Github user davies commented on a diff in the pull request:
https://github.com/apache/spark/pull/13318#discussion_r64781946
--- Diff:
sql/core/src/main/java/org/apache/spark/sql/execution/UnsafeKVExternalSorter.java
---
@@ -86,14 +88,16 @@ public UnsafeKVExternalSorter(
prefixComparator,
/* initialSize */ 4096,
pageSizeBytes,
- keySchema.length() == 1 &&
SortPrefixUtils.canSortFullyWithPrefix(keySchema.apply(0)));
+ canUseRadixSort);
} else {
+ // The array will be used to do in-place sort, which require half of
the space to be empty.
+ assert(map.numKeys() <= map.getArray().size() / 2);
// During spilling, the array in map will not be used, so we can
borrow that and use it
// as the underline array for in-memory sorter (it's always large
enough).
// Since we will not grow the array, it's fine to pass `null` as
consumer.
final UnsafeInMemorySorter inMemSorter = new UnsafeInMemorySorter(
null, taskMemoryManager, recordComparator, prefixComparator,
map.getArray(),
- false /* TODO(ekl) we can only radix sort if the BytesToBytes load
factor is <= 0.5 */);
+ canUseRadixSort);
--- End diff --
We only need to make sure that only half of array could be used. The load
factor could be different. For example, we could move the check into
TungstenAggregate without changing the load factor of BytesToBytesMap.
---
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]
