Use hash join when writing sparkey #5402
Merged
Add this suggestion to a batch that can be applied as a single commit.
This suggestion is invalid because no changes were made to the code.
Suggestions cannot be applied while the pull request is closed.
Suggestions cannot be applied while viewing a subset of changes.
Only one suggestion per line can be applied in a batch.
Add this suggestion to a batch that can be applied as a single commit.
Applying suggestions on deleted lines is not supported.
You must change the existing code in this line in order to create a valid suggestion.
Outdated suggestions cannot be applied.
This suggestion has been applied or marked resolved.
Suggestions cannot be applied from pending reviews.
Suggestions cannot be applied on multi-line comments.
Suggestions cannot be applied while the pull request is queued to merge.
Suggestion cannot be applied right now. Please check back later.
When writing to sparkey,
allShards
represents every expected shard even if there is no corresponding data inshards
for that shard number.shards.rightOuterJoin(allShards)
(added in #5208) fails when a shard contains large amounts of data, leading to the error described in #5300:java.lang.OutOfMemoryError: Required array length 2147483639 + 15534 is too large
.This PR replaces
rightOuterJoin
withhashFullOuterJoin
(note that there is nohashRightOuterJoin
implementation). A hash join is a good fit because the right-hand side contains very little data (only the keys of the shards) and it doesn't need to use an array to represent the large left-hand side's values. As a result, some failing workflows that succeeded in Scio 0.13.* will run successfully again.