You are viewing a plain text version of this content. The canonical link for it is here.

Posted to issues@flink.apache.org by "Timo Walther (Jira)" <ji...@apache.org> on 2026/04/29 15:42:00 UTC

[jira] [Closed] (FLINK-39575) Fix output rowtype in StreamPhysicalProcessTableFunction.toUdfCall when table arg is not first

     [ https://issues.apache.org/jira/browse/FLINK-39575?page=com.atlassian.jira.plugin.system.issuetabpanels:all-tabpanel ]

Timo Walther closed FLINK-39575.
--------------------------------
    Fix Version/s: 2.4.0
       Resolution: Fixed

Fixed in master: f325b4a8d5914d218b394c8a7e2f7e4f0c27a358

> Fix output rowtype in StreamPhysicalProcessTableFunction.toUdfCall when table arg is not first
> ----------------------------------------------------------------------------------------------
>
>                 Key: FLINK-39575
>                 URL: https://issues.apache.org/jira/browse/FLINK-39575
>             Project: Flink
>          Issue Type: Improvement
>          Components: Table SQL / API
>    Affects Versions: 2.4.0
>            Reporter: Gustavo de Morais
>            Assignee: Gustavo de Morais
>            Priority: Major
>              Labels: pull-request-available
>             Fix For: 2.4.0
>
>
> StreamPhysicalProcessTableFunction.toUdfCall always reads staticArgs.get(0) when deciding whether a table operand has the PASS_COLUMNS_THROUGH trait. The lookup should be per-operand: staticArgs.get(operand.i). This is a leftover from the first released versions.
>  
> When the first declared argument is not the table being inspected (e.g. f(Integer i, Row r WITH PASS_COLUMNS_THROUGH)), the trait check uses the wrong static argument. This produces an incorrect prefixOutputSystemFields count and the UDF
>   call's rowtype is not stripped of the system-prefixed columns. The wrong rowtype is then passed to ProcessTableRunnerGenerator, leading to incorrect code generation for the PTF runner.



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