Thread: [GENERAL] RUM-index and support for storing BIGINT as part of a tsvector+timestamp
[GENERAL] RUM-index and support for storing BIGINT as part of a tsvector+timestamp
From
Andreas Joseph Krogh
Date:
Hi PostgresPro-guys.
I've asked this before but didn't get any response, so I'll try again. I know you have a lot on you plate...
On RUM's TODO-list is this:
- Allow multiple additional information (lexemes positions + timestamp).
- Add support for arrays.
Will any of these items support storing BIGINT as part of a tsvector+timestamp in the same index?
so one an have an index like this:
create index rum_idx on message using rum (fts_all some_opclass, folder_id, received_timestamp) WITH (attach=received_timestamp, "to"=fts_all, order_by_attach=t);
Making it possible for a query like this to use one index effectively:
WHERE fts_all @@ to_tsquery('simple', 'andre:*&jose:*') AND folder_id = ANY(ARRAY[2,3]::BIGINT[]) ORDER BY received_timestamp DESC LIMIT 10;
The reason I ask is that storing only tsvector+timestamp often is't enough because often there is some other dimention which is *very* relevant to limit the results (like what folder, or account, to search for email-messages in, instead of searching for all in the system).
Thanks.
--
Andreas Joseph Krogh
CTO / Partner - Visena AS
Mobile: +47 909 56 963
Attachment
Re: [GENERAL] RUM-index and support for storing BIGINT as part of a tsvector+timestamp
From
Andreas Joseph Krogh
Date:
Ping...
Any feedback is welcome, thanks.
På torsdag 12. januar 2017 kl. 13:13:34, skrev Andreas Joseph Krogh <andreas@visena.com>:
Hi PostgresPro-guys.I've asked this before but didn't get any response, so I'll try again. I know you have a lot on you plate...On RUM's TODO-list is this:
- Allow multiple additional information (lexemes positions + timestamp).
- Add support for arrays.
Will any of these items support storing BIGINT as part of a tsvector+timestamp in the same index?so one an have an index like this:create index rum_idx on message using rum (fts_all some_opclass, folder_id, received_timestamp) WITH (attach=received_timestamp, "to"=fts_all, order_by_attach=t);Making it possible for a query like this to use one index effectively:
WHERE fts_all @@ to_tsquery('simple', 'andre:*&jose:*') AND folder_id = ANY(ARRAY[2,3]::BIGINT[]) ORDER BY received_timestamp DESC LIMIT 10;The reason I ask is that storing only tsvector+timestamp often is't enough because often there is some other dimention which is *very* relevant to limit the results (like what folder, or account, to search for email-messages in, instead of searching for all in the system).Thanks.--Andreas Joseph KroghCTO / Partner - Visena ASMobile: +47 909 56 963
--
Andreas Joseph Krogh
CTO / Partner - Visena AS
Mobile: +47 909 56 963