You are viewing a plain text version of this content. The canonical link for it is here.
Posted to issues@flink.apache.org by "Jingsong Lee (Jira)" <ji...@apache.org> on 2019/12/12 12:58:00 UTC
[jira] [Commented] (FLINK-15186) Column names and nullability
mismatch when querying CatalogViews
[ https://issues.apache.org/jira/browse/FLINK-15186?page=com.atlassian.jira.plugin.system.issuetabpanels:comment-tabpanel&focusedCommentId=16994630#comment-16994630 ]
Jingsong Lee commented on FLINK-15186:
--------------------------------------
Hi [~dwysakowicz] [~danny0405], About add a cast:
> we may lose some plan promotion like project merge or transpose or reduction
I don't think so, is our optimizer too weak? I think redundant cast will be removed easily by optimizer and should not affect optimizer.
And I think cast will enhance correctness, image:
* view sql will produce nullable fields
* view schema define fields not null.
* our code generator for cast consider null check, cast null to a not null field will fail.
So add a cast is good to verify the data.
And maybe not only field name, nullable, but also precision, this kind of scene really needs a cast to transform.
So I am +1 for add a cast, What do you think?
> Column names and nullability mismatch when querying CatalogViews
> ----------------------------------------------------------------
>
> Key: FLINK-15186
> URL: https://issues.apache.org/jira/browse/FLINK-15186
> Project: Flink
> Issue Type: Bug
> Components: Table SQL / Legacy Planner, Table SQL / Planner
> Reporter: Rui Li
> Priority: Major
> Fix For: 1.10.0
>
>
> When querying a CatalogView, the column names and nullability we get from the catalog may differ from what we arrive at after compiling the view's query. And we fail to expand the view due to this mismatch.
--
This message was sent by Atlassian Jira
(v8.3.4#803005)