// 1. build full RegularTree (don't build first, following for positions)
// 2. build regular alphabet (equivalense classes)
// 3. build first, following for non-literal part of regular tree
// 4. build non-literal TdfaData
public RegularToTdfaAlgorithm(RegularTree noConstRegTree, Dictionary <string, Action> literalToAction)
{
var equivClasses = noConstRegTree
.GetEquivalenceCsets()
.Union(new [] { NewLines });
var alphabet = new EquivalenceClassesAlphabet(equivClasses);
foreach (var literal in literalToAction.Keys)
{
foreach (char ch in literal)
{
alphabet.AddInputSet(SparseIntSetType.Instance.Of(ch));
}
}
// Step 1. Convert the NFA for non-constant REs to a DFA using the usual
// algorithms for subset construction and state minimization [ASU86, WaG84].
var initialDfa = new RegularToDfaAlgorithm(noConstRegTree, alphabet);
this.data = initialDfa.Data;
#if false
using (var view = new IronText.Diagnostics.GvGraphView(Guid.NewGuid() + ".gv"))
{
data.DescribeGraph(view);
}
#endif
// Step 2. Extend the DFA to a tunnel automaton by setting Tunnel (s) to NoState
// for every state s.
int initialStateCount = data.StateCount;
foreach (var S in data.EnumerateStates())
{
S.Tunnel = NoState;
}
// Step 3: Compute the set of ambiguous states of the tunnel automaton.
this.ambiguous = FindAmbiguousStates();
// Step 4: For every constant RE execute Step 5 which incrementally extends
// the tunnel automaton. Continue with Step 6.
foreach (var pair in literalToAction)
{
ExtendAutomatonWithLiteral(pair.Key, pair.Value);
}
var newlines = alphabet.Encode(NewLines);
// Add new line handling
foreach (var state in data.EnumerateStates())
{
var i = state.Outgoing.FindIndex(t => t.HasAnySymbol(newlines));
if (i < 0)
{
continue;
}
var newlineTransition = state.Outgoing[i];
var to = data.GetState(newlineTransition.To);
TdfaState newlineState;
if (to.IsNewline)
{
continue;
}
else if (data.EnumerateIncoming(to.Index).All(t => t.HasSingleSymbolFrom(newlines)))
{
newlineState = to;
}
else
{
newlineState = new TdfaState(data);
data.AddState(newlineState);
newlineState.Tunnel = newlineTransition.To;
newlineState.IsAccepting = to.IsAccepting;
newlineState.Actions.AddRange(to.Actions);
data.DeleteTransition(state.Index, newlines);
data.AddTransition(state.Index, newlines, newlineState.Index);
}
newlineState.IsNewline = true;
}
}
private void ExtendAutomatonWithLiteral(string literal, int scanAction)
{
var state = data.Start;
var symbols = EnumerateLiteralSymbols(data.Alphabet, literal).GetEnumerator();
bool hasSymbol;
// trace and do nothing
while ((hasSymbol = symbols.MoveNext()))
{
State next = Control(state, symbols.Current);
if (next == NoState || ambiguous.Contains(next))
{
break;
}
state = next;
}
int previous = state;
// trace and duplicate the path
while (hasSymbol)
{
State next = Control(state, symbols.Current);
if (next != NoState)
{
state = next;
var newStateInfo = new TdfaState(data);
int newState = data.AddState(newStateInfo);
data.DeleteTransition(from: previous, symbol: symbols.Current);
data.AddTransition(
from: previous,
symbol: symbols.Current,
to: newState
);
var S = data.GetState(state);
newStateInfo.IsAccepting = S.IsAccepting;
newStateInfo.Actions.AddRange(S.Actions);
newStateInfo.Tunnel = state;
previous = newState;
hasSymbol = symbols.MoveNext();
}
else
{
var S = data.GetState(state);
if (S.Tunnel == NoState)
{
break;
}
state = S.Tunnel;
}
}
// extend the path
for (; hasSymbol; hasSymbol = symbols.MoveNext())
{
var newStateInfo = new TdfaState(data);
int newState = data.AddState(newStateInfo);
data.AddTransition(
from: previous,
symbol: symbols.Current,
to: newState
);
newStateInfo.Tunnel = NoState;
previous = newState;
}
// process new final state
var finalState = data.GetState(previous);
finalState.IsAccepting = true;
finalState.Actions.Insert(0, scanAction);
}
public RegularToDfaAlgorithm(RegularTree regTree, IRegularAlphabet alphabet)
{
this.alphabet = alphabet;
data = new TdfaData(alphabet);
data.AddState(new TdfaState(data) { Positions = regTree.FirstPos });
foreach (var st in data.EnumerateStates())
{
int Sindex = st.Index;
var S = st.Positions;
if (S.Contains(regTree.EoiPosition))
{
data.GetState(Sindex).IsAccepting = true;
}
var actionPosList = new SortedList<int, int>();
foreach (var position in S)
{
var action = regTree.GetPosAction(position);
if (action.HasValue)
{
actionPosList[position] = action.Value;
}
}
if (actionPosList.Count != 0)
{
st.Actions.AddRange(actionPosList.Values);
}
var transitionSymbols = alphabet.SymbolSetType.Union(
st.Positions
.Select(regTree.GetPosSymbols)
.Select(alphabet.Encode));
foreach (var symbol in transitionSymbols)
{
if (symbol == alphabet.EoiSymbol)
{
continue;
}
var U = TdfaData.PositionSetType.Mutable();
foreach (var position in S)
{
var cset = alphabet.Encode(regTree.GetPosSymbols(position));
if (cset.Contains(symbol))
{
U.AddAll(regTree.GetFollowPos(position));
}
}
if (!U.IsEmpty)
{
int Uindex = data.IndexOfState(U);
if (Uindex < 0)
{
Uindex = data.AddState(new TdfaState(data) { Positions = U });
}
data.AddTransition(from: Sindex, symbol: symbol, to: Uindex);
}
}
}
}
public RegularToDfaAlgorithm(RegularTree regTree, IRegularAlphabet alphabet)
{
this.alphabet = alphabet;
data = new TdfaData(alphabet);
data.AddState(new TdfaState(data)
{
Positions = regTree.FirstPos
});
foreach (var st in data.EnumerateStates())
{
int Sindex = st.Index;
var S = st.Positions;
if (S.Contains(regTree.EoiPosition))
{
data.GetState(Sindex).IsAccepting = true;
}
var actionPosList = new SortedList <int, int>();
foreach (var position in S)
{
var action = regTree.GetPosAction(position);
if (action.HasValue)
{
actionPosList[position] = action.Value;
}
}
if (actionPosList.Count != 0)
{
st.Actions.AddRange(actionPosList.Values);
}
var transitionSymbols = alphabet.SymbolSetType.Union(
st.Positions
.Select(regTree.GetPosSymbols)
.Select(alphabet.Encode));
foreach (var symbol in transitionSymbols)
{
if (symbol == alphabet.EoiSymbol)
{
continue;
}
var U = TdfaData.PositionSetType.Mutable();
foreach (var position in S)
{
var cset = alphabet.Encode(regTree.GetPosSymbols(position));
if (cset.Contains(symbol))
{
U.AddAll(regTree.GetFollowPos(position));
}
}
if (!U.IsEmpty)
{
int Uindex = data.IndexOfState(U);
if (Uindex < 0)
{
Uindex = data.AddState(new TdfaState(data)
{
Positions = U
});
}
data.AddTransition(from: Sindex, symbol: symbol, to: Uindex);
}
}
}
}
// 1. build full RegularTree (don't build first, following for positions)
// 2. build regular alphabet (equivalense classes)
// 3. build first, following for non-literal part of regular tree
// 4. build non-literal TdfaData
public RegularToTdfaAlgorithm(RegularTree noConstRegTree, Dictionary<string,Action> literalToAction)
{
var equivClasses = noConstRegTree
.GetEquivalenceCsets()
.Union(new [] { NewLines });
var alphabet = new EquivalenceClassesAlphabet(equivClasses);
foreach (var literal in literalToAction.Keys)
{
foreach (char ch in literal)
{
alphabet.AddInputSet(SparseIntSetType.Instance.Of(ch));
}
}
// Step 1. Convert the NFA for non-constant REs to a DFA using the usual
// algorithms for subset construction and state minimization [ASU86, WaG84].
var initialDfa = new RegularToDfaAlgorithm(noConstRegTree, alphabet);
this.data = initialDfa.Data;
#if false
using (var view = new IronText.Diagnostics.GvGraphView(Guid.NewGuid() + ".gv"))
{
data.DescribeGraph(view);
}
#endif
// Step 2. Extend the DFA to a tunnel automaton by setting Tunnel (s) to NoState
// for every state s.
int initialStateCount = data.StateCount;
foreach (var S in data.EnumerateStates())
{
S.Tunnel = NoState;
}
// Step 3: Compute the set of ambiguous states of the tunnel automaton.
this.ambiguous = FindAmbiguousStates();
// Step 4: For every constant RE execute Step 5 which incrementally extends
// the tunnel automaton. Continue with Step 6.
foreach (var pair in literalToAction)
{
ExtendAutomatonWithLiteral(pair.Key, pair.Value);
}
var newlines = alphabet.Encode(NewLines);
// Add new line handling
foreach (var state in data.EnumerateStates())
{
var i = state.Outgoing.FindIndex(t => t.HasAnySymbol(newlines));
if (i < 0)
{
continue;
}
var newlineTransition = state.Outgoing[i];
var to = data.GetState(newlineTransition.To);
TdfaState newlineState;
if (to.IsNewline)
{
continue;
}
else if (data.EnumerateIncoming(to.Index).All(t => t.HasSingleSymbolFrom(newlines)))
{
newlineState = to;
}
else
{
newlineState = new TdfaState(data);
data.AddState(newlineState);
newlineState.Tunnel = newlineTransition.To;
newlineState.IsAccepting = to.IsAccepting;
newlineState.Actions.AddRange(to.Actions);
data.DeleteTransition(state.Index, newlines);
data.AddTransition(state.Index, newlines, newlineState.Index);
}
newlineState.IsNewline = true;
}
}