Re: GIN Indexes: Extensibility

[Anton Ananich <anton.ananich@xxxxxxxxx>]

Paul,

This is a really valuable idea. It will work in some situations for me. But in other situations I do not know if table will have a key of type int[] or string[] or even mixed. That’s why I’d wish to use JSON arrays and customize sort ordering.

Anyway I appreciate you shared this approach!

Regards,

Anthony Ananich

http://ananich.pro

On Jul 27, 2016, at 18:00, Paul Jungwirth <pj@xxxxxxxxxxxxxxxxxxxxxxxx> wrote:

On 07/27/2016 07:44 AM, Vick Khera wrote:

On Wed, Jul 27, 2016 at 3:28 AM, Anton Ananich <anton.ananich@xxxxxxxxx> wrote:

In my situation this order is invalid. Obviously, year 2016 should go after
2014, like that:

I think you expect JSONB to sort differently than it does. I cannot
imagine what a "natural" ordering of arbitrary JSON objects is.

FWIW, Postgres arrays do sort in the way he's expecting:

paul=# create table t (id integer, v integer[]);
CREATE TABLE
paul=# insert into t values (1, array[2014]), (2, array[2014, 1]), (3, array[2016]);
INSERT 0 3
paul=# select * from t order by v;
id |    v
----+----------
 1 | {2014}
 2 | {2014,1}
 3 | {2016}
(3 rows)

So maybe convert to an array before sorting?

Paul



--
Sent via pgsql-general mailing list (pgsql-general@xxxxxxxxxxxxxx)
To make changes to your subscription:
http://www.postgresql.org/mailpref/pgsql-general