| public SetNumberedStates ( State states, int count ) : void | ||
| states | State | |
| count | int | |
| return | void | |
/// <summary>
/// Converts an incoming utf32 <see cref="Automaton"/> to an equivalent
/// utf8 one. The incoming automaton need not be
/// deterministic. Note that the returned automaton will
/// not in general be deterministic, so you must
/// determinize it if that's needed.
/// </summary>
public Automaton Convert(Automaton utf32)
{
if (utf32.IsSingleton)
{
utf32 = utf32.CloneExpanded();
}
State[] map = new State[utf32.GetNumberedStates().Length];
JCG.List <State> pending = new JCG.List <State>();
State utf32State = utf32.GetInitialState();
pending.Add(utf32State);
Automaton utf8 = new Automaton();
utf8.IsDeterministic = false;
State utf8State = utf8.GetInitialState();
utf8States = new State[5];
utf8StateCount = 0;
utf8State.number = utf8StateCount;
utf8States[utf8StateCount] = utf8State;
utf8StateCount++;
utf8State.Accept = utf32State.Accept;
map[utf32State.number] = utf8State;
while (pending.Count != 0)
{
utf32State = pending[pending.Count - 1];
pending.RemoveAt(pending.Count - 1);
utf8State = map[utf32State.number];
for (int i = 0; i < utf32State.numTransitions; i++)
{
Transition t = utf32State.TransitionsArray[i];
State destUTF32 = t.to;
State destUTF8 = map[destUTF32.number];
if (destUTF8 == null)
{
destUTF8 = NewUTF8State();
destUTF8.accept = destUTF32.accept;
map[destUTF32.number] = destUTF8;
pending.Add(destUTF32);
}
ConvertOneEdge(utf8State, destUTF8, t.min, t.max);
}
}
utf8.SetNumberedStates(utf8States, utf8StateCount);
return(utf8);
}
/// <summary>
/// Determinizes the given automaton.
/// <para/>
/// Worst case complexity: exponential in number of states.
/// </summary>
public static void Determinize(Automaton a)
{
if (a.IsDeterministic || a.IsSingleton)
{
return;
}
State[] allStates = a.GetNumberedStates();
// subset construction
bool initAccept = a.initial.accept;
int initNumber = a.initial.number;
a.initial = new State();
SortedInt32Set.FrozenInt32Set initialset = new SortedInt32Set.FrozenInt32Set(initNumber, a.initial);
Queue <SortedInt32Set.FrozenInt32Set> worklist = new Queue <SortedInt32Set.FrozenInt32Set>(); // LUCENENET specific - Queue is much more performant than LinkedList
IDictionary <SortedInt32Set.FrozenInt32Set, State> newstate = new Dictionary <SortedInt32Set.FrozenInt32Set, State>();
worklist.Enqueue(initialset);
a.initial.accept = initAccept;
newstate[initialset] = a.initial;
int newStateUpto = 0;
State[] newStatesArray = new State[5];
newStatesArray[newStateUpto] = a.initial;
a.initial.number = newStateUpto;
newStateUpto++;
// like Set<Integer,PointTransitions>
PointTransitionSet points = new PointTransitionSet();
// like SortedMap<Integer,Integer>
SortedInt32Set statesSet = new SortedInt32Set(5);
while (worklist.Count > 0)
{
SortedInt32Set.FrozenInt32Set s = worklist.Dequeue();
//worklist.Remove(s);
// Collate all outgoing transitions by min/1+max:
for (int i = 0; i < s.values.Length; i++)
{
State s0 = allStates[s.values[i]];
for (int j = 0; j < s0.numTransitions; j++)
{
points.Add(s0.TransitionsArray[j]);
}
}
if (points.count == 0)
{
// No outgoing transitions -- skip it
continue;
}
points.Sort();
int lastPoint = -1;
int accCount = 0;
State r = s.state;
for (int i = 0; i < points.count; i++)
{
int point = points.points[i].point;
if (statesSet.upto > 0)
{
if (Debugging.AssertsEnabled)
{
Debugging.Assert(lastPoint != -1);
}
statesSet.ComputeHash();
if (!newstate.TryGetValue(statesSet.ToFrozenInt32Set(), out State q) || q == null)
{
q = new State();
SortedInt32Set.FrozenInt32Set p = statesSet.Freeze(q);
worklist.Enqueue(p);
if (newStateUpto == newStatesArray.Length)
{
// LUCENENET: Resize rather than copy
Array.Resize(ref newStatesArray, ArrayUtil.Oversize(1 + newStateUpto, RamUsageEstimator.NUM_BYTES_OBJECT_REF));
}
newStatesArray[newStateUpto] = q;
q.number = newStateUpto;
newStateUpto++;
q.accept = accCount > 0;
newstate[p] = q;
}
else
{
if (Debugging.AssertsEnabled)
{
Debugging.Assert((accCount > 0) == q.accept, "accCount={0} vs existing accept={1} states={2}", accCount, q.accept, statesSet);
}
}
r.AddTransition(new Transition(lastPoint, point - 1, q));
}
// process transitions that end on this point
// (closes an overlapping interval)
Transition[] transitions = points.points[i].ends.transitions;
int limit = points.points[i].ends.count;
for (int j = 0; j < limit; j++)
{
Transition t = transitions[j];
int num = t.to.number;
statesSet.Decr(num);
accCount -= t.to.accept ? 1 : 0;
}
points.points[i].ends.count = 0;
// process transitions that start on this point
// (opens a new interval)
transitions = points.points[i].starts.transitions;
limit = points.points[i].starts.count;
for (int j = 0; j < limit; j++)
{
Transition t = transitions[j];
int num = t.to.number;
statesSet.Incr(num);
accCount += t.to.accept ? 1 : 0;
}
lastPoint = point;
points.points[i].starts.count = 0;
}
points.Reset();
if (Debugging.AssertsEnabled)
{
Debugging.Assert(statesSet.upto == 0, "upto={0}", statesSet.upto);
}
}
a.deterministic = true;
a.SetNumberedStates(newStatesArray, newStateUpto);
}
/// <summary>
/// Determinizes the given automaton.
/// <para/>
/// Worst case complexity: exponential in number of states.
/// </summary>
public static void Determinize(Automaton a)
{
if (a.IsDeterministic || a.IsSingleton)
{
return;
}
State[] allStates = a.GetNumberedStates();
// subset construction
bool initAccept = a.initial.accept;
int initNumber = a.initial.number;
a.initial = new State();
SortedInt32Set.FrozenInt32Set initialset = new SortedInt32Set.FrozenInt32Set(initNumber, a.initial);
LinkedList <SortedInt32Set.FrozenInt32Set> worklist = new LinkedList <SortedInt32Set.FrozenInt32Set>();
IDictionary <SortedInt32Set.FrozenInt32Set, State> newstate = new Dictionary <SortedInt32Set.FrozenInt32Set, State>();
worklist.AddLast(initialset);
a.initial.accept = initAccept;
newstate[initialset] = a.initial;
int newStateUpto = 0;
State[] newStatesArray = new State[5];
newStatesArray[newStateUpto] = a.initial;
a.initial.number = newStateUpto;
newStateUpto++;
// like Set<Integer,PointTransitions>
PointTransitionSet points = new PointTransitionSet();
// like SortedMap<Integer,Integer>
SortedInt32Set statesSet = new SortedInt32Set(5);
// LUCENENET NOTE: The problem here is almost certainly
// due to the conversion to FrozenIntSet along with its
// differing equality checking.
while (worklist.Count > 0)
{
SortedInt32Set.FrozenInt32Set s = worklist.First.Value;
worklist.Remove(s);
// Collate all outgoing transitions by min/1+max:
for (int i = 0; i < s.values.Length; i++)
{
State s0 = allStates[s.values[i]];
for (int j = 0; j < s0.numTransitions; j++)
{
points.Add(s0.TransitionsArray[j]);
}
}
if (points.count == 0)
{
// No outgoing transitions -- skip it
continue;
}
points.Sort();
int lastPoint = -1;
int accCount = 0;
State r = s.state;
for (int i = 0; i < points.count; i++)
{
int point = points.points[i].point;
if (statesSet.upto > 0)
{
Debug.Assert(lastPoint != -1);
statesSet.ComputeHash();
State q;
if (!newstate.TryGetValue(statesSet.ToFrozenInt32Set(), out q) || q == null)
{
q = new State();
SortedInt32Set.FrozenInt32Set p = statesSet.Freeze(q);
worklist.AddLast(p);
if (newStateUpto == newStatesArray.Length)
{
State[] newArray = new State[ArrayUtil.Oversize(1 + newStateUpto, RamUsageEstimator.NUM_BYTES_OBJECT_REF)];
Array.Copy(newStatesArray, 0, newArray, 0, newStateUpto);
newStatesArray = newArray;
}
newStatesArray[newStateUpto] = q;
q.number = newStateUpto;
newStateUpto++;
q.accept = accCount > 0;
newstate[p] = q;
}
else
{
Debug.Assert((accCount > 0) == q.accept, "accCount=" + accCount + " vs existing accept=" + q.accept + " states=" + statesSet);
}
r.AddTransition(new Transition(lastPoint, point - 1, q));
}
// process transitions that end on this point
// (closes an overlapping interval)
Transition[] transitions = points.points[i].ends.transitions;
int limit = points.points[i].ends.count;
for (int j = 0; j < limit; j++)
{
Transition t = transitions[j];
int num = t.to.number;
statesSet.Decr(num);
accCount -= t.to.accept ? 1 : 0;
}
points.points[i].ends.count = 0;
// process transitions that start on this point
// (opens a new interval)
transitions = points.points[i].starts.transitions;
limit = points.points[i].starts.count;
for (int j = 0; j < limit; j++)
{
Transition t = transitions[j];
int num = t.to.number;
statesSet.Incr(num);
accCount += t.to.accept ? 1 : 0;
}
lastPoint = point;
points.points[i].starts.count = 0;
}
points.Reset();
Debug.Assert(statesSet.upto == 0, "upto=" + statesSet.upto);
}
a.deterministic = true;
a.SetNumberedStates(newStatesArray, newStateUpto);
}
/// <summary>
/// Converts an incoming utf32 automaton to an equivalent
/// utf8 one. The incoming automaton need not be
/// deterministic. Note that the returned automaton will
/// not in general be deterministic, so you must
/// determinize it if that's needed.
/// </summary>
public Automaton Convert(Automaton utf32)
{
if (utf32.IsSingleton)
{
utf32 = utf32.CloneExpanded();
}
State[] map = new State[utf32.NumberedStates.Length];
List<State> pending = new List<State>();
State utf32State = utf32.InitialState;
pending.Add(utf32State);
Automaton utf8 = new Automaton();
utf8.Deterministic = false;
State utf8State = utf8.InitialState;
Utf8States = new State[5];
Utf8StateCount = 0;
utf8State.number = Utf8StateCount;
Utf8States[Utf8StateCount] = utf8State;
Utf8StateCount++;
utf8State.Accept = utf32State.Accept;
map[utf32State.number] = utf8State;
while (pending.Count != 0)
{
utf32State = pending[pending.Count - 1];
pending.RemoveAt(pending.Count - 1);
utf8State = map[utf32State.number];
for (int i = 0; i < utf32State.numTransitions; i++)
{
Transition t = utf32State.TransitionsArray[i];
State destUTF32 = t.To;
State destUTF8 = map[destUTF32.number];
if (destUTF8 == null)
{
destUTF8 = NewUTF8State();
destUTF8.accept = destUTF32.accept;
map[destUTF32.number] = destUTF8;
pending.Add(destUTF32);
}
ConvertOneEdge(utf8State, destUTF8, t.Min_Renamed, t.Max_Renamed);
}
}
utf8.SetNumberedStates(Utf8States, Utf8StateCount);
return utf8;
}
