[
https://issues.apache.org/jira/browse/FLINK-8139?page=com.atlassian.jira.plugin.system.issuetabpanels:comment-tabpanel&focusedCommentId=16281628#comment-16281628
]
ASF GitHub Bot commented on FLINK-8139:
---------------------------------------
Github user fhueske commented on the issue:
https://github.com/apache/flink/pull/5065
Hmm, I agree with @twalthr. Adding these checks when a DataSet/DataStream
is imported would prevent valid queries (e.g., those that to simple
transformations). I would not add a config switch to disable the global checks.
Adding the checks to the operators that require properly implemented `equals()`
and `hashCode()` methods sounds like a better approach.
We should also add to the documentation that all objects that are processed
by the Table API / SQL should have equals and hashCode properly implemented.
> Check for proper equals() and hashCode() when registering a table
> -----------------------------------------------------------------
>
> Key: FLINK-8139
> URL: https://issues.apache.org/jira/browse/FLINK-8139
> Project: Flink
> Issue Type: Improvement
> Components: Table API & SQL
> Reporter: Timo Walther
> Assignee: Aegeaner
>
> In the current Table API & SQL implementation we compare {{Row}}s at
> different positions. E.g., for joining we test rows for equality or put them
> into state. A heap state backend requires proper hashCode() and equals() in
> order to work correct. Thus, every type in the Table API needs to have these
> methods implemented.
> We need to check if all fields of a row have implement methods that differ
> from {{Object.equals()}} and {{Object.hashCode()}} via reflections. Both
> coming from TableSource and DataStream/DataSet.
> Additionally, for array types, the {{Row}} class should use
> {{Arrays.deepEquals()}} and {{Arrays.deepHashCode()}} instead of the non-deep
> variants.
--
This message was sent by Atlassian JIRA
(v6.4.14#64029)
