefs > 0 )
602 redFsm->bAnyFromStateActions = true;
603 if ( act->numEofRefs > 0 )
604 redFsm->bAnyEofActions = true;
605 if ( act->numTransRefs > 0 )
606 redFsm->bAnyRegActions = true;
608 /* Recurse through the action's parse tree looking for various things. */
609 analyzeAction( act, act->inlineList );
612 /* Analyze reduced action lists. */
613 for ( ActionTableMap::Iter redAct = redFsm->actionMap; redAct.lte(); redAct++ ) {
614 for ( ActionTable::Iter act = redAct->key; act.lte(); act++ )
615 analyzeActionList( redAct, act->value->inlineList );
618 /* Find states that have transitions with actions that have next
619 * statements. */
620 for ( RedStateList::Iter st = redFsm->stateList; st.lte(); st++ ) {
621 /* Check any actions out of outSinge. */
622 for ( RedTransList::Iter rtel = st->outSingle; rtel.lte(); rtel++ ) {
623 if ( rtel->value->action != 0 && rtel->value->action->anyCurStateRef() )
624 st->bAnyRegCurStateRef = true;
627 /* Check any actions out of outRange. */
628 for ( RedTransList::Iter rtel = st->outRange; rtel.lte(); rtel++ ) {
629 if ( rtel->value->action != 0 && rtel->value->action->anyCurStateRef() )
630 st->bAnyRegCurStateRef = true;
633 /* Check any action out of default. */
634 if ( st->defTrans != 0 && st->defTrans->action != 0 &&
635 st->defTrans->action->anyCurStateRef() )
636 st->bAnyRegCurStateRef = true;
638 if ( st->stateCondList.length() > 0 )
639 redFsm->bAnyConditions = true;
642 /* Assign ids to actions that are referenced. */
643 assignActionIds();
645 /* Set the maximums of various values used for deciding types. */
646 setValueLimits();
649 void CodeGenData::writeStatement( InputLoc &loc, int nargs, char **args )
651 /* FIXME: This should be moved to the virtual functions in the code
652 * generators.
654 * Force a newline. */
655 out << "\n";
656 genLineDirective( out );
658 if ( strcmp( args[0], "data" ) == 0 ) {
659 for ( int i = 1; i < nargs; i++ ) {
660 if ( strcmp( args[i], "noerror" ) == 0 )
661 writeErr = false;
662 else if ( strcmp( args[i], "noprefix" ) == 0 )
663 dataPrefix = false;
664 else if ( strcmp( args[i], "nofinal" ) == 0 )
665 writeFirstFinal = false;
666 else {
667 source_warning(loc) << "unrecognized write option \"" <<
668 args[i] << "\"" << endl;
671 writeData();
673 else if ( strcmp( args[0], "init" ) == 0 ) {
674 for ( int i = 1; i < nargs; i++ ) {
675 if ( strcmp( args[i], "nocs" ) == 0 )
676 writeCS = false;
677 else {
678 source_warning(loc) << "unrecognized write option \"" <<
679 args[i] << "\"" << endl;
682 writeInit();
684 else if ( strcmp( args[0], "exec" ) == 0 ) {
685 for ( int i = 1; i < nargs; i++ ) {
686 if ( strcmp( args[i], "noend" ) == 0 )
687 hasEnd = false;
688 else {
689 source_warning(loc) << "unrecognized write option \"" <<
690 args[i] << "\"" << endl;
693 writeExec();
695 else if ( strcmp( args[0], "eof" ) == 0 ) {
696 for ( int i = 1; i < nargs; i++ ) {
697 source_warning(loc) << "unrecognized write option \"" <<
698 args[i] << "\"" << endl;
700 writeEOF();
702 else if ( strcmp( args[0], "exports" ) == 0 ) {
703 for ( int i = 1; i < nargs; i++ ) {
704 source_warning(loc) << "unrecognized write option \"" <<
705 args[i] << "\"" << endl;
707 writeExports();
709 else {
710 /* EMIT An error here. */
711 source_error(loc) << "unrecognized write command \"" <<
712 args[0] << "\"" << endl;
716 ostream &CodeGenData::source_warning( const InputLoc &loc )
718 cerr << sourceFileName << ":" << loc.line << ":" << loc.col << ": warning: ";
719 return cerr;
722 ostream &CodeGenData::source_error( const InputLoc &loc )
724 gblErrorCount += 1;
725 assert( sourceFileName != 0 );
726 cerr << sourceFileName << ":" << loc.line << ":" << loc.col << ": ";
727 return cerr;