Cheaper sampling for native index population

By doing one pass over the raw index keys in the end
instead of keeping a map of unique values where both
the map as well as materializing the values into it
are expensive.

community/kernel/src/main/java/org/neo4j/kernel/impl/index/schema/DateLayout.java

@@ -25,7 +25,7 @@
 /**
  * {@link Layout} for dates.
  */
-class DateLayout extends IndexLayout<DateIndexKey>
+class DateLayout extends IndexLayout<DateIndexKey,NativeIndexValue>
 {
     DateLayout()
     {

community/kernel/src/main/java/org/neo4j/kernel/impl/index/schema/DurationLayout.java

@@ -25,7 +25,7 @@
 /**
  * {@link Layout} for durations.
  */
-class DurationLayout extends IndexLayout<DurationIndexKey>
+class DurationLayout extends IndexLayout<DurationIndexKey,NativeIndexValue>
 {
     DurationLayout()
     {

community/kernel/src/main/java/org/neo4j/kernel/impl/index/schema/FullScanNonUniqueIndexSampler.java

@@ -25,9 +25,6 @@
 import org.neo4j.cursor.RawCursor;
 import org.neo4j.index.internal.gbptree.GBPTree;
 import org.neo4j.index.internal.gbptree.Hit;
-import org.neo4j.index.internal.gbptree.Layout;
-import org.neo4j.kernel.impl.api.index.sampling.DefaultNonUniqueIndexSampler;
-import org.neo4j.kernel.impl.api.index.sampling.IndexSamplingConfig;
 import org.neo4j.kernel.impl.api.index.sampling.NonUniqueIndexSampler;
 import org.neo4j.storageengine.api.schema.IndexSample;
 
@@ -41,15 +38,12 @@ class FullScanNonUniqueIndexSampler<KEY extends NativeIndexKey<KEY>, VALUE exten
         extends NonUniqueIndexSampler.Adapter
 {
     private final GBPTree<KEY,VALUE> gbpTree;
-    private final Layout<KEY,VALUE> layout;
-    private final IndexSamplingConfig samplingConfig;
+    private final IndexLayout<KEY,VALUE> layout;
 
-    FullScanNonUniqueIndexSampler( GBPTree<KEY,VALUE> gbpTree, Layout<KEY,VALUE> layout,
-            IndexSamplingConfig samplingConfig )
+    FullScanNonUniqueIndexSampler( GBPTree<KEY,VALUE> gbpTree, IndexLayout<KEY,VALUE> layout )
     {
         this.gbpTree = gbpTree;
         this.layout = layout;
-        this.samplingConfig = samplingConfig;
     }
 
     @Override
@@ -61,15 +55,33 @@ public IndexSample result()
         KEY highest = layout.newKey();
         highest.initialize( Long.MAX_VALUE );
         highest.initValuesAsHighest();
+        KEY prev = layout.newKey();
         try ( RawCursor<Hit<KEY,VALUE>,IOException> seek = gbpTree.seek( lowest, highest ) )
         {
-            NonUniqueIndexSampler sampler = new DefaultNonUniqueIndexSampler( samplingConfig.sampleSizeLimit() );
-            while ( seek.next() )
+            long sampledValues = 0;
+            long uniqueValues = 0;
+
+            // Get the first one so that prev gets initialized
+            if ( seek.next() )
             {
-                Hit<KEY,VALUE> hit = seek.get();
-                sampler.include( hit.key().propertiesAsString() );
+                prev = layout.copyKey( seek.get().key(), prev );
+                sampledValues++;
+                uniqueValues++;
+
+                // Then do the rest
+                while ( seek.next() )
+                {
+                    Hit<KEY,VALUE> hit = seek.get();
+                    if ( layout.compareValue( prev, hit.key() ) != 0 )
+                    {
+                        uniqueValues++;
+                        layout.copyKey( hit.key(), prev );
+                    }
+                    // else this is a duplicate of the previous one
+                    sampledValues++;
+                }
             }
-            return sampler.result();
+            return new IndexSample( sampledValues, uniqueValues, sampledValues );
         }
         catch ( IOException e )
         {

community/kernel/src/main/java/org/neo4j/kernel/impl/index/schema/GenericLayout.java

@@ -28,7 +28,7 @@
 
 import static java.util.Comparator.comparing;
 
-class GenericLayout extends IndexLayout<CompositeGenericKey>
+class GenericLayout extends IndexLayout<CompositeGenericKey,NativeIndexValue>
 {
     static final Comparator<Type> TYPE_COMPARATOR = comparing( t -> t.valueGroup );
     private final int numberOfSlots;

community/kernel/src/main/java/org/neo4j/kernel/impl/index/schema/GenericNativeIndexAccessor.java

@@ -23,7 +23,6 @@
 import java.io.IOException;
 
 import org.neo4j.index.internal.gbptree.GBPTree;
-import org.neo4j.index.internal.gbptree.Layout;
 import org.neo4j.index.internal.gbptree.RecoveryCleanupWorkCollector;
 import org.neo4j.io.fs.FileSystemAbstraction;
 import org.neo4j.io.pagecache.PageCache;
@@ -38,7 +37,7 @@ class GenericNativeIndexAccessor extends NativeIndexAccessor<CompositeGenericKey
 {
     private Validator<Value[]> validator;
 
-    GenericNativeIndexAccessor( PageCache pageCache, FileSystemAbstraction fs, File storeFile, Layout<CompositeGenericKey,NativeIndexValue> layout,
+    GenericNativeIndexAccessor( PageCache pageCache, FileSystemAbstraction fs, File storeFile, IndexLayout<CompositeGenericKey,NativeIndexValue> layout,
             RecoveryCleanupWorkCollector recoveryCleanupWorkCollector, IndexProvider.Monitor monitor, StoreIndexDescriptor descriptor,
             IndexSamplingConfig samplingConfig ) throws IOException
     {

community/kernel/src/main/java/org/neo4j/kernel/impl/index/schema/GenericNativeIndexPopulator.java

@@ -21,7 +21,6 @@
 
 import java.io.File;
 
-import org.neo4j.index.internal.gbptree.Layout;
 import org.neo4j.io.fs.FileSystemAbstraction;
 import org.neo4j.io.pagecache.PageCache;
 import org.neo4j.kernel.api.index.IndexProvider;
@@ -31,7 +30,7 @@
 
 class GenericNativeIndexPopulator extends NativeIndexPopulator<CompositeGenericKey,NativeIndexValue>
 {
-    GenericNativeIndexPopulator( PageCache pageCache, FileSystemAbstraction fs, File storeFile, Layout<CompositeGenericKey,NativeIndexValue> layout,
+    GenericNativeIndexPopulator( PageCache pageCache, FileSystemAbstraction fs, File storeFile, IndexLayout<CompositeGenericKey,NativeIndexValue> layout,
             IndexProvider.Monitor monitor, StoreIndexDescriptor descriptor, IndexSamplingConfig samplingConfig )
     {
         super( pageCache, fs, storeFile, layout, monitor, descriptor, samplingConfig );

community/kernel/src/main/java/org/neo4j/kernel/impl/index/schema/GenericNativeIndexProvider.java

@@ -23,7 +23,6 @@
 import java.io.IOException;
 
 import org.neo4j.graphdb.factory.GraphDatabaseSettings;
-import org.neo4j.index.internal.gbptree.Layout;
 import org.neo4j.index.internal.gbptree.RecoveryCleanupWorkCollector;
 import org.neo4j.internal.kernel.api.IndexCapability;
 import org.neo4j.internal.kernel.api.IndexOrder;
@@ -118,21 +117,21 @@ public GenericNativeIndexProvider( int priority, IndexDirectoryStructure.Factory
     }
 
     @Override
-    Layout<CompositeGenericKey,NativeIndexValue> layout( StoreIndexDescriptor descriptor )
+    IndexLayout<CompositeGenericKey,NativeIndexValue> layout( StoreIndexDescriptor descriptor )
     {
         int numberOfSlots = descriptor.properties().length;
         return new GenericLayout( numberOfSlots );
     }
 
     @Override
-    protected IndexPopulator newIndexPopulator( File storeFile, Layout<CompositeGenericKey,NativeIndexValue> layout, StoreIndexDescriptor descriptor,
+    protected IndexPopulator newIndexPopulator( File storeFile, IndexLayout<CompositeGenericKey,NativeIndexValue> layout, StoreIndexDescriptor descriptor,
             IndexSamplingConfig samplingConfig )
     {
         return new GenericNativeIndexPopulator( pageCache, fs, storeFile, layout, monitor, descriptor, samplingConfig );
     }
 
     @Override
-    protected IndexAccessor newIndexAccessor( File storeFile, Layout<CompositeGenericKey,NativeIndexValue> layout, StoreIndexDescriptor descriptor,
+    protected IndexAccessor newIndexAccessor( File storeFile, IndexLayout<CompositeGenericKey,NativeIndexValue> layout, StoreIndexDescriptor descriptor,
             IndexSamplingConfig samplingConfig ) throws IOException
     {
         return new GenericNativeIndexAccessor( pageCache, fs, storeFile, layout, recoveryCleanupWorkCollector, monitor, descriptor, samplingConfig );

community/kernel/src/main/java/org/neo4j/kernel/impl/index/schema/GenericNativeIndexReader.java

@@ -20,7 +20,6 @@
 package org.neo4j.kernel.impl.index.schema;
 
 import org.neo4j.index.internal.gbptree.GBPTree;
-import org.neo4j.index.internal.gbptree.Layout;
 import org.neo4j.internal.kernel.api.IndexOrder;
 import org.neo4j.internal.kernel.api.IndexQuery;
 import org.neo4j.internal.kernel.api.IndexQuery.ExactPredicate;
@@ -38,7 +37,7 @@
 
 class GenericNativeIndexReader extends NativeIndexReader<CompositeGenericKey,NativeIndexValue>
 {
-    GenericNativeIndexReader( GBPTree<CompositeGenericKey,NativeIndexValue> tree, Layout<CompositeGenericKey,NativeIndexValue> layout,
+    GenericNativeIndexReader( GBPTree<CompositeGenericKey,NativeIndexValue> tree, IndexLayout<CompositeGenericKey,NativeIndexValue> layout,
             IndexSamplingConfig samplingConfig, IndexDescriptor descriptor )
     {
         super( tree, layout, samplingConfig, descriptor );

community/kernel/src/main/java/org/neo4j/kernel/impl/index/schema/IndexLayout.java

@@ -22,7 +22,7 @@
 import org.neo4j.index.internal.gbptree.Layout;
 import org.neo4j.io.pagecache.PageCursor;
 
-abstract class IndexLayout<KEY extends NativeIndexKey<KEY>> extends Layout.Adapter<KEY,NativeIndexValue>
+abstract class IndexLayout<KEY extends NativeIndexKey<KEY>, VALUE extends NativeIndexValue> extends Layout.Adapter<KEY,VALUE>
 {
     private final long identifier;
     private final int majorVersion;
@@ -42,9 +42,9 @@ abstract class IndexLayout<KEY extends NativeIndexKey<KEY>> extends Layout.Adapt
     }
 
     @Override
-    public NativeIndexValue newValue()
+    public VALUE newValue()
     {
-        return NativeIndexValue.INSTANCE;
+        return (VALUE) NativeIndexValue.INSTANCE;
     }
 
     @Override

community/kernel/src/main/java/org/neo4j/kernel/impl/index/schema/LocalDateTimeLayout.java

@@ -25,7 +25,7 @@
 /**
  * {@link Layout} for local date times.
  */
-class LocalDateTimeLayout extends IndexLayout<LocalDateTimeIndexKey>
+class LocalDateTimeLayout extends IndexLayout<LocalDateTimeIndexKey,NativeIndexValue>
 {
     LocalDateTimeLayout()
     {

community/kernel/src/main/java/org/neo4j/kernel/impl/index/schema/LocalTimeLayout.java

@@ -25,7 +25,7 @@
 /**
  * {@link Layout} for local times.
  */
-class LocalTimeLayout extends IndexLayout<LocalTimeIndexKey>
+class LocalTimeLayout extends IndexLayout<LocalTimeIndexKey,NativeIndexValue>
 {
     LocalTimeLayout()
     {

community/kernel/src/main/java/org/neo4j/kernel/impl/index/schema/NativeIndex.java

@@ -25,7 +25,6 @@
 import java.util.function.Consumer;
 
 import org.neo4j.index.internal.gbptree.GBPTree;
-import org.neo4j.index.internal.gbptree.Layout;
 import org.neo4j.index.internal.gbptree.RecoveryCleanupWorkCollector;
 import org.neo4j.io.IOUtils;
 import org.neo4j.io.fs.FileSystemAbstraction;
@@ -41,14 +40,14 @@ abstract class NativeIndex<KEY extends NativeIndexKey<KEY>, VALUE extends Native
 {
     final PageCache pageCache;
     final File storeFile;
-    final Layout<KEY,VALUE> layout;
+    final IndexLayout<KEY,VALUE> layout;
     final FileSystemAbstraction fileSystem;
     final IndexDescriptor descriptor;
     private final IndexProvider.Monitor monitor;
 
     protected GBPTree<KEY,VALUE> tree;
 
-    NativeIndex( PageCache pageCache, FileSystemAbstraction fs, File storeFile, Layout<KEY,VALUE> layout, IndexProvider.Monitor monitor,
+    NativeIndex( PageCache pageCache, FileSystemAbstraction fs, File storeFile, IndexLayout<KEY,VALUE> layout, IndexProvider.Monitor monitor,
             StoreIndexDescriptor descriptor )
     {
         this.pageCache = pageCache;

community/kernel/src/main/java/org/neo4j/kernel/impl/index/schema/NativeIndexAccessor.java

@@ -25,7 +25,6 @@
 
 import org.neo4j.graphdb.ResourceIterator;
 import org.neo4j.helpers.collection.BoundedIterable;
-import org.neo4j.index.internal.gbptree.Layout;
 import org.neo4j.index.internal.gbptree.RecoveryCleanupWorkCollector;
 import org.neo4j.io.fs.FileSystemAbstraction;
 import org.neo4j.io.pagecache.IOLimiter;
@@ -48,7 +47,7 @@ public abstract class NativeIndexAccessor<KEY extends NativeIndexKey<KEY>, VALUE
     private final NativeIndexUpdater<KEY,VALUE> singleUpdater;
     final IndexSamplingConfig samplingConfig;
 
-    NativeIndexAccessor( PageCache pageCache, FileSystemAbstraction fs, File storeFile, Layout<KEY,VALUE> layout,
+    NativeIndexAccessor( PageCache pageCache, FileSystemAbstraction fs, File storeFile, IndexLayout<KEY,VALUE> layout,
             RecoveryCleanupWorkCollector recoveryCleanupWorkCollector, IndexProvider.Monitor monitor, StoreIndexDescriptor descriptor,
             IndexSamplingConfig samplingConfig ) throws IOException
     {

community/kernel/src/main/java/org/neo4j/kernel/impl/index/schema/NativeIndexPopulator.java

@@ -32,7 +32,6 @@
 import java.util.function.Consumer;
 
 import org.neo4j.index.internal.gbptree.GBPTree;
-import org.neo4j.index.internal.gbptree.Layout;
 import org.neo4j.index.internal.gbptree.RecoveryCleanupWorkCollector;
 import org.neo4j.index.internal.gbptree.Writer;
 import org.neo4j.io.fs.FileSystemAbstraction;
@@ -45,9 +44,7 @@
 import org.neo4j.kernel.api.index.IndexProvider;
 import org.neo4j.kernel.api.index.IndexUpdater;
 import org.neo4j.kernel.api.index.NodePropertyAccessor;
-import org.neo4j.kernel.impl.api.index.sampling.DefaultNonUniqueIndexSampler;
 import org.neo4j.kernel.impl.api.index.sampling.IndexSamplingConfig;
-import org.neo4j.kernel.impl.api.index.sampling.NonUniqueIndexSampler;
 import org.neo4j.kernel.impl.api.index.sampling.UniqueIndexSampler;
 import org.neo4j.storageengine.api.schema.IndexReader;
 import org.neo4j.storageengine.api.schema.IndexSample;
@@ -75,7 +72,6 @@ public abstract class NativeIndexPopulator<KEY extends NativeIndexKey<KEY>, VALU
     private final KEY treeKey;
     private final VALUE treeValue;
     private final UniqueIndexSampler uniqueSampler;
-    private final NonUniqueIndexSampler nonUniqueSampler;
     final IndexSamplingConfig samplingConfig;
 
     private WorkSync<IndexUpdateApply<KEY,VALUE>,IndexUpdateWork<KEY,VALUE>> additionsWorkSync;
@@ -85,7 +81,7 @@ public abstract class NativeIndexPopulator<KEY extends NativeIndexKey<KEY>, VALU
     private boolean dropped;
     private boolean closed;
 
-    NativeIndexPopulator( PageCache pageCache, FileSystemAbstraction fs, File storeFile, Layout<KEY,VALUE> layout, IndexProvider.Monitor monitor,
+    NativeIndexPopulator( PageCache pageCache, FileSystemAbstraction fs, File storeFile, IndexLayout<KEY,VALUE> layout, IndexProvider.Monitor monitor,
                                 StoreIndexDescriptor descriptor, IndexSamplingConfig samplingConfig )
     {
         super( pageCache, fs, storeFile, layout, monitor, descriptor );
@@ -96,11 +92,9 @@ public abstract class NativeIndexPopulator<KEY extends NativeIndexKey<KEY>, VALU
         {
         case GENERAL:
             uniqueSampler = null;
-            nonUniqueSampler = new DefaultNonUniqueIndexSampler( samplingConfig.sampleSizeLimit() );
             break;
         case UNIQUE:
             uniqueSampler = new UniqueIndexSampler();
-            nonUniqueSampler = null;
             break;
         default:
             throw new IllegalArgumentException( "Unexpected index type " + descriptor.type() );
@@ -366,7 +360,7 @@ public void includeSample( IndexEntryUpdate<?> update )
         switch ( descriptor.type() )
         {
         case GENERAL:
-            updateNonUniqueSample( update );
+            // Don't do anything here, we'll do a scan in the end instead
             break;
         case UNIQUE:
             updateUniqueSample( update );
@@ -393,33 +387,13 @@ private void updateUniqueSample( IndexEntryUpdate<?> update )
         }
     }
 
-    private void updateNonUniqueSample( IndexEntryUpdate<?> update )
-    {
-        String encodedValues = SamplingUtil.encodedStringValuesForSampling( (Object[]) update.values() );
-        switch ( update.updateMode() )
-        {
-        case ADDED:
-            nonUniqueSampler.include( encodedValues );
-            break;
-        case REMOVED:
-            nonUniqueSampler.exclude( encodedValues );
-            break;
-        case CHANGED:
-            nonUniqueSampler.exclude( SamplingUtil.encodedStringValuesForSampling( (Object[]) update.beforeValues() ) );
-            nonUniqueSampler.include( encodedValues );
-            break;
-        default:
-            throw new IllegalArgumentException( "Unsupported update mode type:" + update.updateMode() );
-        }
-    }
-
     @Override
     public IndexSample sampleResult()
     {
         switch ( descriptor.type() )
         {
         case GENERAL:
-            return nonUniqueSampler.result();
+            return new FullScanNonUniqueIndexSampler<>( tree, layout ).result();
         case UNIQUE:
             return uniqueSampler.result();
         default:

community/kernel/src/main/java/org/neo4j/kernel/impl/index/schema/NativeIndexProvider.java

@@ -23,7 +23,6 @@
 import java.io.IOException;
 
 import org.neo4j.index.internal.gbptree.GBPTree;
-import org.neo4j.index.internal.gbptree.Layout;
 import org.neo4j.index.internal.gbptree.MetadataMismatchException;
 import org.neo4j.index.internal.gbptree.RecoveryCleanupWorkCollector;
 import org.neo4j.internal.kernel.api.InternalIndexState;
@@ -63,7 +62,7 @@ protected NativeIndexProvider( IndexProviderDescriptor descriptor, int priority,
         this.readOnly = readOnly;
     }
 
-    abstract Layout<KEY,VALUE> layout( StoreIndexDescriptor descriptor );
+    abstract IndexLayout<KEY,VALUE> layout( StoreIndexDescriptor descriptor );
 
     @Override
     public IndexPopulator getPopulator( StoreIndexDescriptor descriptor, IndexSamplingConfig samplingConfig )
@@ -77,7 +76,7 @@ public IndexPopulator getPopulator( StoreIndexDescriptor descriptor, IndexSampli
         return newIndexPopulator( storeFile, layout( descriptor ), descriptor, samplingConfig );
     }
 
-    protected abstract IndexPopulator newIndexPopulator( File storeFile, Layout<KEY,VALUE> layout, StoreIndexDescriptor descriptor,
+    protected abstract IndexPopulator newIndexPopulator( File storeFile, IndexLayout<KEY,VALUE> layout, StoreIndexDescriptor descriptor,
             IndexSamplingConfig samplingConfig );
 
     @Override
@@ -87,7 +86,7 @@ public IndexAccessor getOnlineAccessor( StoreIndexDescriptor descriptor, IndexSa
         return newIndexAccessor( storeFile, layout( descriptor ), descriptor, samplingConfig );
     }
 
-    protected abstract IndexAccessor newIndexAccessor( File storeFile, Layout<KEY,VALUE> layout, StoreIndexDescriptor descriptor,
+    protected abstract IndexAccessor newIndexAccessor( File storeFile, IndexLayout<KEY,VALUE> layout, StoreIndexDescriptor descriptor,
             IndexSamplingConfig samplingConfig ) throws IOException;
 
     @Override