public SRegex(string pattern, SRegexOptions options = SRegexOptions.Default)
{
var ast = Language.Parse(new SreSyntax(), pattern).Result.Node;
switch (options)
{
case SRegexOptions.ByteCodeCompilation:
var compiler = new NfaVMBytecodeBackend(ast);
matcher = (int[] input) =>
{
var vm = new PikeNfaVM(compiler.Code.ToArray());
vm.Feed(input.Select(ch => (int)ch)).Done();
return vm.HasMatch;
};
break;
case SRegexOptions.ILCompilation:
string methodName = "MatchSrePattern" + PatternID++;
var builder = new CachedMethod<MatchDelegate>(
methodName,
(emit, args) => EmitAst(emit, ast, args[0]));
matcher = builder.Delegate;
break;
case SRegexOptions.NfaCompilation:
var nfa = new Nfa(ast);
matcher = nfa.Match;
break;
case SRegexOptions.DfaCompilation:
var dfa = new RegularToDfaAlgorithm(new RegularTree(ast));
var simulation = new DfaSimulation(dfa.Data);
matcher = input => simulation.Match(input);
break;
}
}
// 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 CreateDfa()
{
var dfa = new RegularToDfaAlgorithm(regularTree);
this.simulation = new DfaSimulation(dfa.Data);
this.serialization = new DfaSerialization(dfa.Data);
// DEBUG: Console.WriteLine(dfa);
}
// 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;
}
}
