use of org.antlr.v4.runtime.atn.SetTransition in project antlr4 by tunnelvisionlabs.
the class ATNOptimizer method optimizeSets.
private static void optimizeSets(Grammar g, ATN atn) {
if (g.isParser()) {
// parser codegen doesn't currently support SetTransition
return;
}
int removedStates = 0;
List<DecisionState> decisions = atn.decisionToState;
for (DecisionState decision : decisions) {
if (decision.ruleIndex >= 0) {
Rule rule = g.getRule(decision.ruleIndex);
if (Character.isLowerCase(rule.name.charAt(0))) {
// parser codegen doesn't currently support SetTransition
continue;
}
}
IntervalSet setTransitions = new IntervalSet();
for (int i = 0; i < decision.getNumberOfTransitions(); i++) {
Transition epsTransition = decision.transition(i);
if (!(epsTransition instanceof EpsilonTransition)) {
continue;
}
if (epsTransition.target.getNumberOfTransitions() != 1) {
continue;
}
Transition transition = epsTransition.target.transition(0);
if (!(transition.target instanceof BlockEndState)) {
continue;
}
if (transition instanceof NotSetTransition) {
// TODO: not yet implemented
continue;
}
if (transition instanceof AtomTransition || transition instanceof RangeTransition || transition instanceof SetTransition) {
setTransitions.add(i);
}
}
// due to min alt resolution policies, can only collapse sequential alts
for (int i = setTransitions.getIntervals().size() - 1; i >= 0; i--) {
Interval interval = setTransitions.getIntervals().get(i);
if (interval.length() <= 1) {
continue;
}
ATNState blockEndState = decision.transition(interval.a).target.transition(0).target;
IntervalSet matchSet = new IntervalSet();
for (int j = interval.a; j <= interval.b; j++) {
Transition matchTransition = decision.transition(j).target.transition(0);
if (matchTransition instanceof NotSetTransition) {
throw new UnsupportedOperationException("Not yet implemented.");
}
IntervalSet set = matchTransition.label();
List<Interval> intervals = set.getIntervals();
int n = intervals.size();
for (int k = 0; k < n; k++) {
Interval setInterval = intervals.get(k);
int a = setInterval.a;
int b = setInterval.b;
if (a != -1 && b != -1) {
for (int v = a; v <= b; v++) {
if (matchSet.contains(v)) {
// TODO: Token is missing (i.e. position in source will not be displayed).
g.tool.errMgr.grammarError(ErrorType.CHARACTERS_COLLISION_IN_SET, g.fileName, null, CharSupport.getANTLRCharLiteralForChar(v), CharSupport.getIntervalSetEscapedString(matchSet));
break;
}
}
}
}
matchSet.addAll(set);
}
Transition newTransition;
if (matchSet.getIntervals().size() == 1) {
if (matchSet.size() == 1) {
newTransition = CodePointTransitions.createWithCodePoint(blockEndState, matchSet.getMinElement());
} else {
Interval matchInterval = matchSet.getIntervals().get(0);
newTransition = CodePointTransitions.createWithCodePointRange(blockEndState, matchInterval.a, matchInterval.b);
}
} else {
newTransition = new SetTransition(blockEndState, matchSet);
}
decision.transition(interval.a).target.setTransition(0, newTransition);
for (int j = interval.a + 1; j <= interval.b; j++) {
Transition removed = decision.removeTransition(interval.a + 1);
atn.removeState(removed.target);
removedStates++;
}
}
}
// System.out.println("ATN optimizer removed " + removedStates + " states by collapsing sets.");
}
use of org.antlr.v4.runtime.atn.SetTransition in project antlr4 by tunnelvisionlabs.
the class LexerATNFactory method charSetLiteral.
/**
* [Aa\t \u1234a-z\]\p{Letter}\-] char sets
*/
@Override
public Handle charSetLiteral(GrammarAST charSetAST) {
ATNState left = newState(charSetAST);
ATNState right = newState(charSetAST);
IntervalSet set = getSetFromCharSetLiteral(charSetAST);
left.addTransition(new SetTransition(right, set));
charSetAST.atnState = left;
return new Handle(left, right);
}
use of org.antlr.v4.runtime.atn.SetTransition in project antlr4 by tunnelvisionlabs.
the class LexerATNFactory method set.
@Override
public Handle set(GrammarAST associatedAST, List<GrammarAST> alts, boolean invert) {
ATNState left = newState(associatedAST);
ATNState right = newState(associatedAST);
IntervalSet set = new IntervalSet();
for (GrammarAST t : alts) {
if (t.getType() == ANTLRParser.RANGE) {
int a = CharSupport.getCharValueFromGrammarCharLiteral(t.getChild(0).getText());
int b = CharSupport.getCharValueFromGrammarCharLiteral(t.getChild(1).getText());
if (checkRange((GrammarAST) t.getChild(0), (GrammarAST) t.getChild(1), a, b)) {
checkSetCollision(associatedAST, set, a, b);
set.add(a, b);
}
} else if (t.getType() == ANTLRParser.LEXER_CHAR_SET) {
set.addAll(getSetFromCharSetLiteral(t));
} else if (t.getType() == ANTLRParser.STRING_LITERAL) {
int c = CharSupport.getCharValueFromGrammarCharLiteral(t.getText());
if (c != -1) {
checkSetCollision(associatedAST, set, c);
set.add(c);
} else {
g.tool.errMgr.grammarError(ErrorType.INVALID_LITERAL_IN_LEXER_SET, g.fileName, t.getToken(), t.getText());
}
} else if (t.getType() == ANTLRParser.TOKEN_REF) {
g.tool.errMgr.grammarError(ErrorType.UNSUPPORTED_REFERENCE_IN_LEXER_SET, g.fileName, t.getToken(), t.getText());
}
}
if (invert) {
left.addTransition(new NotSetTransition(right, set));
} else {
Transition transition;
if (set.getIntervals().size() == 1) {
Interval interval = set.getIntervals().get(0);
transition = CodePointTransitions.createWithCodePointRange(right, interval.a, interval.b);
} else {
transition = new SetTransition(right, set);
}
left.addTransition(transition);
}
associatedAST.atnState = left;
return new Handle(left, right);
}
use of org.antlr.v4.runtime.atn.SetTransition in project antlr4 by tunnelvisionlabs.
the class ParserATNFactory method set.
/**
* From set build single edge graph {@code o->o-set->o}. To conform to
* what an alt block looks like, must have extra state on left.
* This also handles {@code ~A}, converted to {@code ~{A}} set.
*/
@NotNull
@Override
public Handle set(@NotNull GrammarAST associatedAST, @NotNull List<GrammarAST> terminals, boolean invert) {
ATNState left = newState(associatedAST);
ATNState right = newState(associatedAST);
IntervalSet set = new IntervalSet();
for (GrammarAST t : terminals) {
int ttype = g.getTokenType(t.getText());
set.add(ttype);
}
if (invert) {
left.addTransition(new NotSetTransition(right, set));
} else {
left.addTransition(new SetTransition(right, set));
}
associatedAST.atnState = left;
return new Handle(left, right);
}
use of org.antlr.v4.runtime.atn.SetTransition in project antlr4 by antlr.
the class ATNSerializer method serialize.
/**
* Serialize state descriptors, edge descriptors, and decision→state map
* into list of ints:
*
* grammar-type, (ANTLRParser.LEXER, ...)
* max token type,
* num states,
* state-0-type ruleIndex, state-1-type ruleIndex, ... state-i-type ruleIndex optional-arg ...
* num rules,
* rule-1-start-state rule-1-args, rule-2-start-state rule-2-args, ...
* (args are token type,actionIndex in lexer else 0,0)
* num modes,
* mode-0-start-state, mode-1-start-state, ... (parser has 0 modes)
* num unicode-bmp-sets
* bmp-set-0-interval-count intervals, bmp-set-1-interval-count intervals, ...
* num unicode-smp-sets
* smp-set-0-interval-count intervals, smp-set-1-interval-count intervals, ...
* num total edges,
* src, trg, edge-type, edge arg1, optional edge arg2 (present always), ...
* num decisions,
* decision-0-start-state, decision-1-start-state, ...
*
* Convenient to pack into unsigned shorts to make as Java string.
*/
public IntegerList serialize() {
IntegerList data = new IntegerList();
data.add(ATNDeserializer.SERIALIZED_VERSION);
serializeUUID(data, ATNDeserializer.SERIALIZED_UUID);
// convert grammar type to ATN const to avoid dependence on ANTLRParser
data.add(atn.grammarType.ordinal());
data.add(atn.maxTokenType);
int nedges = 0;
// Note that we use a LinkedHashMap as a set to
// maintain insertion order while deduplicating
// entries with the same key.
Map<IntervalSet, Boolean> sets = new LinkedHashMap<>();
// dump states, count edges and collect sets while doing so
IntegerList nonGreedyStates = new IntegerList();
IntegerList precedenceStates = new IntegerList();
data.add(atn.states.size());
for (ATNState s : atn.states) {
if (s == null) {
// might be optimized away
data.add(ATNState.INVALID_TYPE);
continue;
}
int stateType = s.getStateType();
if (s instanceof DecisionState && ((DecisionState) s).nonGreedy) {
nonGreedyStates.add(s.stateNumber);
}
if (s instanceof RuleStartState && ((RuleStartState) s).isLeftRecursiveRule) {
precedenceStates.add(s.stateNumber);
}
data.add(stateType);
if (s.ruleIndex == -1) {
data.add(Character.MAX_VALUE);
} else {
data.add(s.ruleIndex);
}
if (s.getStateType() == ATNState.LOOP_END) {
data.add(((LoopEndState) s).loopBackState.stateNumber);
} else if (s instanceof BlockStartState) {
data.add(((BlockStartState) s).endState.stateNumber);
}
if (s.getStateType() != ATNState.RULE_STOP) {
// the deserializer can trivially derive these edges, so there's no need to serialize them
nedges += s.getNumberOfTransitions();
}
for (int i = 0; i < s.getNumberOfTransitions(); i++) {
Transition t = s.transition(i);
int edgeType = Transition.serializationTypes.get(t.getClass());
if (edgeType == Transition.SET || edgeType == Transition.NOT_SET) {
SetTransition st = (SetTransition) t;
sets.put(st.set, true);
}
}
}
// non-greedy states
data.add(nonGreedyStates.size());
for (int i = 0; i < nonGreedyStates.size(); i++) {
data.add(nonGreedyStates.get(i));
}
// precedence states
data.add(precedenceStates.size());
for (int i = 0; i < precedenceStates.size(); i++) {
data.add(precedenceStates.get(i));
}
int nrules = atn.ruleToStartState.length;
data.add(nrules);
for (int r = 0; r < nrules; r++) {
ATNState ruleStartState = atn.ruleToStartState[r];
data.add(ruleStartState.stateNumber);
if (atn.grammarType == ATNType.LEXER) {
if (atn.ruleToTokenType[r] == Token.EOF) {
data.add(Character.MAX_VALUE);
} else {
data.add(atn.ruleToTokenType[r]);
}
}
}
int nmodes = atn.modeToStartState.size();
data.add(nmodes);
if (nmodes > 0) {
for (ATNState modeStartState : atn.modeToStartState) {
data.add(modeStartState.stateNumber);
}
}
List<IntervalSet> bmpSets = new ArrayList<>();
List<IntervalSet> smpSets = new ArrayList<>();
for (IntervalSet set : sets.keySet()) {
if (!set.isNil() && set.getMaxElement() <= Character.MAX_VALUE) {
bmpSets.add(set);
} else {
smpSets.add(set);
}
}
serializeSets(data, bmpSets, new CodePointSerializer() {
@Override
public void serializeCodePoint(IntegerList data, int cp) {
data.add(cp);
}
});
serializeSets(data, smpSets, new CodePointSerializer() {
@Override
public void serializeCodePoint(IntegerList data, int cp) {
serializeInt(data, cp);
}
});
Map<IntervalSet, Integer> setIndices = new HashMap<>();
int setIndex = 0;
for (IntervalSet bmpSet : bmpSets) {
setIndices.put(bmpSet, setIndex++);
}
for (IntervalSet smpSet : smpSets) {
setIndices.put(smpSet, setIndex++);
}
data.add(nedges);
for (ATNState s : atn.states) {
if (s == null) {
// might be optimized away
continue;
}
if (s.getStateType() == ATNState.RULE_STOP) {
continue;
}
for (int i = 0; i < s.getNumberOfTransitions(); i++) {
Transition t = s.transition(i);
if (atn.states.get(t.target.stateNumber) == null) {
throw new IllegalStateException("Cannot serialize a transition to a removed state.");
}
int src = s.stateNumber;
int trg = t.target.stateNumber;
int edgeType = Transition.serializationTypes.get(t.getClass());
int arg1 = 0;
int arg2 = 0;
int arg3 = 0;
switch(edgeType) {
case Transition.RULE:
trg = ((RuleTransition) t).followState.stateNumber;
arg1 = ((RuleTransition) t).target.stateNumber;
arg2 = ((RuleTransition) t).ruleIndex;
arg3 = ((RuleTransition) t).precedence;
break;
case Transition.PRECEDENCE:
PrecedencePredicateTransition ppt = (PrecedencePredicateTransition) t;
arg1 = ppt.precedence;
break;
case Transition.PREDICATE:
PredicateTransition pt = (PredicateTransition) t;
arg1 = pt.ruleIndex;
arg2 = pt.predIndex;
arg3 = pt.isCtxDependent ? 1 : 0;
break;
case Transition.RANGE:
arg1 = ((RangeTransition) t).from;
arg2 = ((RangeTransition) t).to;
if (arg1 == Token.EOF) {
arg1 = 0;
arg3 = 1;
}
break;
case Transition.ATOM:
arg1 = ((AtomTransition) t).label;
if (arg1 == Token.EOF) {
arg1 = 0;
arg3 = 1;
}
break;
case Transition.ACTION:
ActionTransition at = (ActionTransition) t;
arg1 = at.ruleIndex;
arg2 = at.actionIndex;
if (arg2 == -1) {
arg2 = 0xFFFF;
}
arg3 = at.isCtxDependent ? 1 : 0;
break;
case Transition.SET:
arg1 = setIndices.get(((SetTransition) t).set);
break;
case Transition.NOT_SET:
arg1 = setIndices.get(((SetTransition) t).set);
break;
case Transition.WILDCARD:
break;
}
data.add(src);
data.add(trg);
data.add(edgeType);
data.add(arg1);
data.add(arg2);
data.add(arg3);
}
}
int ndecisions = atn.decisionToState.size();
data.add(ndecisions);
for (DecisionState decStartState : atn.decisionToState) {
data.add(decStartState.stateNumber);
}
//
if (atn.grammarType == ATNType.LEXER) {
data.add(atn.lexerActions.length);
for (LexerAction action : atn.lexerActions) {
data.add(action.getActionType().ordinal());
switch(action.getActionType()) {
case CHANNEL:
int channel = ((LexerChannelAction) action).getChannel();
data.add(channel != -1 ? channel : 0xFFFF);
data.add(0);
break;
case CUSTOM:
int ruleIndex = ((LexerCustomAction) action).getRuleIndex();
int actionIndex = ((LexerCustomAction) action).getActionIndex();
data.add(ruleIndex != -1 ? ruleIndex : 0xFFFF);
data.add(actionIndex != -1 ? actionIndex : 0xFFFF);
break;
case MODE:
int mode = ((LexerModeAction) action).getMode();
data.add(mode != -1 ? mode : 0xFFFF);
data.add(0);
break;
case MORE:
data.add(0);
data.add(0);
break;
case POP_MODE:
data.add(0);
data.add(0);
break;
case PUSH_MODE:
mode = ((LexerPushModeAction) action).getMode();
data.add(mode != -1 ? mode : 0xFFFF);
data.add(0);
break;
case SKIP:
data.add(0);
data.add(0);
break;
case TYPE:
int type = ((LexerTypeAction) action).getType();
data.add(type != -1 ? type : 0xFFFF);
data.add(0);
break;
default:
String message = String.format(Locale.getDefault(), "The specified lexer action type %s is not valid.", action.getActionType());
throw new IllegalArgumentException(message);
}
}
}
// don't adjust the first value since that's the version number
for (int i = 1; i < data.size(); i++) {
if (data.get(i) < Character.MIN_VALUE || data.get(i) > Character.MAX_VALUE) {
throw new UnsupportedOperationException("Serialized ATN data element " + data.get(i) + " element " + i + " out of range " + (int) Character.MIN_VALUE + ".." + (int) Character.MAX_VALUE);
}
int value = (data.get(i) + 2) & 0xFFFF;
data.set(i, value);
}
return data;
}
Aggregations