On Fri, May 03, 2019 at 03:56:41PM -0700, Paul Jungwirth wrote:
> Hello,
>
> I wrote an extension to add a range_agg function with similar behavior to
> existing *_agg functions, and I'm wondering if folks would like to have it
> in core? Here is the repo: https://github.com/pjungwir/range_agg
>
> I'm also working on a patch for temporal foreign keys, and having range_agg
> would make the FK check easier and faster, which is why I'd like to get it
> added. But also it just seems useful, like array_agg, json_agg, etc.
>
> One question is how to aggregate ranges that would leave gaps and/or
> overlaps.
This suggests two different ways to extend ranges over aggregation:
one which is a union of (in general) disjoint intervals, two others
are a union of intervals, each of which has a weight. Please pardon
the ASCII art.
The aggregation of:
[1, 4)
[2, 5)
[8, 10)
could turn into:
{[1,5), [8, 10)} (union without weight)
{{[1,2),1}, {[2,4),2}, {[4,5),1}, {[8,10),1}} (strictly positive weights which don't (in general) cover the space)
{{[1,2),1}, {[2,4),2}, {[4,5),1}, {[5,8),0}, {[8,10),1}} (non-negative weights which guarantee the space is covered)
There is no principled reason to choose one over the others.
> What do people think? I plan to work on a patch regardless, so that I can
> use it for temporal FKs, but I'd appreciate some feedback on the "user
> interface".
I think the cases above, or at least the first two of them, should be
available. They could be called range_agg, weighted_range_agg, and
covering_range_agg.
Best,
David.
--
David Fetter <david(at)fetter(dot)org> http://fetter.org/
Phone: +1 415 235 3778
Remember to vote!
Consider donating to Postgres: http://www.postgresql.org/about/donate