Github user HyukjinKwon commented on the issue:
https://github.com/apache/spark/pull/14649
Hi @liancheng, regardless of the PR description and the implementation, I
would like to describe what I thought and I hope this is sensible in a way.
I thought in the exactly same way as you described. However, if you look at
the codes closely, it actually only takes care of the case when there is
`_metadata` which IIUC includes block offsets and paths. So, this PR tries to
drop the files to touch by this information, which we would anyway touch the
files and then will drop whole row-groups.
So, this is rather about reducing the files to touch within each partition
(as far as I know, we introduced the optimization that collects small files and
put into a partition).
I might be wrong but this was the reason I thought it should be helpful
anyway, in particular, for S3 where touching files itself is expensive.
---
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 infrastruct...@apache.org or file a JIRA ticket
with INFRA.
---
---------------------------------------------------------------------
To unsubscribe, e-mail: reviews-unsubscr...@spark.apache.org
For additional commands, e-mail: reviews-h...@spark.apache.org
