use of org.antlr.v4.runtime.atn.AtomTransition in project antlr4 by antlr.
the class LexerATNFactory method tokenRef.
@Override
public Handle tokenRef(TerminalAST node) {
// Ref to EOF in lexer yields char transition on -1
if (node.getText().equals("EOF")) {
ATNState left = newState(node);
ATNState right = newState(node);
left.addTransition(new AtomTransition(right, IntStream.EOF));
return new Handle(left, right);
}
return _ruleRef(node);
}
use of org.antlr.v4.runtime.atn.AtomTransition in project antlr4 by antlr.
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.AtomTransition in project antlr4 by antlr.
the class ATNDeserializer method deserialize.
@SuppressWarnings("deprecation")
public ATN deserialize(char[] data) {
data = data.clone();
// was implemented.
for (int i = 1; i < data.length; i++) {
data[i] = (char) (data[i] - 2);
}
int p = 0;
int version = toInt(data[p++]);
if (version != SERIALIZED_VERSION) {
String reason = String.format(Locale.getDefault(), "Could not deserialize ATN with version %d (expected %d).", version, SERIALIZED_VERSION);
throw new UnsupportedOperationException(new InvalidClassException(ATN.class.getName(), reason));
}
UUID uuid = toUUID(data, p);
p += 8;
if (!SUPPORTED_UUIDS.contains(uuid)) {
String reason = String.format(Locale.getDefault(), "Could not deserialize ATN with UUID %s (expected %s or a legacy UUID).", uuid, SERIALIZED_UUID);
throw new UnsupportedOperationException(new InvalidClassException(ATN.class.getName(), reason));
}
boolean supportsPrecedencePredicates = isFeatureSupported(ADDED_PRECEDENCE_TRANSITIONS, uuid);
boolean supportsLexerActions = isFeatureSupported(ADDED_LEXER_ACTIONS, uuid);
ATNType grammarType = ATNType.values()[toInt(data[p++])];
int maxTokenType = toInt(data[p++]);
ATN atn = new ATN(grammarType, maxTokenType);
//
// STATES
//
List<Pair<LoopEndState, Integer>> loopBackStateNumbers = new ArrayList<Pair<LoopEndState, Integer>>();
List<Pair<BlockStartState, Integer>> endStateNumbers = new ArrayList<Pair<BlockStartState, Integer>>();
int nstates = toInt(data[p++]);
for (int i = 0; i < nstates; i++) {
int stype = toInt(data[p++]);
// ignore bad type of states
if (stype == ATNState.INVALID_TYPE) {
atn.addState(null);
continue;
}
int ruleIndex = toInt(data[p++]);
if (ruleIndex == Character.MAX_VALUE) {
ruleIndex = -1;
}
ATNState s = stateFactory(stype, ruleIndex);
if (stype == ATNState.LOOP_END) {
// special case
int loopBackStateNumber = toInt(data[p++]);
loopBackStateNumbers.add(new Pair<LoopEndState, Integer>((LoopEndState) s, loopBackStateNumber));
} else if (s instanceof BlockStartState) {
int endStateNumber = toInt(data[p++]);
endStateNumbers.add(new Pair<BlockStartState, Integer>((BlockStartState) s, endStateNumber));
}
atn.addState(s);
}
// delay the assignment of loop back and end states until we know all the state instances have been initialized
for (Pair<LoopEndState, Integer> pair : loopBackStateNumbers) {
pair.a.loopBackState = atn.states.get(pair.b);
}
for (Pair<BlockStartState, Integer> pair : endStateNumbers) {
pair.a.endState = (BlockEndState) atn.states.get(pair.b);
}
int numNonGreedyStates = toInt(data[p++]);
for (int i = 0; i < numNonGreedyStates; i++) {
int stateNumber = toInt(data[p++]);
((DecisionState) atn.states.get(stateNumber)).nonGreedy = true;
}
if (supportsPrecedencePredicates) {
int numPrecedenceStates = toInt(data[p++]);
for (int i = 0; i < numPrecedenceStates; i++) {
int stateNumber = toInt(data[p++]);
((RuleStartState) atn.states.get(stateNumber)).isLeftRecursiveRule = true;
}
}
//
// RULES
//
int nrules = toInt(data[p++]);
if (atn.grammarType == ATNType.LEXER) {
atn.ruleToTokenType = new int[nrules];
}
atn.ruleToStartState = new RuleStartState[nrules];
for (int i = 0; i < nrules; i++) {
int s = toInt(data[p++]);
RuleStartState startState = (RuleStartState) atn.states.get(s);
atn.ruleToStartState[i] = startState;
if (atn.grammarType == ATNType.LEXER) {
int tokenType = toInt(data[p++]);
if (tokenType == 0xFFFF) {
tokenType = Token.EOF;
}
atn.ruleToTokenType[i] = tokenType;
if (!isFeatureSupported(ADDED_LEXER_ACTIONS, uuid)) {
// this piece of unused metadata was serialized prior to the
// addition of LexerAction
int actionIndexIgnored = toInt(data[p++]);
}
}
}
atn.ruleToStopState = new RuleStopState[nrules];
for (ATNState state : atn.states) {
if (!(state instanceof RuleStopState)) {
continue;
}
RuleStopState stopState = (RuleStopState) state;
atn.ruleToStopState[state.ruleIndex] = stopState;
atn.ruleToStartState[state.ruleIndex].stopState = stopState;
}
//
// MODES
//
int nmodes = toInt(data[p++]);
for (int i = 0; i < nmodes; i++) {
int s = toInt(data[p++]);
atn.modeToStartState.add((TokensStartState) atn.states.get(s));
}
//
// SETS
//
List<IntervalSet> sets = new ArrayList<IntervalSet>();
// First, read all sets with 16-bit Unicode code points <= U+FFFF.
p = deserializeSets(data, p, sets, getUnicodeDeserializer(UnicodeDeserializingMode.UNICODE_BMP));
// deserialize sets with 32-bit arguments <= U+10FFFF.
if (isFeatureSupported(ADDED_UNICODE_SMP, uuid)) {
p = deserializeSets(data, p, sets, getUnicodeDeserializer(UnicodeDeserializingMode.UNICODE_SMP));
}
//
// EDGES
//
int nedges = toInt(data[p++]);
for (int i = 0; i < nedges; i++) {
int src = toInt(data[p]);
int trg = toInt(data[p + 1]);
int ttype = toInt(data[p + 2]);
int arg1 = toInt(data[p + 3]);
int arg2 = toInt(data[p + 4]);
int arg3 = toInt(data[p + 5]);
Transition trans = edgeFactory(atn, ttype, src, trg, arg1, arg2, arg3, sets);
// System.out.println("EDGE "+trans.getClass().getSimpleName()+" "+
// src+"->"+trg+
// " "+Transition.serializationNames[ttype]+
// " "+arg1+","+arg2+","+arg3);
ATNState srcState = atn.states.get(src);
srcState.addTransition(trans);
p += 6;
}
// edges for rule stop states can be derived, so they aren't serialized
for (ATNState state : atn.states) {
for (int i = 0; i < state.getNumberOfTransitions(); i++) {
Transition t = state.transition(i);
if (!(t instanceof RuleTransition)) {
continue;
}
RuleTransition ruleTransition = (RuleTransition) t;
int outermostPrecedenceReturn = -1;
if (atn.ruleToStartState[ruleTransition.target.ruleIndex].isLeftRecursiveRule) {
if (ruleTransition.precedence == 0) {
outermostPrecedenceReturn = ruleTransition.target.ruleIndex;
}
}
EpsilonTransition returnTransition = new EpsilonTransition(ruleTransition.followState, outermostPrecedenceReturn);
atn.ruleToStopState[ruleTransition.target.ruleIndex].addTransition(returnTransition);
}
}
for (ATNState state : atn.states) {
if (state instanceof BlockStartState) {
// we need to know the end state to set its start state
if (((BlockStartState) state).endState == null) {
throw new IllegalStateException();
}
// block end states can only be associated to a single block start state
if (((BlockStartState) state).endState.startState != null) {
throw new IllegalStateException();
}
((BlockStartState) state).endState.startState = (BlockStartState) state;
}
if (state instanceof PlusLoopbackState) {
PlusLoopbackState loopbackState = (PlusLoopbackState) state;
for (int i = 0; i < loopbackState.getNumberOfTransitions(); i++) {
ATNState target = loopbackState.transition(i).target;
if (target instanceof PlusBlockStartState) {
((PlusBlockStartState) target).loopBackState = loopbackState;
}
}
} else if (state instanceof StarLoopbackState) {
StarLoopbackState loopbackState = (StarLoopbackState) state;
for (int i = 0; i < loopbackState.getNumberOfTransitions(); i++) {
ATNState target = loopbackState.transition(i).target;
if (target instanceof StarLoopEntryState) {
((StarLoopEntryState) target).loopBackState = loopbackState;
}
}
}
}
//
// DECISIONS
//
int ndecisions = toInt(data[p++]);
for (int i = 1; i <= ndecisions; i++) {
int s = toInt(data[p++]);
DecisionState decState = (DecisionState) atn.states.get(s);
atn.decisionToState.add(decState);
decState.decision = i - 1;
}
//
if (atn.grammarType == ATNType.LEXER) {
if (supportsLexerActions) {
atn.lexerActions = new LexerAction[toInt(data[p++])];
for (int i = 0; i < atn.lexerActions.length; i++) {
LexerActionType actionType = LexerActionType.values()[toInt(data[p++])];
int data1 = toInt(data[p++]);
if (data1 == 0xFFFF) {
data1 = -1;
}
int data2 = toInt(data[p++]);
if (data2 == 0xFFFF) {
data2 = -1;
}
LexerAction lexerAction = lexerActionFactory(actionType, data1, data2);
atn.lexerActions[i] = lexerAction;
}
} else {
// for compatibility with older serialized ATNs, convert the old
// serialized action index for action transitions to the new
// form, which is the index of a LexerCustomAction
List<LexerAction> legacyLexerActions = new ArrayList<LexerAction>();
for (ATNState state : atn.states) {
for (int i = 0; i < state.getNumberOfTransitions(); i++) {
Transition transition = state.transition(i);
if (!(transition instanceof ActionTransition)) {
continue;
}
int ruleIndex = ((ActionTransition) transition).ruleIndex;
int actionIndex = ((ActionTransition) transition).actionIndex;
LexerCustomAction lexerAction = new LexerCustomAction(ruleIndex, actionIndex);
state.setTransition(i, new ActionTransition(transition.target, ruleIndex, legacyLexerActions.size(), false));
legacyLexerActions.add(lexerAction);
}
}
atn.lexerActions = legacyLexerActions.toArray(new LexerAction[legacyLexerActions.size()]);
}
}
markPrecedenceDecisions(atn);
if (deserializationOptions.isVerifyATN()) {
verifyATN(atn);
}
if (deserializationOptions.isGenerateRuleBypassTransitions() && atn.grammarType == ATNType.PARSER) {
atn.ruleToTokenType = new int[atn.ruleToStartState.length];
for (int i = 0; i < atn.ruleToStartState.length; i++) {
atn.ruleToTokenType[i] = atn.maxTokenType + i + 1;
}
for (int i = 0; i < atn.ruleToStartState.length; i++) {
BasicBlockStartState bypassStart = new BasicBlockStartState();
bypassStart.ruleIndex = i;
atn.addState(bypassStart);
BlockEndState bypassStop = new BlockEndState();
bypassStop.ruleIndex = i;
atn.addState(bypassStop);
bypassStart.endState = bypassStop;
atn.defineDecisionState(bypassStart);
bypassStop.startState = bypassStart;
ATNState endState;
Transition excludeTransition = null;
if (atn.ruleToStartState[i].isLeftRecursiveRule) {
// wrap from the beginning of the rule to the StarLoopEntryState
endState = null;
for (ATNState state : atn.states) {
if (state.ruleIndex != i) {
continue;
}
if (!(state instanceof StarLoopEntryState)) {
continue;
}
ATNState maybeLoopEndState = state.transition(state.getNumberOfTransitions() - 1).target;
if (!(maybeLoopEndState instanceof LoopEndState)) {
continue;
}
if (maybeLoopEndState.epsilonOnlyTransitions && maybeLoopEndState.transition(0).target instanceof RuleStopState) {
endState = state;
break;
}
}
if (endState == null) {
throw new UnsupportedOperationException("Couldn't identify final state of the precedence rule prefix section.");
}
excludeTransition = ((StarLoopEntryState) endState).loopBackState.transition(0);
} else {
endState = atn.ruleToStopState[i];
}
// all non-excluded transitions that currently target end state need to target blockEnd instead
for (ATNState state : atn.states) {
for (Transition transition : state.transitions) {
if (transition == excludeTransition) {
continue;
}
if (transition.target == endState) {
transition.target = bypassStop;
}
}
}
// all transitions leaving the rule start state need to leave blockStart instead
while (atn.ruleToStartState[i].getNumberOfTransitions() > 0) {
Transition transition = atn.ruleToStartState[i].removeTransition(atn.ruleToStartState[i].getNumberOfTransitions() - 1);
bypassStart.addTransition(transition);
}
// link the new states
atn.ruleToStartState[i].addTransition(new EpsilonTransition(bypassStart));
bypassStop.addTransition(new EpsilonTransition(endState));
ATNState matchState = new BasicState();
atn.addState(matchState);
matchState.addTransition(new AtomTransition(bypassStop, atn.ruleToTokenType[i]));
bypassStart.addTransition(new EpsilonTransition(matchState));
}
if (deserializationOptions.isVerifyATN()) {
// reverify after modification
verifyATN(atn);
}
}
return atn;
}
use of org.antlr.v4.runtime.atn.AtomTransition in project antlr4 by antlr.
the class ATNPrinter method asString.
public String asString() {
if (start == null)
return null;
marked = new HashSet<ATNState>();
work = new ArrayList<ATNState>();
work.add(start);
StringBuilder buf = new StringBuilder();
ATNState s;
while (!work.isEmpty()) {
s = work.remove(0);
if (marked.contains(s))
continue;
int n = s.getNumberOfTransitions();
// System.out.println("visit "+s+"; edges="+n);
marked.add(s);
for (int i = 0; i < n; i++) {
Transition t = s.transition(i);
if (!(s instanceof RuleStopState)) {
// don't add follow states to work
if (t instanceof RuleTransition)
work.add(((RuleTransition) t).followState);
else
work.add(t.target);
}
buf.append(getStateString(s));
if (t instanceof EpsilonTransition) {
buf.append("->").append(getStateString(t.target)).append('\n');
} else if (t instanceof RuleTransition) {
buf.append("-").append(g.getRule(((RuleTransition) t).ruleIndex).name).append("->").append(getStateString(t.target)).append('\n');
} else if (t instanceof ActionTransition) {
ActionTransition a = (ActionTransition) t;
buf.append("-").append(a.toString()).append("->").append(getStateString(t.target)).append('\n');
} else if (t instanceof SetTransition) {
SetTransition st = (SetTransition) t;
boolean not = st instanceof NotSetTransition;
if (g.isLexer()) {
buf.append("-").append(not ? "~" : "").append(st.toString()).append("->").append(getStateString(t.target)).append('\n');
} else {
buf.append("-").append(not ? "~" : "").append(st.label().toString(g.getVocabulary())).append("->").append(getStateString(t.target)).append('\n');
}
} else if (t instanceof AtomTransition) {
AtomTransition a = (AtomTransition) t;
String label = g.getTokenDisplayName(a.label);
buf.append("-").append(label).append("->").append(getStateString(t.target)).append('\n');
} else {
buf.append("-").append(t.toString()).append("->").append(getStateString(t.target)).append('\n');
}
}
}
return buf.toString();
}
use of org.antlr.v4.runtime.atn.AtomTransition in project antlr4 by antlr.
the class ParserATNFactory method tokenRef.
/** From label {@code A} build graph {@code o-A->o}. */
@Override
public Handle tokenRef(TerminalAST node) {
ATNState left = newState(node);
ATNState right = newState(node);
int ttype = g.getTokenType(node.getText());
left.addTransition(new AtomTransition(right, ttype));
node.atnState = left;
return new Handle(left, right);
}
Aggregations