Check for value with current_setting()
Asked Answered
N

2

25

I'm trying to work with current_setting().

I came up with this:

CREATE OR REPLACE FUNCTION process_audit() RETURNS TRIGGER AS $audit$
    DECLARE
        user_id integer;
    BEGIN
        BEGIN
            user_id := current_setting('hws.current_user_id');
        EXCEPTION WHEN OTHERS THEN
            user_id := NULL;
        END;
        ...
        RETURN NULL;
   END;
$audit$ LANGUAGE plpgsql;

The setting is set via:

SELECT set_config('hws.current_user_id', '5', true); -- true = local setting -> only visible in current transaction

The problem is, that current_setting() throws an exception if the value is not valid. I don't want to use EXCEPTION because I read that exception blocks are expensive.

Is there a way to check if the setting has a value without using exceptions?

Btw: I also tried to read from pg_settings but that doesn't seem to work with local settings.

Narbonne answered 30/7, 2013 at 20:19 Comment(2)
For anyone looking for a complete example, I created a gist of a working an auditing mechanism leveraging SET LOCAL and current_settings. Hope you find it useful :-)Septavalent
@blubb: awesomeSequestration
G
32

9.6 and newer:

PostgreSQL (9.6+) supports current_setting('setting_name', 't') to fetch a setting and return NULL if it's unset. you can combine this with coalesce to supply a default.

9.5 and older:

Per the question, you can do it with a plpgsql function that uses a BEGIN ... EXCEPTION handler, if you don't mind the performance hit and clumsiness. But there's no built-in support.

Goon answered 31/7, 2013 at 3:6 Comment(9)
Did you ever get any traction on this?Sideslip
@toxaq: As far as I understand it, this feature will make it for 9.5 (though it's not in the alpha2 prerelease), in the form of an optional second parameter missing_ok: current_setting('foo', true) will not throw an error in case foo is undefined; see the devel-branch documentation. According to this blog post, the patch is by David Christensen (thanks!).Dinnerware
@toxaq: Ugh; the patch was committed to the master branch, but not to the branch leading to the 9.5 release. So unless someone cherry-picks it for 9.5 (not sure if that can happen?), it seems the feature won't be available before 9.6, next year. :-/Dinnerware
@Dinnerware Gutted. It would make some really nice options available in terms of passing values from an app into PG.Sideslip
@Sideslip Bring it up on -hackers then.Goon
The second missing_ok param is available in 9.6 now FYIHeroic
@Heroic -- this is gold. This quesiton, and the additional parameter is EXACTLY what I was looking for. Thanks.Suture
@CraigRinger I'm curious, why using 't' for the second param instead of true? I'm looking the postgresql.org/docs/current/… for this function. Thank you!Chittagong
@Chittagong Historical habit. Once upon a time IIRC postgres didn't have a true constant.Goon
T
1

Please find the below example which i use while working with current_setting.

CREATE OR REPLACE FUNCTION public.usp_fetch_current_setting()
    RETURNS text
    LANGUAGE plpgsql
    AS $function$
    declare 
        v_appCode text;
    begin
        select current_setting('spm.appCode', 't') into v_appCode;
        return v_appCode;
    END;
$function$
;

while calling:

set session "spm.appCode" = 'spm-alignment-web';
SELECT public.usp_fetch_current_setting();
Trisaccharide answered 9/7, 2021 at 13:26 Comment(0)

© 2022 - 2024 — McMap. All rights reserved.