/// <summary>
/// Returns an automaton that accepts between <paramref name="min"/> and
/// <paramref name="max"/> (including both) concatenated repetitions of the language
/// of the given automaton.
/// <para/>
/// Complexity: linear in number of states and in <paramref name="min"/> and
/// <paramref name="max"/>.
/// </summary>
public static Automaton Repeat(Automaton a, int min, int max)
{
if (min > max)
{
return(BasicAutomata.MakeEmpty());
}
max -= min;
a.ExpandSingleton();
Automaton b;
if (min == 0)
{
b = BasicAutomata.MakeEmptyString();
}
else if (min == 1)
{
b = (Automaton)a.Clone();
}
else
{
IList <Automaton> @as = new List <Automaton>();
while (min-- > 0)
{
@as.Add(a);
}
b = Concatenate(@as);
}
if (max > 0)
{
Automaton d = (Automaton)a.Clone();
while (--max > 0)
{
Automaton c = (Automaton)a.Clone();
foreach (State p in c.GetAcceptStates())
{
p.AddEpsilon(d.initial);
}
d = c;
}
foreach (State p in b.GetAcceptStates())
{
p.AddEpsilon(d.initial);
}
b.deterministic = false;
//b.clearHashCode();
b.ClearNumberedStates();
b.CheckMinimizeAlways();
}
return(b);
}
/// <summary>
/// Returns a (deterministic) automaton that accepts the intersection of the
/// language of <paramref name="a1"/> and the complement of the language of
/// <paramref name="a2"/>. As a side-effect, the automata may be determinized, if not
/// already deterministic.
/// <para/>
/// Complexity: quadratic in number of states (if already deterministic).
/// </summary>
public static Automaton Minus(Automaton a1, Automaton a2)
{
if (BasicOperations.IsEmpty(a1) || a1 == a2)
{
return(BasicAutomata.MakeEmpty());
}
if (BasicOperations.IsEmpty(a2))
{
return(a1.CloneIfRequired());
}
if (a1.IsSingleton)
{
if (BasicOperations.Run(a2, a1.singleton))
{
return(BasicAutomata.MakeEmpty());
}
else
{
return(a1.CloneIfRequired());
}
}
return(Intersection(a1, a2.Complement()));
}
/// <summary>
/// Returns an automaton that accepts the concatenation of the languages of the
/// given automata.
/// <para/>
/// Complexity: linear in number of states.
/// </summary>
public static Automaton Concatenate(Automaton a1, Automaton a2)
{
if (a1.IsSingleton && a2.IsSingleton)
{
return(BasicAutomata.MakeString(a1.singleton + a2.singleton));
}
if (IsEmpty(a1) || IsEmpty(a2))
{
return(BasicAutomata.MakeEmpty());
}
// adding epsilon transitions with the NFA concatenation algorithm
// in this case always produces a resulting DFA, preventing expensive
// redundant determinize() calls for this common case.
bool deterministic = a1.IsSingleton && a2.IsDeterministic;
if (a1 == a2)
{
a1 = a1.CloneExpanded();
a2 = a2.CloneExpanded();
}
else
{
a1 = a1.CloneExpandedIfRequired();
a2 = a2.CloneExpandedIfRequired();
}
foreach (State s in a1.GetAcceptStates())
{
s.accept = false;
s.AddEpsilon(a2.initial);
}
a1.deterministic = deterministic;
//a1.clearHashCode();
a1.ClearNumberedStates();
a1.CheckMinimizeAlways();
return(a1);
}
/// <summary>
/// Returns an automaton that accepts the concatenation of the languages of the
/// given automata.
/// <para/>
/// Complexity: linear in total number of states.
/// </summary>
public static Automaton Concatenate(IList <Automaton> l)
{
if (l.Count == 0)
{
return(BasicAutomata.MakeEmptyString());
}
bool all_singleton = true;
foreach (Automaton a in l)
{
if (!a.IsSingleton)
{
all_singleton = false;
break;
}
}
if (all_singleton)
{
StringBuilder b = new StringBuilder();
foreach (Automaton a in l)
{
b.Append(a.singleton);
}
return(BasicAutomata.MakeString(b.ToString()));
}
else
{
foreach (Automaton a in l)
{
if (BasicOperations.IsEmpty(a))
{
return(BasicAutomata.MakeEmpty());
}
}
JCG.HashSet <int> ids = new JCG.HashSet <int>();
foreach (Automaton a in l)
{
ids.Add(a.GetHashCode());
}
bool has_aliases = ids.Count != l.Count;
Automaton b = l[0];
if (has_aliases)
{
b = b.CloneExpanded();
}
else
{
b = b.CloneExpandedIfRequired();
}
ISet <State> ac = b.GetAcceptStates();
bool first = true;
foreach (Automaton a in l)
{
if (first)
{
first = false;
}
else
{
if (a.IsEmptyString)
{
continue;
}
Automaton aa = a;
if (has_aliases)
{
aa = aa.CloneExpanded();
}
else
{
aa = aa.CloneExpandedIfRequired();
}
ISet <State> ns = aa.GetAcceptStates();
foreach (State s in ac)
{
s.accept = false;
s.AddEpsilon(aa.initial);
if (s.accept)
{
ns.Add(s);
}
}
ac = ns;
}
}
b.deterministic = false;
//b.clearHashCode();
b.ClearNumberedStates();
b.CheckMinimizeAlways();
return(b);
}
}
/// <summary>
/// Returns an automaton that accepts the intersection of the languages of the
/// given automata. Never modifies the input automata languages.
/// <para/>
/// Complexity: quadratic in number of states.
/// </summary>
public static Automaton Intersection(Automaton a1, Automaton a2)
{
if (a1.IsSingleton)
{
if (BasicOperations.Run(a2, a1.singleton))
{
return(a1.CloneIfRequired());
}
else
{
return(BasicAutomata.MakeEmpty());
}
}
if (a2.IsSingleton)
{
if (BasicOperations.Run(a1, a2.singleton))
{
return(a2.CloneIfRequired());
}
else
{
return(BasicAutomata.MakeEmpty());
}
}
if (a1 == a2)
{
return(a1.CloneIfRequired());
}
Transition[][] transitions1 = a1.GetSortedTransitions();
Transition[][] transitions2 = a2.GetSortedTransitions();
Automaton c = new Automaton();
Queue <StatePair> worklist = new Queue <StatePair>(); // LUCENENET specific - Queue is much more performant than LinkedList
Dictionary <StatePair, StatePair> newstates = new Dictionary <StatePair, StatePair>();
StatePair p = new StatePair(c.initial, a1.initial, a2.initial);
worklist.Enqueue(p);
newstates[p] = p;
while (worklist.Count > 0)
{
p = worklist.Dequeue();
p.s.accept = p.s1.accept && p.s2.accept;
Transition[] t1 = transitions1[p.s1.number];
Transition[] t2 = transitions2[p.s2.number];
for (int n1 = 0, b2 = 0; n1 < t1.Length; n1++)
{
while (b2 < t2.Length && t2[b2].max < t1[n1].min)
{
b2++;
}
for (int n2 = b2; n2 < t2.Length && t1[n1].max >= t2[n2].min; n2++)
{
if (t2[n2].max >= t1[n1].min)
{
StatePair q = new StatePair(t1[n1].to, t2[n2].to);
if (!newstates.TryGetValue(q, out StatePair r) || r is null)
{
q.s = new State();
worklist.Enqueue(q);
newstates[q] = q;
r = q;
}
int min = t1[n1].min > t2[n2].min ? t1[n1].min : t2[n2].min;
int max = t1[n1].max < t2[n2].max ? t1[n1].max : t2[n2].max;
p.s.AddTransition(new Transition(min, max, r.s));
}
}
}
}
c.deterministic = a1.deterministic && a2.deterministic;
c.RemoveDeadTransitions();
c.CheckMinimizeAlways();
return(c);
}
private Automaton ToAutomaton(IDictionary <string, Automaton> automata, IAutomatonProvider automaton_provider)
{
IList <Automaton> list;
Automaton a = null;
switch (kind)
{
case Kind.REGEXP_UNION:
list = new List <Automaton>();
FindLeaves(exp1, Kind.REGEXP_UNION, list, automata, automaton_provider);
FindLeaves(exp2, Kind.REGEXP_UNION, list, automata, automaton_provider);
a = BasicOperations.Union(list);
MinimizationOperations.Minimize(a);
break;
case Kind.REGEXP_CONCATENATION:
list = new List <Automaton>();
FindLeaves(exp1, Kind.REGEXP_CONCATENATION, list, automata, automaton_provider);
FindLeaves(exp2, Kind.REGEXP_CONCATENATION, list, automata, automaton_provider);
a = BasicOperations.Concatenate(list);
MinimizationOperations.Minimize(a);
break;
case Kind.REGEXP_INTERSECTION:
a = exp1.ToAutomaton(automata, automaton_provider).Intersection(exp2.ToAutomaton(automata, automaton_provider));
MinimizationOperations.Minimize(a);
break;
case Kind.REGEXP_OPTIONAL:
a = exp1.ToAutomaton(automata, automaton_provider).Optional();
MinimizationOperations.Minimize(a);
break;
case Kind.REGEXP_REPEAT:
a = exp1.ToAutomaton(automata, automaton_provider).Repeat();
MinimizationOperations.Minimize(a);
break;
case Kind.REGEXP_REPEAT_MIN:
a = exp1.ToAutomaton(automata, automaton_provider).Repeat(min);
MinimizationOperations.Minimize(a);
break;
case Kind.REGEXP_REPEAT_MINMAX:
a = exp1.ToAutomaton(automata, automaton_provider).Repeat(min, max);
MinimizationOperations.Minimize(a);
break;
case Kind.REGEXP_COMPLEMENT:
a = exp1.ToAutomaton(automata, automaton_provider).Complement();
MinimizationOperations.Minimize(a);
break;
case Kind.REGEXP_CHAR:
a = BasicAutomata.MakeChar(c);
break;
case Kind.REGEXP_CHAR_RANGE:
a = BasicAutomata.MakeCharRange(from, to);
break;
case Kind.REGEXP_ANYCHAR:
a = BasicAutomata.MakeAnyChar();
break;
case Kind.REGEXP_EMPTY:
a = BasicAutomata.MakeEmpty();
break;
case Kind.REGEXP_STRING:
a = BasicAutomata.MakeString(s);
break;
case Kind.REGEXP_ANYSTRING:
a = BasicAutomata.MakeAnyString();
break;
case Kind.REGEXP_AUTOMATON:
Automaton aa = null;
if (automata != null)
{
aa = automata[s];
}
if (aa == null && automaton_provider != null)
{
try
{
aa = automaton_provider.GetAutomaton(s);
}
catch (System.IO.IOException e)
{
throw new System.ArgumentException(e.ToString(), e);
}
}
if (aa == null)
{
throw new System.ArgumentException("'" + s + "' not found");
}
a = (Automaton)aa.Clone(); // always clone here (ignore allow_mutate)
break;
case Kind.REGEXP_INTERVAL:
a = BasicAutomata.MakeInterval(min, max, digits);
break;
}
return(a);
}
/// <summary>
/// Compute a DFA that accepts all strings within an edit distance of <paramref name="n"/>.
/// <para>
/// All automata have the following properties:
/// <list type="bullet">
/// <item><description>They are deterministic (DFA).</description></item>
/// <item><description>There are no transitions to dead states.</description></item>
/// <item><description>They are not minimal (some transitions could be combined).</description></item>
/// </list>
/// </para>
/// </summary>
public virtual Automaton ToAutomaton(int n)
{
if (n == 0)
{
return(BasicAutomata.MakeString(word, 0, word.Length));
}
if (n >= descriptions.Length)
{
return(null);
}
int range = 2 * n + 1;
ParametricDescription description = descriptions[n];
// the number of states is based on the length of the word and n
State[] states = new State[description.Count];
// create all states, and mark as accept states if appropriate
for (int i = 0; i < states.Length; i++)
{
states[i] = new State();
states[i].number = i;
states[i].Accept = description.IsAccept(i);
}
// create transitions from state to state
for (int k = 0; k < states.Length; k++)
{
int xpos = description.GetPosition(k);
if (xpos < 0)
{
continue;
}
int end = xpos + Math.Min(word.Length - xpos, range);
for (int x = 0; x < alphabet.Length; x++)
{
int ch = alphabet[x];
// get the characteristic vector at this position wrt ch
int cvec = GetVector(ch, xpos, end);
int dest = description.Transition(k, xpos, cvec);
if (dest >= 0)
{
states[k].AddTransition(new Transition(ch, states[dest]));
}
}
// add transitions for all other chars in unicode
// by definition, their characteristic vectors are always 0,
// because they do not exist in the input string.
int dest_ = description.Transition(k, xpos, 0); // by definition
if (dest_ >= 0)
{
for (int r = 0; r < numRanges; r++)
{
states[k].AddTransition(new Transition(rangeLower[r], rangeUpper[r], states[dest_]));
}
}
}
Automaton a = new Automaton(states[0]);
a.IsDeterministic = true;
// we create some useless unconnected states, and its a net-win overall to remove these,
// as well as to combine any adjacent transitions (it makes later algorithms more efficient).
// so, while we could set our numberedStates here, its actually best not to, and instead to
// force a traversal in reduce, pruning the unconnected states while we combine adjacent transitions.
//a.setNumberedStates(states);
a.Reduce();
// we need not trim transitions to dead states, as they are not created.
//a.restoreInvariant();
return(a);
}
public CompiledAutomaton(Automaton automaton, bool?finite, bool simplify)
{
if (simplify)
{
// Test whether the automaton is a "simple" form and
// if so, don't create a runAutomaton. Note that on a
// large automaton these tests could be costly:
if (BasicOperations.IsEmpty(automaton))
{
// matches nothing
Type = AUTOMATON_TYPE.NONE;
Term = null;
CommonSuffixRef = null;
RunAutomaton = null;
sortedTransitions = null;
this.Finite = null;
return;
}
else if (BasicOperations.IsTotal(automaton))
{
// matches all possible strings
Type = AUTOMATON_TYPE.ALL;
Term = null;
CommonSuffixRef = null;
RunAutomaton = null;
sortedTransitions = null;
this.Finite = null;
return;
}
else
{
string commonPrefix;
string singleton;
if (automaton.Singleton == null)
{
commonPrefix = SpecialOperations.GetCommonPrefix(automaton);
if (commonPrefix.Length > 0 && BasicOperations.SameLanguage(automaton, BasicAutomata.MakeString(commonPrefix)))
{
singleton = commonPrefix;
}
else
{
singleton = null;
}
}
else
{
commonPrefix = null;
singleton = automaton.Singleton;
}
if (singleton != null)
{
// matches a fixed string in singleton or expanded
// representation
Type = AUTOMATON_TYPE.SINGLE;
Term = new BytesRef(singleton);
CommonSuffixRef = null;
RunAutomaton = null;
sortedTransitions = null;
this.Finite = null;
return;
}
else if (BasicOperations.SameLanguage(automaton, BasicOperations.Concatenate(BasicAutomata.MakeString(commonPrefix), BasicAutomata.MakeAnyString())))
{
// matches a constant prefix
Type = AUTOMATON_TYPE.PREFIX;
Term = new BytesRef(commonPrefix);
CommonSuffixRef = null;
RunAutomaton = null;
sortedTransitions = null;
this.Finite = null;
return;
}
}
}
Type = AUTOMATON_TYPE.NORMAL;
Term = null;
if (finite == null)
{
this.Finite = SpecialOperations.IsFinite(automaton);
}
else
{
this.Finite = finite;
}
Automaton utf8 = (new UTF32ToUTF8()).Convert(automaton);
if (this.Finite == true)
{
CommonSuffixRef = null;
}
else
{
CommonSuffixRef = SpecialOperations.GetCommonSuffixBytesRef(utf8);
}
RunAutomaton = new ByteRunAutomaton(utf8, true);
sortedTransitions = utf8.GetSortedTransitions();
}