public bool AreEquivalent(Label <SYMBOL> l1, Label <SYMBOL> l2)
{
if (l1.IsIdentity != l2.IsIdentity)
{
Set <SYMBOL> i1 = pa.InclusiveSet(l1.Input);
Set <SYMBOL> i2 = pa.InclusiveSet(l2.Input);
Set <SYMBOL> o = pa.InclusiveSet(l1.IsIdentity ? l2.Output : l1.Output);
return(i1.Count == 1 && i2.Count == 1 && o.Count == 1 && i1 == i2 && i2 == o);
}
if (l1.IsIdentity) // l2.IsIdentity implicit
{
return(pa.AreEquivalent(l1.Input, l2.Input));
}
return(pa.AreEquivalent(l1.Input, l2.Input) && pa.AreEquivalent(l1.Output, l2.Output));
}
/* format spec at http://web.eecs.umich.edu/~radev/NLP-fall2015/resources/fsm_archive/fsm.5.html */
private static void PrintAutomatonFSM(
StreamWriter file, AutomatonType type, string name, Dictionary <int, string> stateNames,
PredicateAlgebra <SYMBOL> algebra, List <int> states, List <SYMBOL> alphabet,
List <Move <ILabel <SYMBOL> > > moves, int initialState, List <int> finalStates,
string stateSymbolsFileName = null, string inputSymbolsFileName = null, string outputSymbolsFileName = null)
{
// map symbols to numeric IDs
var symbolDict = new Dictionary <SYMBOL, int>(alphabet.Count);
int id = 1; // 0 reserved for epsilon
foreach (SYMBOL symbol in alphabet)
{
symbolDict[symbol] = id++;
}
// print state/symbol as number if corresponding symbols file unspecified,
// otherwise print as string and map it to number in symbols file
Func <int, string> state2str = state => (stateSymbolsFileName == null) ? state.ToString() : stateNames[state];
Func <SYMBOL, string> input2str = symbol => (inputSymbolsFileName == null) ? symbolDict[symbol].ToString() : symbol.ToString();
Func <SYMBOL, string> output2str = symbol => (outputSymbolsFileName == null) ? symbolDict[symbol].ToString() : symbol.ToString();
Action <Move <ILabel <SYMBOL> > > printPredicateMove = move => {
int sourceState = move.SourceState;
var predicate = (Predicate <SYMBOL>)move.Label;
int targetState = move.TargetState;
if (predicate == null)
{
file.WriteLine("{0} {1} 0", state2str(sourceState), state2str(targetState));
}
else
{
foreach (SYMBOL symbol in algebra.InclusiveSet(predicate))
{
file.WriteLine("{0} {1} {2}", state2str(sourceState), state2str(targetState), input2str(symbol));
}
}
};
Action <Move <ILabel <SYMBOL> > > printLabelMove = move => {
int sourceState = move.SourceState;
var label = (Label <SYMBOL>)move.Label;
int targetState = move.TargetState;
if (label.IsIdentity)
{
if (label.Input == null)
{
file.WriteLine("{0} {1} 0 0", state2str(sourceState), state2str(targetState));
}
else
{
foreach (SYMBOL symbol in algebra.InclusiveSet(label.Input))
{
file.WriteLine("{0} {1} {2} {3}",
state2str(sourceState), state2str(targetState), input2str(symbol), output2str(symbol));
}
}
}
else
{
if (label.Input == null)
{
if (label.Output == null)
{
file.WriteLine("{0} {1} 0 0", state2str(sourceState), state2str(targetState));
}
else
{
foreach (SYMBOL output in algebra.InclusiveSet(label.Output))
{
file.WriteLine("{0} {1} 0 {2}", state2str(sourceState), state2str(targetState), output2str(output));
}
}
}
else
{
foreach (SYMBOL input in algebra.InclusiveSet(label.Input))
{
if (label.Output == null)
{
file.WriteLine("{0} {1} {2} 0", state2str(sourceState), state2str(targetState), input2str(input));
}
else
{
foreach (SYMBOL output in algebra.InclusiveSet(label.Output))
{
file.WriteLine("{0} {1} {2} {3}",
state2str(sourceState), state2str(targetState), input2str(input), output2str(output));
}
}
}
}
}
};
Action <Move <ILabel <SYMBOL> > > printMove = (type == AutomatonType.SSA) ?
printPredicateMove : printLabelMove;
// print moves
foreach (Move <ILabel <SYMBOL> > move in moves)
{
printMove(move);
}
// print final states
foreach (int finalState in states)
{
file.WriteLine((stateSymbolsFileName == null) ? finalState.ToString() : stateNames[finalState]);
}
if (stateSymbolsFileName != null)
{
var stateFile = new StreamWriter(stateSymbolsFileName);
try {
foreach (KeyValuePair <int, string> item in stateNames)
{
stateFile.WriteLine("{0} {1}", item.Value, item.Key);
}
} finally {
stateFile.Close();
}
}
string[] symbolsFileNames = (type == AutomatonType.SSA) ?
new string[] { inputSymbolsFileName } :
new string[] { inputSymbolsFileName, outputSymbolsFileName };
foreach (string symbolsFileName in symbolsFileNames)
{
if (symbolsFileName != null)
{
var symbolsFile = new StreamWriter(symbolsFileName);
try {
foreach (KeyValuePair <SYMBOL, int> item in symbolDict)
{
symbolsFile.WriteLine("{0} {1}", item.Key, item.Value);
}
} finally {
symbolsFile.Close();
}
}
}
}
/* FSA format spec at http://www.let.rug.nl/~vannoord/Fsa/Manual/node5.html#anc1 */
private static void PrintAutomatonFSA(
StreamWriter file, AutomatonType type, string name, Dictionary <int, string> stateNames,
PredicateAlgebra <SYMBOL> algebra, List <int> states, List <SYMBOL> alphabet,
List <Move <ILabel <SYMBOL> > > moves, int initialState, List <int> finalStates,
string stateSymbolsFileName = null, string inputSymbolsFileName = null, string outputSymbolsFileName = null)
{
var transitions = new List <Move <ILabel <SYMBOL> > >(); // non-epsilon moves
var jumps = new List <Tuple <int, int> >(); // epsilon moves
foreach (Move <ILabel <SYMBOL> > move in moves)
{
if (move.IsEpsilon)
{
jumps.Add(new Tuple <int, int>(move.SourceState, move.TargetState));
}
else
{
transitions.Add(move);
}
}
Func <ILabel <SYMBOL>, string> printPredicate = ilabel => {
var predicate = (Predicate <SYMBOL>)ilabel;
if (predicate == null) // epsilon (only for transducers)
{
return("[]");
}
// add quotes if necessary
Func <SYMBOL, string> sanitizeSymbol = symbol => {
string s = symbol.ToString();
if (char.IsUpper(s[0]) || s.Any(ch => char.IsWhiteSpace(ch)) || (s[0] == '0' && s.Length > 1))
{
return("'" + s + "'");
}
return(s);
};
// `in([a])` may just be written as `a`
if (algebra.InclusiveSet(predicate).Count == 1)
{
return(sanitizeSymbol(algebra.InclusiveSet(predicate).First()));
}
string typeName = (predicate.Type == PredicateType.In) ? "in" : "not_in";
List <string> symbols = predicate.Set.Select(sanitizeSymbol).ToList();
symbols.Sort();
return(string.Format("{0}([{1}])", typeName, string.Join(",", symbols)));
};
Func <ILabel <SYMBOL>, string> printLabel = ilabel => {
var label = (Label <SYMBOL>)ilabel;
return(label.IsIdentity ?
string.Format("$@({0})/$@({0})", printPredicate(label.Input)) :
string.Format("{0}/{1}", printPredicate(label.Input), printPredicate(label.Output)));
};
Func <ILabel <SYMBOL>, string> printILabel = (type == AutomatonType.SSA) ?
printPredicate : printLabel;
file.WriteLine("%% {0} {1}", (type == AutomatonType.SSA) ? "Recognizer" : "Transducer", name ?? "");
file.WriteLine("%% Automatically generated by ARMC.");
file.WriteLine("%% For more info, cf. http://www.let.rug.nl/~vannoord/Fsa/");
file.WriteLine();
file.WriteLine("fa(");
if (type == AutomatonType.SSA)
{
file.WriteLine(" r(fsa_preds),");
}
else
{
file.WriteLine(" t(fsa_preds,fsa_preds),");
}
file.WriteLine(" % number of states");
file.WriteLine(" {0},", states.Count);
file.WriteLine(" % start states");
file.WriteLine(" [ {0} ],", initialState);
file.WriteLine(" % final states");
// put up to 10 final states on one line
int i = 0;
FSAWriteList(
file,
finalStates.GroupBy(state => i++ / 10).Select(group => group.ToList()),
stateList => string.Join(",", stateList)
);
file.WriteLine(" ],");
file.WriteLine(" % moves");
FSAWriteList(file, transitions, (move => string.Format("trans({0},{1},{2})", move.SourceState, printILabel(move.Label), move.TargetState)));
file.WriteLine(" % jumps");
FSAWriteList(file, jumps, (jump => string.Format("jump({0},{1})", jump.Item1, jump.Item2)), false);
file.WriteLine(").");
file.WriteLine();
}
/* format spec at http://www.graphviz.org/doc/info/lang.html */
private static void PrintAutomatonDot(
StreamWriter file, AutomatonType type, string name, Dictionary <int, string> stateNames,
PredicateAlgebra <SYMBOL> algebra, List <int> states, List <SYMBOL> alphabet,
List <Move <ILabel <SYMBOL> > > moves, int initialState, List <int> finalStates,
string stateSymbolsFileName = null, string inputSymbolsFileName = null, string outputSymbolsFileName = null)
{
// group labels for moves with same source and target state (reduce edges in graph)
var transGroups = moves.GroupBy(
move => new Tuple <int, int>(move.SourceState, move.TargetState),
move => move.Label,
(key, labels) => new Tuple <int, List <ILabel <SYMBOL> >, int>(key.Item1, labels.ToList(), key.Item2)
);
// create subscripts and superscripts based on LaTeX-like markup in state names
// e.g. q_0 becomes <q<sub>0</sub>>, q_{42} becomes <q<sub>42</sub>>, q_M^2 becomes <q<sub>M</sub><sup>2</sup>>
var regexSubSingle = new Regex(@"_([^{])");
var regexSubGroup = new Regex(@"_{([^}]*)}");
var regexSuperSingle = new Regex(@"\^([^{])");
var regexSuperGroup = new Regex(@"\^{([^}]*)}");
Func <int, string> stateToHTML = delegate(int state) {
string stateName = stateNames[state];
stateName = regexSubSingle.Replace(stateName, "<sub>$1</sub>");
stateName = regexSubGroup.Replace(stateName, "<sub>$1</sub>");
stateName = regexSuperSingle.Replace(stateName, "<sup>$1</sup>");
stateName = regexSuperGroup.Replace(stateName, "<sup>$1</sup>");
return(stateName);
};
Func <SYMBOL, string> printSymbol = symbol => symbol.ToString()
.Replace("<", "<").Replace(">", ">")
.Replace("[", "[").Replace("]", "]");
Func <ILabel <SYMBOL>, string> printPredicate = ilabel => {
var predicate = (Predicate <SYMBOL>)ilabel;
if (predicate == null)
{
return("ε");
}
if (algebra.InclusiveSet(predicate).Count == 1)
{
return(string.Format("<i>{0}</i>", printSymbol(algebra.InclusiveSet(predicate).First())));
}
string typeSymbol = (predicate.Type == PredicateType.In) ? "∈" : "∉"; // unicode math set symbols
// sort symbols and put in italics
var symbols = new List <SYMBOL>(predicate.Set);
symbols.Sort();
var symbolsFormatted = new List <string>(symbols.Select(symbol => string.Format("<i>{0}</i>", symbol)));
Func <int, int, int> ceilDiv = (x, y) => (x - 1) / y + 1;
int groupSize = 5;
if (symbols.Count > 1)
{
while (ceilDiv(symbols.Count, groupSize - 1) == ceilDiv(symbols.Count, groupSize))
{
groupSize--;
}
}
int i = 0;
string joined = string.Join(",<br/>", symbols
.Select(printSymbol)
.Select(symbol => string.Format("<i>{0}</i>", symbol))
.GroupBy(symbol => i++ / groupSize)
.Select(group => string.Join(", ", group.ToList()))
);
return(typeSymbol + "{" + joined + "}");
};
Func <ILabel <SYMBOL>, string> printLabel = ilabel => {
var label = (Label <SYMBOL>)ilabel;
return(label.IsIdentity ?
string.Format("{0}<b>/</b>𝜄", printPredicate(label.Input)) :
string.Format("{0}<b>/</b>{1}", printPredicate(label.Input), printPredicate(label.Output)));
};
Func <ILabel <SYMBOL>, string> printILabel = (type == AutomatonType.SSA) ?
printPredicate : printLabel;
file.WriteLine("digraph {");
file.WriteLine(" rankdir=LR;"); // left-to-right direction more readable for automata
if (name != null) // print automaton name
{
file.WriteLine(" label=<{0}:>;", name);
file.WriteLine(" labelloc=top;");
file.WriteLine(" labeljust=left;");
}
// final states have double circle
file.Write(" node [shape=doublecircle];");
foreach (int finalState in finalStates)
{
file.Write(" {0};", finalState);
}
file.WriteLine();
file.WriteLine(" node [shape=circle];");
// invisible zero-width dummy node as source of arrow to initial state
file.WriteLine(" dummy_node [style=invis,width=0,fixedsize=true,label=\"\"];");
file.WriteLine(" dummy_node -> {0} [len=0.2,penwidth=2.0];", initialState);
foreach (Tuple <int, List <ILabel <SYMBOL> >, int> trans in transGroups)
{
int sourceState = trans.Item1;
List <ILabel <SYMBOL> > labels = trans.Item2;
int targetState = trans.Item3;
file.WriteLine(" {0} -> {1} [label=<{2}>];",
sourceState, targetState, string.Join(",<br/>", labels.Select(printILabel))
);
}
foreach (int state in states)
{
file.WriteLine(" {0} [label=<{1}>];", state, stateToHTML(state));
}
file.WriteLine("}");
file.WriteLine();
}