Improvements to query optimizer

elki-clustering/src/main/java/elki/clustering/hierarchical/BufferedSearchSingleLink.java

@@ -95,8 +95,6 @@ public TypeInformation[] getInputTypeRestriction() {
   public ClusterMergeHistory run(Relation<O> relation) {
     DBIDEnum ids = DBIDUtil.ensureEnum(relation.getDBIDs());
     ClusterMergeHistoryBuilder builder = new ClusterMergeHistoryBuilder(ids, distance.isSquared());
-    // Create one for testing we have a suitable index.
-    // TODO: enforce a well-tuned VP-tree?
     PrioritySearcher<DBIDRef> pq = new QueryBuilder<>(relation, distance).priorityByDBID();
     if(pq instanceof LinearScanPrioritySearcher || pq instanceof LinearScanEuclideanPrioritySearcher) {
       throw new UnsupportedOperationException("No index acceleration available. This will be very slow.");
@@ -213,7 +211,7 @@ private void initializeHeap() {
         if(builder.getSize(ca) > 1) {
           continue; // duplicate
         }
-        DoubleIntegerMinHeap h = heaps[a] = new DoubleIntegerMinHeap((int) Math.sqrt(ids.size()));
+        DoubleIntegerMinHeap h = heaps[a] = new DoubleIntegerMinHeap();
         double t = Double.POSITIVE_INFINITY;
         for(pq.search(ita); pq.valid() && pq.allLowerBound() < t; pq.advance()) {
           final int b = ids.index(pq);
@@ -229,7 +227,7 @@ private void initializeHeap() {
             continue outer;
           }
           h.add(d, b);
-          t = h.peekKey();
+          pq.decreaseCutoff(t = h.peekKey());
         }
         if(!h.isEmpty()) {
           heap.add(t, a);
@@ -257,7 +255,7 @@ private void refillNeighbors(int a, int ca) {
           continue;
         }
         h.add(pq.computeExactDistance(), b);
-        t = h.peekKey();
+        pq.decreaseCutoff(t = h.peekKey());
       }
       threshold[a] = pq.allLowerBound();
     }

elki-clustering/src/main/java/elki/clustering/hierarchical/HeapOfSearchersSingleLink.java

@@ -100,8 +100,8 @@ public ClusterMergeHistory run(Relation<O> relation) {
     DBIDEnum ids = DBIDUtil.ensureEnum(relation.getDBIDs());
     ClusterMergeHistoryBuilder builder = new ClusterMergeHistoryBuilder(ids, distance.isSquared());
     // Create one for testing we have a suitable index.
-    // TODO: enforce a well-tuned VP-tree?
-    PrioritySearcher<DBIDRef> pq = new QueryBuilder<>(relation, distance).priorityByDBID();
+    PrioritySearcher<DBIDRef> pq = new QueryBuilder<>(relation, distance) //
+        .lowSelectivity().priorityByDBID();
     if(pq instanceof LinearScanPrioritySearcher || pq instanceof LinearScanEuclideanPrioritySearcher) {
       throw new UnsupportedOperationException("No index acceleration available. This will be very slow.");
     }

elki-clustering/src/main/java/elki/clustering/hierarchical/IncrementalNearestNeighborChain.java

@@ -107,8 +107,8 @@ public IncrementalNearestNeighborChain(GeometricLinkage linkage) {
   public ClusterMergeHistory run(Relation<O> relation) {
     DBIDEnum ids = DBIDUtil.ensureEnum(relation.getDBIDs());
     ClusterMergeHistoryBuilder builder = new ClusterMergeHistoryBuilder(ids, true);
-    // TODO: ensure we have good page size!
-    PrioritySearcher<O> pq = new QueryBuilder<>(relation, SquaredEuclideanDistance.STATIC).priorityByObject();
+    PrioritySearcher<O> pq = new QueryBuilder<>(relation, SquaredEuclideanDistance.STATIC) //
+        .lowSelectivity().priorityByObject();
     if(pq instanceof LinearScanPrioritySearcher || pq instanceof LinearScanEuclideanPrioritySearcher) {
       throw new UnsupportedOperationException("No index acceleration available. This will be very slow.");
     }

elki-clustering/src/main/java/elki/clustering/hierarchical/RestartingSearchSingleLink.java

@@ -101,8 +101,8 @@ public ClusterMergeHistory run(Relation<O> relation) {
     DBIDEnum ids = DBIDUtil.ensureEnum(relation.getDBIDs());
     ClusterMergeHistoryBuilder builder = new ClusterMergeHistoryBuilder(ids, distance.isSquared());
     // Create one for testing we have a suitable index.
-    // TODO: enforce a well-tuned VP-tree?
-    PrioritySearcher<DBIDRef> pq = new QueryBuilder<>(relation, distance).priorityByDBID();
+    PrioritySearcher<DBIDRef> pq = new QueryBuilder<>(relation, distance) //
+        .lowSelectivity().priorityByDBID();
     if(pq instanceof LinearScanPrioritySearcher || pq instanceof LinearScanEuclideanPrioritySearcher) {
       throw new UnsupportedOperationException("No index acceleration available. This will be very slow.");
     }

elki-core-api/src/main/java/elki/database/query/knn/SquaredKNNSearcher.java

@@ -0,0 +1,52 @@
+/*
+ * This file is part of ELKI:
+ * Environment for Developing KDD-Applications Supported by Index-Structures
+ * 
+ * Copyright (C) 2024
+ * ELKI Development Team
+ * 
+ * This program is free software: you can redistribute it and/or modify
+ * it under the terms of the GNU Affero General Public License as published by
+ * the Free Software Foundation, either version 3 of the License, or
+ * (at your option) any later version.
+ * 
+ * This program is distributed in the hope that it will be useful,
+ * but WITHOUT ANY WARRANTY; without even the implied warranty of
+ * MERCHANTABILITY or FITNESS FOR A PARTICULAR PURPOSE. See the
+ * GNU Affero General Public License for more details.
+ * 
+ * You should have received a copy of the GNU Affero General Public License
+ * along with this program. If not, see <http://www.gnu.org/licenses/>.
+ */
+package elki.database.query.knn;
+
+import elki.database.ids.KNNList;
+
+/**
+ * Wrapper class to allow using an Euclidean search tree with squared Euclidean
+ * distance.
+ * 
+ * @author Erich Schubert
+ *
+ * @param <O> Data type
+ */
+public class SquaredKNNSearcher<O> implements KNNSearcher<O> {
+  /**
+   * Wrapped searcher
+   */
+  private KNNSearcher<O> inner;
+
+  /**
+   * Constructor.
+   *
+   * @param inner Inner searcher
+   */
+  public SquaredKNNSearcher(KNNSearcher<O> inner) {
+    this.inner = inner;
+  }
+
+  @Override
+  public KNNList getKNN(O query, int k) {
+    return inner.getKNN(query, k).map(x -> x * x);
+  }
+}

elki-core-api/src/main/java/elki/database/query/range/SquaredRangeSearcher.java

@@ -0,0 +1,59 @@
+/*
+ * This file is part of ELKI:
+ * Environment for Developing KDD-Applications Supported by Index-Structures
+ * 
+ * Copyright (C) 2024
+ * ELKI Development Team
+ * 
+ * This program is free software: you can redistribute it and/or modify
+ * it under the terms of the GNU Affero General Public License as published by
+ * the Free Software Foundation, either version 3 of the License, or
+ * (at your option) any later version.
+ * 
+ * This program is distributed in the hope that it will be useful,
+ * but WITHOUT ANY WARRANTY; without even the implied warranty of
+ * MERCHANTABILITY or FITNESS FOR A PARTICULAR PURPOSE. See the
+ * GNU Affero General Public License for more details.
+ * 
+ * You should have received a copy of the GNU Affero General Public License
+ * along with this program. If not, see <http://www.gnu.org/licenses/>.
+ */
+package elki.database.query.range;
+
+import elki.database.ids.DoubleDBIDListMIter;
+import elki.database.ids.ModifiableDoubleDBIDList;
+
+/**
+ * Wrapper class to allow using an Euclidean search tree with squared Euclidean
+ * distance.
+ * 
+ * @author Erich Schubert
+ *
+ * @param <O> Data type
+ */
+public class SquaredRangeSearcher<O> implements RangeSearcher<O> {
+  /**
+   * Wrapped searcher
+   */
+  private RangeSearcher<O> inner;
+
+  /**
+   * Constructor.
+   *
+   * @param inner Inner searcher
+   */
+  public SquaredRangeSearcher(RangeSearcher<O> inner) {
+    this.inner = inner;
+  }
+
+  @Override
+  public ModifiableDoubleDBIDList getRange(O query, double range, ModifiableDoubleDBIDList result) {
+    int oldsize = result.size();
+    inner.getRange(query, Math.sqrt(range), result);
+    for(DoubleDBIDListMIter iter = result.iter().seek(oldsize); iter.valid(); iter.advance()) {
+      double d = iter.doubleValue();
+      iter.setDouble(d * d);
+    }
+    return result;
+  }
+}