[
https://issues.apache.org/jira/browse/PHOENIX-6151?page=com.atlassian.jira.plugin.system.issuetabpanels:comment-tabpanel&focusedCommentId=17200265#comment-17200265
]
Josh Elser commented on PHOENIX-6151:
-------------------------------------
bq. Instead of shading packages explicitly, try the approach of shading
everything, and only keeping unshaded the libraries that cannot be shaded.
+1. Giving you some more history. I remember when we were first trying to shade
phoenix jars. We approached it with a mindset of "who knows what's going to
break, let's shade as much as we can"
We had some pain in making tiny-changes as it broke for the "next user". Aside
from the known painful dependencies (e.g. Hadoop and Protobuf), I can't think
of an example where our shading _caused_ a problem. I think this "inversion"
makes sense, given where we've come in the past years. Draft looks good so far!
> Switch phoenix-client to shade-by-default mode
> ----------------------------------------------
>
> Key: PHOENIX-6151
> URL: https://issues.apache.org/jira/browse/PHOENIX-6151
> Project: Phoenix
> Issue Type: Improvement
> Components: core
> Affects Versions: 5.1.0, 4.16.0
> Reporter: Istvan Toth
> Assignee: Istvan Toth
> Priority: Major
> Time Spent: 10m
> Remaining Estimate: 0h
>
> Phoenix-client contains Phoenix, plus most of a Hadoop+Hbase stack.
> We keep running into difficulties, where phoenix-client conflicts with user
> code, or in the case of connectors with different components.
> Instead of shading packages explicitly, try the approach of shading
> everything, and only keeping unshaded the libraries that cannot be shaded.
> This is the approach taken by Hadoop.
--
This message was sent by Atlassian Jira
(v8.3.4#803005)
