Updated RUM?

On Fri, Jun 24, 2016 at 2:36 PM, Andreas Joseph Krogh <andreas@visena.com>
wrote:

Hi.

@PostgresPro; Will there be any updates to the rum git-repo any time soon?

Sure.

I understand you're working hard on this, care to share some
vision/roadmap? I'm specifically interested in if RUM will support the same
datatypes as GIN supports (arrays, JSON etc.), and what opclasses you plan
to implement (for equalty and sorting). Will there be some kind of
btree_rum like there is for btree_gin, to add RUM-support for the types
there exists Btree-support for?

We are testing fts+timestamp combination right now. The challenge is to
have posting list/tree sorted by addinfo (timestamp) not by item pointer.
It works already, but we are not quite satisfied with generated WAL traffic
and are researching some options in improving posting tree building and
generic wal.

We have in mind some sort of generalization like anyarray, but don't have
any real ideas.

På mandag 27. juni 2016 kl. 13:20:03, skrev Oleg Bartunov <obartunov@gmail.com
<mailto:obartunov@gmail.com>>:
    On Fri, Jun 24, 2016 at 2:36 PM, Andreas Joseph Krogh <andreas@visena.com
<mailto:andreas@visena.com>> wrote: Hi.
 
@PostgresPro; Will there be any updates to the rum git-repo any time soon?
 
Sure.
 
 
I understand you're working hard on this, care to share some vision/roadmap?
I'm specifically interested in if RUM will support the same datatypes as GIN
supports (arrays, JSON etc.), and what opclasses you plan to implement (for
equalty and sorting). Will there be some kind of btree_rum like there is for
btree_gin, to add RUM-support for the types there exists Btree-support for?
 
We are testing fts+timestamp combination right now.  The challenge is to have
posting list/tree sorted by addinfo (timestamp) not by item pointer. It works
already, but we are not quite satisfied with generated WAL traffic and are
researching some options in improving posting tree building and generic wal.
 
We have in mind some sort of generalization like anyarray, but don't have any
real ideas.

 
Cool!
 
Will RUM support having ts_vector and bigint in the same index, like GIN does
(utilizine btree_gin), then sorting by some 3rd timestamp-column?
 
This will enable queries like:
 
create table delivery( id BIGSERIAL primary key, fts_all TSVECTOR not null,
folder_idBIGINT NOT NULL, sent TIMESTAMP not null, message varchar not null );
 
SELECT del.id <http://del.id/&gt; , del.sent FROM delivery del WHERE del.fts_all
@@to_tsquery('simple', 'hi:*') AND del.folder_id = ANY(ARRAY[2,3]::BIGINT[])
ORDER BYdel.sent DESC LIMIT 101 OFFSET 0; Having this query use one RUM-index
with optimized sorting by timestamp would be extremely useful.

 
Thanks.
 
-- Andreas Joseph Krogh
CTO / Partner - Visena AS
Mobile: +47 909 56 963
andreas@visena.com <mailto:andreas@visena.com>
www.visena.com <https://www.visena.com&gt;
<https://www.visena.com&gt;