[ 
https://issues.apache.org/jira/browse/FLINK-32740?page=com.atlassian.jira.plugin.system.issuetabpanels:comment-tabpanel&focusedCommentId=17750730#comment-17750730
 ] 

Junrui Li commented on FLINK-32740:
-----------------------------------

Hi, [~paul8263] 

I understand that your idea is to predefine a field as null. If the extracted 
value is equal to this field, the configuration statement will be skipped. I 
have a little concern about this: how can we let users avoid using predefined 
fields as value? This will lead to behavior changes that the user is not aware 
of. For example, the user configured "key: null" in the past, but this 
configuration will no longer take effect after this change. 
Do you have any good way to solve this problem?

> Introduce literal null value for flink-conf.yaml
> ------------------------------------------------
>
>                 Key: FLINK-32740
>                 URL: https://issues.apache.org/jira/browse/FLINK-32740
>             Project: Flink
>          Issue Type: Improvement
>          Components: API / Core
>    Affects Versions: 1.15.4
>            Reporter: Yao Zhang
>            Priority: Minor
>
> Hi community,
> Currently in flink-conf.yaml we only consider simplified YAML syntax like key 
> value pairs. And it might not be the right timing to indroduce YAML parser. 
> As [FLINK-23620 |https://issues.apache.org/jira/browse/FLINK-23620] has been 
> stated, there might be some keys that violate the YAML naming conventions.
> The current situation is, if we want to unset the value (or set the value as 
> its default), what we could do is to remove the key value pair completely or 
> leave the value blank (e.g. `rest.port: `). It might be inconvenient or less 
> readable for conf file that controlled by other applications.
> So is it necessary to add some special literal values that will be translated 
> to the real null value? For YAML we usually see ~, null or empty value as 
> null value. If necessary I would like to do this.



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

Reply via email to