I agree that we can use $.C.D & C.D directly but if the key is not present
and I have several such keys then it is better to not create so many
columns.

On Tue, Sep 18, 2012 at 5:47 AM, Connell, Chuck <chuck.conn...@nuance.com>wrote:

>  OK, but keep in mind that you could use
> get_json_object(whole-json-field, $.C.D). If using a serde, then you can
> say just C.D directly.****
>
> ** **
>
> Chuck Connell****
>
> Nuance R&D Data Team****
>
> Burlington, MA****
>
> ** **
>
> ** **
>
> *From:* Mark Golden [mailto:mark.golden1...@gmail.com]
> *Sent:* Monday, September 17, 2012 7:31 PM
>
> *To:* Connell, Chuck
> *Cc:* user@hive.apache.org
> *Subject:* Re: hive json serde****
>
>  ** **
>
> Thanks Chuck.
>
> I tried both json_tuple & get_json_objects on top of the serde table and
> it works too. I have nested jsons and some of the keys do not have a
> specific schema (You can throw whatever you want there) so if I want to
> query from those keys then I have to use get_json_object on top of the
> serde table. If I have the following json
>
> { "A" : "I love hive", "B" : "User mailing list is great", "C" : { "D" :
> "chuck", "E" : "is awesome and thanks for your help" } }
> { "A" : "I love hive", "B" : "User mailing list is great", "C" : { "G" :
> "It will rain in NY tonight" } }
> { "A" : "I love hive", "B" : "User mailing list is great", "C" : { "H" :
> "No it will snow in NY" } }
>
> If I create a table on top of json like this then the serde will work,
> json_tuple & get_json_object will also work but to query the keys inside
> "C" I need to do get_json_object on "C" something like
>
> select *
> from my_serde_table
> where get_json_object(C, "$.D") ilike "chuck"
>
> I like serde table because it is much cleaner than create a table with
> (value string) and then doing get_json_object or json_tuple and extract all
> the columns out.
>
> I'm exposing this table to other users in my company so I want to have the
> serde table. Thanks for your help.****
>
> On Mon, Sep 17, 2012 at 3:49 PM, Connell, Chuck <chuck.conn...@nuance.com>
> wrote:****
>
> A few things that should help...
>
> - There are three ways to handle JSON within a Hive table,
> get_json_object, json_tuple, and JSON SerDe. These are separate mechanisms.
> So if you want to use get_json_object, you don't need to load a SerDe.
>
> - With a SerDe, Hive understands JSON fields natively, so you can just use
> the plain field names, no special function call at all.
>
> - I'm not sure if auxpath accepts a relative path, but I know the full
> path works.
>
> Chuck
>
>
> ****
>  ------------------------------
>
> *From:* Mark Golden [mark.golden1...@gmail.com]
> *Sent:* Monday, September 17, 2012 5:58 PM
> *To:* Connell, Chuck
> *Cc:* user@hive.apache.org****
>
>
> *Subject:* Re: hive json serde****
>
> ** **
>
> It works now. Looks like there is a bug in the code.
>
> if you do hive --auxpath ./serde then I get an error but if I get the full
> path as
> hive --auxpath /var/lib/hdfs/serde/ then get_json_object() works.
>
> Thanks for your help. Appreciate it.
>
> CCing hive user group as well.
>
> ****
>
> On Mon, Sep 17, 2012 at 2:54 PM, Mark Golden <mark.golden1...@gmail.com>
> wrote:****
>
> :-) Thanks. I downloaded and got it compiled as well.
>
> My table has nested JSON values so when I tried to do
> get_json_object(<column>,'$.<key>') this serde doesn't work in that case.
> It throws an error. I'll email the developer too.
>
> When you tried did you try this on a cluster of nodes or on a single
> instance? ** **
>
> ** **
>
> On Mon, Sep 17, 2012 at 2:38 PM, Connell, Chuck <chuck.conn...@nuance.com>
> wrote:****
>
> Go to the Downloads link on the project page, near the upper right.
>
> ****
>  ------------------------------
>
> *From:* Mark Golden [mark.golden1...@gmail.com]
> *Sent:* Monday, September 17, 2012 5:32 PM
> *To:* Connell, Chuck
> *Subject:* Re: hive json serde****
>
> Hey Chuck,
>
> Where did you download the prebuild jar from?****
>
> On Mon, Sep 17, 2012 at 12:10 PM, Connell, Chuck <chuck.conn...@nuance.com>
> wrote:****
>
> I used his pre-built jar. No need to compile anything.****
>
>  ****
>
> Be sure to add “ --auxpath /path/to/jar/dir “ to the Hive command line.***
> *
>
>  ****
>
> Chuck****
>
>  ****
>
>  ****
>
>  ****
>
> *From:* Connell, Chuck [mailto:chuck.conn...@nuance.com]
> *Sent:* Monday, September 17, 2012 3:06 PM
> *To:* user@hive.apache.org
> *Subject:* RE: hive json serde****
>
>  ****
>
> I just finished testing this one. No problems found. The developer is also
> quite responsive to issues raised.  I encouraged him to submit it to the
> Hive dev team as core code.****
>
>  ****
>
> https://github.com/rcongiu/Hive-JSON-Serde/****
>
>  ****
>
> Chuck Connell****
>
> Nuance R&D Data Team****
>
> Burlington, MA****
>
>  ****
>
>  ****
>
> *From:* Mark Golden 
> [mailto:mark.golden1...@gmail.com<mark.golden1...@gmail.com>]
>
> *Sent:* Monday, September 17, 2012 2:56 PM
> *To:* user@hive.apache.org
> *Subject:* hive json serde****
>
>  ****
>
> Hi,
>
> I'm looking for a hive json serde that works well. Any suggestions? I did
> a quick google and found a few but none of them are complete.
>
> Thanks,
> Mark.****
>
> ** **
>
> ** **
>
> ** **
>
> ** **
>

Reply via email to