org.antlr.v4.runtime.ParserInterpreter Maven / Gradle / Ivy
Go to download
Show more of this group Show more artifacts with this name
Show all versions of antlr4-runtime Show documentation
Show all versions of antlr4-runtime Show documentation
The ANTLR 4 Runtime (Optimized)
/*
* Copyright (c) 2012 The ANTLR Project. All rights reserved.
* Use of this file is governed by the BSD-3-Clause license that
* can be found in the LICENSE.txt file in the project root.
*/
package org.antlr.v4.runtime;
import org.antlr.v4.runtime.atn.ATN;
import org.antlr.v4.runtime.atn.ATNState;
import org.antlr.v4.runtime.atn.ActionTransition;
import org.antlr.v4.runtime.atn.AtomTransition;
import org.antlr.v4.runtime.atn.DecisionState;
import org.antlr.v4.runtime.atn.LoopEndState;
import org.antlr.v4.runtime.atn.ParserATNSimulator;
import org.antlr.v4.runtime.atn.PrecedencePredicateTransition;
import org.antlr.v4.runtime.atn.PredicateTransition;
import org.antlr.v4.runtime.atn.RuleStartState;
import org.antlr.v4.runtime.atn.RuleTransition;
import org.antlr.v4.runtime.atn.StarLoopEntryState;
import org.antlr.v4.runtime.atn.Transition;
import org.antlr.v4.runtime.misc.NotNull;
import org.antlr.v4.runtime.misc.Tuple;
import org.antlr.v4.runtime.misc.Tuple2;
import java.util.ArrayDeque;
import java.util.BitSet;
import java.util.Collection;
import java.util.Deque;
/** A parser simulator that mimics what ANTLR's generated
* parser code does. A ParserATNSimulator is used to make
* predictions via adaptivePredict but this class moves a pointer through the
* ATN to simulate parsing. ParserATNSimulator just
* makes us efficient rather than having to backtrack, for example.
*
* This properly creates parse trees even for left recursive rules.
*
* We rely on the left recursive rule invocation and special predicate
* transitions to make left recursive rules work.
*
* See TestParserInterpreter for examples.
*/
public class ParserInterpreter extends Parser {
protected final String grammarFileName;
protected final ATN atn;
/** This identifies StarLoopEntryState's that begin the (...)*
* precedence loops of left recursive rules.
*/
protected final BitSet pushRecursionContextStates;
@Deprecated
protected final String[] tokenNames;
protected final String[] ruleNames;
@NotNull
private final Vocabulary vocabulary;
/** This stack corresponds to the _parentctx, _parentState pair of locals
* that would exist on call stack frames with a recursive descent parser;
* in the generated function for a left-recursive rule you'd see:
*
* private EContext e(int _p) throws RecognitionException {
* ParserRuleContext _parentctx = _ctx; // Pair.a
* int _parentState = getState(); // Pair.b
* ...
* }
*
* Those values are used to create new recursive rule invocation contexts
* associated with left operand of an alt like "expr '*' expr".
*/
protected final Deque> _parentContextStack =
new ArrayDeque>();
/** We need a map from (decision,inputIndex)->forced alt for computing ambiguous
* parse trees. For now, we allow exactly one override.
*/
protected int overrideDecision = -1;
protected int overrideDecisionInputIndex = -1;
protected int overrideDecisionAlt = -1;
protected boolean overrideDecisionReached = false; // latch and only override once; error might trigger infinite loop
/** What is the current context when we override a decisions? This tells
* us what the root of the parse tree is when using override
* for an ambiguity/lookahead check.
*/
protected InterpreterRuleContext overrideDecisionRoot = null;
protected InterpreterRuleContext rootContext;
/** A copy constructor that creates a new parser interpreter by reusing
* the fields of a previous interpreter.
*
* @param old The interpreter to copy
*
* @since 4.5
*/
public ParserInterpreter(@NotNull ParserInterpreter old) {
super(old.getInputStream());
this.grammarFileName = old.grammarFileName;
this.atn = old.atn;
this.pushRecursionContextStates = old.pushRecursionContextStates;
this.tokenNames = old.tokenNames;
this.ruleNames = old.ruleNames;
this.vocabulary = old.vocabulary;
setInterpreter(new ParserATNSimulator(this, atn));
}
/**
* @deprecated Use {@link #ParserInterpreter(String, Vocabulary, Collection, ATN, TokenStream)} instead.
*/
@Deprecated
public ParserInterpreter(String grammarFileName, Collection tokenNames,
Collection ruleNames, ATN atn, TokenStream input) {
this(grammarFileName, VocabularyImpl.fromTokenNames(tokenNames.toArray(new String[tokenNames.size()])), ruleNames, atn, input);
}
public ParserInterpreter(String grammarFileName, @NotNull Vocabulary vocabulary,
Collection ruleNames, ATN atn, TokenStream input)
{
super(input);
this.grammarFileName = grammarFileName;
this.atn = atn;
this.tokenNames = new String[atn.maxTokenType];
for (int i = 0; i < tokenNames.length; i++) {
tokenNames[i] = vocabulary.getDisplayName(i);
}
this.ruleNames = ruleNames.toArray(new String[ruleNames.size()]);
this.vocabulary = vocabulary;
// identify the ATN states where pushNewRecursionContext() must be called
this.pushRecursionContextStates = new BitSet(atn.states.size());
for (ATNState state : atn.states) {
if (!(state instanceof StarLoopEntryState)) {
continue;
}
if (((StarLoopEntryState)state).precedenceRuleDecision) {
this.pushRecursionContextStates.set(state.stateNumber);
}
}
// get atn simulator that knows how to do predictions
setInterpreter(new ParserATNSimulator(this, atn));
}
@Override
public void reset() {
super.reset();
overrideDecisionReached = false;
overrideDecisionRoot = null;
}
@Override
public ATN getATN() {
return atn;
}
@Override
@Deprecated
public String[] getTokenNames() {
return tokenNames;
}
@Override
public Vocabulary getVocabulary() {
return vocabulary;
}
@Override
public String[] getRuleNames() {
return ruleNames;
}
@Override
public String getGrammarFileName() {
return grammarFileName;
}
/** Begin parsing at startRuleIndex */
public ParserRuleContext parse(int startRuleIndex) {
RuleStartState startRuleStartState = atn.ruleToStartState[startRuleIndex];
rootContext = createInterpreterRuleContext(null, ATNState.INVALID_STATE_NUMBER, startRuleIndex);
if (startRuleStartState.isPrecedenceRule) {
enterRecursionRule(rootContext, startRuleStartState.stateNumber, startRuleIndex, 0);
}
else {
enterRule(rootContext, startRuleStartState.stateNumber, startRuleIndex);
}
while ( true ) {
ATNState p = getATNState();
switch ( p.getStateType() ) {
case ATNState.RULE_STOP :
// pop; return from rule
if ( _ctx.isEmpty() ) {
if (startRuleStartState.isPrecedenceRule) {
ParserRuleContext result = _ctx;
Tuple2 parentContext = _parentContextStack.pop();
unrollRecursionContexts(parentContext.getItem1());
return result;
}
else {
exitRule();
return rootContext;
}
}
visitRuleStopState(p);
break;
default :
try {
visitState(p);
}
catch (RecognitionException e) {
setState(atn.ruleToStopState[p.ruleIndex].stateNumber);
getContext().exception = e;
getErrorHandler().reportError(this, e);
recover(e);
}
break;
}
}
}
@Override
public void enterRecursionRule(ParserRuleContext localctx, int state, int ruleIndex, int precedence) {
_parentContextStack.push(Tuple.create(_ctx, localctx.invokingState));
super.enterRecursionRule(localctx, state, ruleIndex, precedence);
}
protected ATNState getATNState() {
return atn.states.get(getState());
}
protected void visitState(ATNState p) {
int predictedAlt = 1;
if (p.getNumberOfTransitions() > 1) {
predictedAlt = visitDecisionState((DecisionState) p);
}
Transition transition = p.transition(predictedAlt - 1);
switch (transition.getSerializationType()) {
case Transition.EPSILON:
if ( pushRecursionContextStates.get(p.stateNumber) &&
!(transition.target instanceof LoopEndState))
{
// We are at the start of a left recursive rule's (...)* loop
// and we're not taking the exit branch of loop.
InterpreterRuleContext localctx =
createInterpreterRuleContext(_parentContextStack.peek().getItem1(),
_parentContextStack.peek().getItem2(),
_ctx.getRuleIndex());
pushNewRecursionContext(localctx,
atn.ruleToStartState[p.ruleIndex].stateNumber,
_ctx.getRuleIndex());
}
break;
case Transition.ATOM:
match(((AtomTransition)transition).label);
break;
case Transition.RANGE:
case Transition.SET:
case Transition.NOT_SET:
if (!transition.matches(_input.LA(1), Token.MIN_USER_TOKEN_TYPE, 65535)) {
recoverInline();
}
matchWildcard();
break;
case Transition.WILDCARD:
matchWildcard();
break;
case Transition.RULE:
RuleStartState ruleStartState = (RuleStartState)transition.target;
int ruleIndex = ruleStartState.ruleIndex;
InterpreterRuleContext newctx = createInterpreterRuleContext(_ctx, p.stateNumber, ruleIndex);
if (ruleStartState.isPrecedenceRule) {
enterRecursionRule(newctx, ruleStartState.stateNumber, ruleIndex, ((RuleTransition)transition).precedence);
}
else {
enterRule(newctx, transition.target.stateNumber, ruleIndex);
}
break;
case Transition.PREDICATE:
PredicateTransition predicateTransition = (PredicateTransition)transition;
if (!sempred(_ctx, predicateTransition.ruleIndex, predicateTransition.predIndex)) {
throw new FailedPredicateException(this);
}
break;
case Transition.ACTION:
ActionTransition actionTransition = (ActionTransition)transition;
action(_ctx, actionTransition.ruleIndex, actionTransition.actionIndex);
break;
case Transition.PRECEDENCE:
if (!precpred(_ctx, ((PrecedencePredicateTransition)transition).precedence)) {
throw new FailedPredicateException(this, String.format("precpred(_ctx, %d)", ((PrecedencePredicateTransition)transition).precedence));
}
break;
default:
throw new UnsupportedOperationException("Unrecognized ATN transition type.");
}
setState(transition.target.stateNumber);
}
/** Method visitDecisionState() is called when the interpreter reaches
* a decision state (instance of DecisionState). It gives an opportunity
* for subclasses to track interesting things.
*/
protected int visitDecisionState(DecisionState p) {
int edge = 1;
int predictedAlt;
getErrorHandler().sync(this);
int decision = p.decision;
if (decision == overrideDecision && _input.index() == overrideDecisionInputIndex && !overrideDecisionReached) {
predictedAlt = overrideDecisionAlt;
overrideDecisionReached = true;
}
else {
predictedAlt = getInterpreter().adaptivePredict(_input, decision, _ctx);
}
return predictedAlt;
}
/** Provide simple "factory" for InterpreterRuleContext's.
* @since 4.5.1
*/
protected InterpreterRuleContext createInterpreterRuleContext(
ParserRuleContext parent,
int invokingStateNumber,
int ruleIndex)
{
return new InterpreterRuleContext(parent, invokingStateNumber, ruleIndex);
}
protected void visitRuleStopState(ATNState p) {
RuleStartState ruleStartState = atn.ruleToStartState[p.ruleIndex];
if (ruleStartState.isPrecedenceRule) {
Tuple2 parentContext = _parentContextStack.pop();
unrollRecursionContexts(parentContext.getItem1());
setState(parentContext.getItem2());
}
else {
exitRule();
}
RuleTransition ruleTransition = (RuleTransition)atn.states.get(getState()).transition(0);
setState(ruleTransition.followState.stateNumber);
}
/** Override this parser interpreters normal decision-making process
* at a particular decision and input token index. Instead of
* allowing the adaptive prediction mechanism to choose the
* first alternative within a block that leads to a successful parse,
* force it to take the alternative, 1..n for n alternatives.
*
* As an implementation limitation right now, you can only specify one
* override. This is sufficient to allow construction of different
* parse trees for ambiguous input. It means re-parsing the entire input
* in general because you're never sure where an ambiguous sequence would
* live in the various parse trees. For example, in one interpretation,
* an ambiguous input sequence would be matched completely in expression
* but in another it could match all the way back to the root.
*
* s : e '!'? ;
* e : ID
* | ID '!'
* ;
*
* Here, x! can be matched as (s (e ID) !) or (s (e ID !)). In the first
* case, the ambiguous sequence is fully contained only by the root.
* In the second case, the ambiguous sequences fully contained within just
* e, as in: (e ID !).
*
* Rather than trying to optimize this and make
* some intelligent decisions for optimization purposes, I settled on
* just re-parsing the whole input and then using
* {link Trees#getRootOfSubtreeEnclosingRegion} to find the minimal
* subtree that contains the ambiguous sequence. I originally tried to
* record the call stack at the point the parser detected and ambiguity but
* left recursive rules create a parse tree stack that does not reflect
* the actual call stack. That impedance mismatch was enough to make
* it it challenging to restart the parser at a deeply nested rule
* invocation.
*
* Only parser interpreters can override decisions so as to avoid inserting
* override checking code in the critical ALL(*) prediction execution path.
*
* @since 4.5
*/
public void addDecisionOverride(int decision, int tokenIndex, int forcedAlt) {
overrideDecision = decision;
overrideDecisionInputIndex = tokenIndex;
overrideDecisionAlt = forcedAlt;
}
public InterpreterRuleContext getOverrideDecisionRoot() {
return overrideDecisionRoot;
}
/** Rely on the error handler for this parser but, if no tokens are consumed
* to recover, add an error node. Otherwise, nothing is seen in the parse
* tree.
*/
protected void recover(RecognitionException e) {
int i = _input.index();
getErrorHandler().recover(this, e);
if ( _input.index()==i ) {
// no input consumed, better add an error node
if ( e instanceof InputMismatchException ) {
InputMismatchException ime = (InputMismatchException)e;
Token tok = e.getOffendingToken();
int expectedTokenType = Token.INVALID_TYPE;
if ( !ime.getExpectedTokens().isNil() ) {
expectedTokenType = ime.getExpectedTokens().getMinElement(); // get any element
}
Token errToken =
getTokenFactory().create(Tuple.create(tok.getTokenSource(), tok.getTokenSource().getInputStream()),
expectedTokenType, tok.getText(),
Token.DEFAULT_CHANNEL,
-1, -1, // invalid start/stop
tok.getLine(), tok.getCharPositionInLine());
_ctx.addErrorNode(createErrorNode(_ctx,errToken));
}
else { // NoViableAlt
Token tok = e.getOffendingToken();
Token errToken =
getTokenFactory().create(Tuple.create(tok.getTokenSource(), tok.getTokenSource().getInputStream()),
Token.INVALID_TYPE, tok.getText(),
Token.DEFAULT_CHANNEL,
-1, -1, // invalid start/stop
tok.getLine(), tok.getCharPositionInLine());
_ctx.addErrorNode(createErrorNode(_ctx,errToken));
}
}
}
protected Token recoverInline() {
return _errHandler.recoverInline(this);
}
/** Return the root of the parse, which can be useful if the parser
* bails out. You still can access the top node. Note that,
* because of the way left recursive rules add children, it's possible
* that the root will not have any children if the start rule immediately
* called and left recursive rule that fails.
*
* @since 4.5.1
*/
public InterpreterRuleContext getRootContext() {
return rootContext;
}
}