On Thu, Jun 07, 2001 at 12:50:21AM -0400, Bruce Momjian wrote:
> >
> > Ok, now I am awake. How about this one?
> >
> > reset all: command line and .conf options change defaults
> > on RESET ALL those are restored.
> >
> > show all: GUC + non-GUC.
>
> I have applied this patch with minor modifications to make it more
> meaningful. I added meaningful default values if not set, and changed
> function names of set/reset/parse_date to *_datestyle to match the
> actual meaning of the value.
Fine.
> There is no clear distinction between set-able and non-setable because
> RESET ALL just calls the GUC code to set variables as though the backend
> was just starting up. Seems quite difficult to make that distinction in
> the code and I am not sure it is worth it.
It is not needed, as on non-SETable values will have
value == default_val. Same goes for permissions, if user
has no right to SET a value, the reset will do nothing.
At least in my understanding it is ok already.
Although - one thing that is worth adding is checking
on RESET if var has changed - if not then do nothing.
_And_ if it is, the RESET should go through set_, so the
callbacks get called too (ATM they are not). This means
splitting the set_config_option() up, set_config_option_real()
should take then the actual record.
The SHOW output should be reworked some time, but this is not
possible until we have tuples-returning-funtions.
--
marko
