public static ISet<CharacterClassElement> NormaliseAST(RegexNode rootNode)
{
//walk the tree and find all char classes
var matchClassLink = new Dictionary<CharacterClass, CharacterClassMatchNode>();
Action<RegexNode> classFinder = null;
classFinder = (rxNode) =>
{
var curAltNode = rxNode.FirstAlternative;
while (curAltNode != null)
{
var curMatchNode = curAltNode.FirstFactor;
while (curMatchNode != null)
{
if (curMatchNode is GroupMatchNode)
{
var g = (GroupMatchNode) curMatchNode;
classFinder(g.Body);
}
else if (curMatchNode is CharacterClassMatchNode)
{
var cn = (CharacterClassMatchNode) curMatchNode;
matchClassLink[cn.MatchingCharacterClass] = cn;
}
curMatchNode = curMatchNode.Next;
}
curAltNode = curAltNode.Next;
}
};
classFinder(rootNode);
//that's filled, now map them
//but hide a [min-max] class to make sre the alphabet is total
var allClass = new CharacterClass();
allClass.Elements.Add(new CharacterClassElement(char.MinValue, char.MaxValue));
//matchClassLink[allClass] = null;
var rStruct = NormaliseCharacterClasses(matchClassLink.Keys.ToList());
foreach (var cClass in matchClassLink.Keys)
{
//if (cClass != allClass) //skip the one we added
matchClassLink[cClass].MatchingCharacterClass = rStruct.Mapping[cClass];
}
return rStruct.Alphabet;
}
public GroupMatchNode(RegexNode body = null, bool capturing = true)
{
Body = body;
Capturing = capturing;
}
public NFAGraph(RegexNode rxNode, StateFactory fac = null)
{
Alphabet = CharacterClassMapper.NormaliseAST(rxNode);
StateFac = fac ?? new StateFactory();
AdjList = new Dictionary<NFAState, IDictionary<CharacterClassElement, ISet<NFAState>>>();
//begin from the start
StartState = MakeState();
StartState.Tags.Add(FAStateTags.Start);
FinalState = MakeState();
FinalState.Tags.Add(FAStateTags.Final);
//every alternative is connected to this
for (var altNode = rxNode.FirstAlternative; altNode != null; altNode = altNode.Next)
{
//node for this alt
var altPosNode = MakeState();
LinkStates(StartState, altPosNode, Empty);
NFAState oldPosNode = null;
//Iter the and matches
for (var matchNode = altNode.FirstFactor; matchNode != null; matchNode = matchNode.Next)
{
oldPosNode = altPosNode;
if (matchNode is CharacterClassMatchNode)
{
var ccn = (CharacterClassMatchNode) matchNode;
var thisState = MakeState();
LinkStates(altPosNode, thisState, ccn.MatchingCharacterClass);
altPosNode = thisState;
}
else if (matchNode is GroupMatchNode)
{
var gn = (GroupMatchNode) matchNode;
//Create a NFA for the group and join it in
var groupGraph = new NFAGraph(gn.Body, StateFac);
//Merge all their states into ours
foreach (var oState in groupGraph.AdjList.Keys)
AdjList[oState] = groupGraph.AdjList[oState];
LinkStates(altPosNode, groupGraph.StartState, Empty);
//tag it
groupGraph.StartState.Tags.Remove(FAStateTags.Start);
groupGraph.StartState.Tags.Add(FAStateTags.PushSubmatch);
groupGraph.FinalState.Tags.Remove(FAStateTags.Final); //destroys final tag as intended
groupGraph.FinalState.Tags.Add(FAStateTags.PopSubmatch);
altPosNode = groupGraph.FinalState;
}
//operators?
switch (matchNode.OpType)
{
case UnaryOperatorType.NoneMany: //kleene star
//link the state we just made back via e
LinkStates(altPosNode, oldPosNode, Empty);
var newTermination = MakeState();
LinkStates(oldPosNode, newTermination, Empty);
oldPosNode = altPosNode;
altPosNode = newTermination;
break;
case UnaryOperatorType.Optional:
LinkStates(oldPosNode, altPosNode, Empty); //skip path we just made
break;
case UnaryOperatorType.OneMany:
LinkStates(altPosNode, oldPosNode, Empty); //make a cycle
break;
}
}
//join it into the end
LinkStates(altPosNode, FinalState, Empty);
}
}
