[
https://issues.apache.org/jira/browse/FLINK-33947?page=com.atlassian.jira.plugin.system.issuetabpanels:comment-tabpanel&focusedCommentId=17800774#comment-17800774
]
RocMarshal edited comment on FLINK-33947 at 12/27/23 2:09 PM:
--------------------------------------------------------------
Before officially starting the repair, there are two issues worth discussing:
- First:
Is the *_prefix_* attribute allowed to be {*}_null_{*}?
What are the risks in using *_prefix_* if it is not allowed to be {*}_null_{*}?
Or will there be any defects?
>From the current effect, if the *_prefix_* is {*}_null_{*}, methods such as
>*_hashCode()_* will result in *_NPE_*
Based on the feedback from the current community regarding the use of this
class, there should be no instances of using a *_null_* {_}*prefix*{_}, unless
there are no calls to methods that may result in _*NPE*_
There are two corresponding alternative solutions here:
1. It is not allowed for _*prefix*_ to be a *_null_* object, so there will be
no corresponding method to cause _*NPE*_
2. Alternatively, allow the _*prefix*_ to be a _*null*_ object, but we need to
design logic to handle _*null*_ situations for methods that may cause _*NPE*_
to avoid the occurrence of _*NPE*_
- Secondly:
Regarding the mapping bug of _*prefix*_ values in various methods,
Only exists in the _*removeKey* *removeConfiguration*_ methods. We just need to
fix and add tests
was (Author: rocmarshal):
Before officially starting the repair, there are two issues worth discussing:
- First:
Is the *_prefix_* attribute allowed to be {*}_null_{*}?
What are the risks in using *_prefix_* if it is not allowed to be {*}_null_{*}?
Or will there be any defects?
>From the current effect, if the *_prefix_* is {*}_null_{*}, methods such as
>*_hashCode()_* will result in *_NPE_*
Based on the feedback from the current community regarding the use of this
class, there should be no instances of using a *_null_* {_}*prefix*{_}, unless
there are no calls to methods that may result in _*NPE*_
There are two corresponding alternative solutions here:
1. It is not allowed for _*prefix*_ to be a *_null_* object, so there will be
no corresponding method to cause _*NPE*_
2. Alternatively, allow the _*prefix*_ to be a _*null*_ object, but we need to
design logic to handle _*null*_ situations for methods that may cause _*NPE*_
to avoid the occurrence of _*NPE*_
By the way, it's best to modify the _*prefix*_ as _*final*_
- Secondly:
Regarding the mapping bug of _*prefix*_ values in various methods,
Only exists in the _*removeKey* *removeConfiguration*_ methods. We just need to
fix and add tests
> Fix bugs in DelegatingConfiguration missed the prefix mapping
> --------------------------------------------------------------
>
> Key: FLINK-33947
> URL: https://issues.apache.org/jira/browse/FLINK-33947
> Project: Flink
> Issue Type: Bug
> Components: API / Core
> Reporter: RocMarshal
> Assignee: RocMarshal
> Priority: Major
> Fix For: 1.19.0, 1.17.3, 1.18.2
>
>
> It was resulted from
> [https://github.com/apache/flink/pull/23994#issuecomment-1869905090]
> - Check and confirm other potential bug points
> - Fix the bugs about prefix key mapping when operating.
--
This message was sent by Atlassian Jira
(v8.20.10#820010)
