Re: foreign key from array element

Hi Chris,

thank you very much for taking the time to read the article and get
into the features proposed with our patch.

On Tue, 18 Sep 2012 17:17:56 -0700, Chris Travers
<chris(dot)travers(at)gmail(dot)com> wrote:
> So those are the cautions and why I don't think a feature like this
> is suitable for routine usage, but truth be told a lot of the
> object-relational features are definitely not for routine usage and
> make a mess of things if people use them just because they can.  I
> use table inheritance and I totally understand a lot of people's
> hostility towards this feature.  Again, anytime you break 1NF you
> should probably have a really good reason.  I don't think this
> changes here.

I agree with you that this feature won't (and probably shouldn't)
change modelling approaches in the majority of the cases. But will bring
new opportunities, therefore make PostgreSQL even more versatile. I
still believe that in some cases - not just indistinctively -
aggregation in object oriented modelling can definitely be logically
modelled using arrays, with referential integrity guaranteed by this
feature.

> However, after thinking about the feature overnight, I can see a
> number of use cases for it, ranging from recording something like
> race
> results (where update contention is definitionally not an issue
> because the record of an event aren't supposed to change) to sanity
> checks in materialized views, and there are probably additional uses
> that are not apparent yet.

I totally agree with you. This is exactly what we (as a community) need
to do now as far as this feature is concerned. We need to have a larger
use base and from there fully understand what the community needs. For
instance, for 9.2 we had already developed actions on update and delete
operations - assuming generic use cases. We have preferred for now to
take out that part and start with a simpler patch where actions are
forbidden. Through community feedback we found a name for the feature
that was commonly accepted (we had called them EACH FOREIGN KEYS last
year), and came up with an easy to understand syntax (and a better
naming). It was important not to go too far down an unexplored
territory. :)

> So yeah, as far as the feature goes, as documented, I haven't tried
> it fully yet (expect to do so this weekend), but it looks useful at
> least in some cases.

Thank you. That's really much appreciated.

Cheers,
Gabriel
--
Gabriele Bartolini - 2ndQuadrant Italia
PostgreSQL Training, Services and Support
Gabriele(dot)Bartolini(at)2ndQuadrant(dot)it - www.2ndQuadrant.it