[ 
https://issues.apache.org/jira/browse/FLINK-30687?page=com.atlassian.jira.plugin.system.issuetabpanels:comment-tabpanel&focusedCommentId=17677122#comment-17677122
 ] 

tanjialiang edited comment on FLINK-30687 at 1/16/23 2:58 AM:
--------------------------------------------------------------

I try to found out how to fix this, i found in 
org.apache.flink.table.planner.codegen.agg.AggsHandlerCodeGenerator#createFilterExpression
 the filterArg must be >0, but ‘SELECT COUNT( * ) FILTER (WHERE cluster_id = 1) 
FROM ${table}’ 's filterArg is 0. I fixed the condition in 'filterArg >= 0', 
and it work.

!image-2023-01-16-10-54-04-673.png!

 

So i want to ask why we can not filterArg=0? Did it some scenes we can not use? 
Because 'inputFieldTypes' is a 'scala.collection.Seq' which index is start from 
0.


was (Author: JIRAUSER279823):
I try to found out how to fix this, i found in 
org.apache.flink.table.planner.codegen.agg.AggsHandlerCodeGenerator#createFilterExpression
 the filterArg must be >0, but ‘SELECT COUNT(*) FILTER (WHERE cluster_id = 1) 
FROM ${table}’ 's filterArg is 0. I fixed the condition in 'filterArg >= 0', 
and it work.

!image-2023-01-16-10-54-04-673.png!

 

So i want to ask why we can not filterArg=0? Did it some scenes we can not use? 
Because 'inputFieldTypes' is a 'scala.collection.Seq' which index is start from 
0.

> FILTER not effect in count(*)
> -----------------------------
>
>                 Key: FLINK-30687
>                 URL: https://issues.apache.org/jira/browse/FLINK-30687
>             Project: Flink
>          Issue Type: Bug
>          Components: Table SQL / Planner
>    Affects Versions: 1.16.0
>            Reporter: tanjialiang
>            Priority: Major
>         Attachments: image-2023-01-16-10-54-04-673.png
>
>
> When i try to using Flink SQL like this, i found 'FILTER(WHERE class_id = 1)' 
> is not effect.
> {code:java}
> CREATE TABLE student
> (
>     id INT NOT NULL,
>     name STRING,
>     class_id INT NOT NULL
> )
> WITH (
>     'connector' = 'jdbc',
>     'url' = 'jdbc:mysql://localhost:3306/test',
>     'table-name' = 'student',
>     'username' = 'root',
>     'password' = '12345678'
> );
> SELECT COUNT(*) FILTER (WHERE class_id = 1) FROM student;
> or
> SELECT COUNT(1) FILTER (WHERE class_id = 1) FROM student;{code}
>  
> But when i tried Flink SQL like this, it worked.
> {code:java}
> SELECT COUNT(*) FROM student WHERE class_id = 1;
> or 
> SELECT COUNT(class_id) FILTER (WHERE class_id = 1) FROM student;{code}
>  
> By the way, mysql connector has a bug and fixed in FLINK-27268. Maybe you try 
> this demo should cherry-pick this PR first.



--
This message was sent by Atlassian Jira
(v8.20.10#820010)

Reply via email to