[ 
https://issues.apache.org/jira/browse/CALCITE-7123?page=com.atlassian.jira.plugin.system.issuetabpanels:comment-tabpanel&focusedCommentId=18011952#comment-18011952
 ] 

Julian Hyde commented on CALCITE-7123:
--------------------------------------

My requests above were a little terse, so I'll expand a little. Converting your 
initial test case you a minimal, reproducible test case, working if possible on 
a standard data set, is The Work. (I should have added 'search for related 
cases' to my requests.) That work needs to be done, at some point, by someone, 
before the issue can be marked resolved. Once done, it gives a beautiful test 
case that will explain the problem to anyone who is interested in the future, 
and it almost certainly tells you which lines of code need to be fixed.

[~mbudiu] is being kind. If it is literally impossible to reproduce on a 
built-in table, you can of course create a new table for the test case. But if 
it only reproduces on a column with {{TIMESTAMP}} data type, or on a table 
backed by a CSV file, those are very important observations, and the sooner we 
know, the better.

> Query doesn't work as expected when using COALESCE or CASE in the group by 
> query
> --------------------------------------------------------------------------------
>
>                 Key: CALCITE-7123
>                 URL: https://issues.apache.org/jira/browse/CALCITE-7123
>             Project: Calcite
>          Issue Type: Bug
>          Components: avatica
>    Affects Versions: 1.39.0
>            Reporter: Priyanka
>            Priority: Major
>         Attachments: image-2025-08-03-21-12-40-517.png, 
> image-2025-08-03-21-13-52-179.png, image-2025-08-03-21-18-10-506.png
>
>
> Running the following query directly on the Druid UI returns the expected 
> results:
> {code:java}
> SELECT
>                           FLOOR("__time" TO DAY)  AS "createtime.day", 
> COALESCE("parentName", "stockName") AS "unitName",
>                           COUNT(*) AS "count"
>                         FROM
>                           "druid"."events"
>                         WHERE  "customerName" = 'x' AND  "__time" >= 
> TIMESTAMP '2025-05-31 22:00:00.000' AND  "__time" <= TIMESTAMP '2025-06-12 
> 21:59:59.000'
>                             AND (FALSE = true  OR "fleetName" IN ('y'))
>                             AND (FALSE = false  OR "eventTemplateName" IN 
> (null))
>                             AND (FALSE = false  OR "consistNumber" IN (null))
>                             AND (FALSE = false  OR ("parentName" IS NOT NULL 
> AND "stockName" IN (null)))
>                             AND (FALSE = false  OR (COALESCE("parentName", 
> "stockName") IN (null)))
>                             AND (FALSE = false  OR "locationName" IN (null))
>                             AND (FALSE = false  OR "eventCategory" IN (null))
>                             AND (FALSE = false OR "eventPriorityName" IN 
> (null))
>                             AND (FALSE = false  OR null)
>                             AND (FALSE = false  OR "active" = null)
>                         GROUP BY FLOOR("__time" TO DAY) , 
> COALESCE("parentName", "stockName") {code}
> But when it goes through Calcite, it returns the stockName always as the 
> unitName, even if the parentName is available.
> Testing with CASE instead of COALESCE or using a CAST for the type also 
> didn't have any effect with Calcite.



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

Reply via email to