PHOENIX-6098 IndexPredicateAnalyzer wrongly handles pushdown predicat…

[brfrn169]

…es and residual predicates

[stoty]

You're going to need to rebase this for the tests to pass, to get the changes from Phoenix-6100, @brfrn169

[ankitsinghal]

It seems HIVE-11398 introduced the change where GenericUDFOPAnd can expect more than 2 arguments but Phoenix still uses Hive 1.2.1 version(https://github.com/apache/phoenix-connectors/blob/master/pom.xml#L75) which doesn't have this change. so is the problem really can affect us?

[stoty]

@richardantal has recently resurrected the Hive 3 connector from the old master branch with various fixes,
and is working on backporting it to Hive 2 for 4.x.
So even if this is not relevant for the old Hive 1 connector in phoenix-connectors, I think that it is still needed for the new Hive3/Hive2 connector.

[ankitsinghal]

yes, that makes sense, thanks @stoty for clarification, I didn't know that we have Hive3 connector separately. Changes look good to me for phoenix-hive3 (and for hive2 in 4.x branch) @brfrn169

[brfrn169]

@ankitsinghal @stoty Ah, okay. Thank you for pointing that out. I will modify this patch and apply the fix to phoenix-hive3.

[brfrn169]

@ankitsinghal @stoty @joshelser I just modified the patch. Can you please review this?

Also, I posted the steps to reproduce the issue to the Jira:
https://issues.apache.org/jira/browse/PHOENIX-6098?focusedCommentId=17183656&page=com.atlassian.jira.plugin.system.issuetabpanels%3Acomment-tabpanel#comment-17183656

[ankitsinghal]

+1,

@richardantal , should we have a backport ticket for this?

[richardantal]

@ankitsinghal It is not necessary if the same code works, because the common part will be unified to reduce code duplication. I am currently working on it.