[jira] [Commented] (KAFKA-2593) KeyValueStores should not require use of the context's default serializers and deserializers

["Randall Hauch (JIRA)" <ji...@apache.org>]

    [ https://issues.apache.org/jira/browse/KAFKA-2593?page=com.atlassian.jira.plugin.system.issuetabpanels:comment-tabpanel&focusedCommentId=14947520#comment-14947520 ] 

Randall Hauch commented on KAFKA-2593:
--------------------------------------

~guozhang, I've replaced the static factory methods in each of the key-value store implementations with a builder class. (See the PR for details.)

> KeyValueStores should not require use of the context's default serializers and deserializers
> --------------------------------------------------------------------------------------------
>
>                 Key: KAFKA-2593
>                 URL: https://issues.apache.org/jira/browse/KAFKA-2593
>             Project: Kafka
>          Issue Type: Sub-task
>          Components: kafka streams
>            Reporter: Randall Hauch
>            Assignee: Randall Hauch
>             Fix For: 0.9.0.0
>
>
> Currently the {{InMemoryKeyValueStore}} is only able to use the key and value serializers and deserializers (aka, "serdes") from the {{ProcessingContext}}. This means that a {{Processor}} implementation that wants to use the {{InMemoryKeyValueStore}} can only do this if the key and value types match those set up as the default serdes in the topology's configuration.
> Additionally, the {{RocksDBKeyValueStore}} is only capable of {{byte[]}} keys and values.
> Both of these key-value stores should allow the component using them to specify the serdes for both the keys and values. As a convenience, the current behavior should still be supported, as should a way to infer the serdes for the "built-in" serializers and deserializers (e.g., strings, integers, longs, and byte arrays).



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