Github user rxin commented on a diff in the pull request:
https://github.com/apache/spark/pull/7860#discussion_r36046315
--- Diff:
sql/core/src/test/scala/org/apache/spark/sql/execution/UnsafeFixedWidthAggregationMapSuite.scala
---
@@ -152,26 +185,47 @@ class UnsafeFixedWidthAggregationMapSuite
false // disable perf metrics
)
- val rand = new Random(42)
- val groupKeys: Set[String] = Seq.fill(512) {
- Seq.fill(rand.nextInt(100))(rand.nextPrintableChar()).mkString
- }.toSet
- groupKeys.foreach { keyString =>
+ val keys = randomStrings(1024).take(512)
+ keys.foreach { keyString =>
val buf =
map.getAggregationBuffer(InternalRow(UTF8String.fromString(keyString)))
buf.setInt(0, keyString.length)
assert(buf != null)
}
+ // Convert the map into a sorter
+ val sorter = map.destructAndCreateExternalSorter()
+
+ withClue(s"destructAndCreateExternalSorter should release memory used
by the map") {
+ // 4096 * 16 is the initial size allocated for the pointer/prefix
array in the in-mem sorter.
+ assert(shuffleMemoryManager.getMemoryConsumptionForThisTask() ===
+ initialMemoryConsumption + 4096 * 16)
+ }
+
+ // Add more keys to the sorter and make sure the results come out
sorted.
+ val additionalKeys = randomStrings(1024)
+ val keyConverter = UnsafeProjection.create(groupKeySchema)
+ val valueConverter = UnsafeProjection.create(aggBufferSchema)
+
+ additionalKeys.zipWithIndex.foreach { case (str, i) =>
+ val k = InternalRow(UTF8String.fromString(str))
+ val v = InternalRow(str.length)
+ sorter.insertKV(keyConverter.apply(k), valueConverter.apply(v))
+
+ if ((i % 100) == 0) {
+ shuffleMemoryManager.markAsOutOfMemory()
+ sorter.closeCurrentPage()
--- End diff --
If we don't close it, they can continue to write to it, and won't spill.
---
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]
