Github user srowen commented on the pull request:
https://github.com/apache/spark/pull/4780#issuecomment-76214312
Since the PR is the implementation of an issue resolution, if the
discussion is about the implementation it can happen here on the PR.
Shading isn't the issue in the sense that it wouldn't affect whether you
can access un-shaded fastutil classes. So, that's bad if parquet-column brings
back in all of fastutil, since it has caused a problem with assembly size in
the past. It looks like Spark SQL uses it. This is a separate issue at the
moment.
It shouldn't matter what Spark does with Spark's copy of
clearspring-analytics, in the sense that you shouldn't care or rely on what
Spark does or doesn't use. In practice it can matter, since, if your code uses
CA, and it finds CA in the Spark classloader, then that copy can't see fastutil
that your usage and your copy of CA (which you should be bundling with your
app) can see.
Are you packaging CA yourself, are you marking all the Spark deps as
provided? I see you're already using the `userClassPathFirst` settings.
---
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]
