[
https://issues.apache.org/jira/browse/FLINK-16577?page=com.atlassian.jira.plugin.system.issuetabpanels:comment-tabpanel&focusedCommentId=17058370#comment-17058370
]
Shuo Cheng edited comment on FLINK-16577 at 3/13/20, 2:45 AM:
--------------------------------------------------------------
I think one possible solution is when we get value from RexLiteral using
FlinkRelOptUtil#getLiteralValue(..), we do not get value by its `concrete
type`, but by its `broad type`. e.g., for RexLiteral(1), now we get
"Integer(1)", after modifying, it'll be 'BigDecimal(1)'.
was (Author: icshuo):
I think one possible solution is when we get value from RexLiteral, we do not
get the
> Exception will be thrown when computing columnInterval relmetadata in some
> case
> -------------------------------------------------------------------------------
>
> Key: FLINK-16577
> URL: https://issues.apache.org/jira/browse/FLINK-16577
> Project: Flink
> Issue Type: Bug
> Components: Table SQL / Planner
> Affects Versions: 1.10.0
> Reporter: Shuo Cheng
> Priority: Major
> Fix For: 1.11.0
>
> Attachments: image-2020-03-13-10-32-35-375.png,
> image-2020-03-13-10-38-17-001.png
>
>
> Consider the following SQL
>
> {code:java}
> // a: INT, c: LONG
> SELECT
> c, SUM(a)
> FROM T
> WHERE a > 0.1 AND a < 1
> GROUP BY c{code}
>
> Here the sql type of 0.1 is Decimal and 1 is Integer, and they are both in
> NUMERIC type family, and do not trigger type coercion, so the plan is:
> {code:java}
> FlinkLogicalAggregate(group=[{0}], EXPR$1=[SUM($1)])
> +- FlinkLogicalCalc(select=[c, a], where=[AND(>(a, 0.1:DECIMAL(2, 1)), <(a,
> 1))])
> +- FlinkLogicalTableSourceScan(table=[[...]], fields=[a, b, c])
> {code}
> When we calculate the filtered column interval of calc, it'll lead to
> validation exception of `FiniteValueInterval`:
> !image-2020-03-13-10-32-35-375.png!
--
This message was sent by Atlassian Jira
(v8.3.4#803005)
