[jira] [Commented] (GEODE-5112) As a user I would like my JSON documents to be integrated better with my POJOs

["Jens Deppe (JIRA)" <ji...@apache.org>]

    [ https://issues.apache.org/jira/browse/GEODE-5112?page=com.atlassian.jira.plugin.system.issuetabpanels:comment-tabpanel&focusedCommentId=16444655#comment-16444655 ] 

Jens Deppe commented on GEODE-5112:
-----------------------------------

Somewhat related is GEODE-629 which is aiming to completely remove use of org.json, AKA geode-json. Most of the work is targeted at internal components but where it would overlap with this Jira is where gfsh is used to query/retrieve region entries.

> As a user I would like my JSON documents to be integrated better with my POJOs
> ------------------------------------------------------------------------------
>
>                 Key: GEODE-5112
>                 URL: https://issues.apache.org/jira/browse/GEODE-5112
>             Project: Geode
>          Issue Type: New Feature
>          Components: rest (dev)
>            Reporter: Bruce Schuchardt
>            Priority: Major
>
> Geode's handling of the "@type" field in a JSON document is poorly integrated with the PDX registry.
> If I have a POJO class that is auto-serializable and I use this class name in my "@type" field of JSON document two different PDX types are created.
> If my JSON document is deserialized Geode first converts the PdxInstance back into a JSON document and then uses a Jackson-databind ObjectMapper to convert it to my POJO class.  That's very inefficient.
> If my auto-serialized POJO is deserialized it uses much more efficient PDX deserialization to instantiate my object.
> Why not look for "@type" when serializing my JSON document in the first place and use the class name to look up my existing PdxType in the registry?  Use that PdxType to serialize the JSON document.
> This is probably a lot harder than I imagine but the current JSON serialization seems bolted onto the PDX system and causes PdxType explosion if my documents are missing fields or has fields in different order.  Tying my documents to my auto-serializable POJO class could help reign this in.



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