Github user davies commented on the pull request:
https://github.com/apache/spark/pull/8226#issuecomment-136876278
Currently, We can't return a JoinedRow with two UnsafeRow in it (because
JoinedRow.getUTFString() does not copy the underlying bytes, also for complex
types).
As @chenghao-intel said, switching to GenerateUnsafeRowJoiner could improve
the copy, but actually the copy could be avoided at all. For example, we could
have a UnsafeJoinedRow, which is similar to JoinedRow, but also have some
features as UnsafeRow (fast serialization).
Another direction we could explore is that we could compile the whole query
plan into one function to avoid the materialization of rows, that will be
different story.
---
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]
