The reason we developed the JsonSerDe in HCatalog at that point of
time was because I was more conversant/comfortable with HCatSchema and
HCatRecord when I wrote it, and it was mostly a translation attempt of
what JsonLoader was doing.

HCATALOG-272 ( https://issues.apache.org/jira/browse/HCATALOG-272 )
was opened to track the decoupling of JsonSerDe from HCatalog
dependencies. I'd updated that jira with a patch that did that, so
after a code review (and hive-specific tests), we should be able to
open up a jira on hive for that. I'd appreciate any additional eyes on
it. :)

Thanks,
-Sushanth

On Wed, Aug 1, 2012 at 9:53 AM, Alan Gates <[email protected]> wrote:
> Our plan has been to convert this to Hive all along.  It was originally done 
> as a storage driver and was converted in situ to a serde when we changed from 
> storage drivers to serdes.  I agree it makes sense to wait on the 
> documentation until we get it moved.
>
> Alan.
>
> On Aug 1, 2012, at 8:21 AM, Travis Crawford wrote:
>
>> Going through some patch available issues this morning I came across
>> one that proposes adding documentation for using HCatalog features
>> from Hive, specifically the JsonSerDe
>> (https://issues.apache.org/jira/browse/HCATALOG-397).
>>
>> Is there a good reason for having JsonSerDe in HCatalog, or would it
>> make sense to migrate that to Hive? A quick search shows a few other
>> people have implemented json serde's too. HCat's implementation
>> converts directly between HCatRecord and JSON so it couldn't be
>> directly moved, but I'm mainly curious if at a high level moving this
>> to Hive would make sense long-term. If we want to migrate that to Hive
>> I think it would be better for users if we document after migrating to
>> reduce confusion.
>>
>> --travis
>

Reply via email to