Github user tejasapatil commented on a diff in the pull request:
https://github.com/apache/spark/pull/16909#discussion_r100818119
--- Diff:
sql/core/src/main/scala/org/apache/spark/sql/execution/ExternalAppendOnlyUnsafeRowArray.scala
---
@@ -0,0 +1,179 @@
+/*
+ * 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.sql.execution
+
+import java.util.ConcurrentModificationException
+
+import scala.collection.mutable.ArrayBuffer
+
+import org.apache.spark.{SparkEnv, TaskContext}
+import org.apache.spark.internal.Logging
+import org.apache.spark.sql.catalyst.InternalRow
+import org.apache.spark.sql.catalyst.expressions.UnsafeRow
+import org.apache.spark.util.collection.unsafe.sort.{UnsafeExternalSorter,
UnsafeSorterIterator}
+
+/**
+ * An append-only array for [[UnsafeRow]]s that spills content to disk
when there is insufficient
+ * space for it to grow.
+ *
+ * Setting spill threshold faces following trade-off:
+ *
+ * - If the spill threshold is too high, the in-memory array may occupy
more memory than is
+ * available, resulting in OOM.
+ * - If the spill threshold is too low, we spill frequently and incur
unnecessary disk writes.
+ * This may lead to a performance regression compared to the normal case
of using an
+ * [[ArrayBuffer]] or [[Array]].
+ */
+private[sql] class ExternalAppendOnlyUnsafeRowArray(numRowsSpillThreshold:
Int) extends Logging {
+ private val inMemoryBuffer: ArrayBuffer[UnsafeRow] =
ArrayBuffer.empty[UnsafeRow]
+
+ private var spillableArray: UnsafeExternalSorter = null
+ private var numElements = 0
+
+ // A counter to keep track of total additions made to this array since
its creation.
+ // This helps to invalidate iterators when there are changes done to the
backing array.
+ private var modCount: Long = 0
--- End diff --
No. There is more than that. Apart from `clear()`, even `add()` would be
regarded as a modification.
The inbuilt iterators (from both `ArrayBuffer` and `UnsafeExternalSorter`)
once created do not see any new data added. Clients of
`ExternalAppendOnlyUnsafeRowArray ` may not realise this and think that they
have read all the data. To prevent that, I have to have such mechanism to
invalidate existing iterators.
```
val buffer = ArrayBuffer.empty[Int]
buffer.append(1)
val iterator = buffer.iterator
assert(iterator.hasNext)
assert(iterator.next() == 1)
buffer.append(2)
assert(iterator.hasNext) // <-- THIS FAILS
```
Also, when `add()` transparently switches the backing storage from
`ArrayBuffer` => `UnsafeExternalSorter` and there was an open iterator created
over the `ArrayBuffer`, it will lead to `IndexOutOfBoundsException`
```
val buffer = ArrayBuffer.empty[Int]
buffer.append(1)
buffer.append(2)
val iterator = buffer.iterator
buffer.clear()
assert(iterator.hasNext) // <-- THIS FAILS
}
```
---
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.
---
---------------------------------------------------------------------
To unsubscribe, e-mail: reviews-unsubscr...@spark.apache.org
For additional commands, e-mail: reviews-h...@spark.apache.org
