org.antlr.runtime.debug.Profiler Maven / Gradle / Ivy
Go to download
Show more of this group Show more artifacts with this name
Show all versions of antlr-runtime Show documentation
Show all versions of antlr-runtime Show documentation
A framework for constructing recognizers, compilers, and translators from grammatical descriptions containing Java, C#, C++, or Python actions.
/*
[The "BSD license"]
Copyright (c) 2005-2009 Terence Parr
All rights reserved.
Redistribution and use in source and binary forms, with or without
modification, are permitted provided that the following conditions
are met:
1. Redistributions of source code must retain the above copyright
notice, this list of conditions and the following disclaimer.
2. Redistributions in binary form must reproduce the above copyright
notice, this list of conditions and the following disclaimer in the
documentation and/or other materials provided with the distribution.
3. The name of the author may not be used to endorse or promote products
derived from this software without specific prior written permission.
THIS SOFTWARE IS PROVIDED BY THE AUTHOR ``AS IS'' AND ANY EXPRESS OR
IMPLIED WARRANTIES, INCLUDING, BUT NOT LIMITED TO, THE IMPLIED WARRANTIES
OF MERCHANTABILITY AND FITNESS FOR A PARTICULAR PURPOSE ARE DISCLAIMED.
IN NO EVENT SHALL THE AUTHOR BE LIABLE FOR ANY DIRECT, INDIRECT,
INCIDENTAL, SPECIAL, EXEMPLARY, OR CONSEQUENTIAL DAMAGES (INCLUDING, BUT
NOT LIMITED TO, PROCUREMENT OF SUBSTITUTE GOODS OR SERVICES; LOSS OF USE,
DATA, OR PROFITS; OR BUSINESS INTERRUPTION) HOWEVER CAUSED AND ON ANY
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.
*/
package org.antlr.runtime.debug;
import org.antlr.runtime.*;
import org.antlr.runtime.misc.DoubleKeyMap;
import java.util.*;
/** Using the debug event interface, track what is happening in the parser
* and record statistics about the runtime.
*/
public class Profiler extends BlankDebugEventListener {
public static final String DATA_SEP = "\t";
public static final String newline = System.getProperty("line.separator");
static boolean dump = false;
public static class ProfileStats {
public String Version;
public String name;
public int numRuleInvocations;
public int numUniqueRulesInvoked;
public int numDecisionEvents;
public int numDecisionsCovered;
public int numDecisionsThatPotentiallyBacktrack;
public int numDecisionsThatDoBacktrack;
public int maxRuleInvocationDepth;
public float avgkPerDecisionEvent;
public float avgkPerBacktrackingDecisionEvent;
public float averageDecisionPercentBacktracks;
public int numBacktrackOccurrences; // doesn't count gated DFA edges
public int numFixedDecisions;
public int minDecisionMaxFixedLookaheads;
public int maxDecisionMaxFixedLookaheads;
public int avgDecisionMaxFixedLookaheads;
public int stddevDecisionMaxFixedLookaheads;
public int numCyclicDecisions;
public int minDecisionMaxCyclicLookaheads;
public int maxDecisionMaxCyclicLookaheads;
public int avgDecisionMaxCyclicLookaheads;
public int stddevDecisionMaxCyclicLookaheads;
// int Stats.min(toArray(decisionMaxSynPredLookaheads);
// int Stats.max(toArray(decisionMaxSynPredLookaheads);
// int Stats.avg(toArray(decisionMaxSynPredLookaheads);
// int Stats.stddev(toArray(decisionMaxSynPredLookaheads);
public int numSemanticPredicates;
public int numTokens;
public int numHiddenTokens;
public int numCharsMatched;
public int numHiddenCharsMatched;
public int numReportedErrors;
public int numMemoizationCacheHits;
public int numMemoizationCacheMisses;
public int numGuessingRuleInvocations;
public int numMemoizationCacheEntries;
}
public static class DecisionDescriptor {
public int decision;
public String fileName;
public String ruleName;
public int line;
public int pos;
public boolean couldBacktrack;
public int n;
public float avgk; // avg across all decision events
public int maxk;
public int numBacktrackOccurrences;
public int numSemPredEvals;
}
// all about a specific exec of a single decision
public static class DecisionEvent {
public DecisionDescriptor decision;
public int startIndex;
public int k;
public boolean backtracks; // doesn't count gated DFA edges
public boolean evalSemPred;
public long startTime;
public long stopTime;
public int numMemoizationCacheHits;
public int numMemoizationCacheMisses;
}
/** Because I may change the stats, I need to track that for later
* computations to be consistent.
*/
public static final String Version = "3";
public static final String RUNTIME_STATS_FILENAME = "runtime.stats";
/** Ack, should not store parser; can't do remote stuff. Well, we pass
* input stream around too so I guess it's ok.
*/
public DebugParser parser = null;
// working variables
protected int ruleLevel = 0;
//protected int decisionLevel = 0;
protected Token lastRealTokenTouchedInDecision;
protected Set uniqueRules = new HashSet();
protected Stack currentGrammarFileName = new Stack();
protected Stack currentRuleName = new Stack();
protected Stack currentLine = new Stack();
protected Stack currentPos = new Stack();
// Vector
//protected Vector decisions = new Vector(200); // need setSize
protected DoubleKeyMap decisions =
new DoubleKeyMap();
// Record a DecisionData for each decision we hit while parsing
protected List decisionEvents = new ArrayList();
protected Stack decisionStack = new Stack();
protected int backtrackDepth;
ProfileStats stats = new ProfileStats();
public Profiler() {
}
public Profiler(DebugParser parser) {
this.parser = parser;
}
public void enterRule(String grammarFileName, String ruleName) {
// System.out.println("enterRule "+grammarFileName+":"+ruleName);
ruleLevel++;
stats.numRuleInvocations++;
uniqueRules.add(grammarFileName+":"+ruleName);
stats.maxRuleInvocationDepth = Math.max(stats.maxRuleInvocationDepth, ruleLevel);
currentGrammarFileName.push( grammarFileName );
currentRuleName.push( ruleName );
}
public void exitRule(String grammarFileName, String ruleName) {
ruleLevel--;
currentGrammarFileName.pop();
currentRuleName.pop();
}
/** Track memoization; this is not part of standard debug interface
* but is triggered by profiling. Code gen inserts an override
* for this method in the recognizer, which triggers this method.
* Called from alreadyParsedRule().
*/
public void examineRuleMemoization(IntStream input,
int ruleIndex,
int stopIndex, // index or MEMO_RULE_UNKNOWN...
String ruleName)
{
if (dump) System.out.println("examine memo "+ruleName+" at "+input.index()+": "+stopIndex);
if ( stopIndex==BaseRecognizer.MEMO_RULE_UNKNOWN ) {
//System.out.println("rule "+ruleIndex+" missed @ "+input.index());
stats.numMemoizationCacheMisses++;
stats.numGuessingRuleInvocations++; // we'll have to enter
currentDecision().numMemoizationCacheMisses++;
}
else {
// regardless of rule success/failure, if in cache, we have a cache hit
//System.out.println("rule "+ruleIndex+" hit @ "+input.index());
stats.numMemoizationCacheHits++;
currentDecision().numMemoizationCacheHits++;
}
}
/** Warning: doesn't track success/failure, just unique recording event */
public void memoize(IntStream input,
int ruleIndex,
int ruleStartIndex,
String ruleName)
{
// count how many entries go into table
if (dump) System.out.println("memoize "+ruleName);
stats.numMemoizationCacheEntries++;
}
@Override
public void location(int line, int pos) {
currentLine.push(line);
currentPos.push(pos);
}
public void enterDecision(int decisionNumber, boolean couldBacktrack) {
lastRealTokenTouchedInDecision = null;
stats.numDecisionEvents++;
int startingLookaheadIndex = parser.getTokenStream().index();
TokenStream input = parser.getTokenStream();
if ( dump ) System.out.println("enterDecision canBacktrack="+couldBacktrack+" "+ decisionNumber +
" backtrack depth " + backtrackDepth +
" @ " + input.get(input.index()) +
" rule " +locationDescription());
String g = (String) currentGrammarFileName.peek();
DecisionDescriptor descriptor = decisions.get(g, decisionNumber);
if ( descriptor == null ) {
descriptor = new DecisionDescriptor();
decisions.put(g, decisionNumber, descriptor);
descriptor.decision = decisionNumber;
descriptor.fileName = (String)currentGrammarFileName.peek();
descriptor.ruleName = (String)currentRuleName.peek();
descriptor.line = (Integer)currentLine.peek();
descriptor.pos = (Integer)currentPos.peek();
descriptor.couldBacktrack = couldBacktrack;
}
descriptor.n++;
DecisionEvent d = new DecisionEvent();
decisionStack.push(d);
d.decision = descriptor;
d.startTime = System.currentTimeMillis();
d.startIndex = startingLookaheadIndex;
}
public void exitDecision(int decisionNumber) {
DecisionEvent d = decisionStack.pop();
d.stopTime = System.currentTimeMillis();
int lastTokenIndex = lastRealTokenTouchedInDecision.getTokenIndex();
int numHidden = getNumberOfHiddenTokens(d.startIndex, lastTokenIndex);
int depth = lastTokenIndex - d.startIndex - numHidden + 1; // +1 counts consuming start token as 1
d.k = depth;
d.decision.maxk = Math.max(d.decision.maxk, depth);
if (dump) System.out.println("exitDecision "+decisionNumber+" in "+d.decision.ruleName+
" lookahead "+d.k +" max token "+lastRealTokenTouchedInDecision);
decisionEvents.add(d); // done with decision; track all
}
public void consumeToken(Token token) {
if (dump) System.out.println("consume token "+token);
if ( !inDecision() ) {
stats.numTokens++;
return;
}
if ( lastRealTokenTouchedInDecision==null ||
lastRealTokenTouchedInDecision.getTokenIndex() < token.getTokenIndex() )
{
lastRealTokenTouchedInDecision = token;
}
DecisionEvent d = currentDecision();
// compute lookahead depth
int thisRefIndex = token.getTokenIndex();
int numHidden = getNumberOfHiddenTokens(d.startIndex, thisRefIndex);
int depth = thisRefIndex - d.startIndex - numHidden + 1; // +1 counts consuming start token as 1
//d.maxk = Math.max(d.maxk, depth);
if (dump) System.out.println("consume "+thisRefIndex+" "+depth+" tokens ahead in "+
d.decision.ruleName+"-"+d.decision.decision+" start index "+d.startIndex);
}
/** The parser is in a decision if the decision depth > 0. This
* works for backtracking also, which can have nested decisions.
*/
public boolean inDecision() {
return decisionStack.size()>0;
}
public void consumeHiddenToken(Token token) {
//System.out.println("consume hidden token "+token);
if ( !inDecision() ) stats.numHiddenTokens++;
}
/** Track refs to lookahead if in a fixed/nonfixed decision.
*/
public void LT(int i, Token t) {
if ( inDecision() && i>0 ) {
DecisionEvent d = currentDecision();
if (dump) System.out.println("LT("+i+")="+t+" index "+t.getTokenIndex()+" relative to "+d.decision.ruleName+"-"+
d.decision.decision+" start index "+d.startIndex);
if ( lastRealTokenTouchedInDecision==null ||
lastRealTokenTouchedInDecision.getTokenIndex() < t.getTokenIndex() )
{
lastRealTokenTouchedInDecision = t;
if (dump) System.out.println("set last token "+lastRealTokenTouchedInDecision);
}
// get starting index off stack
// int stackTop = lookaheadStack.size()-1;
// Integer startingIndex = (Integer)lookaheadStack.get(stackTop);
// // compute lookahead depth
// int thisRefIndex = parser.getTokenStream().index();
// int numHidden =
// getNumberOfHiddenTokens(startingIndex.intValue(), thisRefIndex);
// int depth = i + thisRefIndex - startingIndex.intValue() - numHidden;
// /*
// System.out.println("LT("+i+") @ index "+thisRefIndex+" is depth "+depth+
// " max is "+maxLookaheadInCurrentDecision);
// */
// if ( depth>maxLookaheadInCurrentDecision ) {
// maxLookaheadInCurrentDecision = depth;
// }
// d.maxk = currentDecision()/
}
}
/** Track backtracking decisions. You'll see a fixed or cyclic decision
* and then a backtrack.
*
* enter rule
* ...
* enter decision
* LA and possibly consumes (for cyclic DFAs)
* begin backtrack level
* mark m
* rewind m
* end backtrack level, success
* exit decision
* ...
* exit rule
*/
public void beginBacktrack(int level) {
if (dump) System.out.println("enter backtrack "+level);
backtrackDepth++;
DecisionEvent e = currentDecision();
if ( e.decision.couldBacktrack ) {
stats.numBacktrackOccurrences++;
e.decision.numBacktrackOccurrences++;
e.backtracks = true;
}
}
/** Successful or not, track how much lookahead synpreds use */
public void endBacktrack(int level, boolean successful) {
if (dump) System.out.println("exit backtrack "+level+": "+successful);
backtrackDepth--;
}
@Override
public void mark(int i) {
if (dump) System.out.println("mark "+i);
}
@Override
public void rewind(int i) {
if (dump) System.out.println("rewind "+i);
}
@Override
public void rewind() {
if (dump) System.out.println("rewind");
}
protected DecisionEvent currentDecision() {
return decisionStack.peek();
}
public void recognitionException(RecognitionException e) {
stats.numReportedErrors++;
}
public void semanticPredicate(boolean result, String predicate) {
stats.numSemanticPredicates++;
if ( inDecision() ) {
DecisionEvent d = currentDecision();
d.evalSemPred = true;
d.decision.numSemPredEvals++;
if (dump) System.out.println("eval "+predicate+" in "+d.decision.ruleName+"-"+
d.decision.decision);
}
}
public void terminate() {
for (DecisionEvent e : decisionEvents) {
//System.out.println("decision "+e.decision.decision+": k="+e.k);
e.decision.avgk += e.k;
stats.avgkPerDecisionEvent += e.k;
if ( e.backtracks ) { // doesn't count gated syn preds on DFA edges
stats.avgkPerBacktrackingDecisionEvent += e.k;
}
}
stats.averageDecisionPercentBacktracks = 0.0f;
for (DecisionDescriptor d : decisions.values()) {
stats.numDecisionsCovered++;
d.avgk /= (double)d.n;
if ( d.couldBacktrack ) {
stats.numDecisionsThatPotentiallyBacktrack++;
float percentBacktracks = d.numBacktrackOccurrences / (float)d.n;
//System.out.println("dec "+d.decision+" backtracks "+percentBacktracks*100+"%");
stats.averageDecisionPercentBacktracks += percentBacktracks;
}
// ignore rules that backtrack along gated DFA edges
if ( d.numBacktrackOccurrences > 0 ) {
stats.numDecisionsThatDoBacktrack++;
}
}
stats.averageDecisionPercentBacktracks /= stats.numDecisionsThatPotentiallyBacktrack;
stats.averageDecisionPercentBacktracks *= 100; // it's a percentage
stats.avgkPerDecisionEvent /= stats.numDecisionEvents;
stats.avgkPerBacktrackingDecisionEvent /= (double)stats.numBacktrackOccurrences;
System.err.println(toString());
System.err.println(getDecisionStatsDump());
// String stats = toNotifyString();
// try {
// Stats.writeReport(RUNTIME_STATS_FILENAME,stats);
// }
// catch (IOException ioe) {
// System.err.println(ioe);
// ioe.printStackTrace(System.err);
// }
}
public void setParser(DebugParser parser) {
this.parser = parser;
}
// R E P O R T I N G
public String toNotifyString() {
StringBuffer buf = new StringBuffer();
buf.append(Version);
buf.append('\t');
buf.append(parser.getClass().getName());
// buf.append('\t');
// buf.append(numRuleInvocations);
// buf.append('\t');
// buf.append(maxRuleInvocationDepth);
// buf.append('\t');
// buf.append(numFixedDecisions);
// buf.append('\t');
// buf.append(Stats.min(decisionMaxFixedLookaheads));
// buf.append('\t');
// buf.append(Stats.max(decisionMaxFixedLookaheads));
// buf.append('\t');
// buf.append(Stats.avg(decisionMaxFixedLookaheads));
// buf.append('\t');
// buf.append(Stats.stddev(decisionMaxFixedLookaheads));
// buf.append('\t');
// buf.append(numCyclicDecisions);
// buf.append('\t');
// buf.append(Stats.min(decisionMaxCyclicLookaheads));
// buf.append('\t');
// buf.append(Stats.max(decisionMaxCyclicLookaheads));
// buf.append('\t');
// buf.append(Stats.avg(decisionMaxCyclicLookaheads));
// buf.append('\t');
// buf.append(Stats.stddev(decisionMaxCyclicLookaheads));
// buf.append('\t');
// buf.append(numBacktrackDecisions);
// buf.append('\t');
// buf.append(Stats.min(toArray(decisionMaxSynPredLookaheads)));
// buf.append('\t');
// buf.append(Stats.max(toArray(decisionMaxSynPredLookaheads)));
// buf.append('\t');
// buf.append(Stats.avg(toArray(decisionMaxSynPredLookaheads)));
// buf.append('\t');
// buf.append(Stats.stddev(toArray(decisionMaxSynPredLookaheads)));
// buf.append('\t');
// buf.append(numSemanticPredicates);
// buf.append('\t');
// buf.append(parser.getTokenStream().size());
// buf.append('\t');
// buf.append(numHiddenTokens);
// buf.append('\t');
// buf.append(numCharsMatched);
// buf.append('\t');
// buf.append(numHiddenCharsMatched);
// buf.append('\t');
// buf.append(numberReportedErrors);
// buf.append('\t');
// buf.append(numMemoizationCacheHits);
// buf.append('\t');
// buf.append(numMemoizationCacheMisses);
// buf.append('\t');
// buf.append(numGuessingRuleInvocations);
// buf.append('\t');
// buf.append(numMemoizationCacheEntries);
return buf.toString();
}
public String toString() {
return toString(getReport());
}
public ProfileStats getReport() {
// TokenStream input = parser.getTokenStream();
// for (int i=0; i