I've been playing around with making it possible to create user defined guc
variables. This has been discussed, at least in passing, before. And it is
even anticipated in guc.c as a possible future feature:
/* * Build the sorted array. This is split out so that it could be * re-executed after startup (eg, we could allow
loadablemodules to * add vars, and then we'd need to re-sort). */
It is a feature that would be nice to have, so that, for example, a user
defined variable named "my_classpath" could be created to point to the java
CLASSPATH needed by a custom C function.
So far I have this much working:
- A new backend function, pg_create_user_setting(name, value, islocal) is used to "register" the setting.
- SHOW ALL, SHOW, current_setting(), and pg_show_all_settings()) will display it just like any other setting
- Similarly, SET and set_config() will change it.
I still need to make the user defined settings survive being saved by ALTER
USER or ALTER DATABASE. I'm also thinking about a corresponding grammar
addition, something along the lines of:
CREATE SETTING name WITH VALUE value;
This would effectively perform: SELECT pg_create_user_setting(name, value, false);
I'm wondering whether it would be "a good thing" or "a bad thing" to have
unrecognized settings found in postgresql.conf be registered as user defined
settings?
Any comments, concerns, or objections?
Thanks,
Joe
