/// <summary>
/// Builds a DFA for some string, and checks all Lev automata
/// up to some maximum distance.
/// </summary>
private void AssertLev(string s, int maxDistance)
{
LevenshteinAutomata builder = new LevenshteinAutomata(s, false);
LevenshteinAutomata tbuilder = new LevenshteinAutomata(s, true);
Automaton[] automata = new Automaton[maxDistance + 1];
Automaton[] tautomata = new Automaton[maxDistance + 1];
for (int n = 0; n < automata.Length; n++)
{
automata[n] = builder.ToAutomaton(n);
tautomata[n] = tbuilder.ToAutomaton(n);
Assert.IsNotNull(automata[n]);
Assert.IsNotNull(tautomata[n]);
Assert.IsTrue(automata[n].Deterministic);
Assert.IsTrue(tautomata[n].Deterministic);
Assert.IsTrue(SpecialOperations.IsFinite(automata[n]));
Assert.IsTrue(SpecialOperations.IsFinite(tautomata[n]));
AutomatonTestUtil.AssertNoDetachedStates(automata[n]);
AutomatonTestUtil.AssertNoDetachedStates(tautomata[n]);
// check that the dfa for n-1 accepts a subset of the dfa for n
if (n > 0)
{
Assert.IsTrue(automata[n - 1].SubsetOf(automata[n]));
Assert.IsTrue(automata[n - 1].SubsetOf(tautomata[n]));
Assert.IsTrue(tautomata[n - 1].SubsetOf(automata[n]));
Assert.IsTrue(tautomata[n - 1].SubsetOf(tautomata[n]));
Assert.AreNotSame(automata[n - 1], automata[n]);
}
// check that Lev(N) is a subset of LevT(N)
Assert.IsTrue(automata[n].SubsetOf(tautomata[n]));
// special checks for specific n
switch (n)
{
case 0:
// easy, matches the string itself
Assert.IsTrue(BasicOperations.SameLanguage(BasicAutomata.MakeString(s), automata[0]));
Assert.IsTrue(BasicOperations.SameLanguage(BasicAutomata.MakeString(s), tautomata[0]));
break;
case 1:
// generate a lev1 naively, and check the accepted lang is the same.
Assert.IsTrue(BasicOperations.SameLanguage(NaiveLev1(s), automata[1]));
Assert.IsTrue(BasicOperations.SameLanguage(NaiveLev1T(s), tautomata[1]));
break;
default:
AssertBruteForce(s, automata[n], n);
AssertBruteForceT(s, tautomata[n], n);
break;
}
}
}
/// <summary>
/// Builds a DFA for some string, and checks all Lev automata
/// up to some maximum distance.
/// </summary>
private void AssertLev(string s, int maxDistance)
{
LevenshteinAutomata builder = new LevenshteinAutomata(s, false);
LevenshteinAutomata tbuilder = new LevenshteinAutomata(s, true);
Automaton[] automata = new Automaton[maxDistance + 1];
Automaton[] tautomata = new Automaton[maxDistance + 1];
for (int n = 0; n < automata.Length; n++)
{
automata[n] = builder.ToAutomaton(n);
tautomata[n] = tbuilder.ToAutomaton(n);
Assert.IsNotNull(automata[n]);
Assert.IsNotNull(tautomata[n]);
Assert.IsTrue(automata[n].Deterministic);
Assert.IsTrue(tautomata[n].Deterministic);
Assert.IsTrue(SpecialOperations.IsFinite(automata[n]));
Assert.IsTrue(SpecialOperations.IsFinite(tautomata[n]));
AutomatonTestUtil.AssertNoDetachedStates(automata[n]);
AutomatonTestUtil.AssertNoDetachedStates(tautomata[n]);
// check that the dfa for n-1 accepts a subset of the dfa for n
if (n > 0)
{
Assert.IsTrue(automata[n - 1].SubsetOf(automata[n]));
Assert.IsTrue(automata[n - 1].SubsetOf(tautomata[n]));
Assert.IsTrue(tautomata[n - 1].SubsetOf(automata[n]));
Assert.IsTrue(tautomata[n - 1].SubsetOf(tautomata[n]));
Assert.AreNotSame(automata[n - 1], automata[n]);
}
// check that Lev(N) is a subset of LevT(N)
Assert.IsTrue(automata[n].SubsetOf(tautomata[n]));
// special checks for specific n
switch (n)
{
case 0:
// easy, matches the string itself
Assert.IsTrue(BasicOperations.SameLanguage(BasicAutomata.MakeString(s), automata[0]));
Assert.IsTrue(BasicOperations.SameLanguage(BasicAutomata.MakeString(s), tautomata[0]));
break;
case 1:
// generate a lev1 naively, and check the accepted lang is the same.
Assert.IsTrue(BasicOperations.SameLanguage(NaiveLev1(s), automata[1]));
Assert.IsTrue(BasicOperations.SameLanguage(NaiveLev1T(s), tautomata[1]));
break;
default:
AssertBruteForce(s, automata[n], n);
AssertBruteForceT(s, tautomata[n], n);
break;
}
}
}
internal Automaton ToLevenshteinAutomata(Automaton automaton)
{
var @ref = SpecialOperations.GetFiniteStrings(automaton, -1);
Automaton[] subs = new Automaton[@ref.Count];
int upto = 0;
foreach (IntsRef path in @ref)
{
if (path.Length <= nonFuzzyPrefix || path.Length < minFuzzyLength)
{
subs[upto] = BasicAutomata.MakeString(path.Ints, path.Offset, path.Length);
upto++;
}
else
{
Automaton prefix = BasicAutomata.MakeString(path.Ints, path.Offset, nonFuzzyPrefix);
int[] ints = new int[path.Length - nonFuzzyPrefix];
Array.Copy(path.Ints, path.Offset + nonFuzzyPrefix, ints, 0, ints.Length);
// TODO: maybe add alphaMin to LevenshteinAutomata,
// and pass 1 instead of 0? We probably don't want
// to allow the trailing dedup bytes to be
// edited... but then 0 byte is "in general" allowed
// on input (but not in UTF8).
LevenshteinAutomata lev = new LevenshteinAutomata(ints, unicodeAware ? char.MAX_CODE_POINT : 255, transpositions);
Automaton levAutomaton = lev.ToAutomaton(maxEdits);
Automaton combined = BasicOperations.Concatenate(Arrays.AsList(prefix, levAutomaton));
combined.Deterministic = true; // its like the special case in concatenate itself, except we cloneExpanded already
subs[upto] = combined;
upto++;
}
}
if (subs.Length == 0)
{
// automaton is empty, there is no accepted paths through it
return BasicAutomata.MakeEmpty(); // matches nothing
}
else if (subs.Length == 1)
{
// no synonyms or anything: just a single path through the tokenstream
return subs[0];
}
else
{
// multiple paths: this is really scary! is it slow?
// maybe we should not do this and throw UOE?
Automaton a = BasicOperations.Union(Arrays.AsList(subs));
// TODO: we could call toLevenshteinAutomata() before det?
// this only happens if you have multiple paths anyway (e.g. synonyms)
BasicOperations.Determinize(a);
return a;
}
}