Github user sameeragarwal commented on a diff in the pull request:
https://github.com/apache/spark/pull/11870#discussion_r57622786
--- Diff:
sql/core/src/main/scala/org/apache/spark/sql/execution/joins/HashedRelation.scala
---
@@ -347,29 +293,53 @@ private[joins] object UnsafeHashedRelation {
keyGenerator: UnsafeProjection,
sizeEstimate: Int): HashedRelation = {
- // Use a Java hash table here because unsafe maps expect fixed size
records
- // TODO: Use BytesToBytesMap for memory efficiency
- val hashTable = new JavaHashMap[UnsafeRow,
CompactBuffer[UnsafeRow]](sizeEstimate)
+ val taskMemoryManager = if (TaskContext.get() != null) {
+ TaskContext.get().taskMemoryManager()
+ } else {
+ new TaskMemoryManager(
+ new StaticMemoryManager(
+ new SparkConf().set("spark.memory.offHeap.enabled", "false"),
+ Long.MaxValue,
+ Long.MaxValue,
+ 1),
+ 0)
+ }
+ val pageSizeBytes =
Option(SparkEnv.get).map(_.memoryManager.pageSizeBytes)
+ .getOrElse(new SparkConf().getSizeAsBytes("spark.buffer.pageSize",
"16m"))
+
+ val binaryMap = new BytesToBytesMap(
+ taskMemoryManager,
+ (sizeEstimate * 1.5 + 1).toInt, // reduce hash collision
+ pageSizeBytes)
// Create a mapping of buildKeys -> rows
+ var numFields = 0
+ // Whether all the keys are unique or not
+ var allUnique: Boolean = true
while (input.hasNext) {
- val unsafeRow = input.next().asInstanceOf[UnsafeRow]
- val rowKey = keyGenerator(unsafeRow)
- if (!rowKey.anyNull) {
- val existingMatchList = hashTable.get(rowKey)
- val matchList = if (existingMatchList == null) {
- val newMatchList = new CompactBuffer[UnsafeRow]()
- hashTable.put(rowKey.copy(), newMatchList)
- newMatchList
- } else {
- existingMatchList
+ val row = input.next().asInstanceOf[UnsafeRow]
+ numFields = row.numFields()
+ val key = keyGenerator(row)
+ if (!key.anyNull) {
+ val loc = binaryMap.lookup(key.getBaseObject, key.getBaseOffset,
key.getSizeInBytes)
+ if (loc.isDefined) {
+ allUnique = false
+ }
+ val success = loc.append(
+ key.getBaseObject, key.getBaseOffset, key.getSizeInBytes,
+ row.getBaseObject, row.getBaseOffset, row.getSizeInBytes)
+ if (!success) {
+ binaryMap.free()
+ throw new SparkException("There is no enough memory to build
hash map")
--- End diff --
nit: not
---
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]
