Github user aarondav commented on a diff in the pull request:
https://github.com/apache/spark/pull/2753#discussion_r19383720
--- Diff:
core/src/main/scala/org/apache/spark/network/netty/NettyBlockFetcher.scala ---
@@ -0,0 +1,92 @@
+/*
+ * 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.spark.network.netty
+
+import java.nio.ByteBuffer
+import java.util
+
+import org.apache.spark.Logging
+import org.apache.spark.network.BlockFetchingListener
+import org.apache.spark.serializer.Serializer
+import org.apache.spark.network.buffer.ManagedBuffer
+import org.apache.spark.network.client.{RpcResponseCallback,
ChunkReceivedCallback, SluiceClient}
+import org.apache.spark.storage.BlockId
+import org.apache.spark.util.Utils
+
+/**
+ * Responsible for holding the state for a request for a single set of
blocks. This assumes that
+ * the chunks will be returned in the same order as requested, and that
there will be exactly
+ * one chunk per block.
+ *
+ * Upon receipt of any block, the listener will be called back. Upon
failure part way through,
+ * the listener will receive a failure callback for each outstanding block.
+ */
+class NettyBlockFetcher(
+ serializer: Serializer,
+ client: SluiceClient,
+ blockIds: Seq[String],
+ listener: BlockFetchingListener)
+ extends Logging {
+
+ require(blockIds.nonEmpty)
+
+ val ser = serializer.newInstance()
+
+ var streamHandle: ShuffleStreamHandle = _
+
+ val chunkCallback = new ChunkReceivedCallback {
+ // On receipt of a chunk, pass it upwards as a block.
+ def onSuccess(chunkIndex: Int, buffer: ManagedBuffer): Unit =
Utils.logUncaughtExceptions {
+ buffer.retain()
+ listener.onBlockFetchSuccess(blockIds(chunkIndex), buffer)
+ }
+
+ // On receipt of a failure, fail every block from chunkIndex onwards.
+ def onFailure(chunkIndex: Int, e: Throwable): Unit = {
+ blockIds.drop(chunkIndex).foreach { blockId =>
+ listener.onBlockFetchFailure(blockId, e);
+ }
+ }
+ }
+
+ // Send the RPC to open the given set of blocks. This will return a
ShuffleStreamHandle.
+
client.sendRpc(ser.serialize(OpenBlocks(blockIds.map(BlockId.apply))).array(),
--- End diff --
Ahh, gotcha, totally misunderstood your initial concern. Sure, I can put it
in a method. This is not a common pattern in Spark because Scala makes it so
convenient to make constructors complicated, but I do agree it's better design.
The reason it's a class onto itself is because there is state related to
the fetching of a single set of blocks. Previously, it was more complicated
because we were doing chunking over large blocks -- I got rid of that to
minimize changes, but I don't want all that logic and state to be buried inside
the NettyBlockTransferService code, especially once we add it back in.
---
If your project is set up for it, you can reply to this email and have your
reply appear on GitHub as well. If your project does not have this feature
enabled and wishes so, or if the feature is enabled but not working, please
contact infrastructure at [email protected] or file a JIRA ticket
with INFRA.
---
---------------------------------------------------------------------
To unsubscribe, e-mail: [email protected]
For additional commands, e-mail: [email protected]
