Apache Lucene open-source search software
LUCENE-10517: Improve performance of SortedSetDV faceting by iterating on class types (#812)
Chris Hegarty3bcc40efe91d
diff --git a/lucene/facet/src/java/org/apache/lucene/facet/StringValueFacetCounts.java b/lucene/facet/src/java/org/apache/lucene/facet/StringValueFacetCounts.javaindex 6876d36c9a7..3486dbf66e1 100644--- a/lucene/facet/src/java/org/apache/lucene/facet/StringValueFacetCounts.java+++ b/lucene/facet/src/java/org/apache/lucene/facet/StringValueFacetCounts.java@@ -306,7 +306,12 @@ public class StringValueFacetCounts extends Facets {assert ordinalMap == null;LeafReaderContext context = leaves.get(0);- countOneSegment(docValues, context.ord, null, context.reader().getLiveDocs());+ Bits liveDocs = context.reader().getLiveDocs();+ if (liveDocs == null) {+ countOneSegmentNHLD(docValues, context.ord);+ } else {+ countOneSegment(docValues, context.ord, null, liveDocs);+ }} else {// Since we have more than one segment, we should have a non-null ordinalMap and docValues// should be a MultiSortedSetDocValues instance:@@ -318,7 +323,12 @@ public class StringValueFacetCounts extends Facets {for (int i = 0; i < numLeaves; i++) {LeafReaderContext context = leaves.get(i);- countOneSegment(multiValues.values[i], context.ord, null, context.reader().getLiveDocs());+ Bits liveDocs = context.reader().getLiveDocs();+ if (liveDocs == null) {+ countOneSegmentNHLD(multiValues.values[i], context.ord);+ } else {+ countOneSegment(multiValues.values[i], context.ord, null, liveDocs);+ }}}}@@ -339,7 +349,8 @@ public class StringValueFacetCounts extends Facets {// all doc values for this segment:DocIdSetIterator it;if (hits == null) {- it = (liveDocs != null) ? FacetUtils.liveDocsDISI(valuesIt, liveDocs) : valuesIt;+ assert liveDocs != null;+ it = FacetUtils.liveDocsDISI(valuesIt, liveDocs);} else {it = ConjunctionUtils.intersectIterators(Arrays.asList(hits.bits.iterator(), valuesIt));}@@ -440,6 +451,86 @@ public class StringValueFacetCounts extends Facets {}}+ // Variant of countOneSegment, that has No Hits or Live Docs+ private void countOneSegmentNHLD(SortedSetDocValues multiValues, int segmentOrd)+ throws IOException {++ // It's slightly more efficient to work against SortedDocValues if the field is actually+ // single-valued (see: LUCENE-5309)+ SortedDocValues singleValues = DocValues.unwrapSingleton(multiValues);++ if (ordinalMap == null) {+ // If there's no ordinal map we don't need to map segment ordinals to globals, so counting+ // is very straight-forward:+ if (singleValues != null) {+ for (int doc = singleValues.nextDoc();+ doc != DocIdSetIterator.NO_MORE_DOCS;+ doc = singleValues.nextDoc()) {+ increment(singleValues.ordValue());+ totalDocCount++;+ }+ } else {+ for (int doc = multiValues.nextDoc();+ doc != DocIdSetIterator.NO_MORE_DOCS;+ doc = multiValues.nextDoc()) {+ boolean countedDocInTotal = false;+ for (int term = (int) multiValues.nextOrd();+ term != SortedSetDocValues.NO_MORE_ORDS;+ term = (int) multiValues.nextOrd()) {+ increment(term);+ if (countedDocInTotal == false) {+ totalDocCount++;+ countedDocInTotal = true;+ }+ }+ }+ }+ } else {+ // We need to map segment ordinals to globals. We have two different approaches to this+ // depending on how many hits we have to count relative to how many unique doc val ordinals+ // there are in this segment:+ final LongValues ordMap = ordinalMap.getGlobalOrds(segmentOrd);+ int segmentCardinality = (int) multiValues.getValueCount();++ // First count in seg-ord space.+ // At this point, we're either counting all ordinals or our heuristic suggests that+ // we expect to visit a large percentage of the unique ordinals (lots of hits relative+ // to the segment cardinality), so we count the segment densely:+ final int[] segCounts = new int[segmentCardinality];+ if (singleValues != null) {+ for (int doc = singleValues.nextDoc();+ doc != DocIdSetIterator.NO_MORE_DOCS;+ doc = singleValues.nextDoc()) {+ segCounts[singleValues.ordValue()]++;+ totalDocCount++;+ }+ } else {+ for (int doc = multiValues.nextDoc();+ doc != DocIdSetIterator.NO_MORE_DOCS;+ doc = multiValues.nextDoc()) {+ boolean countedDocInTotal = false;+ for (int term = (int) multiValues.nextOrd();+ term != SortedSetDocValues.NO_MORE_ORDS;+ term = (int) multiValues.nextOrd()) {+ segCounts[term]++;+ if (countedDocInTotal == false) {+ totalDocCount++;+ countedDocInTotal = true;+ }+ }+ }+ }++ // Then, migrate to global ords:+ for (int ord = 0; ord < segmentCardinality; ord++) {+ int count = segCounts[ord];+ if (count != 0) {+ increment((int) ordMap.get(ord), count);+ }+ }+ }+ }+private void increment(int ordinal) {increment(ordinal, 1);}diff --git a/lucene/facet/src/java/org/apache/lucene/facet/sortedset/ConcurrentSortedSetDocValuesFacetCounts.java b/lucene/facet/src/java/org/apache/lucene/facet/sortedset/ConcurrentSortedSetDocValuesFacetCounts.javaindex f2ef14254e4..d28182bc36d 100644--- a/lucene/facet/src/java/org/apache/lucene/facet/sortedset/ConcurrentSortedSetDocValuesFacetCounts.java+++ b/lucene/facet/src/java/org/apache/lucene/facet/sortedset/ConcurrentSortedSetDocValuesFacetCounts.java@@ -258,15 +258,39 @@ public class ConcurrentSortedSetDocValuesFacetCounts extends Facets {if (hits != null && hits.totalHits < numSegOrds / 10) {// Remap every ord to global ord as we iterate:if (singleValues != null) {- for (int doc = it.nextDoc(); doc != DocIdSetIterator.NO_MORE_DOCS; doc = it.nextDoc()) {- counts.incrementAndGet((int) ordMap.get(singleValues.ordValue()));+ if (singleValues == it) {+ for (int doc = singleValues.nextDoc();+ doc != DocIdSetIterator.NO_MORE_DOCS;+ doc = singleValues.nextDoc()) {+ counts.incrementAndGet((int) ordMap.get(singleValues.ordValue()));+ }+ } else {+ for (int doc = it.nextDoc();+ doc != DocIdSetIterator.NO_MORE_DOCS;+ doc = it.nextDoc()) {+ counts.incrementAndGet((int) ordMap.get(singleValues.ordValue()));+ }}} else {- for (int doc = it.nextDoc(); doc != DocIdSetIterator.NO_MORE_DOCS; doc = it.nextDoc()) {- int term = (int) multiValues.nextOrd();- while (term != SortedSetDocValues.NO_MORE_ORDS) {- counts.incrementAndGet((int) ordMap.get(term));- term = (int) multiValues.nextOrd();+ if (multiValues == it) {+ for (int doc = multiValues.nextDoc();+ doc != DocIdSetIterator.NO_MORE_DOCS;+ doc = multiValues.nextDoc()) {+ for (int term = (int) multiValues.nextOrd();+ term != SortedSetDocValues.NO_MORE_ORDS;+ term = (int) multiValues.nextOrd()) {+ counts.incrementAndGet((int) ordMap.get(term));+ }+ }+ } else {+ for (int doc = it.nextDoc();+ doc != DocIdSetIterator.NO_MORE_DOCS;+ doc = it.nextDoc()) {+ for (int term = (int) multiValues.nextOrd();+ term != SortedSetDocValues.NO_MORE_ORDS;+ term = (int) multiValues.nextOrd()) {+ counts.incrementAndGet((int) ordMap.get(term));+ }}}}@@ -275,15 +299,39 @@ public class ConcurrentSortedSetDocValuesFacetCounts extends Facets {// First count in seg-ord space:final int[] segCounts = new int[numSegOrds];if (singleValues != null) {- for (int doc = it.nextDoc(); doc != DocIdSetIterator.NO_MORE_DOCS; doc = it.nextDoc()) {- segCounts[singleValues.ordValue()]++;+ if (singleValues == it) {+ for (int doc = singleValues.nextDoc();+ doc != DocIdSetIterator.NO_MORE_DOCS;+ doc = singleValues.nextDoc()) {+ segCounts[singleValues.ordValue()]++;+ }+ } else {+ for (int doc = it.nextDoc();+ doc != DocIdSetIterator.NO_MORE_DOCS;+ doc = it.nextDoc()) {+ segCounts[singleValues.ordValue()]++;+ }}} else {- for (int doc = it.nextDoc(); doc != DocIdSetIterator.NO_MORE_DOCS; doc = it.nextDoc()) {- int term = (int) multiValues.nextOrd();- while (term != SortedSetDocValues.NO_MORE_ORDS) {- segCounts[term]++;- term = (int) multiValues.nextOrd();+ if (multiValues == it) {+ for (int doc = multiValues.nextDoc();+ doc != DocIdSetIterator.NO_MORE_DOCS;+ doc = multiValues.nextDoc()) {+ for (int term = (int) multiValues.nextOrd();+ term != SortedSetDocValues.NO_MORE_ORDS;+ term = (int) multiValues.nextOrd()) {+ segCounts[term]++;+ }+ }+ } else {+ for (int doc = it.nextDoc();+ doc != DocIdSetIterator.NO_MORE_DOCS;+ doc = it.nextDoc()) {+ for (int term = (int) multiValues.nextOrd();+ term != SortedSetDocValues.NO_MORE_ORDS;+ term = (int) multiValues.nextOrd()) {+ segCounts[term]++;+ }}}}@@ -300,15 +348,35 @@ public class ConcurrentSortedSetDocValuesFacetCounts extends Facets {// No ord mapping (e.g., single segment index):// just aggregate directly into counts:if (singleValues != null) {- for (int doc = it.nextDoc(); doc != DocIdSetIterator.NO_MORE_DOCS; doc = it.nextDoc()) {- counts.incrementAndGet(singleValues.ordValue());+ if (singleValues == it) {+ for (int doc = singleValues.nextDoc();+ doc != DocIdSetIterator.NO_MORE_DOCS;+ doc = singleValues.nextDoc()) {+ counts.incrementAndGet(singleValues.ordValue());+ }+ } else {+ for (int doc = it.nextDoc(); doc != DocIdSetIterator.NO_MORE_DOCS; doc = it.nextDoc()) {+ counts.incrementAndGet(singleValues.ordValue());+ }}} else {- for (int doc = it.nextDoc(); doc != DocIdSetIterator.NO_MORE_DOCS; doc = it.nextDoc()) {- int term = (int) multiValues.nextOrd();- while (term != SortedSetDocValues.NO_MORE_ORDS) {- counts.incrementAndGet(term);- term = (int) multiValues.nextOrd();+ if (multiValues == it) {+ for (int doc = multiValues.nextDoc();+ doc != DocIdSetIterator.NO_MORE_DOCS;+ doc = multiValues.nextDoc()) {+ for (int term = (int) multiValues.nextOrd();+ term != SortedSetDocValues.NO_MORE_ORDS;+ term = (int) multiValues.nextOrd()) {+ counts.incrementAndGet(term);+ }+ }+ } else {+ for (int doc = it.nextDoc(); doc != DocIdSetIterator.NO_MORE_DOCS; doc = it.nextDoc()) {+ for (int term = (int) multiValues.nextOrd();+ term != SortedSetDocValues.NO_MORE_ORDS;+ term = (int) multiValues.nextOrd()) {+ counts.incrementAndGet(term);+ }}}}diff --git a/lucene/facet/src/java/org/apache/lucene/facet/sortedset/SortedSetDocValuesFacetCounts.java b/lucene/facet/src/java/org/apache/lucene/facet/sortedset/SortedSetDocValuesFacetCounts.javaindex e97ead7019e..53e11b47e0a 100644--- a/lucene/facet/src/java/org/apache/lucene/facet/sortedset/SortedSetDocValuesFacetCounts.java+++ b/lucene/facet/src/java/org/apache/lucene/facet/sortedset/SortedSetDocValuesFacetCounts.java@@ -278,6 +278,83 @@ public class SortedSetDocValuesFacetCounts extends Facets {return childOrdsResult.dimCount;}+ // Variant of countOneSegment, that has No Hits or Live Docs+ private void countOneSegmentNHLD(OrdinalMap ordinalMap, LeafReader reader, int segOrd)+ throws IOException {+ SortedSetDocValues multiValues = DocValues.getSortedSet(reader, field);+ if (multiValues == null) {+ // nothing to count+ return;+ }++ // It's slightly more efficient to work against SortedDocValues if the field is actually+ // single-valued (see: LUCENE-5309)+ SortedDocValues singleValues = DocValues.unwrapSingleton(multiValues);++ // TODO: yet another option is to count all segs+ // first, only in seg-ord space, and then do a+ // merge-sort-PQ in the end to only "resolve to+ // global" those seg ords that can compete, if we know+ // we just want top K? ie, this is the same algo+ // that'd be used for merging facets across shards+ // (distributed faceting). but this has much higher+ // temp ram req'ts (sum of number of ords across all+ // segs)+ if (ordinalMap != null) {+ final LongValues ordMap = ordinalMap.getGlobalOrds(segOrd);+ int numSegOrds = (int) multiValues.getValueCount();++ // First count in seg-ord space:+ final int[] segCounts = new int[numSegOrds];+ if (singleValues != null) {+ for (int doc = singleValues.nextDoc();+ doc != DocIdSetIterator.NO_MORE_DOCS;+ doc = singleValues.nextDoc()) {+ segCounts[singleValues.ordValue()]++;+ }+ } else {+ for (int doc = multiValues.nextDoc();+ doc != DocIdSetIterator.NO_MORE_DOCS;+ doc = multiValues.nextDoc()) {+ int term = (int) multiValues.nextOrd();+ while (term != SortedSetDocValues.NO_MORE_ORDS) {+ segCounts[term]++;+ term = (int) multiValues.nextOrd();+ }+ }+ }++ // Then, migrate to global ords:+ for (int ord = 0; ord < numSegOrds; ord++) {+ int count = segCounts[ord];+ if (count != 0) {+ // ordinalMap.getGlobalOrd(segOrd, ord));+ counts[(int) ordMap.get(ord)] += count;+ }+ }+ } else {+ // No ord mapping (e.g., single segment index):+ // just aggregate directly into counts:+ if (singleValues != null) {+ for (int doc = singleValues.nextDoc();+ doc != DocIdSetIterator.NO_MORE_DOCS;+ doc = singleValues.nextDoc()) {+ counts[singleValues.ordValue()]++;+ }+ } else {+ for (int doc = multiValues.nextDoc();+ doc != DocIdSetIterator.NO_MORE_DOCS;+ doc = multiValues.nextDoc()) {+ int term = (int) multiValues.nextOrd();+ while (term != SortedSetDocValues.NO_MORE_ORDS) {+ counts[term]++;+ term = (int) multiValues.nextOrd();+ }+ }+ }+ }+ }+private void countOneSegment(OrdinalMap ordinalMap, LeafReader reader, int segOrd, MatchingDocs hits, Bits liveDocs)throws IOException {@@ -294,7 +371,8 @@ public class SortedSetDocValuesFacetCounts extends Facets {DocIdSetIterator it;if (hits == null) {- it = (liveDocs != null) ? FacetUtils.liveDocsDISI(valuesIt, liveDocs) : valuesIt;+ assert liveDocs != null;+ it = FacetUtils.liveDocsDISI(valuesIt, liveDocs);} else {it = ConjunctionUtils.intersectIterators(Arrays.asList(hits.bits.iterator(), valuesIt));}@@ -420,8 +498,13 @@ public class SortedSetDocValuesFacetCounts extends Facets {for (LeafReaderContext context : state.getReader().leaves()) {- countOneSegment(- ordinalMap, context.reader(), context.ord, null, context.reader().getLiveDocs());+ Bits liveDocs = context.reader().getLiveDocs();+ if (liveDocs == null) {+ countOneSegmentNHLD(ordinalMap, context.reader(), context.ord);+ } else {+ countOneSegment(+ ordinalMap, context.reader(), context.ord, null, context.reader().getLiveDocs());+ }}}diff --git a/lucene/facet/src/java/org/apache/lucene/facet/taxonomy/FastTaxonomyFacetCounts.java b/lucene/facet/src/java/org/apache/lucene/facet/taxonomy/FastTaxonomyFacetCounts.javaindex 812a20c346e..82c3b2a3af7 100644--- a/lucene/facet/src/java/org/apache/lucene/facet/taxonomy/FastTaxonomyFacetCounts.java+++ b/lucene/facet/src/java/org/apache/lucene/facet/taxonomy/FastTaxonomyFacetCounts.java@@ -126,23 +126,41 @@ public class FastTaxonomyFacetCounts extends IntTaxonomyFacets {NumericDocValues singleValued = DocValues.unwrapSingleton(multiValued);if (singleValued != null) {- for (int doc = singleValued.nextDoc();- doc != DocIdSetIterator.NO_MORE_DOCS;- doc = singleValued.nextDoc()) {- if (liveDocs != null && liveDocs.get(doc) == false) {- continue;+ if (liveDocs == null) {+ for (int doc = singleValued.nextDoc();+ doc != DocIdSetIterator.NO_MORE_DOCS;+ doc = singleValued.nextDoc()) {+ values[(int) singleValued.longValue()]++;+ }+ } else {+ for (int doc = singleValued.nextDoc();+ doc != DocIdSetIterator.NO_MORE_DOCS;+ doc = singleValued.nextDoc()) {+ if (liveDocs.get(doc) == false) {+ continue;+ }+ values[(int) singleValued.longValue()]++;}- values[(int) singleValued.longValue()]++;}} else {- for (int doc = multiValued.nextDoc();- doc != DocIdSetIterator.NO_MORE_DOCS;- doc = multiValued.nextDoc()) {- if (liveDocs != null && liveDocs.get(doc) == false) {- continue;+ if (liveDocs == null) {+ for (int doc = multiValued.nextDoc();+ doc != DocIdSetIterator.NO_MORE_DOCS;+ doc = multiValued.nextDoc()) {+ for (int i = 0; i < multiValued.docValueCount(); i++) {+ values[(int) multiValued.nextValue()]++;+ }}- for (int i = 0; i < multiValued.docValueCount(); i++) {- values[(int) multiValued.nextValue()]++;+ } else {+ for (int doc = multiValued.nextDoc();+ doc != DocIdSetIterator.NO_MORE_DOCS;+ doc = multiValued.nextDoc()) {+ if (liveDocs.get(doc) == false) {+ continue;+ }+ for (int i = 0; i < multiValued.docValueCount(); i++) {+ values[(int) multiValued.nextValue()]++;+ }}}}