/*
* Check and apply implicit and explicit lexical precedence rules. Display
* errors and infos for the human user during the process.
*
* @param automaton
* is the automaton to check. In order to have the explicit
* priorities applied, it is required that the automaton is
* tagged with the acceptation of the LexerExpression.
* @return a new automaton where only the right acceptation tags remains.
*/
public Automaton checkAndApplyLexerPrecedence(
Automaton automaton, Trace trace, Strictness strictness) {
automaton = automaton.minimal();
Map<State, String> words = automaton.collectShortestWords();
Map<Acceptation, Set<State>> accepts = automaton.collectAcceptationStates();
// Associate each acceptation with the ones it share at least a common
// state.
Map<Acceptation, Set<Acceptation>> conflicts = new HashMap<Acceptation, Set<Acceptation>>();
// Associate each acceptation with the ones it supersedes.
Map<Acceptation, Set<Acceptation>> priorities = new HashMap<Acceptation, Set<Acceptation>>();
// Fill the priorities structure with the implicit inclusion rule
for (Acceptation acc1 : automaton.getAcceptations()) {
if (acc1 == Acceptation.ACCEPT) {
continue;
}
// FIXME: empty LexerExpressions are not detected here since
// their acceptation tag is not in the automaton.
// Collect all the conflicts
Set<State> set1 = accepts.get(acc1);
Set<Acceptation> confs = new TreeSet<Acceptation>();
for (State s : set1) {
confs.addAll(s.getAcceptations());
}
conflicts.put(acc1, confs);
// Check for implicit priority for each conflict
for (Acceptation acc2 : confs) {
if (acc2 == Acceptation.ACCEPT) {
continue;
}
if (acc1 == acc2) {
continue;
}
Set<State> set2 = accepts.get(acc2);
if (set2.equals(set1)) {
if (!conflicts.containsKey(acc2)) {
throw SemanticException.genericError(
"The " + acc1.getName() + " and " + acc2.getName() + " tokens are equivalent.");
}
} else if (set2.containsAll(set1)) {
addPriority(priorities, acc1, acc2);
State example = null;
for (State s : set2) {
if (!set1.contains(s)) {
example = s;
break;
}
}
// Note: Since set1 is strictly included in set2, example
// cannot be null
trace.verboseln(
" The "
+ acc1.getName()
+ " token is included in the "
+ acc2.getName()
+ " token. (Example of divergence: '"
+ words.get(example)
+ "'.)");
}
}
}
// Collect new acceptation states and see if a conflict still exists
Map<State, Acceptation> newAccepts = new HashMap<State, Acceptation>();
for (State s : automaton.getStates()) {
if (s.getAcceptations().isEmpty()) {
continue;
}
Acceptation candidate = s.getAcceptations().first();
for (Acceptation challenger : s.getAcceptations()) {
if (candidate == challenger) {
continue;
}
if (hasPriority(priorities, candidate, challenger)) {
// nothing. keep the candidate
} else if (hasPriority(priorities, challenger, candidate)) {
candidate = challenger;
} else {
throw SemanticException.genericError(
"The "
+ candidate.getName()
+ " token and the "
+ challenger.getName()
+ " token conflict on the string '"
+ words.get(s)
+ "'. You should specify a precedence between them.");
}
}
newAccepts.put(s, candidate);
}
// Ask for a new automaton with the correct acceptation states.
return automaton.resetAcceptations(newAccepts);
}
public void compileLexer(Trace trace, Strictness strictness) {
Automaton automaton;
if (this.globalAnonymousContext != null) {
automaton = this.globalAnonymousContext.computeAutomaton().minimal().longest().minimal();
automaton = checkAndApplyLexerPrecedence(automaton, trace, strictness).minimal();
this.lexer.setAutomaton(automaton.withMarkers().minimal());
} else {
throw new InternalException("not implemented");
}
for (Context context : this.namedContexts) {
context.computeAutomaton();
}
// Look for useless LexerExpression
for (LexerExpression lexerExpression : this.lexer.getExpressions()) {
// If their is no automaton saved it means that the LexerExpression
// was not used to build the big automaton.
if (lexerExpression.getSavedAutomaton() == null) {
if (strictness == Strictness.STRICT) {
throw SemanticException.genericError(
"The " + lexerExpression.getExpressionName() + " expression is useless.");
} else {
trace.verboseln(
" The " + lexerExpression.getExpressionName() + " expression is useless.");
}
}
}
for (LexerExpression lexerExpression : this.lexer.getExpressions()) {
// Note: getting the automaton forces the validation of the semantic
// validity of (eg. cirularity)
Automaton lexerAutomaton = lexerExpression.getAutomaton();
}
for (LexerExpression lexerExpression :
this.globalAnonymousContext.getLexerExpressionTokensAndIgnored()) {
// Note: The big automaton has to be minimal (thus with the unused
// acceptations removed)
if (!automaton.getAcceptations().contains(lexerExpression.getAcceptation())) {
if (strictness == Strictness.STRICT) {
throw SemanticException.genericError(
"The " + lexerExpression.getExpressionName() + " token does not match anything.");
} else {
trace.verboseln(
" The " + lexerExpression.getExpressionName() + " token does not match anything.");
}
}
Automaton expressionAutomaton = lexerExpression.getAutomaton();
for (RichSymbol richSymbol : expressionAutomaton.getStartState().getTransitions().keySet()) {
if (richSymbol.isLookahead()) {
// We have a lookahead transition from the start state.
// Note: this works since the result of getAutomaton() is
// minimized.
throw SemanticException.genericError(
"The " + lexerExpression.getExpressionName() + " token matches the empty string.");
}
}
}
}