Minor search controller changes (#36479)

This commit contains a few minor changes to our search code:

- adjust the visibility of a couple of methods in our search code to package private from public or protected.
- make some of the `SearchPhaseController` methods static where possible
- rename one of the `SearchPhaseController#reducedQueryPhase` methods (used only for scroll requests) to `reducedScrollQueryPhase` without the `isScrollRequest` argument which was always set to `true`
- replace leniency in `SearchPhaseController#setShardIndex` with an assert to make sure that we never set the shard index twice
- remove two null checks where the checked field can never be null
- resolve an unchecked warning
- replace `List#toArray` invocation that creates an array providing the true size with array creation of length 0
- correct a couple of typos in comments

Author: javanna

server/src/main/java/org/elasticsearch/action/search/AbstractSearchAsyncAction.java

@@ -72,7 +72,7 @@ abstract class AbstractSearchAsyncAction<Result extends SearchPhaseResult> exten
     private final TransportSearchAction.SearchTimeProvider timeProvider;
     private final SearchResponse.Clusters clusters;
 
-    protected AbstractSearchAsyncAction(String name, Logger logger, SearchTransportService searchTransportService,
+    AbstractSearchAsyncAction(String name, Logger logger, SearchTransportService searchTransportService,
                                         BiFunction<String, String, Transport.Connection> nodeIdToConnection,
                                         Map<String, AliasFilter> aliasFilter, Map<String, Float> concreteIndexBoosts,
                                         Map<String, Set<String>> indexRoutings,

server/src/main/java/org/elasticsearch/action/search/InitialSearchPhase.java

@@ -42,7 +42,7 @@
  * and collect the results. If a shard request returns a failure this class handles the advance to the next replica of the shard until
  * the shards replica iterator is exhausted. Each shard is referenced by position in the {@link GroupShardsIterator} which is later
  * referred to as the {@code shardIndex}.
- * The fan out and collect algorithm is traditionally used as the initial phase which can either be a query execution or collection
+ * The fan out and collect algorithm is traditionally used as the initial phase which can either be a query execution or collection of
  * distributed frequencies
  */
 abstract class InitialSearchPhase<FirstResult extends SearchPhaseResult> extends SearchPhase {
@@ -261,7 +261,6 @@ private void successfulShardExecution(SearchShardIterator shardsIt) {
         }
     }
 
-
     /**
      * Executed once all shard results have been received and processed
      * @see #onShardFailure(int, SearchShardTarget, Exception)
@@ -301,7 +300,7 @@ protected abstract void executePhaseOnShard(SearchShardIterator shardIt, ShardRo
     abstract static class SearchPhaseResults<Result extends SearchPhaseResult> {
         private final int numShards;
 
-        protected SearchPhaseResults(int numShards) {
+        SearchPhaseResults(int numShards) {
             this.numShards = numShards;
         }
         /**

server/src/main/java/org/elasticsearch/action/search/SearchPhaseController.java

@@ -21,7 +21,6 @@
 
 import com.carrotsearch.hppc.IntArrayList;
 import com.carrotsearch.hppc.ObjectObjectHashMap;
-
 import org.apache.lucene.index.Term;
 import org.apache.lucene.search.CollectionStatistics;
 import org.apache.lucene.search.FieldDoc;
@@ -151,8 +150,8 @@ private static long optionalSum(long left, long right) {
      * @param from the offset into the search results top docs
      * @param size the number of hits to return from the merged top docs
      */
-    public SortedTopDocs sortDocs(boolean ignoreFrom, Collection<? extends SearchPhaseResult> results,
-                               final Collection<TopDocs> bufferedTopDocs, final TopDocsStats topDocsStats, int from, int size) {
+    static SortedTopDocs sortDocs(boolean ignoreFrom, Collection<? extends SearchPhaseResult> results,
+                                  final Collection<TopDocs> bufferedTopDocs, final TopDocsStats topDocsStats, int from, int size) {
         if (results.isEmpty()) {
             return SortedTopDocs.EMPTY;
         }
@@ -210,7 +209,7 @@ public SortedTopDocs sortDocs(boolean ignoreFrom, Collection<? extends SearchPha
             }
             final boolean isSortedByField;
             final SortField[] sortFields;
-            if (mergedTopDocs != null && mergedTopDocs instanceof TopFieldDocs) {
+            if (mergedTopDocs instanceof TopFieldDocs) {
                 TopFieldDocs fieldDocs = (TopFieldDocs) mergedTopDocs;
                 isSortedByField = (fieldDocs instanceof CollapseTopFieldDocs &&
                     fieldDocs.fields.length == 1 && fieldDocs.fields[0].getType() == SortField.Type.SCORE) == false;
@@ -226,11 +225,10 @@ public SortedTopDocs sortDocs(boolean ignoreFrom, Collection<? extends SearchPha
         }
     }
 
-    TopDocs mergeTopDocs(Collection<TopDocs> results, int topN, int from) {
+    static TopDocs mergeTopDocs(Collection<TopDocs> results, int topN, int from) {
         if (results.isEmpty()) {
             return null;
         }
-        assert results.isEmpty() == false;
         final boolean setShardIndex = false;
         final TopDocs topDocs = results.stream().findFirst().get();
         final TopDocs mergedTopDocs;
@@ -255,12 +253,8 @@ TopDocs mergeTopDocs(Collection<TopDocs> results, int topN, int from) {
     }
 
     private static void setShardIndex(TopDocs topDocs, int shardIndex) {
+        assert topDocs.scoreDocs.length == 0 || topDocs.scoreDocs[0].shardIndex == -1 : "shardIndex is already set";
         for (ScoreDoc doc : topDocs.scoreDocs) {
-            if (doc.shardIndex != -1) {
-                // once there is a single shard index initialized all others will be initialized too
-                // there are many asserts down in lucene land that this is actually true. we can shortcut it here.
-                return;
-            }
             doc.shardIndex = shardIndex;
         }
     }
@@ -279,7 +273,6 @@ public ScoreDoc[] getLastEmittedDocPerShard(ReducedQueryPhase reducedQueryPhase,
             }
         }
         return lastEmittedDocPerShard;
-
     }
 
     /**
@@ -396,16 +389,15 @@ private SearchHits getHits(ReducedQueryPhase reducedQueryPhase, boolean ignoreFr
                 hits.add(searchHit);
             }
         }
-        return new SearchHits(hits.toArray(new SearchHit[hits.size()]), reducedQueryPhase.totalHits,
-            reducedQueryPhase.maxScore);
+        return new SearchHits(hits.toArray(new SearchHit[0]), reducedQueryPhase.totalHits, reducedQueryPhase.maxScore);
     }
 
     /**
      * Reduces the given query results and consumes all aggregations and profile results.
      * @param queryResults a list of non-null query shard results
      */
-    public ReducedQueryPhase reducedQueryPhase(Collection<? extends SearchPhaseResult> queryResults, boolean isScrollRequest) {
-        return reducedQueryPhase(queryResults, isScrollRequest, true);
+    public ReducedQueryPhase reducedScrollQueryPhase(Collection<? extends SearchPhaseResult> queryResults) {
+        return reducedQueryPhase(queryResults, true, true);
     }
 
     /**
@@ -417,7 +409,6 @@ public ReducedQueryPhase reducedQueryPhase(Collection<? extends SearchPhaseResul
         return reducedQueryPhase(queryResults, null, new ArrayList<>(), new TopDocsStats(trackTotalHits), 0, isScrollRequest);
     }
 
-
     /**
      * Reduces the given query results and consumes all aggregations and profile results.
      * @param queryResults a list of non-null query shard results
@@ -500,14 +491,12 @@ private ReducedQueryPhase reducedQueryPhase(Collection<? extends SearchPhaseResu
         final InternalAggregations aggregations = aggregationsList.isEmpty() ? null : reduceAggs(aggregationsList,
             firstResult.pipelineAggregators(), reduceContext);
         final SearchProfileShardResults shardResults = profileResults.isEmpty() ? null : new SearchProfileShardResults(profileResults);
-        final SortedTopDocs scoreDocs = this.sortDocs(isScrollRequest, queryResults, bufferedTopDocs, topDocsStats, from, size);
+        final SortedTopDocs scoreDocs = sortDocs(isScrollRequest, queryResults, bufferedTopDocs, topDocsStats, from, size);
         return new ReducedQueryPhase(topDocsStats.totalHits, topDocsStats.fetchHits, topDocsStats.maxScore,
             timedOut, terminatedEarly, suggest, aggregations, shardResults, scoreDocs.scoreDocs, scoreDocs.sortFields,
-            firstResult != null ? firstResult.sortValueFormats() : null,
-            numReducePhases, scoreDocs.isSortedByField, size, from, firstResult == null);
+            firstResult.sortValueFormats(), numReducePhases, scoreDocs.isSortedByField, size, from, false);
     }
 
-
     /**
      * Performs an intermediate reduce phase on the aggregations. For instance with this reduce phase never prune information
      * that relevant for the final reduce step. For final reduce see {@link #reduceAggs(List, List, ReduceContext)}
@@ -518,7 +507,7 @@ private InternalAggregations reduceAggsIncrementally(List<InternalAggregations>
             null, reduceContext);
     }
 
-    private InternalAggregations reduceAggs(List<InternalAggregations> aggregationsList,
+    private static InternalAggregations reduceAggs(List<InternalAggregations> aggregationsList,
                                             List<SiblingPipelineAggregator> pipelineAggregators, ReduceContext reduceContext) {
         InternalAggregations aggregations = InternalAggregations.reduce(aggregationsList, reduceContext);
         if (pipelineAggregators != null) {
@@ -649,7 +638,6 @@ private QueryPhaseResultConsumer(SearchPhaseController controller, int expectedR
             this.hasTopDocs = hasTopDocs;
             this.hasAggs = hasAggs;
             this.bufferSize = bufferSize;
-
         }
 
         @Override
@@ -667,7 +655,7 @@ private synchronized void consumeInternal(QuerySearchResult querySearchResult) {
                     aggsBuffer[0] = reducedAggs;
                 }
                 if (hasTopDocs) {
-                    TopDocs reducedTopDocs = controller.mergeTopDocs(Arrays.asList(topDocsBuffer),
+                    TopDocs reducedTopDocs = mergeTopDocs(Arrays.asList(topDocsBuffer),
                         // we have to merge here in the same way we collect on a shard
                         querySearchResult.from() + querySearchResult.size(), 0);
                     Arrays.fill(topDocsBuffer, null);
@@ -683,7 +671,7 @@ private synchronized void consumeInternal(QuerySearchResult querySearchResult) {
             if (hasTopDocs) {
                 final TopDocs topDocs = querySearchResult.consumeTopDocs(); // can't be null
                 topDocsStats.add(topDocs);
-                SearchPhaseController.setShardIndex(topDocs, querySearchResult.getShardIndex());
+                setShardIndex(topDocs, querySearchResult.getShardIndex());
                 topDocsBuffer[i] = topDocs;
             }
         }
@@ -696,7 +684,6 @@ private synchronized List<TopDocs> getRemainingTopDocs() {
             return hasTopDocs ? Arrays.asList(topDocsBuffer).subList(0, index) : null;
         }
 
-
         @Override
         public ReducedQueryPhase reduce() {
             return controller.reducedQueryPhase(results.asList(), getRemainingAggs(), getRemainingTopDocs(), topDocsStats,
@@ -730,7 +717,7 @@ InitialSearchPhase.ArraySearchPhaseResults<SearchPhaseResult> newSearchPhaseResu
                 return new QueryPhaseResultConsumer(this, numShards, request.getBatchedReduceSize(), hasTopDocs, hasAggs);
             }
         }
-        return new InitialSearchPhase.ArraySearchPhaseResults(numShards) {
+        return new InitialSearchPhase.ArraySearchPhaseResults<SearchPhaseResult>(numShards) {
             @Override
             public ReducedQueryPhase reduce() {
                 return reducedQueryPhase(results.asList(), isScrollRequest, trackTotalHits);

server/src/main/java/org/elasticsearch/action/search/SearchScrollQueryAndFetchAsyncAction.java

@@ -52,7 +52,7 @@ protected void executeInitialPhase(Transport.Connection connection, InternalScro
 
     @Override
     protected SearchPhase moveToNextPhase(BiFunction<String, String, DiscoveryNode> clusterNodeLookup) {
-        return sendResponsePhase(searchPhaseController.reducedQueryPhase(queryFetchResults.asList(), true), queryFetchResults);
+        return sendResponsePhase(searchPhaseController.reducedScrollQueryPhase(queryFetchResults.asList()), queryFetchResults);
     }
 
     @Override

server/src/main/java/org/elasticsearch/action/search/SearchScrollQueryThenFetchAsyncAction.java

@@ -69,8 +69,8 @@ protected SearchPhase moveToNextPhase(BiFunction<String, String, DiscoveryNode>
         return new SearchPhase("fetch") {
             @Override
             public void run() throws IOException {
-                final SearchPhaseController.ReducedQueryPhase reducedQueryPhase = searchPhaseController.reducedQueryPhase(
-                    queryResults.asList(), true);
+                final SearchPhaseController.ReducedQueryPhase reducedQueryPhase = searchPhaseController.reducedScrollQueryPhase(
+                    queryResults.asList());
                 if (reducedQueryPhase.scoreDocs.length == 0) {
                     sendResponse(reducedQueryPhase, fetchResults);
                     return;

server/src/main/java/org/elasticsearch/search/SearchPhaseResult.java

@@ -25,7 +25,7 @@
 import org.elasticsearch.transport.TransportResponse;
 
 /**
- * This class is a base class for all search releated results. It contains the shard target it
+ * This class is a base class for all search related results. It contains the shard target it
  * was executed against, a shard index used to reference the result on the coordinating node
  * and a request ID that is used to reference the request context on the executing node. The
  * request ID is particularly important since it is used to reference and maintain a context

server/src/main/java/org/elasticsearch/search/aggregations/metrics/tophits/InternalTopHits.java

@@ -63,7 +63,6 @@ public InternalTopHits(StreamInput in) throws IOException {
         from = in.readVInt();
         size = in.readVInt();
         topDocs = Lucene.readTopDocs(in);
-        assert topDocs != null;
         searchHits = SearchHits.readSearchHits(in);
     }
 

server/src/test/java/org/elasticsearch/action/search/SearchPhaseControllerTests.java

@@ -27,6 +27,8 @@
 import org.elasticsearch.common.util.concurrent.AtomicArray;
 import org.elasticsearch.index.Index;
 import org.elasticsearch.search.DocValueFormat;
+import org.elasticsearch.search.SearchHit;
+import org.elasticsearch.search.SearchHits;
 import org.elasticsearch.search.SearchPhaseResult;
 import org.elasticsearch.search.SearchShardTarget;
 import org.elasticsearch.search.aggregations.AggregationBuilders;
@@ -35,8 +37,6 @@
 import org.elasticsearch.search.aggregations.metrics.max.InternalMax;
 import org.elasticsearch.search.builder.SearchSourceBuilder;
 import org.elasticsearch.search.fetch.FetchSearchResult;
-import org.elasticsearch.search.SearchHit;
-import org.elasticsearch.search.SearchHits;
 import org.elasticsearch.search.internal.InternalSearchResponse;
 import org.elasticsearch.search.query.QuerySearchResult;
 import org.elasticsearch.search.suggest.Suggest;
@@ -70,7 +70,7 @@ public void setup() {
             (b) -> new InternalAggregation.ReduceContext(BigArrays.NON_RECYCLING_INSTANCE, null, b));
     }
 
-    public void testSort() throws Exception {
+    public void testSort() {
         List<CompletionSuggestion> suggestions = new ArrayList<>();
         for (int i = 0; i < randomIntBetween(1, 5); i++) {
             suggestions.add(new CompletionSuggestion(randomAlphaOfLength(randomIntBetween(1, 5)), randomIntBetween(1, 20), false));
@@ -85,7 +85,7 @@ public void testSort() throws Exception {
             size = first.get().queryResult().size();
         }
         int accumulatedLength = Math.min(queryResultSize, getTotalQueryHits(results));
-        ScoreDoc[] sortedDocs = searchPhaseController.sortDocs(true, results.asList(), null, new SearchPhaseController.TopDocsStats(),
+        ScoreDoc[] sortedDocs = SearchPhaseController.sortDocs(true, results.asList(), null, new SearchPhaseController.TopDocsStats(),
             from, size)
             .scoreDocs;
         for (Suggest.Suggestion<?> suggestion : reducedSuggest(results)) {
@@ -110,12 +110,12 @@ public void testSortIsIdempotent() throws Exception {
             size = first.get().queryResult().size();
         }
         SearchPhaseController.TopDocsStats topDocsStats = new SearchPhaseController.TopDocsStats();
-        ScoreDoc[] sortedDocs = searchPhaseController.sortDocs(ignoreFrom, results.asList(), null, topDocsStats, from, size).scoreDocs;
+        ScoreDoc[] sortedDocs = SearchPhaseController.sortDocs(ignoreFrom, results.asList(), null, topDocsStats, from, size).scoreDocs;
 
         results = generateSeededQueryResults(randomSeed, nShards, Collections.emptyList(), queryResultSize,
             useConstantScore);
         SearchPhaseController.TopDocsStats topDocsStats2 = new SearchPhaseController.TopDocsStats();
-        ScoreDoc[] sortedDocs2 = searchPhaseController.sortDocs(ignoreFrom, results.asList(), null, topDocsStats2, from, size).scoreDocs;
+        ScoreDoc[] sortedDocs2 = SearchPhaseController.sortDocs(ignoreFrom, results.asList(), null, topDocsStats2, from, size).scoreDocs;
         assertEquals(sortedDocs.length, sortedDocs2.length);
         for (int i = 0; i < sortedDocs.length; i++) {
             assertEquals(sortedDocs[i].doc, sortedDocs2[i].doc);