* Thom Brown (t...@linux.com) wrote: > On 24 November 2014 at 20:40, Stephen Frost <sfr...@snowman.net> wrote: > > I will point out that this use of COMMENT is novel though, no? Comments > > are normally handled as "COMMENT ON blah IS 'whatever';" ALTER SYSTEM > > is certainly special but I'm not sure I like the idea of having some > > commands which support in-command COMMENT while others don't. > > I typed that out in my original email, thought about it, then removed it > because I decided that perhaps it isn't the same class as comment as > COMMENT ON uses. That affects objects, whereas this would apply to > individual config parameters within a file. Also bear in mind that if > someone runs: > > SHOW maintenance_work_mem; > > And sees "4GB", they may decide to add a comment based on that, even though > the source of that setting isn't postgresql.auto.conf.
I'd be fine with supporting two distinct object type or perhaps one object type and two sub types to allow those to be different (no clue what the actual syntax would be..). I'd actually prefer that these comments be represented in both pg_description as well as being in the actual config file- otherwise how are you going to be able to view what's there from the database before you go and change it? The fact that the information is also persisted into the actual .auto.conf file is a convenience for admins who happen to be looking there but I'd consider what's in pg_description to be the canonical source. Thanks, Stephen
Description: Digital signature