[
https://issues.apache.org/jira/browse/ACCUMULO-2486?page=com.atlassian.jira.plugin.system.issuetabpanels:comment-tabpanel&focusedCommentId=13938469#comment-13938469
]
Christopher Tubbs commented on ACCUMULO-2486:
---------------------------------------------
No strong opinion here. It seems like it makes sense in many cases, such as in
the example, but we should think about the implications on the API semantics
whenever we do something like that. In this case, it makes sense because we
expect users to not just use the interfaces... we also expect that they'll
subclass the iterators.
> Use covariant return types in SKVI.deepCopy
> -------------------------------------------
>
> Key: ACCUMULO-2486
> URL: https://issues.apache.org/jira/browse/ACCUMULO-2486
> Project: Accumulo
> Issue Type: Improvement
> Reporter: Mike Drob
> Fix For: 1.7.0
>
>
> Starting with Java 5, subclasses with method overrides can declare covariant
> return types. In general, I believe that we want to do this, as it provides
> stronger type guarantees with not a lot of extra cost.
> Example:
> {code}
> public interface SortedKeyValueIterator {
> SortedKeyValueIterator deepCopy(IteratorEnvironment env);
> }
> public interface InterruptibleIterator extends SortedKeyValueIterator {
> @Override
> InterruptibleIterator deepCopy(IteratorEnvironment env);
> }
> {code}
> I believe this change will be fully backwards compatible with the existing
> APIs.
--
This message was sent by Atlassian JIRA
(v6.2#6252)
