Github user lianhuiwang commented on the pull request:
https://github.com/apache/spark/pull/10024#issuecomment-211953148
@squito Yes, I think your understanding is correct. this PR only support
that a Spillables will be called once. The code 'val sort = new Spillable()
sort.iterator() sort.iterator()' will be wrong.
for ExternalSorter.iterator, I find that is just called by
BlockStoreShuffleReader.Read(). Every Reader will create new ExternalSorter, So
i think it's ok for it.
for ExternalAppendOnlyMap, It has two places that use it.They are
Aggregator and CoGroupedRDD. we also can find that it create new
ExternalAppendOnlyMap before call one iterator(). So i think it is also ok for
it.
The input and output of them are Iterator and Iterator cannot be cached, so
it will not affect when rdds are cached.
BTW: this PR has been running for a long time on our many online spark jobs.
If i missed some points, please put forward. It is worth mentioning that
---
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]
