Github user mridulm commented on a diff in the pull request:
https://github.com/apache/spark/pull/151#discussion_r10638435
--- Diff:
core/src/main/scala/org/apache/spark/deploy/yarn/YarnAllocationHandlerMacro.scala
---
@@ -0,0 +1,46 @@
+/*
+ * 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.deploy.yarn
+
+import scala.language.experimental.macros
+import scala.reflect.macros.Context
+
+private[yarn] object YarnAllocationHandlerMacro {
+ def getAMResp(resp: Any): Any = macro getAMRespImpl
+
+ /**
+ * From Hadoop CDH 4.4.0+ (2.1.0-beta),
+ * AMResponse is merged into AllocateResponse,
+ * so we don't need to call getAMResponse(), just use AllocateResponse
directly.
+ * This macro will test the existence of AMResponse,
+ * and generate diffenert expressions.
+ *
+ * This macro now is only used in spark's alpha version of yarn api.
+ * It stays in the core project, for the two-stage compiling of
+ * the scala macro system.
+ */
+ def getAMRespImpl(c: Context)(resp: c.Expr[Any]) = {
+ try {
+ import c.universe._
+ c.mirror.staticClass("org.apache.hadoop.yarn.api.records.AMResponse")
+ c.Expr[Any](Apply(Select(resp.tree, newTermName("getAMResponse")),
List()))
+ } catch {
+ case _: Throwable => resp
+ }
--- End diff --
I would look at this as a very uncommon and temporary situation.
We would need to continue supporting the old api for only a little while
longer - after which we can remove the reflection stuff; and ideally, yarn
should converge on a stable api soon enough.
So until then, unfortunately, we have instability in our impl.
On Sat, Mar 15, 2014 at 10:32 PM, gzm55 <notificati...@github.com> wrote:
> In
>
core/src/main/scala/org/apache/spark/deploy/yarn/YarnAllocationHandlerMacro.scala:
>
> > + * so we don't need to call getAMResponse(), just use
AllocateResponse directly.
> > + * This macro will test the existence of AMResponse,
> > + * and generate diffenert expressions.
> > + *
> > + * This macro now is only used in spark's alpha version of yarn api.
> > + * It stays in the core project, for the two-stage compiling of
> > + * the scala macro system.
> > + */
> > + def getAMRespImpl(c: Context)(resp: c.Expr[Any]) = {
> > + try {
> > + import c.universe._
> > +
c.mirror.staticClass("org.apache.hadoop.yarn.api.records.AMResponse")
> > + c.Expr[Any](Apply(Select(resp.tree,
newTermName("getAMResponse")), List()))
> > + } catch {
> > + case _: Throwable => resp
> > + }
>
> I tried to use runtime refleciton, and found some difficults. because
> amResp will have different type, I've to give it a duck type like
>
> AnyRef {
> def getAllocatedContainers ...
> def getAvailableResources ...
> def getCompletedContainersStatuses ...
> }
>
> or have to cache four reflection methods (the upper three plus
> getAMResponse()). in the future, each time we introduce a method of
> AMResponse/AllocateResponse, we'll have to make additional cache for it.
>
> --
> Reply to this email directly or view it on
GitHub<https://github.com/apache/spark/pull/151/files#r10638121>
> .
>
---
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 infrastruct...@apache.org or file a JIRA ticket
with INFRA.
---
