Github user steveloughran commented on the pull request:
https://github.com/apache/spark/pull/12076#issuecomment-204324891
1. Hive uses Kryo "the guava of serialization" internally; I don't know the
specifics, but its not insignificant.
1. they moved ahead of spark's version to fix some bugs; that's the kryo
which repackaged things.
1. Kryo types were used across methods called from Spark; the unshading was
needed to get things to link, same with the switch from 2.22 to 2.21 (and a
change to some files). It was less traumatic to push Hive back slightly than to
try to force an update into unknown territory for chill & dependencies, even if
it were possible.
Upgrading to hive 2 would finally get everything in sync: ideally eliminate
the need to have a custom hive JAR at all. Essentially all spark needs is a
version of Hive with parameters that can be exchanged across all needed
methods, with the conflict packages shaded, and the non-conflict packages
generally omitted
If you want to find out where things don't link, edit the spark pom to pull
in org.apache.hive/hive and see what breaks.
(of course, the other thing that would be nice would be for Hive to make it
possible to subclass their thrift service cleanly; I've got a PoC of that
somewhere ... it's not hard to add)
---
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]
