Subject | CHECK Constraints (was: Re: [firebird-support] Re: Checking periods don't overlap) |
---|---|
Author | johnsparrowuk |
Post date | 2004-06-15T16:26:38Z |
Grin, well I did start that way! (I won't mention 'dirty reads'
again, promise!)
I wanted to have period records that can't overlap, as in:
Check not exists(select idcode from T where new.startdate > enddate
or new.enddate < startdate and new.idcode <> idcode)
Loads of people (including me!) have suggested cunning work-arounds
to prevent overlapping. The simplest of which is to use
table_stability isolation.
That's fine. I'd just hope there will be a more simple (or less
blocking) solution in the future (like a check constraint).
Esoteric multi-gen solutions are great, but they're going to turn
newbies off.
John
--- In firebird-support@yahoogroups.com, "Ivan Prenosil"
<Ivan.Prenosil@s...> wrote:
again, promise!)
I wanted to have period records that can't overlap, as in:
Check not exists(select idcode from T where new.startdate > enddate
or new.enddate < startdate and new.idcode <> idcode)
Loads of people (including me!) have suggested cunning work-arounds
to prevent overlapping. The simplest of which is to use
table_stability isolation.
That's fine. I'd just hope there will be a more simple (or less
blocking) solution in the future (like a check constraint).
Esoteric multi-gen solutions are great, but they're going to turn
newbies off.
John
--- In firebird-support@yahoogroups.com, "Ivan Prenosil"
<Ivan.Prenosil@s...> wrote:
> > So you don't see a problem with the behaviour of checkconstraints at
> > the moment?problem
>
> I am not against any useful feature. If you have some suggestion
> how check constraints could be enhanced just tell us.
> Don't worry how much you know abou implementation - somebody
> who knows enough will say whether such proposal is feasible or not.
> (and sometimes even discussion about feature that can't work
> can lead to better ideas that will)
>
> In each case it will be better to start by exact defining the
> and specifying required behaviour than by directly suggesting
> solution (e.g. dirty-reads).
>
> Ivan