You are viewing a plain text version of this content. The canonical link for it is here.
Posted to dev@storm.apache.org by "Rick Kellogg (JIRA)" <ji...@apache.org> on 2015/10/09 02:13:26 UTC

[jira] [Updated] (STORM-84) Don't use hashCode for fields grouping

     [ https://issues.apache.org/jira/browse/STORM-84?page=com.atlassian.jira.plugin.system.issuetabpanels:all-tabpanel ]

Rick Kellogg updated STORM-84:
------------------------------
    Component/s: storm-core

> Don't use hashCode for fields grouping
> --------------------------------------
>
>                 Key: STORM-84
>                 URL: https://issues.apache.org/jira/browse/STORM-84
>             Project: Apache Storm
>          Issue Type: Improvement
>          Components: storm-core
>            Reporter: James Xu
>
> https://github.com/nathanmarz/storm/issues/244
> The contract of hashCode does not guarantee that it will be consistent across JVM's, so it's just a coincidence that it happens to work. We need a new way to do hashcodes. Hopefully there's an open source library that already solves this problem, otherwise we'll have to build a solution for this to get hashcodes for arbitrary objects.
> A tempting way to do it would be to hash the serialized form of an object, but this has two issues:
> 1. Equal objects are not guaranteed to have same serialized form (e.g., it represents a set and elements are serialized out of order
> 2. We don't want to incur the serialization cost when sending between tasks colocated in the same worker



--
This message was sent by Atlassian JIRA
(v6.3.4#6332)