/**
* Returns the action with highest priority in the specified
* set of states.
*
* @param set the set of states for which to determine the action
*/
private Action getAction(StateSet set)
{
states.reset(set);
Action maxAction = null;
Out.debug("Determining action of : " + set);
while (states.hasMoreElements())
{
Action currentAction = action[states.nextElement()];
if (currentAction != null)
{
if (maxAction == null)
{
maxAction = currentAction;
}
else
{
maxAction = maxAction.getHigherPriority(currentAction);
}
}
}
return(maxAction);
}
private void increaseSize(int length)
{
length = Math.Max(length + 1, 4 * p.Length);
Out.debug("increasing length to " + length); //$NON-NLS-1$
int[] pn = new int[length];
int[] qn = new int[length];
Array.Copy(p, 0, pn, 0, p.Length);
Array.Copy(q, 0, qn, 0, q.Length);
p = pn;
q = qn;
}
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;
}
}
/**
* Stores a new macro and its definition.
*
* @param name the name of the new macro
* @param definition the definition of the new macro
*
* @return <code>true</code>, iff the macro name has not been
* stored before.
*/
public bool insert(String name, RegExp definition)
{
#if DEBUG_TRACE
log.WriteLine("insert(String name = \"{0}\", RegExp definition = {1})", name, definition);
#endif // DEBUG_TRACE
if (Options.DEBUG)
{
Out.debug("inserting macro " + name + " with definition :" + Out.NL + definition); //$NON-NLS-1$ //$NON-NLS-2$
}
used[name] = false;
bool new_entry = !macros.ContainsKey(name);
macros[name] = definition;
return(new_entry);
}
public int insert(ArrayList stateList, Action action)
{
if (Options.DEBUG)
{
Out.debug("Inserting eofrule with statelist :" + Out.NL + stateList); //$NON-NLS-1$
Out.debug("and action code :" + Out.NL + action.content + Out.NL); //$NON-NLS-1$
}
states.Add(stateList);
regExps.Add(null);
actions.Add(action);
BOL.Add(null);
look.Add(null);
lines.Add(null);
return(states.Count - 1);
}
public int insert(int line, ArrayList stateList, RegExp regExp, Action action,
Boolean isBOL, RegExp lookAhead)
{
if (Options.DEBUG)
{
Out.debug("Inserting regular expression with statelist :" + Out.NL + stateList); //$NON-NLS-1$
Out.debug("and action code :" + Out.NL + action.content + Out.NL); //$NON-NLS-1$
Out.debug("expression :" + Out.NL + regExp); //$NON-NLS-1$
}
states.Add(stateList);
regExps.Add(regExp);
actions.Add(action);
BOL.Add(isBOL);
look.Add(lookAhead);
lines.Add(line);
return(states.Count - 1);
}
public void addTransition(int start, int input, int dest)
{
Out.debug("Adding transition (" + start + ", " + input + ", " + dest + ")");
int maxS = Math.Max(start, dest) + 1;
ensureCapacity(maxS);
if (maxS > numStates)
{
numStates = maxS;
}
if (table[start][input] != null)
{
table[start][input].addState(dest);
}
else
{
table[start][input] = new StateSet(estSize, dest);
}
}
public bool [] [] old_minimize()
{
int i, j;
char c;
Out.print(numStates + " states before minimization, ");
if (numStates == 0)
{
Out.error(ErrorMessages.ZERO_STATES);
throw new GeneratorException();
}
if (Options.no_minimize)
{
Out.println("minimization skipped.");
return(null);
}
// notequiv[i][j] == true <=> state i and state j are equivalent
bool [] [] equiv = new bool [numStates] [];
// list[i][j] contains all pairs of states that have to be marked "not equivalent"
// if states i and j are recognized to be not equivalent
StatePairList [] [] list = new StatePairList [numStates] [];
// construct a triangular matrix equiv[i][j] with j < i
// and mark pairs (final state, not final state) as not equivalent
for (i = 1; i < numStates; i++)
{
list[i] = new StatePairList[i];
equiv[i] = new bool [i];
for (j = 0; j < i; j++)
{
// i and j are equivalent, iff :
// i and j are both final and their actions are equivalent and have same pushback behaviour or
// i and j are both not final
if (isFinal[i] && isFinal[j] && (isPushback[i] == isPushback[j]) && (isLookEnd[i] == isLookEnd[j]))
{
equiv[i][j] = action[i].isEquiv(action[j]) && !Options.emit_csharp; // C# #line directives get messed up by merged states
}
else
{
equiv[i][j] = !isFinal[j] && !isFinal[i] && (isPushback[i] == isPushback[j]) && (isLookEnd[i] == isLookEnd[j]);
}
}
}
for (i = 1; i < numStates; i++)
{
Out.debug("Testing state " + i);
for (j = 0; j < i; j++)
{
if (equiv[i][j])
{
for (c = (char)0; c < numInput; c++)
{
if (equiv[i][j])
{
int p = table[i][c];
int q = table[j][c];
if (p < q)
{
int t = p;
p = q;
q = t;
}
if (p >= 0 || q >= 0)
{
// Out.debug("Testing input '"+c+"' for ("+i+","+j+")");
// Out.debug("Target states are ("+p+","+q+")");
if (p != q && (p == -1 || q == -1 || !equiv[p][q]))
{
equiv[i][j] = false;
if (list[i][j] != null)
{
list[i][j].markAll(list, equiv);
}
}
// printTable(equiv);
} // if (p >= 0) ..
} // if (equiv[i][j]
} // for (char c = 0; c < numInput ..
// if i and j are still marked equivalent..
if (equiv[i][j])
{
// Out.debug("("+i+","+j+") are still marked equivalent");
for (c = (char)0; c < numInput; c++)
{
int p = table[i][c];
int q = table[j][c];
if (p < q)
{
int t = p;
p = q;
q = t;
}
if (p != q && p >= 0 && q >= 0)
{
if (list[p][q] == null)
{
list[p][q] = new StatePairList();
}
list[p][q].addPair(i, j);
}
}
}
else
{
// Out.debug("("+i+","+j+") are not equivalent");
}
} // of first if (equiv[i][j])
} // of for j
} // of for i
// }
// printTable(equiv);
return(equiv);
}
/**
* 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));
}
/**
* Returns an DFA that accepts the same language as this NFA.
* This DFA is usualy not minimal.
*/
public DFA getDFA()
{
Hashtable dfaStates = new PrettyHashtable(numStates);
ArrayList dfaVector = new PrettyArrayList(numStates);
DFA dfa = new DFA(2 * numLexStates, numInput);
int numDFAStates = 0;
int currentDFAState = 0;
Out.println("Converting NFA to DFA : ");
epsilonFill();
StateSet currentState, newState;
for (int i = 0; i < 2 * numLexStates; i++)
{
newState = epsilon[i];
dfaStates[newState] = new Integer(numDFAStates);
dfaVector.Add(newState);
dfa.setLexState(i, numDFAStates);
dfa.setFinal(numDFAStates, containsFinal(newState));
dfa.setPushback(numDFAStates, containsPushback(newState));
dfa.setAction(numDFAStates, getAction(newState));
numDFAStates++;
}
numDFAStates--;
if (Options.DEBUG)
{
Out.debug("DFA start states are :" + Out.NL + dfaStates + Out.NL + Out.NL + "ordered :" + Out.NL + dfaVector);
}
currentDFAState = 0;
StateSet tempStateSet = NFA.tempStateSet;
StateSetEnumerator states = NFA.states;
// will be reused
newState = new StateSet(numStates);
while (currentDFAState <= numDFAStates)
{
currentState = (StateSet)dfaVector[currentDFAState];
for (char input = (char)0; input < numInput; input++)
{
// newState = DFAEdge(currentState, input);
// inlining DFAEdge for performance:
// Out.debug("Calculating DFAEdge for state set "+currentState+" and input '"+input+"'");
tempStateSet.clear();
states.reset(currentState);
while (states.hasMoreElements())
{
tempStateSet.add(table[states.nextElement()][input]);
}
newState.copy(tempStateSet);
states.reset(tempStateSet);
while (states.hasMoreElements())
{
newState.add(epsilon[states.nextElement()]);
}
// Out.debug("DFAEdge is : "+newState);
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!");
dfa.addTransition(currentDFAState, input, nextDFAState.intValue());
}
else
{
if (Options.progress)
{
Out.print(".");
}
// Out.debug("NOT FOUND!");
// Out.debug("Table was "+dfaStates);
numDFAStates++;
// make a new copy of newState to store in dfaStates
StateSet storeState = new StateSet(newState);
dfaStates[storeState] = new Integer(numDFAStates);
dfaVector.Add(storeState);
dfa.addTransition(currentDFAState, input, numDFAStates);
dfa.setFinal(numDFAStates, containsFinal(storeState));
dfa.setPushback(numDFAStates, containsPushback(storeState));
dfa.setAction(numDFAStates, getAction(storeState));
}
}
}
currentDFAState++;
}
if (Options.verbose)
{
Out.println("");
}
return(dfa);
}
/**
* Constructs an NFA for regExp such that the NFA has
*
* exactly one start state,
* exactly one end state,
* no transitions leading out of the end state
* no transitions leading into the start state
*
* @param regExp the regular expression to construct the
* NFA for
*
* @return a pair of integers denoting the index of start
* and end state of the NFA.
*/
public IntPair insertNFA(RegExp regExp)
{
IntPair nfa1, nfa2;
int start, end;
RegExp2 r;
if (Options.DEBUG)
{
Out.debug("Inserting RegExp : " + regExp);
}
if (regExp.isCharClass(macros))
{
start = numStates;
end = numStates + 1;
ensureCapacity(end + 1);
if (end + 1 > numStates)
{
numStates = end + 1;
}
insertNFA(regExp, start, end);
return(new IntPair(start, end));
}
switch (regExp.type)
{
case sym.BAR:
r = (RegExp2)regExp;
nfa1 = insertNFA(r.r1);
nfa2 = insertNFA(r.r2);
start = nfa2.end + 1;
end = nfa2.end + 2;
addEpsilonTransition(start, nfa1.start);
addEpsilonTransition(start, nfa2.start);
addEpsilonTransition(nfa1.end, end);
addEpsilonTransition(nfa2.end, end);
return(new IntPair(start, end));
case sym.CONCAT:
r = (RegExp2)regExp;
nfa1 = insertNFA(r.r1);
nfa2 = insertNFA(r.r2);
addEpsilonTransition(nfa1.end, nfa2.start);
return(new IntPair(nfa1.start, nfa2.end));
case sym.STAR:
nfa1 = insertNFA((RegExp)((RegExp1)regExp).content);
start = nfa1.end + 1;
end = nfa1.end + 2;
addEpsilonTransition(nfa1.end, end);
addEpsilonTransition(start, nfa1.start);
addEpsilonTransition(start, end);
addEpsilonTransition(nfa1.end, nfa1.start);
return(new IntPair(start, end));
case sym.PLUS:
nfa1 = insertNFA((RegExp)((RegExp1)regExp).content);
start = nfa1.end + 1;
end = nfa1.end + 2;
addEpsilonTransition(nfa1.end, end);
addEpsilonTransition(start, nfa1.start);
addEpsilonTransition(nfa1.end, nfa1.start);
return(new IntPair(start, end));
case sym.QUESTION:
nfa1 = insertNFA((RegExp)((RegExp1)regExp).content);
addEpsilonTransition(nfa1.start, nfa1.end);
return(new IntPair(nfa1.start, nfa1.end));
case sym.BANG:
return(complement(insertNFA((RegExp)((RegExp1)regExp).content)));
case sym.TILDE:
nfa1 = insertNFA((RegExp)((RegExp1)regExp).content);
start = nfa1.end + 1;
int s1 = start + 1;
int s2 = s1 + 1;
end = s2 + 1;
for (int i = 0; i < numInput; i++)
{
addTransition(s1, i, s1);
addTransition(s2, i, s2);
}
addEpsilonTransition(start, s1);
addEpsilonTransition(s1, nfa1.start);
addEpsilonTransition(nfa1.end, s2);
addEpsilonTransition(s2, end);
nfa1 = complement(new IntPair(start, end));
nfa2 = insertNFA((RegExp)((RegExp1)regExp).content);
addEpsilonTransition(nfa1.end, nfa2.start);
return(new IntPair(nfa1.start, nfa2.end));
case sym.STRING:
return(insertStringNFA(false, (String)((RegExp1)regExp).content));
case sym.STRING_I:
return(insertStringNFA(true, (String)((RegExp1)regExp).content));
case sym.MACROUSE:
return(insertNFA(macros.getDefinition((String)((RegExp1)regExp).content)));
}
throw new Exception("Unknown expression type " + regExp.type + " in NFA construction");
}