Github user JoshRosen commented on a diff in the pull request:
https://github.com/apache/spark/pull/7904#discussion_r36579502
--- Diff:
sql/core/src/main/scala/org/apache/spark/sql/execution/joins/RowIterator.scala
---
@@ -0,0 +1,63 @@
+/*
+ * 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.joins
+
+import java.util.NoSuchElementException
+
+import org.apache.spark.sql.catalyst.InternalRow
+
+private[sql] abstract class RowIterator {
+ def advanceNext(): Boolean
+ def getRow: InternalRow
+ def toScala: Iterator[InternalRow] = new RowIteratorToScala(this)
+}
+
+object RowIterator {
+ def fromScala(scalaIter: Iterator[InternalRow]): RowIterator = {
+ scalaIter match {
+ case wrappedRowIter: RowIteratorToScala => wrappedRowIter.rowIter
+ case _ => new RowIteratorFromScala(scalaIter)
+ }
+ }
+}
+
+private final class RowIteratorToScala(val rowIter: RowIterator) extends
Iterator[InternalRow] {
+ private [this] var _hasNext: Boolean = rowIter.advanceNext()
+ override def hasNext: Boolean = _hasNext
+ override def next(): InternalRow = {
+ if (!_hasNext) throw new NoSuchElementException
+ val row: InternalRow = rowIter.getRow.copy()
--- End diff --
I'm going to comment this file to explain what's going on here, but in a
nutshell:
The Scala iterator exposes separate `hasNext()` and `next()` methods. This
creates the potential for a user to call `hasNext()` two times in a row without
calling `next()` in between those calls. For some iterators, like
`iterator.filter(...)`, non-trivial work must be performed to figure out
whether there's a next tuple and this may cause our mutable row objects to be
mutated. This is the reason that we needed the copying in the `.collect()` call
in `leftOuterIterator`, even after changing that method to call `collect` on an
iterator instead of an iterable.
This wrapper class, `RowIteratorToScala`, makes this safe by adding the
required copying at the last possible moment.
In the long run, we can probably remove this copying. The
`RowIteratorFromScala` is designed to help with this: that wrapper class takes
a Scala iterator and hides it behind our `RowIterator` interface, which
prevents the double `hasNext` issue. Therefore, if an operator can use
`RowIterators` then it doesn't need to immediately copy its incoming rows. This
is the reason that I was able to safely remove the immediate
`left.execute.map(_.copy())` and `right.execute.map(_.copy())` calls from
SortMergeJoin.
---
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]
