/** Checks if a PTA constructed is consistent with provided sequences. */
private void checkPTAConsistency(
PTASequenceEngine engine, Set<List<Label>> sequences, boolean accept) {
SequenceSet initSeq = engine.new SequenceSet();
initSeq.setIdentity();
int leafNumber = engine.getDebugDataMapDepth(null).size();
// Now we check the consistency
for (List<Label> seq : sequences) {
SequenceSet endOfSeq =
initSeq.crossWithSequence(seq); // this is aimed to follow a path in a PTA
// to a state which is not necessarily a tail-state (hence no use calling
// getDebugDataMapDepth(null),
// but can inadvertently add a new sequence, hence we have to check that it has to happened at
// the end.
Map<String, String> map = engine.getDebugDataMapDepth(endOfSeq);
assert map.size() == 1 : "expected size of 1, got " + map.size();
String attrs = map.values().iterator().next();
// For reject-sequences,
// If the end of the sequence is not a leaf, it should be considered an accept-sequence, hence
// throw.
// If the end of the sequence is a leaf but should be returned, this is also an
// accept-sequence (see the sequence engine construction above) throw.
// The only remaining case is that the end is a leaf and should not be returned - this is ok.
if (!accept && !attrs.equals(DebugDataValues.booleanToString(true, false)))
throw new IllegalArgumentException(
"reject-sequence " + seq + " is present in PTA with a positive ending");
// For accept-sequences, the only erroneous case is when a leaf is not returned.
// (we assume that non-leaf is always returned.)
if (accept && attrs.equals(DebugDataValues.booleanToString(true, false)))
throw new IllegalArgumentException(
"reject-sequence " + seq + " is present in PTA with a negative ending");
}
Assert.assertEquals(leafNumber, engine.getDebugDataMapDepth(null).size());
}
public static PTASequenceEngine buildPTA(
Set<List<Label>> plusStrings, Set<List<Label>> minusStrings) {
final Boolean accept = new Boolean(true), reject = new Boolean(false);
boolean theOnlyStateReject = false;
for (List<Label> seq : minusStrings)
if (seq.isEmpty()) {
theOnlyStateReject = true;
break;
}
final Boolean rejectAllStates = new Boolean(theOnlyStateReject);
final AtomicBoolean statesAccept = new AtomicBoolean(true);
PTASequenceEngine allSequences = new PTASequenceEngine();
// Here we are building a PTA which consists of accept states for plusStrings and reject-states
// for minusStrings. This way if a plus string is a prefix of a plusString, the prefix will
// be labelled with Boolean(true) states and the suffix - with Boolean(false) states.
// Note that the suffix may contain many states.
allSequences.init(
new PTASequenceSetAutomaton() {
@Override
public Object getTheOnlyState() {
return statesAccept.get() ? accept : reject;
}
@Override
public boolean shouldBeReturned(Object elem) {
return elem != null && ((Boolean) elem).booleanValue();
}
@Override
public boolean isAccept(@SuppressWarnings("unused") Object elem) {
return !rejectAllStates.booleanValue();
}
});
SequenceSet initSeq = allSequences.new SequenceSet();
initSeq.setIdentity();
initSeq.cross(plusStrings);
statesAccept.getAndSet(false);
initSeq.cross(minusStrings);
return allSequences;
}
