On 2 Dec 2013, at 19:27, Albe Laurenz <laurenz.a...@wien.gv.at> wrote:
> What strikes me is that since foreign key constraints are implemented
> as triggers in PostgreSQL, this solution would probably not have many
> performance benefits over a self-written trigger that implements the
> same functionality. Since you need two triggers for your example,
> the performance might even be worse than a single self-written trigger.
Well, the main cost on insert in the FK table should be looking for matching
rows in the referenced tables, which the patch avoids for non-matching rows. So
while you’ll get the overhead of N triggers firing, you only pay the expected
query cost (which will even use a partial index if you’ve got one set up). Each
of the referenced tables is only involved in one FK, so there’s no difference
in cost there.
> Now performance isn't everything, but that would mean that the benefit
> of your proposal is entirely on the usability side.
Well, I don’t think there’s much of a performance hit, and I don’t think any of
the alternatives would perform much better in practice, but certainly
performance wasn’t a motivating factor for this feature, it was a) correctness
and b) avoiding the ugliness of the existing solutions.
> I personally don't think that it is so difficult to write a trigger
> for that functionality yourself, but I guess that the argument for
> this feature rests on how coveted such a functionality would be
> (to justify the trade-off in code complexity).
The patch is pretty small so far - and more than half of it is regression
tests. So there’s not much extra code complexity IMO. I wouldn’t want to touch
the FK system with anything but the lightest touch.
Sent via pgsql-hackers mailing list (email@example.com)
To make changes to your subscription: