[GitHub] [beam] mxm commented on a change in pull request #11874: Flink Batch Runner: optimize GBK translation for non-merging windows.

[GitBox <gi...@apache.org>]

mxm commented on a change in pull request #11874:
URL: https://github.com/apache/beam/pull/11874#discussion_r433295709



##########
File path: runners/flink/src/main/java/org/apache/beam/runners/flink/FlinkBatchTransformTranslators.java
##########
@@ -105,40 +111,53 @@
   // --------------------------------------------------------------------------------------------
 
   @SuppressWarnings("rawtypes")
-  private static final Map<String, FlinkBatchPipelineTranslator.BatchTransformTranslator>
+  private static final Map<String, List<FlinkBatchPipelineTranslator.BatchTransformTranslator>>
       TRANSLATORS = new HashMap<>();
 
-  static {
-    TRANSLATORS.put(PTransformTranslation.IMPULSE_TRANSFORM_URN, new ImpulseTranslatorBatch());
+  private static void registerTranslator(
+      String urn, FlinkBatchPipelineTranslator.BatchTransformTranslator<?> translator) {
+    if (!TRANSLATORS.containsKey(urn)) {
+      TRANSLATORS.put(urn, new ArrayList<>());
+    }
+    TRANSLATORS.get(urn).add(translator);
+  }
 
-    TRANSLATORS.put(
+  static {
+    registerTranslator(PTransformTranslation.IMPULSE_TRANSFORM_URN, new ImpulseTranslatorBatch());
+    registerTranslator(

Review comment:
       I think we can get rid of these helper calls and keep the existing code.

##########
File path: runners/flink/src/main/java/org/apache/beam/runners/flink/FlinkBatchTransformTranslators.java
##########
@@ -105,40 +111,53 @@
   // --------------------------------------------------------------------------------------------
 
   @SuppressWarnings("rawtypes")
-  private static final Map<String, FlinkBatchPipelineTranslator.BatchTransformTranslator>
+  private static final Map<String, List<FlinkBatchPipelineTranslator.BatchTransformTranslator>>

Review comment:
       We might want to use `Multimap`.

##########
File path: runners/flink/src/main/java/org/apache/beam/runners/flink/FlinkBatchTransformTranslators.java
##########
@@ -105,40 +111,53 @@
   // --------------------------------------------------------------------------------------------
 
   @SuppressWarnings("rawtypes")
-  private static final Map<String, FlinkBatchPipelineTranslator.BatchTransformTranslator>
+  private static final Map<String, List<FlinkBatchPipelineTranslator.BatchTransformTranslator>>
       TRANSLATORS = new HashMap<>();
 
-  static {
-    TRANSLATORS.put(PTransformTranslation.IMPULSE_TRANSFORM_URN, new ImpulseTranslatorBatch());
+  private static void registerTranslator(
+      String urn, FlinkBatchPipelineTranslator.BatchTransformTranslator<?> translator) {
+    if (!TRANSLATORS.containsKey(urn)) {
+      TRANSLATORS.put(urn, new ArrayList<>());

Review comment:
       Again, multimap would handle this.

##########
File path: runners/flink/src/main/java/org/apache/beam/runners/flink/translation/functions/FlinkNonMergingReduceFunction.java
##########
@@ -0,0 +1,112 @@
+/*
+ * Licensed to the Apache Software Foundation (ASF) under one
+ * or more contributor license agreements.  See the NOTICE file
+ * distributed with this work for additional information
+ * regarding copyright ownership.  The ASF licenses this file
+ * to you under the Apache License, Version 2.0 (the
+ * "License"); you may not use this file except in compliance
+ * with the License.  You may obtain a copy of the License at
+ *
+ *     http://www.apache.org/licenses/LICENSE-2.0
+ *
+ * Unless required by applicable law or agreed to in writing, software
+ * distributed under the License is distributed on an "AS IS" BASIS,
+ * WITHOUT WARRANTIES OR CONDITIONS OF ANY KIND, either express or implied.
+ * See the License for the specific language governing permissions and
+ * limitations under the License.
+ */
+package org.apache.beam.runners.flink.translation.functions;
+
+import java.util.ArrayList;
+import java.util.Iterator;
+import java.util.List;
+import java.util.Objects;
+import java.util.concurrent.atomic.AtomicBoolean;
+import org.apache.beam.sdk.transforms.windowing.BoundedWindow;
+import org.apache.beam.sdk.transforms.windowing.PaneInfo;
+import org.apache.beam.sdk.util.WindowedValue;
+import org.apache.beam.sdk.values.KV;
+import org.apache.beam.sdk.values.WindowingStrategy;
+import org.apache.beam.vendor.guava.v26_0_jre.com.google.common.collect.Iterables;
+import org.apache.beam.vendor.guava.v26_0_jre.com.google.common.collect.Iterators;
+import org.apache.beam.vendor.guava.v26_0_jre.com.google.common.collect.PeekingIterator;
+import org.apache.flink.api.common.functions.GroupReduceFunction;
+import org.apache.flink.util.Collector;
+import org.joda.time.Instant;
+
+/**
+ * Reduce function for non-merging GBK implementation. Implementation tries to return non-iterable
+ * results when possible, so we do not have to materialize all values for a single key in memory.
+ *
+ * @param <K> Key type.
+ * @param <InputT> Input type.
+ */
+public class FlinkNonMergingReduceFunction<K, InputT>
+    implements GroupReduceFunction<
+        WindowedValue<KV<K, InputT>>, WindowedValue<KV<K, Iterable<InputT>>>> {
+
+  private static class OnceIterable<T> implements Iterable<T> {
+
+    private final Iterator<T> iterator;
+
+    private final AtomicBoolean used = new AtomicBoolean(false);
+
+    OnceIterable(Iterator<T> iterator) {
+      this.iterator = iterator;
+    }
+
+    @Override
+    public Iterator<T> iterator() {
+      if (used.compareAndSet(false, true)) {
+        return iterator;
+      }
+      throw new IllegalStateException(
+          "GBK result is not re-iterable. You can enable re-iterations by setting '--reIterableGroupByKeyResult=true'.");

Review comment:
       ```suggestion
             "GBK result is not re-iterable. You can enable re-iterations by setting '--reIterableGroupByKeyResult'.");
   ```

##########
File path: runners/flink/src/main/java/org/apache/beam/runners/flink/FlinkBatchTranslationContext.java
##########
@@ -68,13 +71,18 @@ public FlinkBatchTranslationContext(ExecutionEnvironment env, PipelineOptions op
     this.danglingDataSets = new HashMap<>();
   }
 
+  void init(Pipeline pipeline) {
+    pipeline.traverseTopologically(countingPipelineVisitor);
+    pipeline.traverseTopologically(lookupPipelineVisitor);
+  }
+
   // ------------------------------------------------------------------------
 
-  public Map<PValue, DataSet<?>> getDanglingDataSets() {
+  Map<PValue, DataSet<?>> getDanglingDataSets() {

Review comment:
       Can you revert these? This makes integration hard for external Runners.




----------------------------------------------------------------
This is an automated message from the Apache Git Service.
To respond to the message, please log on to GitHub and use the
URL above to go to the specific comment.

For queries about this service, please contact Infrastructure at:
users@infra.apache.org