flink-issues mailing list archives

Site index · List index
Message view « Date » · « Thread »
Top « Date » · « Thread »
From "Stephan Ewen (JIRA)" <j...@apache.org>
Subject [jira] [Commented] (FLINK-9393) LocatableInputSplit#hashCode should take hostnames into account
Date Thu, 17 May 2018 17:57:00 GMT

    [ https://issues.apache.org/jira/browse/FLINK-9393?page=com.atlassian.jira.plugin.system.issuetabpanels:comment-tabpanel&focusedCommentId=16479459#comment-16479459
] 

Stephan Ewen commented on FLINK-9393:
-------------------------------------

This is not a bug, right?

hashCode() is allowed to be worse than equals(), this is not really wrong. It only needs to
not be different when objects are equal.
In how the splits get generated, splitNumber is a quite good hash code, and fast (compared
to mixing in the host name string hashes).

Not sure we need/should actually change this. This issue falls a bit under the category
  - might have a good reason to be like it is
  - is not a bug
  - any change might counter the original intention.

I personally would favor to focus on other issues...

> LocatableInputSplit#hashCode should take hostnames into account
> ---------------------------------------------------------------
>
>                 Key: FLINK-9393
>                 URL: https://issues.apache.org/jira/browse/FLINK-9393
>             Project: Flink
>          Issue Type: Bug
>            Reporter: Ted Yu
>            Assignee: vinoyang
>            Priority: Major
>
> Currently:
> {code}
>   public int hashCode() {
>     return this.splitNumber;
> {code}
> This is not symmetrical with {{equals}} method where hostnames are compared.
> LocatableInputSplit#hashCode should take hostnames into account.



--
This message was sent by Atlassian JIRA
(v7.6.3#76005)

Mime
View raw message