public void TestSimpleRangeVersusChar()
{
Grammar g = new Grammar(
"lexer grammar t;\n" +
"A : 'a'..'z' '@' | 'k' '$' ;" );
g.CreateLookaheadDFAs();
string expecting =
".s0-'k'->.s1" + NewLine +
".s0-{'a'..'j', 'l'..'z'}->:s2=>1" + NewLine +
".s1-'$'->:s3=>2" + NewLine +
".s1-'@'->:s2=>1" + NewLine;
checkDecision( g, 1, expecting, null );
}
//throws Exception
protected void checkDecision( Grammar g,
int decision,
string expecting,
int[] expectingUnreachableAlts )
{
Antlr3.AntlrTool tool = new Antlr3.AntlrTool();
// mimic actions of org.antlr.Tool first time for grammar g
if ( g.CodeGenerator == null )
{
CodeGenerator generator = new CodeGenerator( tool, g, "Java" );
g.CodeGenerator = generator;
g.BuildNFA();
g.CreateLookaheadDFAs( false );
}
DFA dfa = g.GetLookaheadDFA( decision );
Assert.IsNotNull(dfa, "unknown decision #" + decision);
FASerializer serializer = new FASerializer( g );
string result = serializer.Serialize( dfa.StartState );
//System.out.print(result);
var nonDetAlts = dfa.UnreachableAlts;
//System.out.println("alts w/o predict state="+nonDetAlts);
// first make sure nondeterministic alts are as expected
if ( expectingUnreachableAlts == null )
{
if ( nonDetAlts != null && nonDetAlts.Count != 0 )
{
Console.Error.WriteLine( "nondeterministic alts (should be empty): " + ( (IList)nonDetAlts ).ToElementString() );
}
Assert.AreEqual(0, nonDetAlts != null ? nonDetAlts.Count : 0, "unreachable alts mismatch");
}
else
{
for ( int i = 0; i < expectingUnreachableAlts.Length; i++ )
{
Assert.IsTrue(nonDetAlts != null ? nonDetAlts.Contains(expectingUnreachableAlts[i]) : false, "unreachable alts mismatch");
}
}
Assert.AreEqual( expecting, result );
}
public void TestRecursive()
{
// this is cool because the 3rd alt includes !(all other possibilities)
Grammar g = new Grammar(
"lexer grammar duh;\n" +
"SUBTEMPLATE\n" +
" : '{'\n" +
" ( SUBTEMPLATE\n" +
" | ESC\n" +
" | ~('}'|'\\\\'|'{')\n" +
" )*\n" +
" '}'\n" +
" ;\n" +
"fragment\n" +
"ESC : '\\\\' . ;" );
g.CreateLookaheadDFAs();
string expecting =
".s0-'\\\\'->:s2=>2" + NewLine +
".s0-'{'->:s1=>1" + NewLine +
".s0-'}'->:s4=>4" + NewLine +
".s0-{'\\u0000'..'[', ']'..'z', '|', '~'..'\\uFFFF'}->:s3=>3" + NewLine;
checkDecision( g, 1, expecting, null );
}
public void TestRecursive2()
{
// this is also cool because it resolves \\ to be ESC alt; it's just
// less efficient of a DFA
Grammar g = new Grammar(
"lexer grammar duh;\n" +
"SUBTEMPLATE\n" +
" : '{'\n" +
" ( SUBTEMPLATE\n" +
" | ESC\n" +
" | ~('}'|'{')\n" +
" )*\n" +
" '}'\n" +
" ;\n" +
"fragment\n" +
"ESC : '\\\\' . ;" );
g.CreateLookaheadDFAs();
string expecting =
".s0-'\\\\'->.s3" + NewLine +
".s0-'{'->:s2=>1" + NewLine +
".s0-'}'->:s1=>4" + NewLine +
".s0-{'\\u0000'..'[', ']'..'z', '|', '~'..'\\uFFFF'}->:s5=>3" + NewLine +
".s3-'\\\\'->:s8=>2" + NewLine +
".s3-'{'->:s7=>2" + NewLine +
".s3-'}'->.s4" + NewLine +
".s3-{'\\u0000'..'[', ']'..'z', '|', '~'..'\\uFFFF'}->:s6=>2" + NewLine +
".s4-'\\u0000'..'\\uFFFF'->:s6=>2" + NewLine +
".s4-<EOT>->:s5=>3" + NewLine;
checkDecision( g, 1, expecting, null );
}
public void TestNotTokenInLexer()
{
Grammar g = new Grammar(
"lexer grammar T;\n" +
"A : 'x' ('a' | ~B {;}) ;\n" +
"B : 'a' ;\n" );
g.CreateLookaheadDFAs();
string expecting =
".s0-'a'->:s1=>1" + NewLine +
".s0-{'\\u0000'..'`', 'b'..'\\uFFFF'}->:s2=>2" + NewLine;
checkDecision( g, 1, expecting, null );
}
public void TestRangeWithDisjointSet()
{
Grammar g = new Grammar(
"lexer grammar t;\n" +
"A : 'a'..'z' '@'\n" +
" | ('k'|'9'|'p') '$'\n" +
" ;\n" );
g.CreateLookaheadDFAs();
// must break up a..z into {'a'..'j', 'l'..'o', 'q'..'z'}
string expecting =
".s0-'9'->:s3=>2" + NewLine +
".s0-{'a'..'j', 'l'..'o', 'q'..'z'}->:s2=>1" + NewLine +
".s0-{'k', 'p'}->.s1" + NewLine +
".s1-'$'->:s3=>2" + NewLine +
".s1-'@'->:s2=>1" + NewLine;
checkDecision( g, 1, expecting, null );
}
public void TestDisjointSetCollidingWithTwoRanges()
{
Grammar g = new Grammar(
"lexer grammar t;\n" +
"A : ('a'..'z'|'0'..'9') '@'\n" +
" | ('k'|'9'|'p') '$'\n" +
" ;\n" );
g.CreateLookaheadDFAs( false );
// must break up a..z into {'a'..'j', 'l'..'o', 'q'..'z'} and 0..9
// into 0..8
string expecting =
".s0-{'0'..'8', 'a'..'j', 'l'..'o', 'q'..'z'}->:s2=>1" + NewLine +
".s0-{'9', 'k', 'p'}->.s1" + NewLine +
".s1-'$'->:s3=>2" + NewLine +
".s1-'@'->:s2=>1" + NewLine;
checkDecision( g, 1, expecting, null );
}
public void TestNotSetFragmentInLexer()
{
Grammar g = new Grammar(
"lexer grammar T;\n" +
"A : B | ~B {;} ;\n" +
"fragment B : 'a'|'b' ;\n" );
g.CreateLookaheadDFAs();
string expecting =
".s0-'a'..'b'->:s1=>1" + NewLine +
".s0-{'\\u0000'..'`', 'c'..'\\uFFFF'}->:s2=>2" + NewLine;
checkDecision( g, 1, expecting, null );
}
public void TestPredsUsedAfterRecursionOverflow()
{
// analysis must bail out due to non-LL(*) nature (ovf)
// retries with k=1 (but with LL(*) algorithm not optimized version
// as it has preds)
Grammar g = new Grammar(
"parser grammar P;\n" +
"s : {p1}? e '.' | {p2}? e ':' ;\n" +
"e : '(' e ')' | INT ;\n" );
string expecting =
".s0-'('->.s1" + NewLine +
".s0-INT->.s4" + NewLine +
".s1-{p1}?->:s2=>1" + NewLine +
".s1-{p2}?->:s3=>2" + NewLine +
".s4-{p1}?->:s2=>1" + NewLine +
".s4-{p2}?->:s3=>2" + NewLine;
DecisionProbe.verbose = true; // make sure we get all error info
ErrorQueue equeue = new ErrorQueue();
ErrorManager.SetErrorListener( equeue );
CodeGenerator generator = new CodeGenerator( newTool(), g, "Java" );
g.CodeGenerator = generator;
if ( g.NumberOfDecisions == 0 )
{
g.BuildNFA();
g.CreateLookaheadDFAs( false );
}
Assert.AreEqual(0, equeue.size(), "unexpected number of expected problems");
checkDecision( g, 1, expecting, null, null, null, null, null, 0, false );
}
//throws Exception
protected void checkDecision( Grammar g,
int decision,
string expecting,
int[] expectingUnreachableAlts,
int[] expectingNonDetAlts,
string expectingAmbigInput,
int[] expectingInsufficientPredAlts,
int[] expectingDanglingAlts,
int expectingNumWarnings,
bool hasPredHiddenByAction )
{
DecisionProbe.verbose = true; // make sure we get all error info
ErrorQueue equeue = new ErrorQueue();
ErrorManager.SetErrorListener( equeue );
CodeGenerator generator = new CodeGenerator( newTool(), g, "Java" );
g.CodeGenerator = generator;
// mimic actions of org.antlr.Tool first time for grammar g
if ( g.NumberOfDecisions == 0 )
{
g.BuildNFA();
g.CreateLookaheadDFAs( false );
}
if ( equeue.size() != expectingNumWarnings )
{
Console.Error.WriteLine( "Warnings issued: " + equeue );
}
Assert.AreEqual(expectingNumWarnings, equeue.size(), "unexpected number of expected problems");
DFA dfa = g.GetLookaheadDFA( decision );
FASerializer serializer = new FASerializer( g );
string result = serializer.Serialize( dfa.StartState );
//System.out.print(result);
var unreachableAlts = dfa.UnreachableAlts;
// make sure unreachable alts are as expected
if ( expectingUnreachableAlts != null )
{
BitSet s = new BitSet();
s.AddAll( expectingUnreachableAlts );
BitSet s2 = new BitSet();
s2.AddAll( unreachableAlts );
Assert.AreEqual(s, s2, "unreachable alts mismatch");
}
else
{
Assert.AreEqual(0, unreachableAlts != null ? unreachableAlts.Count : 0, "unreachable alts mismatch");
}
// check conflicting input
if ( expectingAmbigInput != null )
{
// first, find nondet message
Message msg = getNonDeterminismMessage( equeue.warnings );
Assert.IsNotNull(msg, "no nondeterminism warning?");
Assert.IsTrue(msg is GrammarNonDeterminismMessage, "expecting nondeterminism; found " + msg.GetType().Name);
GrammarNonDeterminismMessage nondetMsg =
getNonDeterminismMessage( equeue.warnings );
var labels =
nondetMsg.probe.GetSampleNonDeterministicInputSequence( nondetMsg.problemState );
string input = nondetMsg.probe.GetInputSequenceDisplay( labels );
Assert.AreEqual( expectingAmbigInput, input );
}
// check nondet alts
if ( expectingNonDetAlts != null )
{
GrammarNonDeterminismMessage nondetMsg =
getNonDeterminismMessage( equeue.warnings );
Assert.IsNotNull(nondetMsg, "found no nondet alts; expecting: " + str(expectingNonDetAlts));
var nonDetAlts =
nondetMsg.probe.GetNonDeterministicAltsForState( nondetMsg.problemState );
// compare nonDetAlts with expectingNonDetAlts
BitSet s = new BitSet();
s.AddAll( expectingNonDetAlts );
BitSet s2 = new BitSet();
s2.AddAll( nonDetAlts );
Assert.AreEqual(s, s2, "nondet alts mismatch");
Assert.AreEqual(hasPredHiddenByAction, nondetMsg.problemState.Dfa.HasPredicateBlockedByAction, "mismatch between expected hasPredHiddenByAction");
}
else
{
// not expecting any nondet alts, make sure there are none
GrammarNonDeterminismMessage nondetMsg =
getNonDeterminismMessage( equeue.warnings );
Assert.IsNull(nondetMsg, "found nondet alts, but expecting none");
}
if ( expectingInsufficientPredAlts != null )
{
GrammarInsufficientPredicatesMessage insuffPredMsg =
getGrammarInsufficientPredicatesMessage( equeue.warnings );
Assert.IsNotNull(insuffPredMsg, "found no GrammarInsufficientPredicatesMessage alts; expecting: " + str(expectingNonDetAlts));
var locations = insuffPredMsg.altToLocations;
var actualAlts = locations.Keys;
BitSet s = new BitSet();
s.AddAll( expectingInsufficientPredAlts );
BitSet s2 = new BitSet();
s2.AddAll( actualAlts );
Assert.AreEqual(s, s2, "mismatch between insufficiently covered alts");
Assert.AreEqual(hasPredHiddenByAction, insuffPredMsg.problemState.Dfa.HasPredicateBlockedByAction, "mismatch between expected hasPredHiddenByAction");
}
else
{
// not expecting any nondet alts, make sure there are none
GrammarInsufficientPredicatesMessage nondetMsg =
getGrammarInsufficientPredicatesMessage( equeue.warnings );
if ( nondetMsg != null )
{
Console.Out.WriteLine( equeue.warnings );
}
Assert.IsNull(nondetMsg, "found insufficiently covered alts, but expecting none");
}
Assert.AreEqual( expecting, result );
}
public void TestLeftRecursivePred()
{
// No analysis possible. but probably good to fail. Not sure we really want
// left-recursion even if guarded with pred.
Grammar g = new Grammar(
"parser grammar P;\n" +
"s : a ;\n" +
"a : {p1}? a | ID ;\n" );
DecisionProbe.verbose = true; // make sure we get all error info
ErrorQueue equeue = new ErrorQueue();
ErrorManager.SetErrorListener( equeue );
CodeGenerator generator = new CodeGenerator( newTool(), g, "Java" );
g.CodeGenerator = generator;
if ( g.NumberOfDecisions == 0 )
{
g.BuildNFA();
g.CreateLookaheadDFAs( false );
}
DFA dfa = g.GetLookaheadDFA( 1 );
Assert.AreEqual( null, dfa ); // can't analyze.
/*
String expecting =
".s0-ID->.s1" + NewLine +
".s1-{p1}?->:s2=>1" + NewLine +
".s1-{true}?->:s3=>2" + NewLine;
String result = serializer.serialize(dfa.startState);
Assert.AreEqual(expecting, result);
*/
Assert.AreEqual(1, equeue.size(), "unexpected number of expected problems");
Message msg = equeue.errors[0];
Assert.IsTrue(msg is LeftRecursionCyclesMessage, "warning must be a left recursion msg");
}
//throws Exception
protected void checkDecision( Grammar g,
int decision,
string expecting,
int[] expectingUnreachableAlts,
int[] expectingNonDetAlts,
string expectingAmbigInput,
int[] expectingDanglingAlts,
int expectingNumWarnings )
{
DecisionProbe.verbose = true; // make sure we get all error info
ErrorQueue equeue = new ErrorQueue();
ErrorManager.SetErrorListener( equeue );
// mimic actions of org.antlr.Tool first time for grammar g
if ( g.NumberOfDecisions == 0 )
{
g.BuildNFA();
g.CreateLookaheadDFAs( false );
}
CodeGenerator generator = new CodeGenerator( newTool(), g, "Java" );
g.CodeGenerator = generator;
if ( equeue.size() != expectingNumWarnings )
{
Console.Error.WriteLine( "Warnings issued: " + equeue );
}
Assert.AreEqual(expectingNumWarnings, equeue.size(), "unexpected number of expected problems");
DFA dfa = g.GetLookaheadDFA( decision );
Assert.IsNotNull( dfa, "no DFA for decision " + decision );
FASerializer serializer = new FASerializer( g );
string result = serializer.Serialize( dfa.StartState );
var unreachableAlts = dfa.UnreachableAlts;
// make sure unreachable alts are as expected
if ( expectingUnreachableAlts != null )
{
BitSet s = new BitSet();
s.AddAll( expectingUnreachableAlts );
BitSet s2 = new BitSet();
s2.AddAll( unreachableAlts );
Assert.AreEqual(s, s2, "unreachable alts mismatch");
}
else
{
Assert.AreEqual(0, unreachableAlts != null ? unreachableAlts.Count : 0, "number of unreachable alts");
}
// check conflicting input
if ( expectingAmbigInput != null )
{
// first, find nondet message
Message msg = (Message)equeue.warnings[0];
Assert.IsTrue(msg is GrammarNonDeterminismMessage, "expecting nondeterminism; found " + msg.GetType().Name);
GrammarNonDeterminismMessage nondetMsg =
getNonDeterminismMessage( equeue.warnings );
var labels =
nondetMsg.probe.GetSampleNonDeterministicInputSequence( nondetMsg.problemState );
string input = nondetMsg.probe.GetInputSequenceDisplay( labels );
Assert.AreEqual( expectingAmbigInput, input );
}
// check nondet alts
if ( expectingNonDetAlts != null )
{
RecursionOverflowMessage recMsg = null;
GrammarNonDeterminismMessage nondetMsg =
getNonDeterminismMessage( equeue.warnings );
IList<int> nonDetAlts = null;
if ( nondetMsg != null )
{
nonDetAlts =
nondetMsg.probe.GetNonDeterministicAltsForState( nondetMsg.problemState );
}
else
{
recMsg = getRecursionOverflowMessage( equeue.warnings );
if ( recMsg != null )
{
//nonDetAlts = new ArrayList(recMsg.alts);
}
}
// compare nonDetAlts with expectingNonDetAlts
BitSet s = new BitSet();
s.AddAll( expectingNonDetAlts );
BitSet s2 = new BitSet();
s2.AddAll( nonDetAlts );
Assert.AreEqual(s, s2, "nondet alts mismatch");
Assert.IsTrue(nondetMsg != null || recMsg != null, "found no nondet alts; expecting: " + str(expectingNonDetAlts));
}
else
{
// not expecting any nondet alts, make sure there are none
GrammarNonDeterminismMessage nondetMsg =
getNonDeterminismMessage( equeue.warnings );
Assert.IsNull(nondetMsg, "found nondet alts, but expecting none");
}
Assert.AreEqual( expecting, result );
}
protected void assertRecursionOverflow( Grammar g,
IList expectedTargetRules,
int expectedAlt )
{
DecisionProbe.verbose = true; // make sure we get all error info
ErrorQueue equeue = new ErrorQueue();
ErrorManager.SetErrorListener( equeue );
// mimic actions of org.antlr.Tool first time for grammar g
if ( g.NumberOfDecisions == 0 )
{
g.BuildNFA();
g.CreateLookaheadDFAs( false );
}
RecursionOverflowMessage msg = getRecursionOverflowMessage( equeue.errors );
Assert.IsNotNull(msg, "missing expected recursion overflow msg" + msg);
Assert.AreEqual(expectedTargetRules.ToElementString(), msg.targetRules.ToList().ToElementString(), "target rules mismatch");
Assert.AreEqual(expectedAlt, msg.alt, "mismatched alt");
}
protected void assertNonLLStar( Grammar g, IList<int> expectedBadAlts )
{
DecisionProbe.verbose = true; // make sure we get all error info
ErrorQueue equeue = new ErrorQueue();
ErrorManager.SetErrorListener( equeue );
// mimic actions of org.antlr.Tool first time for grammar g
if ( g.NumberOfDecisions == 0 )
{
g.BuildNFA();
g.CreateLookaheadDFAs( false );
}
NonRegularDecisionMessage msg = getNonRegularDecisionMessage( equeue.errors );
Assert.IsNotNull(msg, "expected fatal non-LL(*) msg");
List<int> alts = new List<int>();
alts.AddRange( msg.altsWithRecursion );
alts.Sort();
//Collections.sort( alts );
//Assert.AreEqual( expectedBadAlts, alts );
Assert.IsTrue( expectedBadAlts.SequenceEqual( alts ) );
}
public void TestLexerDelegatorRuleOverridesDelegateLeavingNoRules()
{
// M.Tokens has nothing to predict tokens from S. Should
// not include S.Tokens alt in this case?
string slave =
"lexer grammar S;\n" +
"A : 'a' {System.out.println(\"S.A\");} ;\n";
mkdir( tmpdir );
writeFile( tmpdir, "S.g", slave );
string master =
"lexer grammar M;\n" +
"import S;\n" +
"A : 'a' {System.out.println(\"M.A\");} ;\n" +
"WS : (' '|'\\n') {skip();} ;\n";
writeFile( tmpdir, "M.g", master );
ErrorQueue equeue = new ErrorQueue();
ErrorManager.SetErrorListener( equeue );
AntlrTool antlr = newTool( new string[] { "-lib", tmpdir } );
CompositeGrammar composite = new CompositeGrammar();
Grammar g = new Grammar( antlr, tmpdir + "/M.g", composite );
composite.SetDelegationRoot( g );
g.ParseAndBuildAST();
composite.AssignTokenTypes();
composite.DefineGrammarSymbols();
composite.CreateNFAs();
g.CreateLookaheadDFAs( false );
// predict only alts from M not S
string expectingDFA =
".s0-'a'->.s1\n" +
".s0-{'\\n', ' '}->:s3=>2\n" +
".s1-<EOT>->:s2=>1\n";
Antlr3.Analysis.DFA dfa = g.GetLookaheadDFA( 1 );
FASerializer serializer = new FASerializer( g );
string result = serializer.Serialize( dfa.startState );
assertEquals( expectingDFA, result );
// must not be a "unreachable alt: Tokens" error
assertEquals( "unexpected errors: " + equeue, 0, equeue.errors.Count );
}
