[ 
https://issues.apache.org/jira/browse/FLINK-34669?page=com.atlassian.jira.plugin.system.issuetabpanels:all-tabpanel
 ]

Jane Chan updated FLINK-34669:
------------------------------
    Affects Version/s: 1.20.0

> Optimization of Arch Rules for Connector Constraints and Violation File 
> Updates
> -------------------------------------------------------------------------------
>
>                 Key: FLINK-34669
>                 URL: https://issues.apache.org/jira/browse/FLINK-34669
>             Project: Flink
>          Issue Type: Improvement
>          Components: Test Infrastructure
>    Affects Versions: 1.20.0
>            Reporter: Jane Chan
>            Priority: Major
>
> Description:
> I have identified potential areas for optimization within our Arch rules that 
> could improve our development workflow. This originated from the discussion 
> for PR https://github.com/apache/flink/pull/24492
> 1. Connector Constraints:
> Our current Arch rule, `CONNECTOR_CLASSES_ONLY_DEPEND_ON_PUBLIC_API`, was 
> implemented to prevent internal code changes in Flink from affecting the 
> compilation of connectors in external repositories. This rule is crucial for 
> connectors that are external, but it may be unnecessarily restrictive for the 
> filesystem connector, which remains within the same code repository as Flink. 
> Maybe we should consider excluding the filesystem connector from this rule to 
> better reflect its status as an internal component.
> 2. Preconditions Class Promotion:
> The majority of Arch rule violations for connectors are related to the use of 
> `Preconditions#checkX`. This consistent pattern of violations prompts the 
> question of whether we should reclassify `Preconditions` from its current 
> internal status to a `Public` or `PublicEvolving` interface, allowing broader 
> and more official usage within our codebase.
> 3. Violation File Updates:
> Updating the violation file following the `freeze.refreeze=true` process 
> outlined in the readme proves to be difficult. The diffs generated include 
> the line numbers, which complicates the review process, especially when 
> substantial changes are submitted. Reviewers face a considerable challenge in 
> distinguishing between meaningful changes and mere line number alterations. 
> To alleviate this issue, I suggest that we modify the process so that line 
> numbers are not included in the violation file diffs, streamlining reviews 
> and commits.



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

Reply via email to