[ 
https://issues.apache.org/jira/browse/CALCITE-2807?page=com.atlassian.jira.plugin.system.issuetabpanels:comment-tabpanel&focusedCommentId=16845293#comment-16845293
 ] 

Laurent Goujon commented on CALCITE-2807:
-----------------------------------------

I was not necessary thinking of Github guidelines but guidelines outline in the 
git documentation 
(https://git-scm.com/book/en/v2/Distributed-Git-Contributing-to-a-Project in 
section Commit Guidelines), which is kind of captured by 
https://tbaggery.com/2008/04/19/a-note-about-git-commit-messages.html. Looks 
like Linux Kernel development is happy with a summary line with more than 50 
characters (but less than 70-75). Calcite guidelines just mention "The first 
line of the commit message must be a concise and useful description of the 
change."

Not really surprise by Linus' comments regarding Github UI (I have my own beef 
with some of the UI choices too).

> Identify expanded IS NOT DISTINCT FROM expression when pushing down filter 
> past join
> ------------------------------------------------------------------------------------
>
>                 Key: CALCITE-2807
>                 URL: https://issues.apache.org/jira/browse/CALCITE-2807
>             Project: Calcite
>          Issue Type: Improvement
>          Components: core
>            Reporter: Laurent Goujon
>            Priority: Major
>              Labels: pull-request-available
>          Time Spent: 10m
>  Remaining Estimate: 0h
>
> IS NOT DISTINCT FROM expressions in join condition might actually be 
> considered as equi-join conditions, and RelOptUtil#splitJoinConditions() has 
> support for it. But some other join related functions/rules don't.
> One of them is RelOptUtil#pushDownJoinConditions (used by 
> JoinPushExpressionsRule) which tries to push filter expressions below the 
> join, but ends up modifying the join expression in a way which makes identify 
> an IDNF condition impossible later.
> For example expression OR(AND(IS_NULL($1), IS_NULL($4)), EQUALS($1,$4)) will 
> be changed into OR(AND($3, $6), EQUALS($1, $5)) which makes it 
> harder/impossible for RelOptUtil#splitJoinConditions() to identify an IS NOT 
> DISTINCT FROM equi-join condition.
> This is a variant of CALCITE-2803



--
This message was sent by Atlassian JIRA
(v7.6.3#76005)

Reply via email to