public void addRegExp(int regExpNum)
{
if (Options.DEBUG)
{
Out.debug("Adding nfa for regexp " + regExpNum + " :" + Out.NL + regExps.getRegExp(regExpNum));
}
IntPair nfa = insertNFA(regExps.getRegExp(regExpNum));
IEnumerator lexStates = regExps.getStates(regExpNum).GetEnumerator();
if (!lexStates.MoveNext())
{
lexStates = scanner.states.getInclusiveStates();
}
lexStates.Reset();
while (lexStates.MoveNext())
{
int stateNum = (int)lexStates.Current;
if (!regExps.isBOL(regExpNum))
{
addEpsilonTransition(2 * stateNum, nfa.start);
}
addEpsilonTransition(2 * stateNum + 1, nfa.start);
}
if (regExps.getLookAhead(regExpNum) != null)
{
IntPair look = insertNFA(regExps.getLookAhead(regExpNum));
addEpsilonTransition(nfa.end, look.start);
Action a = regExps.getAction(regExpNum);
a.setLookAction(true);
isPushback[nfa.end] = true;
action[look.end] = a;
isFinal[look.end] = true;
}
else
{
action[nfa.end] = regExps.getAction(regExpNum);
isFinal[nfa.end] = true;
}
}
/**
* Constructs an NFA accepting the complement of the language
* of a given NFA.
*
* Converts the NFA into a DFA, then negates that DFA.
* Exponential state blowup possible and common.
*
* @param the NFA to construct the complement for.
*
* @return a pair of integers denoting the index of start
* and end state of the complement NFA.
*/
private IntPair complement(IntPair nfa)
{
if (Options.DEBUG)
{
Out.debug("complement for " + nfa);
Out.debug("NFA is :" + Out.NL + this);
}
int dfaStart = nfa.end + 1;
// fixme: only need epsilon closure of states reachable from nfa.start
epsilonFill();
Hashtable dfaStates = new PrettyHashtable(numStates);
ArrayList dfaVector = new PrettyArrayList(numStates);
int numDFAStates = 0;
int currentDFAState = 0;
StateSet currentState, newState;
newState = epsilon[nfa.start];
dfaStates[newState] = new Integer(numDFAStates);
dfaVector.Add(newState);
if (Options.DEBUG)
{
Out.debug("pos DFA start state is :" + Out.NL + dfaStates + Out.NL + Out.NL + "ordered :" + Out.NL + dfaVector);
}
currentDFAState = 0;
while (currentDFAState <= numDFAStates)
{
currentState = (StateSet)dfaVector[currentDFAState];
for (char input = (char)0; input < numInput; input++)
{
newState = DFAEdge(currentState, input);
if (newState.containsElements())
{
// Out.debug("DFAEdge for input "+(int)input+" and state set "+currentState+" is "+newState);
// Out.debug("Looking for state set "+newState);
Integer nextDFAState = (Integer)dfaStates[newState];
if (nextDFAState != null)
{
// Out.debug("FOUND!");
addTransition(dfaStart + currentDFAState, input, dfaStart + nextDFAState.intValue());
}
else
{
if (Options.dump)
{
Out.print("+");
}
// Out.debug("NOT FOUND!");
// Out.debug("Table was "+dfaStates);
numDFAStates++;
dfaStates[newState] = new Integer(numDFAStates);
dfaVector.Add(newState);
addTransition(dfaStart + currentDFAState, input, dfaStart + numDFAStates);
}
}
}
currentDFAState++;
}
// We have a dfa accepting the positive regexp.
// Now the complement:
if (Options.DEBUG)
{
Out.debug("dfa finished, nfa is now :" + Out.NL + this);
}
int start = dfaStart + numDFAStates + 1;
int error = dfaStart + numDFAStates + 2;
int end = dfaStart + numDFAStates + 3;
addEpsilonTransition(start, dfaStart);
for (int i = 0; i < numInput; i++)
{
addTransition(error, i, error);
}
addEpsilonTransition(error, end);
for (int s = 0; s <= numDFAStates; s++)
{
currentState = (StateSet)dfaVector[s];
currentDFAState = dfaStart + s;
// if it was not a final state, it is now in the complement
if (!currentState.isElement(nfa.end))
{
addEpsilonTransition(currentDFAState, end);
}
// all inputs not present (formerly leading to an implicit error)
// now lead to an explicit (final) state accepting everything.
for (int i = 0; i < numInput; i++)
{
if (table[currentDFAState][i] == null)
{
addTransition(currentDFAState, i, error);
}
}
}
// eliminate transitions leading to dead states
if (live == null || live.Length < numStates)
{
live = new bool [2 * numStates];
visited = new bool [2 * numStates];
}
_end = end;
_dfaStates = dfaVector;
_dfaStart = dfaStart;
removeDead(dfaStart);
if (Options.DEBUG)
{
Out.debug("complement finished, nfa (" + start + "," + end + ") is now :" + this);
}
return(new IntPair(start, end));
}
