pnowojski commented on a change in pull request #6417: [FLINK-9913][runtime]
Improve output serialization only once in RecordWriter
URL: https://github.com/apache/flink/pull/6417#discussion_r209289603
##########
File path:
flink-runtime/src/test/java/org/apache/flink/runtime/io/network/api/writer/RecordWriterTest.java
##########
@@ -377,6 +388,86 @@ public void testBroadcastEmitBufferIndependence() throws
Exception {
assertEquals("Buffer 2 shares the same reader index as buffer
1", 0, buffer2.getReaderIndex());
}
+ /**
+ * Tests that records are broadcast via {@link ChannelSelector} and
+ * {@link RecordWriter#emit(IOReadableWritable)}.
+ */
+ @Test
+ public void testEmitRecordWithBroadcastPartitioner() throws Exception {
+ emitRecordWithBroadcastPartitionerOrBroadcastEmitRecord(false);
+ }
+
+ /**
+ * Tests that records are broadcast via {@link
RecordWriter#broadcastEmit(IOReadableWritable)}.
+ */
+ @Test
+ public void testBroadcastEmitRecord() throws Exception {
+ emitRecordWithBroadcastPartitionerOrBroadcastEmitRecord(true);
+ }
+
+ /**
+ * The results of emitting records via BroadcastPartitioner or
broadcasting records directly are the same,
+ * that is all the target channels can receive the whole outputs.
+ *
+ * @param isBroadcastEmit whether using {@link
RecordWriter#broadcastEmit(IOReadableWritable)} or not
+ */
+ private void
emitRecordWithBroadcastPartitionerOrBroadcastEmitRecord(boolean
isBroadcastEmit) throws Exception {
+ final int numChannels = 4;
+ final int bufferSize = 32;
+ final int numValues = 8;
+ final int serializationLength = 4;
+
+ @SuppressWarnings("unchecked")
+ final Queue<BufferConsumer>[] queues = new Queue[numChannels];
+ for (int i = 0; i < numChannels; i++) {
+ queues[i] = new ArrayDeque<>();
+ }
+
+ final TestPooledBufferProvider bufferProvider = new
TestPooledBufferProvider(Integer.MAX_VALUE, bufferSize);
+ final ResultPartitionWriter partitionWriter = new
CollectingPartitionWriter(queues, bufferProvider);
+ final RecordWriter<SerializationTestType> writer = new
RecordWriter<>(partitionWriter, new Broadcast<>());
+ final RecordDeserializer<SerializationTestType> deserializer =
new SpillingAdaptiveSpanningRecordDeserializer<>(
+ new String[]{ tempFolder.getRoot().getAbsolutePath() });
+
+ final ArrayDeque<SerializationTestType> serializedRecords = new
ArrayDeque<>();
+ final Iterable<SerializationTestType> records =
Util.randomRecords(numValues, SerializationTestTypeFactory.INT);
+ for (SerializationTestType record : records) {
+ serializedRecords.add(record);
+
+ if (isBroadcastEmit) {
+ writer.broadcastEmit(record);
+ } else {
+ writer.emit(record);
+ }
+ }
+
+ final int requiredBuffers = numValues / (bufferSize / (4 +
serializationLength));
+ for (int i = 0; i < numChannels; i++) {
Review comment:
can you somehow extract common logic of this method and
`SpanningRecordSerializationTest#testSerializationRoundTrip(Iterable<SerializationTestType>,
int, RecordSerializer<SerializationTestType>,
RecordDeserializer<SerializationTestType>)`? They share a lot of core.
----------------------------------------------------------------
This is an automated message from the Apache Git Service.
To respond to the message, please log on GitHub and use the
URL above to go to the specific comment.
For queries about this service, please contact Infrastructure at:
us...@infra.apache.org
With regards,
Apache Git Services
