Class TRegexBacktrackingNFAExecutorNode

java.lang.Object

com.oracle.truffle.api.nodes.Node

com.oracle.truffle.regex.tregex.nodes.TRegexExecutorBaseNode

com.oracle.truffle.regex.tregex.nodes.TRegexExecutorNode

com.oracle.truffle.regex.tregex.nodes.nfa.TRegexBacktrackerSubExecutorNode

com.oracle.truffle.regex.tregex.nodes.nfa.TRegexBacktrackingNFAExecutorNode

All Implemented Interfaces:

com.oracle.truffle.api.instrumentation.InstrumentableNode, com.oracle.truffle.api.nodes.NodeInterface, Cloneable

public final class TRegexBacktrackingNFAExecutorNode
extends TRegexBacktrackerSubExecutorNode

This regex executor uses a backtracking algorithm on the NFA. It is used for all expressions that cannot be matched with the DFA, such as expressions with backreferences.

Nested Class Summary

Nested classes/interfaces inherited from class com.oracle.truffle.api.nodes.Node

com.oracle.truffle.api.nodes.Node.Child, com.oracle.truffle.api.nodes.Node.Children

Nested classes/interfaces inherited from interface com.oracle.truffle.api.instrumentation.InstrumentableNode

com.oracle.truffle.api.instrumentation.InstrumentableNode.WrapperNode

Field Summary

Fields inherited from class com.oracle.truffle.regex.tregex.nodes.nfa.TRegexBacktrackerSubExecutorNode

NO_SUB_EXECUTORS, subExecutors

Fields inherited from class com.oracle.truffle.regex.tregex.nodes.TRegexExecutorNode

ASTRAL_PROBABILITY, BMP_PROBABILITY, CONTINUE_PROBABILITY, EXIT_PROBABILITY, LATIN1_PROBABILITY

Constructor Summary

Constructors

Constructor	Description
TRegexBacktrackingNFAExecutorNode(RegexAST ast, PureNFA nfa, int numberOfStates, int numberOfTransitions, TRegexExecutorBaseNode[] subExecutors, boolean mustAdvance, CompilationBuffer compilationBuffer)

Method Summary

Modifier and Type	Method	Description
TRegexExecutorLocals	createLocals(com.oracle.truffle.api.strings.TruffleString input, int fromIndex, int maxIndex, int regionFrom, int regionTo, int index)
Object	execute(com.oracle.truffle.api.frame.VirtualFrame frame, TRegexExecutorLocals abstractLocals, com.oracle.truffle.api.strings.TruffleString.CodeRange codeRange)
protected static int	getBackRefBoundary(TRegexBacktrackingNFAExecutorLocals locals, PureNFATransition transition, int cgIndex, int index)
String	getName()
int	getNumberOfStates()
boolean	isBackreferenceIgnoreCaseMultiCharExpansion()
boolean	isBackrefWithNullTargetFails()	Should a backreference to an unmatched capture group succeed or fail?
boolean	isEmptyChecksOnMandatoryLoopIterations()
boolean	isForward()
boolean	isLoneSurrogates()
boolean	isLoopbackInitialState()
boolean	isMonitorCaptureGroupsInEmptyCheck()	Should the empty check in exitZeroWidth quantifier guards also check the contents of capture groups? If the capture groups were modified, the empty check passes, even if only the empty string was matched.
boolean	isMustAdvance()
boolean	isRecursiveBackreferences()
boolean	isTrackLastGroup()
boolean	isTransitionMatchesStepByStep()	When generating NFAs for Ruby regular expressions, the sequence of quantifier guards on a single transition can become quite complex.
boolean	isUseMergeExplode()
protected boolean	lookAroundExecutorIsLiteral(PureNFAState s)
boolean	returnsFirstGroup()	Should the reported lastGroup point to the first group that *begins* instead of the last group that *ends*? This is needed when executing Python lookbehind expressions.
protected int[]	runSubMatcher(com.oracle.truffle.api.frame.VirtualFrame frame, TRegexBacktrackingNFAExecutorLocals subLocals, com.oracle.truffle.api.strings.TruffleString.CodeRange codeRange, PureNFAState subMatcherState)
TRegexBacktrackerSubExecutorNode	shallowCopy()
protected static boolean	subMatchFailed(PureNFAState curState, Object subMatchResult)
protected boolean	transitionMatches(com.oracle.truffle.api.frame.VirtualFrame frame, TRegexBacktrackingNFAExecutorLocals locals, com.oracle.truffle.api.strings.TruffleString.CodeRange codeRange, PureNFATransition transition, int index, boolean atEnd, int c)
protected boolean	tryUpdateState(com.oracle.truffle.api.frame.VirtualFrame frame, TRegexBacktrackingNFAExecutorLocals locals, com.oracle.truffle.api.strings.TruffleString.CodeRange codeRange, PureNFATransition transition, int index, boolean atEnd, int c)	This method composes transitionMatches(com.oracle.truffle.api.frame.VirtualFrame, com.oracle.truffle.regex.tregex.nodes.nfa.TRegexBacktrackingNFAExecutorLocals, com.oracle.truffle.api.strings.TruffleString.CodeRange, com.oracle.truffle.regex.tregex.nfa.PureNFATransition, int, boolean, int) with updateState(com.oracle.truffle.regex.tregex.nodes.nfa.TRegexBacktrackingNFAExecutorLocals, com.oracle.truffle.regex.tregex.nfa.PureNFATransition, int).
protected void	updateState(TRegexBacktrackingNFAExecutorLocals locals, PureNFATransition transition, int index)
boolean	writesCaptureGroups()	Returns true if this executor may write any new capture group boundaries.

Methods inherited from class com.oracle.truffle.regex.tregex.nodes.TRegexExecutorNode

countUpTo, getAstralProfile, getBMPProfile, getMaxIndex, getMinIndex, getNumberOfCaptureGroups, getNumberOfTransitions, getSource, inputAdvance, inputAtBegin, inputAtEnd, inputGetCodePointSize, inputGetCodePointSize, inputHasNext, inputHasNext, inputHasNext, inputIncNextIndexRaw, inputIncNextIndexRaw, inputIncRaw, inputIncRaw, inputIncRaw, inputIncRaw, inputIncRaw, inputIncRaw, inputReadAndDecode, inputReadAndDecode, inputReadRaw, inputReadRaw, inputReadRaw, inputReadRaw, inputSkip, inputSkipIntl, inputSkipReverse, inputUTF16IsHighSurrogate, inputUTF16IsLowSurrogate, inputUTF16ToCodePoint, isSimpleCG, rewindUpTo

Methods inherited from class com.oracle.truffle.regex.tregex.nodes.TRegexExecutorBaseNode

createWrapper, getEncoding, isBooleanMatch, isInstrumentable, isTrivial, isUTF16, isUTF32, isUTF8, unwrap

Methods inherited from class com.oracle.truffle.api.nodes.Node

accept, adoptChildren, atomic, atomic, copy, deepCopy, getChildren, getCost, getDebugProperties, getDescription, getEncapsulatingSourceSection, getLock, getParent, getRootNode, getSourceSection, insert, insert, isAdoptable, isSafelyReplaceableBy, notifyInserted, onReplace, replace, replace, reportPolymorphicSpecialize, toString

Methods inherited from class java.lang.Object

clone, equals, finalize, getClass, hashCode, notify, notifyAll, wait, wait, wait

Methods inherited from interface com.oracle.truffle.api.instrumentation.InstrumentableNode

findNearestNodeAt, findNearestNodeAt, getNodeObject, hasTag, materializeInstrumentableNodes

Constructor Details

TRegexBacktrackingNFAExecutorNode

public TRegexBacktrackingNFAExecutorNode(RegexAST ast,
 PureNFA nfa,
 int numberOfStates,
 int numberOfTransitions,
 TRegexExecutorBaseNode[] subExecutors,
 boolean mustAdvance,
 CompilationBuffer compilationBuffer)

Method Details

shallowCopy

public TRegexBacktrackerSubExecutorNode shallowCopy()

Specified by:

shallowCopy in class TRegexBacktrackerSubExecutorNode

getNumberOfStates

public int getNumberOfStates()

Specified by:

getNumberOfStates in class TRegexExecutorBaseNode

writesCaptureGroups

public boolean writesCaptureGroups()

Description copied from class: TRegexExecutorBaseNode

Returns true if this executor may write any new capture group boundaries.

Specified by:

writesCaptureGroups in class TRegexExecutorBaseNode

getName

public String getName()

Specified by:

getName in class TRegexExecutorBaseNode

isForward

public boolean isForward()

Specified by:

isForward in class TRegexExecutorBaseNode

isBackrefWithNullTargetFails

public boolean isBackrefWithNullTargetFails()

Should a backreference to an unmatched capture group succeed or fail?

isMonitorCaptureGroupsInEmptyCheck

public boolean isMonitorCaptureGroupsInEmptyCheck()

Should the empty check in exitZeroWidth quantifier guards also check the contents of capture groups? If the capture groups were modified, the empty check passes, even if only the empty string was matched.

isTransitionMatchesStepByStep

public boolean isTransitionMatchesStepByStep()

When generating NFAs for Ruby regular expressions, the sequence of quantifier guards on a single transition can become quite complex. This necessitates evaluating the effects of each guard one by one in order to arrive at the correct answer. This flag controls whether such detailed handling of the quantifiers is to be used or not.

isEmptyChecksOnMandatoryLoopIterations

public boolean isEmptyChecksOnMandatoryLoopIterations()

isTrackLastGroup

public boolean isTrackLastGroup()

isMustAdvance

public boolean isMustAdvance()

isLoneSurrogates

public boolean isLoneSurrogates()

isLoopbackInitialState

public boolean isLoopbackInitialState()

isUseMergeExplode

public boolean isUseMergeExplode()

isRecursiveBackreferences

public boolean isRecursiveBackreferences()

isBackreferenceIgnoreCaseMultiCharExpansion

public boolean isBackreferenceIgnoreCaseMultiCharExpansion()

createLocals

public TRegexExecutorLocals createLocals(com.oracle.truffle.api.strings.TruffleString input,
 int fromIndex,
 int maxIndex,
 int regionFrom,
 int regionTo,
 int index)

Specified by:

createLocals in class TRegexExecutorBaseNode

execute

public Object execute(com.oracle.truffle.api.frame.VirtualFrame frame,
 TRegexExecutorLocals abstractLocals,
 com.oracle.truffle.api.strings.TruffleString.CodeRange codeRange)

Specified by:

execute in class TRegexExecutorBaseNode

returnsFirstGroup

public boolean returnsFirstGroup()

Should the reported lastGroup point to the first group that *begins* instead of the last group that *ends*? This is needed when executing Python lookbehind expressions. The semantics of the lastGroup field should correspond to the left-to-right evaluation of lookbehind assertions in Python, but we run lookbehinds in the right-to-left direction.

lookAroundExecutorIsLiteral

protected boolean lookAroundExecutorIsLiteral(PureNFAState s)

runSubMatcher

protected int[] runSubMatcher(com.oracle.truffle.api.frame.VirtualFrame frame,
 TRegexBacktrackingNFAExecutorLocals subLocals,
 com.oracle.truffle.api.strings.TruffleString.CodeRange codeRange,
 PureNFAState subMatcherState)

subMatchFailed

protected static boolean subMatchFailed(PureNFAState curState,
 Object subMatchResult)

transitionMatches

protected boolean transitionMatches(com.oracle.truffle.api.frame.VirtualFrame frame,
 TRegexBacktrackingNFAExecutorLocals locals,
 com.oracle.truffle.api.strings.TruffleString.CodeRange codeRange,
 PureNFATransition transition,
 int index,
 boolean atEnd,
 int c)

getBackRefBoundary

protected static int getBackRefBoundary(TRegexBacktrackingNFAExecutorLocals locals,
 PureNFATransition transition,
 int cgIndex,
 int index)

updateState

protected void updateState(TRegexBacktrackingNFAExecutorLocals locals,
 PureNFATransition transition,
 int index)

tryUpdateState

protected boolean tryUpdateState(com.oracle.truffle.api.frame.VirtualFrame frame,
 TRegexBacktrackingNFAExecutorLocals locals,
 com.oracle.truffle.api.strings.TruffleString.CodeRange codeRange,
 PureNFATransition transition,
 int index,
 boolean atEnd,
 int c)

This method composes transitionMatches(com.oracle.truffle.api.frame.VirtualFrame, com.oracle.truffle.regex.tregex.nodes.nfa.TRegexBacktrackingNFAExecutorLocals, com.oracle.truffle.api.strings.TruffleString.CodeRange, com.oracle.truffle.regex.tregex.nfa.PureNFATransition, int, boolean, int) with updateState(com.oracle.truffle.regex.tregex.nodes.nfa.TRegexBacktrackingNFAExecutorLocals, com.oracle.truffle.regex.tregex.nfa.PureNFATransition, int). It is somewhat equivalent to the following:

 if (transitionMatches(locals, compactString, transition, index, atEnd, c)) {
     updateState(locals, transition, index);
     return true;
 } else {
     return false;
 }
 

The key difference is that in this method, the updateState(com.oracle.truffle.regex.tregex.nodes.nfa.TRegexBacktrackingNFAExecutorLocals, com.oracle.truffle.regex.tregex.nfa.PureNFATransition, int) effects of quantifier guards are evaluated in parallel with their transitionMatches(com.oracle.truffle.api.frame.VirtualFrame, com.oracle.truffle.regex.tregex.nodes.nfa.TRegexBacktrackingNFAExecutorLocals, com.oracle.truffle.api.strings.TruffleString.CodeRange, com.oracle.truffle.regex.tregex.nfa.PureNFATransition, int, boolean, int) assertions. This more detailed behavior is necessary when working with NFA transitions that have complex chains of quantifier guards, such as the ones used in Ruby.

NB: This method writes to locals even if it returns false.