Message-ID: <2ab176f136f0fda560f63cd091875bf21c4fe7df.camel@j-davis.com>
Subject: Re: range_agg
From: Jeff Davis <pgsql@j-davis.com>
To: Paul A Jungwirth <pj@illuminatedcomputing.com>, David Fetter
	 <david@fetter.org>
Cc: Pgsql Hackers <pgsql-hackers@postgresql.org>
Date: Sat, 06 Jul 2019 12:26:06 -0700
In-Reply-To: 
 <CA+renyXwdgfJJVgAFBK2vfvrXrseLy6pq-mnhvKAdfazwEB30w@mail.gmail.com>
References: <16d71dc8-34cf-5ebd-1ce5-ccd93c0a14f9@illuminatedcomputing.com>
	 <ba323e489b98029be5ba2d5c5cbc99795f333f48.camel@j-davis.com>
	 <CA+renyX7jsWnyry6SOn-WSgdgQLtyFJ7SLWBVs+oatwQsMzvog@mail.gmail.com>
	 <20190705174531.GE24679@fetter.org>
	 <CA+renyXwdgfJJVgAFBK2vfvrXrseLy6pq-mnhvKAdfazwEB30w@mail.gmail.com>
Content-Type: text/plain; charset="UTF-8"
Mime-Version: 1.0
Content-Transfer-Encoding: 7bit
Precedence: bulk

On Fri, 2019-07-05 at 10:57 -0700, Paul A Jungwirth wrote:
> I take it that a multirange contains of *disjoint* ranges, so instead
> of {[1,2), [2,3), [6,7)} you'd have {[1,3), [6,7)}. Jeff does that
> match your expectation?

Yes.

> I just realized that since weighted_range_agg and covering_range_agg
> return tuples of (anyrange, integer) (maybe other numeric types
> too?),
> their elements are *not ranges*, so they couldn't return a
> multirange.
> They would have to return an array of those tuples.

I think you are right. I was originally thinking a multirange and an
array of weights would work, but the multirange would coalesce adjacent
ranges because it would have no way to know they have different
weights.

Regards,
	Jeff Davis