LUCENE-9983: Stop sorting determinize powersets unnecessarily (#163)

* LUCENE-9983: Stop sorting determinize powersets unnecessarily

Author: zhaih

lucene/CHANGES.txt

@@ -368,6 +368,8 @@ Improvements
   cases like det(rev(regexp("(.*a){2000}"))) that spend lots of effort but
   result in smallish eventual returned automata.  (Robert Muir, Mike McCandless)
 
+* LUCENE-9983: Stop sorting determinize powersets unnecessarily. (Patrick Zhai)
+
 Optimizations
 ---------------------
 * LUCENE-9996: Improved memory efficiency of IndexWriter's RAM buffer, in

lucene/core/src/java/org/apache/lucene/util/automaton/FrozenIntSet.java

@@ -20,13 +20,13 @@
 
 final class FrozenIntSet extends IntSet {
   final int[] values;
-  final int hashCode;
   final int state;
+  final long hashCode;
 
-  FrozenIntSet(int[] values, int hashCode, int state) {
+  FrozenIntSet(int[] values, long hashCode, int state) {
     this.values = values;
-    this.hashCode = hashCode;
     this.state = state;
+    this.hashCode = hashCode;
   }
 
   @Override
@@ -40,7 +40,7 @@ int size() {
   }
 
   @Override
-  public int hashCode() {
+  long longHashCode() {
     return hashCode;
   }
 

lucene/core/src/java/org/apache/lucene/util/automaton/IntSet.java

@@ -36,15 +36,19 @@ abstract class IntSet {
    */
   abstract int size();
 
+  abstract long longHashCode();
+
   @Override
-  public abstract int hashCode();
+  public int hashCode() {
+    return Long.hashCode(longHashCode());
+  }
 
   @Override
   public boolean equals(Object o) {
     if (this == o) return true;
     if (!(o instanceof IntSet)) return false;
     IntSet that = (IntSet) o;
-    return hashCode() == that.hashCode()
+    return longHashCode() == that.longHashCode()
         && Arrays.equals(getArray(), 0, size(), that.getArray(), 0, that.size());
   }
 }

lucene/core/src/java/org/apache/lucene/util/automaton/Operations.java

@@ -47,6 +47,7 @@
 import org.apache.lucene.util.IntsRef;
 import org.apache.lucene.util.IntsRefBuilder;
 import org.apache.lucene.util.RamUsageEstimator;
+import org.apache.lucene.util.hppc.BitMixer;
 
 /**
  * Automata operations.
@@ -690,7 +691,7 @@ public static Automaton determinize(Automaton a, int workLimit) {
     // a.writeDot("/l/la/lucene/core/detin.dot");
 
     // Same initial values and state will always have the same hashCode
-    FrozenIntSet initialset = new FrozenIntSet(new int[] {0}, 683, 0);
+    FrozenIntSet initialset = new FrozenIntSet(new int[] {0}, BitMixer.mix(0) + 1, 0);
 
     // Create state 0:
     b.createState();
@@ -706,8 +707,8 @@ public static Automaton determinize(Automaton a, int workLimit) {
     // like Set<Integer,PointTransitions>
     final PointTransitionSet points = new PointTransitionSet();
 
-    // like SortedMap<Integer,Integer>
-    final SortedIntSet statesSet = new SortedIntSet(5);
+    // like HashMap<Integer,Integer>, maps state to its count
+    final StateSet statesSet = new StateSet(5);
 
     Transition t = new Transition();
 
@@ -759,11 +760,9 @@ public static Automaton determinize(Automaton a, int workLimit) {
 
         final int point = points.points[i].point;
 
-        if (statesSet.upto > 0) {
+        if (statesSet.size() > 0) {
           assert lastPoint != -1;
 
-          statesSet.computeHash();
-
           Integer q = newstate.get(statesSet);
           if (q == null) {
             q = b.createState();
@@ -812,7 +811,7 @@ public static Automaton determinize(Automaton a, int workLimit) {
         points.points[i].starts.next = 0;
       }
       points.reset();
-      assert statesSet.upto == 0 : "upto=" + statesSet.upto;
+      assert statesSet.size() == 0 : "size=" + statesSet.size();
     }
 
     Automaton result = b.finish();