diff --git a/runtime-testsuite/test/org/antlr/v4/test/runtime/descriptors/PerformanceDescriptors.java b/runtime-testsuite/test/org/antlr/v4/test/runtime/descriptors/PerformanceDescriptors.java index c0a676690..c499af7b8 100644 --- a/runtime-testsuite/test/org/antlr/v4/test/runtime/descriptors/PerformanceDescriptors.java +++ b/runtime-testsuite/test/org/antlr/v4/test/runtime/descriptors/PerformanceDescriptors.java @@ -3,6 +3,8 @@ package org.antlr.v4.test.runtime.descriptors; import org.antlr.v4.test.runtime.BaseParserTestDescriptor; import org.antlr.v4.test.runtime.CommentHasStringValue; +import java.util.Arrays; + public class PerformanceDescriptors { /* * This is a regression test for antlr/antlr4#192 "Poor performance of @@ -105,7 +107,7 @@ public class PerformanceDescriptors { @Override public boolean ignore(String targetName) { - return !targetName.equals("Java"); + return !Arrays.asList("Java", "Python2", "Python3", "Node").contains(targetName); } } diff --git a/runtime/JavaScript/src/antlr4/atn/ATNDeserializer.js b/runtime/JavaScript/src/antlr4/atn/ATNDeserializer.js index 548549826..4fb9fa9f2 100644 --- a/runtime/JavaScript/src/antlr4/atn/ATNDeserializer.js +++ b/runtime/JavaScript/src/antlr4/atn/ATNDeserializer.js @@ -485,7 +485,7 @@ ATNDeserializer.prototype.stateIsEndStateFor = function(state, idx) { // // Analyze the {@link StarLoopEntryState} states in the specified ATN to set -// the {@link StarLoopEntryState//precedenceRuleDecision} field to the +// the {@link StarLoopEntryState//isPrecedenceDecision} field to the // correct value. // // @param atn The ATN. @@ -505,7 +505,7 @@ ATNDeserializer.prototype.markPrecedenceDecisions = function(atn) { if (maybeLoopEndState instanceof LoopEndState) { if ( maybeLoopEndState.epsilonOnlyTransitions && (maybeLoopEndState.transitions[0].target instanceof RuleStopState)) { - state.precedenceRuleDecision = true; + state.isPrecedenceDecision = true; } } } diff --git a/runtime/JavaScript/src/antlr4/atn/ATNState.js b/runtime/JavaScript/src/antlr4/atn/ATNState.js index e311b120c..04fb32894 100644 --- a/runtime/JavaScript/src/antlr4/atn/ATNState.js +++ b/runtime/JavaScript/src/antlr4/atn/ATNState.js @@ -303,7 +303,7 @@ function StarLoopEntryState() { this.stateType = ATNState.STAR_LOOP_ENTRY; this.loopBackState = null; // Indicates whether this state can benefit from a precedence DFA during SLL decision making. - this.precedenceRuleDecision = null; + this.isPrecedenceDecision = null; return this; } diff --git a/runtime/JavaScript/src/antlr4/atn/ParserATNSimulator.js b/runtime/JavaScript/src/antlr4/atn/ParserATNSimulator.js index b740ed92a..3bb12eaea 100644 --- a/runtime/JavaScript/src/antlr4/atn/ParserATNSimulator.js +++ b/runtime/JavaScript/src/antlr4/atn/ParserATNSimulator.js @@ -262,6 +262,7 @@ var Set = Utils.Set; var BitSet = Utils.BitSet; var DoubleDict = Utils.DoubleDict; var ATN = require('./ATN').ATN; +var ATNState = require('./ATNState').ATNState; var ATNConfig = require('./ATNConfig').ATNConfig; var ATNConfigSet = require('./ATNConfigSet').ATNConfigSet; var Token = require('./../Token').Token; @@ -359,16 +360,7 @@ ParserATNSimulator.prototype.adaptivePredict = function(input, decision, outerCo " exec LA(1)==" + this.getLookaheadName(input) + ", outerContext=" + outerContext.toString(this.parser.ruleNames)); } - // If this is not a precedence DFA, we check the ATN start state - // to determine if this ATN start state is the decision for the - // closure block that determines whether a precedence rule - // should continue or complete. - // - if (!dfa.precedenceDfa && (dfa.atnStartState instanceof StarLoopEntryState)) { - if (dfa.atnStartState.precedenceRuleDecision) { - dfa.setPrecedenceDfa(true); - } - } + var fullCtx = false; var s0_closure = this.computeStartState(dfa.atnStartState, RuleContext.EMPTY, fullCtx); @@ -379,6 +371,7 @@ ParserATNSimulator.prototype.adaptivePredict = function(input, decision, outerCo // appropriate start state for the precedence level rather // than simply setting DFA.s0. // + dfa.s0.configs = s0_closure; // not used for prediction but useful to know start configs anyway s0_closure = this.applyPrecedenceFilter(s0_closure); s0 = this.addDFAState(dfa, new DFAState(null, s0_closure)); dfa.setPrecedenceStartState(this.parser.getPrecedence(), s0); @@ -1293,6 +1286,9 @@ ParserATNSimulator.prototype.closure_ = function(config, configs, closureBusy, c // both epsilon transitions and non-epsilon transitions. } for(var i = 0;i=0) { return this.parser.ruleNames[index]; diff --git a/runtime/JavaScript/src/antlr4/dfa/DFA.js b/runtime/JavaScript/src/antlr4/dfa/DFA.js index d93d9c5ab..1ffae26e6 100644 --- a/runtime/JavaScript/src/antlr4/dfa/DFA.js +++ b/runtime/JavaScript/src/antlr4/dfa/DFA.js @@ -29,6 +29,7 @@ // THIS SOFTWARE, EVEN IF ADVISED OF THE POSSIBILITY OF SUCH DAMAGE. var DFAState = require('./DFAState').DFAState; +var StarLoopEntryState = require('../atn/ATNState').StarLoopEntryState; var ATNConfigSet = require('./../atn/ATNConfigSet').ATNConfigSet; var DFASerializer = require('./DFASerializer').DFASerializer; var LexerDFASerializer = require('./DFASerializer').LexerDFASerializer; @@ -58,6 +59,17 @@ function DFA(atnStartState, decision) { // {@code false}. This is the backing field for {@link //isPrecedenceDfa}, // {@link //setPrecedenceDfa}. this.precedenceDfa = false; + if (atnStartState instanceof StarLoopEntryState) + { + if (atnStartState.isPrecedenceDecision) { + this.precedenceDfa = true; + precedenceState = new DFAState(null, new ATNConfigSet()); + precedenceState.edges = []; + precedenceState.isAcceptState = false; + precedenceState.requiresFullContext = false; + this.s0 = precedenceState; + } + } return this; } @@ -125,7 +137,7 @@ DFA.prototype.setPrecedenceDfa = function(precedenceDfa) { if (this.precedenceDfa!==precedenceDfa) { this._states = new DFAStatesSet(); if (precedenceDfa) { - var precedenceState = new DFAState(new ATNConfigSet()); + var precedenceState = new DFAState(null, new ATNConfigSet()); precedenceState.edges = []; precedenceState.isAcceptState = false; precedenceState.requiresFullContext = false; diff --git a/runtime/Python2/src/antlr4/ParserInterpreter.py b/runtime/Python2/src/antlr4/ParserInterpreter.py index ed059f9d5..f3b473032 100644 --- a/runtime/Python2/src/antlr4/ParserInterpreter.py +++ b/runtime/Python2/src/antlr4/ParserInterpreter.py @@ -69,7 +69,7 @@ class ParserInterpreter(Parser): for state in atn.states: if not isinstance(state, StarLoopEntryState): continue - if state.precedenceRuleDecision: + if state.isPrecedenceDecision: self.pushRecursionContextStates.add(state.stateNumber) # get atn simulator that knows how to do predictions self._interp = ParserATNSimulator(self, atn, self.decisionToDFA, self.sharedContextCache) diff --git a/runtime/Python2/src/antlr4/atn/ATNDeserializer.py b/runtime/Python2/src/antlr4/atn/ATNDeserializer.py index 8745806be..20c561f41 100644 --- a/runtime/Python2/src/antlr4/atn/ATNDeserializer.py +++ b/runtime/Python2/src/antlr4/atn/ATNDeserializer.py @@ -368,7 +368,7 @@ class ATNDeserializer (object): # # Analyze the {@link StarLoopEntryState} states in the specified ATN to set - # the {@link StarLoopEntryState#precedenceRuleDecision} field to the + # the {@link StarLoopEntryState#isPrecedenceDecision} field to the # correct value. # # @param atn The ATN. @@ -387,7 +387,7 @@ class ATNDeserializer (object): if isinstance(maybeLoopEndState, LoopEndState): if maybeLoopEndState.epsilonOnlyTransitions and \ isinstance(maybeLoopEndState.transitions[0].target, RuleStopState): - state.precedenceRuleDecision = True + state.isPrecedenceDecision = True def verifyATN(self, atn): if not self.deserializationOptions.verifyATN: diff --git a/runtime/Python2/src/antlr4/atn/ATNState.py b/runtime/Python2/src/antlr4/atn/ATNState.py index d10e22ddb..038377704 100644 --- a/runtime/Python2/src/antlr4/atn/ATNState.py +++ b/runtime/Python2/src/antlr4/atn/ATNState.py @@ -262,7 +262,7 @@ class StarLoopEntryState(DecisionState): self.stateType = self.STAR_LOOP_ENTRY self.loopBackState = None # Indicates whether this state can benefit from a precedence DFA during SLL decision making. - self.precedenceRuleDecision = None + self.isPrecedenceDecision = None # Mark the end of a * or + loop. class LoopEndState(ATNState): diff --git a/runtime/Python2/src/antlr4/atn/ParserATNSimulator.py b/runtime/Python2/src/antlr4/atn/ParserATNSimulator.py index ee18bad5e..d48610249 100755 --- a/runtime/Python2/src/antlr4/atn/ParserATNSimulator.py +++ b/runtime/Python2/src/antlr4/atn/ParserATNSimulator.py @@ -266,7 +266,7 @@ from antlr4.atn.ATN import ATN from antlr4.atn.ATNConfig import ATNConfig from antlr4.atn.ATNConfigSet import ATNConfigSet from antlr4.atn.ATNSimulator import ATNSimulator -from antlr4.atn.ATNState import StarLoopEntryState, RuleStopState +from antlr4.atn.ATNState import StarLoopEntryState, RuleStopState, ATNState from antlr4.atn.PredictionMode import PredictionMode from antlr4.atn.SemanticContext import SemanticContext, AND, andContext, orContext from antlr4.atn.Transition import Transition, RuleTransition, ActionTransition, AtomTransition, SetTransition, NotSetTransition @@ -341,15 +341,6 @@ class ParserATNSimulator(ATNSimulator): " exec LA(1)==" + self.getLookaheadName(input) + ", outerContext=" + outerContext.toString(self.parser.literalNames, None)) - # If this is not a precedence DFA, we check the ATN start state - # to determine if this ATN start state is the decision for the - # closure block that determines whether a precedence rule - # should continue or complete. - # - if not dfa.precedenceDfa and isinstance(dfa.atnStartState, StarLoopEntryState): - if dfa.atnStartState.precedenceRuleDecision: - dfa.setPrecedenceDfa(True) - fullCtx = False s0_closure = self.computeStartState(dfa.atnStartState, ParserRuleContext.EMPTY, fullCtx) @@ -360,6 +351,7 @@ class ParserATNSimulator(ATNSimulator): # appropriate start state for the precedence level rather # than simply setting DFA.s0. # + dfa.s0.configs = s0_closure # not used for prediction but useful to know start configs anyway s0_closure = self.applyPrecedenceFilter(s0_closure) s0 = self.addDFAState(dfa, DFAState(configs=s0_closure)) dfa.setPrecedenceStartState(self.parser.getPrecedence(), s0) @@ -1168,7 +1160,13 @@ class ParserATNSimulator(ATNSimulator): # make sure to not return here, because EOF transitions can act as # both epsilon transitions and non-epsilon transitions. + first = True for t in p.transitions: + if first: + first = False + if self.canDropLoopEntryEdgeInLeftRecursiveRule(config): + continue + continueCollecting = collectPredicates and not isinstance(t, ActionTransition) c = self.getEpsilonTarget(config, t, continueCollecting, depth == 0, fullCtx, treatEofAsEpsilon) if c is not None: @@ -1205,6 +1203,161 @@ class ParserATNSimulator(ATNSimulator): self.closureCheckingStopState(c, configs, closureBusy, continueCollecting, fullCtx, newDepth, treatEofAsEpsilon) + + + # Implements first-edge (loop entry) elimination as an optimization + # during closure operations. See antlr/antlr4#1398. + # + # The optimization is to avoid adding the loop entry config when + # the exit path can only lead back to the same + # StarLoopEntryState after popping context at the rule end state + # (traversing only epsilon edges, so we're still in closure, in + # this same rule). + # + # We need to detect any state that can reach loop entry on + # epsilon w/o exiting rule. We don't have to look at FOLLOW + # links, just ensure that all stack tops for config refer to key + # states in LR rule. + # + # To verify we are in the right situation we must first check + # closure is at a StarLoopEntryState generated during LR removal. + # Then we check that each stack top of context is a return state + # from one of these cases: + # + # 1. 'not' expr, '(' type ')' expr. The return state points at loop entry state + # 2. expr op expr. The return state is the block end of internal block of (...)* + # 3. 'between' expr 'and' expr. The return state of 2nd expr reference. + # That state points at block end of internal block of (...)*. + # 4. expr '?' expr ':' expr. The return state points at block end, + # which points at loop entry state. + # + # If any is true for each stack top, then closure does not add a + # config to the current config set for edge[0], the loop entry branch. + # + # Conditions fail if any context for the current config is: + # + # a. empty (we'd fall out of expr to do a global FOLLOW which could + # even be to some weird spot in expr) or, + # b. lies outside of expr or, + # c. lies within expr but at a state not the BlockEndState + # generated during LR removal + # + # Do we need to evaluate predicates ever in closure for this case? + # + # No. Predicates, including precedence predicates, are only + # evaluated when computing a DFA start state. I.e., only before + # the lookahead (but not parser) consumes a token. + # + # There are no epsilon edges allowed in LR rule alt blocks or in + # the "primary" part (ID here). If closure is in + # StarLoopEntryState any lookahead operation will have consumed a + # token as there are no epsilon-paths that lead to + # StarLoopEntryState. We do not have to evaluate predicates + # therefore if we are in the generated StarLoopEntryState of a LR + # rule. Note that when making a prediction starting at that + # decision point, decision d=2, compute-start-state performs + # closure starting at edges[0], edges[1] emanating from + # StarLoopEntryState. That means it is not performing closure on + # StarLoopEntryState during compute-start-state. + # + # How do we know this always gives same prediction answer? + # + # Without predicates, loop entry and exit paths are ambiguous + # upon remaining input +b (in, say, a+b). Either paths lead to + # valid parses. Closure can lead to consuming + immediately or by + # falling out of this call to expr back into expr and loop back + # again to StarLoopEntryState to match +b. In this special case, + # we choose the more efficient path, which is to take the bypass + # path. + # + # The lookahead language has not changed because closure chooses + # one path over the other. Both paths lead to consuming the same + # remaining input during a lookahead operation. If the next token + # is an operator, lookahead will enter the choice block with + # operators. If it is not, lookahead will exit expr. Same as if + # closure had chosen to enter the choice block immediately. + # + # Closure is examining one config (some loopentrystate, some alt, + # context) which means it is considering exactly one alt. Closure + # always copies the same alt to any derived configs. + # + # How do we know this optimization doesn't mess up precedence in + # our parse trees? + # + # Looking through expr from left edge of stat only has to confirm + # that an input, say, a+b+c; begins with any valid interpretation + # of an expression. The precedence actually doesn't matter when + # making a decision in stat seeing through expr. It is only when + # parsing rule expr that we must use the precedence to get the + # right interpretation and, hence, parse tree. + # + # @since 4.6 + # + def canDropLoopEntryEdgeInLeftRecursiveRule(self, config): + # return False + p = config.state + # First check to see if we are in StarLoopEntryState generated during + # left-recursion elimination. For efficiency, also check if + # the context has an empty stack case. If so, it would mean + # global FOLLOW so we can't perform optimization + # Are we the special loop entry/exit state? or SLL wildcard + if p.stateType != ATNState.STAR_LOOP_ENTRY \ + or not p.isPrecedenceDecision \ + or config.context.isEmpty() \ + or config.context.hasEmptyPath(): + return False + + # Require all return states to return back to the same rule + # that p is in. + numCtxs = len(config.context) + for i in range(0, numCtxs): # for each stack context + returnState = self.atn.states[config.context.getReturnState(i)] + if returnState.ruleIndex != p.ruleIndex: + return False + + decisionStartState = p.transitions[0].target + blockEndStateNum = decisionStartState.endState.stateNumber + blockEndState = self.atn.states[blockEndStateNum] + + # Verify that the top of each stack context leads to loop entry/exit + # state through epsilon edges and w/o leaving rule. + for i in range(0, numCtxs): # for each stack context + returnStateNumber = config.context.getReturnState(i) + returnState = self.atn.states[returnStateNumber] + # all states must have single outgoing epsilon edge + if len(returnState.transitions) != 1 or not returnState.transitions[0].isEpsilon: + return False + + # Look for prefix op case like 'not expr', (' type ')' expr + returnStateTarget = returnState.transitions[0].target + if returnState.stateType == ATNState.BLOCK_END and returnStateTarget is p: + continue + + # Look for 'expr op expr' or case where expr's return state is block end + # of (...)* internal block; the block end points to loop back + # which points to p but we don't need to check that + if returnState is blockEndState: + continue + + # Look for ternary expr ? expr : expr. The return state points at block end, + # which points at loop entry state + if returnStateTarget is blockEndState: + continue + + # Look for complex prefix 'between expr and expr' case where 2nd expr's + # return state points at block end state of (...)* internal block + if returnStateTarget.stateType == ATNState.BLOCK_END \ + and len(returnStateTarget.transitions) == 1 \ + and returnStateTarget.transitions[0].isEpsilon \ + and returnStateTarget.transitions[0].target is p: + continue + + # anything else ain't conforming + return False + + return True + + def getRuleName(self, index): if self.parser is not None and index>=0: return self.parser.ruleNames[index] diff --git a/runtime/Python2/src/antlr4/dfa/DFA.py b/runtime/Python2/src/antlr4/dfa/DFA.py index 4ed5fb6e9..014dd18f9 100644 --- a/runtime/Python2/src/antlr4/dfa/DFA.py +++ b/runtime/Python2/src/antlr4/dfa/DFA.py @@ -27,6 +27,7 @@ # THEORY OF LIABILITY, WHETHER IN CONTRACT, STRICT LIABILITY, OR TORT # (INCLUDING NEGLIGENCE OR OTHERWISE) ARISING IN ANY WAY OUT OF THE USE OF # THIS SOFTWARE, EVEN IF ADVISED OF THE POSSIBILITY OF SUCH DAMAGE. +from antlr4.atn.ATNState import StarLoopEntryState from antlr4.atn.ATNConfigSet import ATNConfigSet from antlr4.dfa.DFAState import DFAState @@ -48,6 +49,15 @@ class DFA(object): # {@link #setPrecedenceDfa}. self.precedenceDfa = False + if isinstance(atnStartState, StarLoopEntryState): + if atnStartState.isPrecedenceDecision: + self.precedenceDfa = True + precedenceState = DFAState(configs=ATNConfigSet()) + precedenceState.edges = [] + precedenceState.isAcceptState = False + precedenceState.requiresFullContext = False + self.s0 = precedenceState + # Get the start state for a specific precedence value. # @@ -111,7 +121,7 @@ class DFA(object): if self.precedenceDfa != precedenceDfa: self._states = dict() if precedenceDfa: - precedenceState = DFAState(ATNConfigSet()) + precedenceState = DFAState(configs=ATNConfigSet()) precedenceState.edges = [] precedenceState.isAcceptState = False precedenceState.requiresFullContext = False diff --git a/runtime/Python3/src/antlr4/ParserInterpreter.py b/runtime/Python3/src/antlr4/ParserInterpreter.py index 1b6a990ef..91b9d56b2 100644 --- a/runtime/Python3/src/antlr4/ParserInterpreter.py +++ b/runtime/Python3/src/antlr4/ParserInterpreter.py @@ -71,7 +71,7 @@ class ParserInterpreter(Parser): for state in atn.states: if not isinstance(state, StarLoopEntryState): continue - if state.precedenceRuleDecision: + if state.isPrecedenceDecision: self.pushRecursionContextStates.add(state.stateNumber) # get atn simulator that knows how to do predictions self._interp = ParserATNSimulator(self, atn, self.decisionToDFA, self.sharedContextCache) diff --git a/runtime/Python3/src/antlr4/atn/ATNDeserializer.py b/runtime/Python3/src/antlr4/atn/ATNDeserializer.py index 2c225c3ac..ded137c73 100644 --- a/runtime/Python3/src/antlr4/atn/ATNDeserializer.py +++ b/runtime/Python3/src/antlr4/atn/ATNDeserializer.py @@ -368,7 +368,7 @@ class ATNDeserializer (object): # # Analyze the {@link StarLoopEntryState} states in the specified ATN to set - # the {@link StarLoopEntryState#precedenceRuleDecision} field to the + # the {@link StarLoopEntryState#isPrecedenceDecision} field to the # correct value. # # @param atn The ATN. @@ -387,7 +387,7 @@ class ATNDeserializer (object): if isinstance(maybeLoopEndState, LoopEndState): if maybeLoopEndState.epsilonOnlyTransitions and \ isinstance(maybeLoopEndState.transitions[0].target, RuleStopState): - state.precedenceRuleDecision = True + state.isPrecedenceDecision = True def verifyATN(self, atn:ATN): if not self.deserializationOptions.verifyATN: diff --git a/runtime/Python3/src/antlr4/atn/ATNSimulator.py b/runtime/Python3/src/antlr4/atn/ATNSimulator.py index ac5328757..bdde71a0d 100644 --- a/runtime/Python3/src/antlr4/atn/ATNSimulator.py +++ b/runtime/Python3/src/antlr4/atn/ATNSimulator.py @@ -37,7 +37,7 @@ from antlr4.dfa.DFAState import DFAState class ATNSimulator(object): # Must distinguish between missing edge and edge we know leads nowhere#/ - ERROR = DFAState(ATNConfigSet()) + ERROR = DFAState(configs=ATNConfigSet()) ERROR.stateNumber = 0x7FFFFFFF # The context cache maps all PredictionContext objects that are == diff --git a/runtime/Python3/src/antlr4/atn/ATNState.py b/runtime/Python3/src/antlr4/atn/ATNState.py index 6dd89f020..ae31e374d 100644 --- a/runtime/Python3/src/antlr4/atn/ATNState.py +++ b/runtime/Python3/src/antlr4/atn/ATNState.py @@ -261,7 +261,7 @@ class StarLoopEntryState(DecisionState): self.stateType = self.STAR_LOOP_ENTRY self.loopBackState = None # Indicates whether this state can benefit from a precedence DFA during SLL decision making. - self.precedenceRuleDecision = None + self.isPrecedenceDecision = None # Mark the end of a * or + loop. class LoopEndState(ATNState): diff --git a/runtime/Python3/src/antlr4/atn/ParserATNSimulator.py b/runtime/Python3/src/antlr4/atn/ParserATNSimulator.py index 7cb5867b8..b085530b4 100644 --- a/runtime/Python3/src/antlr4/atn/ParserATNSimulator.py +++ b/runtime/Python3/src/antlr4/atn/ParserATNSimulator.py @@ -346,15 +346,6 @@ class ParserATNSimulator(ATNSimulator): " exec LA(1)==" + self.getLookaheadName(input) + ", outerContext=" + outerContext.toString(self.parser.literalNames, None)) - # If this is not a precedence DFA, we check the ATN start state - # to determine if this ATN start state is the decision for the - # closure block that determines whether a precedence rule - # should continue or complete. - # - if not dfa.precedenceDfa and isinstance(dfa.atnStartState, StarLoopEntryState): - if dfa.atnStartState.precedenceRuleDecision: - dfa.setPrecedenceDfa(True) - fullCtx = False s0_closure = self.computeStartState(dfa.atnStartState, ParserRuleContext.EMPTY, fullCtx) @@ -365,6 +356,7 @@ class ParserATNSimulator(ATNSimulator): # appropriate start state for the precedence level rather # than simply setting DFA.s0. # + dfa.s0.configs = s0_closure # not used for prediction but useful to know start configs anyway s0_closure = self.applyPrecedenceFilter(s0_closure) s0 = self.addDFAState(dfa, DFAState(configs=s0_closure)) dfa.setPrecedenceStartState(self.parser.getPrecedence(), s0) @@ -1173,7 +1165,13 @@ class ParserATNSimulator(ATNSimulator): # make sure to not return here, because EOF transitions can act as # both epsilon transitions and non-epsilon transitions. + first = True for t in p.transitions: + if first: + first = False + if self.canDropLoopEntryEdgeInLeftRecursiveRule(config): + continue + continueCollecting = collectPredicates and not isinstance(t, ActionTransition) c = self.getEpsilonTarget(config, t, continueCollecting, depth == 0, fullCtx, treatEofAsEpsilon) if c is not None: @@ -1210,6 +1208,161 @@ class ParserATNSimulator(ATNSimulator): self.closureCheckingStopState(c, configs, closureBusy, continueCollecting, fullCtx, newDepth, treatEofAsEpsilon) + + + # Implements first-edge (loop entry) elimination as an optimization + # during closure operations. See antlr/antlr4#1398. + # + # The optimization is to avoid adding the loop entry config when + # the exit path can only lead back to the same + # StarLoopEntryState after popping context at the rule end state + # (traversing only epsilon edges, so we're still in closure, in + # this same rule). + # + # We need to detect any state that can reach loop entry on + # epsilon w/o exiting rule. We don't have to look at FOLLOW + # links, just ensure that all stack tops for config refer to key + # states in LR rule. + # + # To verify we are in the right situation we must first check + # closure is at a StarLoopEntryState generated during LR removal. + # Then we check that each stack top of context is a return state + # from one of these cases: + # + # 1. 'not' expr, '(' type ')' expr. The return state points at loop entry state + # 2. expr op expr. The return state is the block end of internal block of (...)* + # 3. 'between' expr 'and' expr. The return state of 2nd expr reference. + # That state points at block end of internal block of (...)*. + # 4. expr '?' expr ':' expr. The return state points at block end, + # which points at loop entry state. + # + # If any is true for each stack top, then closure does not add a + # config to the current config set for edge[0], the loop entry branch. + # + # Conditions fail if any context for the current config is: + # + # a. empty (we'd fall out of expr to do a global FOLLOW which could + # even be to some weird spot in expr) or, + # b. lies outside of expr or, + # c. lies within expr but at a state not the BlockEndState + # generated during LR removal + # + # Do we need to evaluate predicates ever in closure for this case? + # + # No. Predicates, including precedence predicates, are only + # evaluated when computing a DFA start state. I.e., only before + # the lookahead (but not parser) consumes a token. + # + # There are no epsilon edges allowed in LR rule alt blocks or in + # the "primary" part (ID here). If closure is in + # StarLoopEntryState any lookahead operation will have consumed a + # token as there are no epsilon-paths that lead to + # StarLoopEntryState. We do not have to evaluate predicates + # therefore if we are in the generated StarLoopEntryState of a LR + # rule. Note that when making a prediction starting at that + # decision point, decision d=2, compute-start-state performs + # closure starting at edges[0], edges[1] emanating from + # StarLoopEntryState. That means it is not performing closure on + # StarLoopEntryState during compute-start-state. + # + # How do we know this always gives same prediction answer? + # + # Without predicates, loop entry and exit paths are ambiguous + # upon remaining input +b (in, say, a+b). Either paths lead to + # valid parses. Closure can lead to consuming + immediately or by + # falling out of this call to expr back into expr and loop back + # again to StarLoopEntryState to match +b. In this special case, + # we choose the more efficient path, which is to take the bypass + # path. + # + # The lookahead language has not changed because closure chooses + # one path over the other. Both paths lead to consuming the same + # remaining input during a lookahead operation. If the next token + # is an operator, lookahead will enter the choice block with + # operators. If it is not, lookahead will exit expr. Same as if + # closure had chosen to enter the choice block immediately. + # + # Closure is examining one config (some loopentrystate, some alt, + # context) which means it is considering exactly one alt. Closure + # always copies the same alt to any derived configs. + # + # How do we know this optimization doesn't mess up precedence in + # our parse trees? + # + # Looking through expr from left edge of stat only has to confirm + # that an input, say, a+b+c; begins with any valid interpretation + # of an expression. The precedence actually doesn't matter when + # making a decision in stat seeing through expr. It is only when + # parsing rule expr that we must use the precedence to get the + # right interpretation and, hence, parse tree. + # + # @since 4.6 + # + def canDropLoopEntryEdgeInLeftRecursiveRule(self, config): + # return False + p = config.state + # First check to see if we are in StarLoopEntryState generated during + # left-recursion elimination. For efficiency, also check if + # the context has an empty stack case. If so, it would mean + # global FOLLOW so we can't perform optimization + # Are we the special loop entry/exit state? or SLL wildcard + if p.stateType != ATNState.STAR_LOOP_ENTRY \ + or not p.isPrecedenceDecision \ + or config.context.isEmpty() \ + or config.context.hasEmptyPath(): + return False + + # Require all return states to return back to the same rule + # that p is in. + numCtxs = len(config.context) + for i in range(0, numCtxs): # for each stack context + returnState = self.atn.states[config.context.getReturnState(i)] + if returnState.ruleIndex != p.ruleIndex: + return False + + decisionStartState = p.transitions[0].target + blockEndStateNum = decisionStartState.endState.stateNumber + blockEndState = self.atn.states[blockEndStateNum] + + # Verify that the top of each stack context leads to loop entry/exit + # state through epsilon edges and w/o leaving rule. + for i in range(0, numCtxs): # for each stack context + returnStateNumber = config.context.getReturnState(i) + returnState = self.atn.states[returnStateNumber] + # all states must have single outgoing epsilon edge + if len(returnState.transitions) != 1 or not returnState.transitions[0].isEpsilon: + return False + + # Look for prefix op case like 'not expr', (' type ')' expr + returnStateTarget = returnState.transitions[0].target + if returnState.stateType == ATNState.BLOCK_END and returnStateTarget is p: + continue + + # Look for 'expr op expr' or case where expr's return state is block end + # of (...)* internal block; the block end points to loop back + # which points to p but we don't need to check that + if returnState is blockEndState: + continue + + # Look for ternary expr ? expr : expr. The return state points at block end, + # which points at loop entry state + if returnStateTarget is blockEndState: + continue + + # Look for complex prefix 'between expr and expr' case where 2nd expr's + # return state points at block end state of (...)* internal block + if returnStateTarget.stateType == ATNState.BLOCK_END \ + and len(returnStateTarget.transitions) == 1 \ + and returnStateTarget.transitions[0].isEpsilon \ + and returnStateTarget.transitions[0].target is p: + continue + + # anything else ain't conforming + return False + + return True + + def getRuleName(self, index:int): if self.parser is not None and index>=0: return self.parser.ruleNames[index] diff --git a/runtime/Python3/src/antlr4/dfa/DFA.py b/runtime/Python3/src/antlr4/dfa/DFA.py index 80a48b56e..19881b3fd 100644 --- a/runtime/Python3/src/antlr4/dfa/DFA.py +++ b/runtime/Python3/src/antlr4/dfa/DFA.py @@ -27,6 +27,7 @@ # THEORY OF LIABILITY, WHETHER IN CONTRACT, STRICT LIABILITY, OR TORT # (INCLUDING NEGLIGENCE OR OTHERWISE) ARISING IN ANY WAY OUT OF THE USE OF # THIS SOFTWARE, EVEN IF ADVISED OF THE POSSIBILITY OF SUCH DAMAGE. +from antlr4.atn.ATNState import StarLoopEntryState from antlr4.atn.ATNConfigSet import ATNConfigSet from antlr4.atn.ATNState import DecisionState @@ -49,6 +50,15 @@ class DFA(object): # {@link #setPrecedenceDfa}. self.precedenceDfa = False + if isinstance(atnStartState, StarLoopEntryState): + if atnStartState.isPrecedenceDecision: + self.precedenceDfa = True + precedenceState = DFAState(configs=ATNConfigSet()) + precedenceState.edges = [] + precedenceState.isAcceptState = False + precedenceState.requiresFullContext = False + self.s0 = precedenceState + # Get the start state for a specific precedence value. # @@ -112,7 +122,7 @@ class DFA(object): if self.precedenceDfa != precedenceDfa: self._states = dict() if precedenceDfa: - precedenceState = DFAState(ATNConfigSet()) + precedenceState = DFAState(configs=ATNConfigSet()) precedenceState.edges = [] precedenceState.isAcceptState = False precedenceState.requiresFullContext = False