Subject: Re: range_agg
To: Pavel Stehule <pavel.stehule@gmail.com>
Cc: Alvaro Herrera <alvherre@2ndquadrant.com>, Jeff Davis
 <pgsql@j-davis.com>, Pgsql Hackers <pgsql-hackers@postgresql.org>
References: 
 <CA+renyWOSQ5P-=KCqti4WN1hdS9h9O5kFSvXqodq49XaR43HkA@mail.gmail.com>
 <20190926211349.GA3375@alvherre.pgsql>
 <CA+renyWY-vfLVM8NkHj11xXyCNYJA7=1sd5_Bk+B+DL66y6-WQ@mail.gmail.com>
 <CA+renyW9KU=1mxyoNMuHPnBx4j5eFw+KKPgEMLcbKRUVekrPoA@mail.gmail.com>
 <CAFj8pRCYd8O32UWC5ZRWOhioOx4Z=AOnr9VO8fSTq6mdW--zOw@mail.gmail.com>
 <CA+renyXY4tY5vpw0Ha3xqPgL4PFz_uAWVMUCHWU4AQ4mTPp30A@mail.gmail.com>
 <CA+renyVNd0mGfbO-i2akkcQE9YrwNzAjVLchAcBEFZF+fiyJQw@mail.gmail.com>
 <CAFj8pRAUso-5-+f-q-nword4F=QZOZuJaKdaqH_y_Ybom+fYsg@mail.gmail.com>
From: Paul Jungwirth <pj@illuminatedcomputing.com>
Message-ID: <5b52f2f4-a0ec-76b1-8a73-98b8a7270bd6@illuminatedcomputing.com>
Date: Thu, 21 Nov 2019 12:15:28 -0800
User-Agent: Mozilla/5.0 (X11; Linux x86_64; rv:60.0) Gecko/20100101
 Thunderbird/60.9.0
MIME-Version: 1.0
In-Reply-To: 
 <CAFj8pRAUso-5-+f-q-nword4F=QZOZuJaKdaqH_y_Ybom+fYsg@mail.gmail.com>
Content-Type: text/plain; charset=utf-8; format=flowed
Content-Language: en-US
Content-Transfer-Encoding: 8bit
Precedence: bulk

On 11/21/19 1:06 AM, Pavel Stehule wrote:
> 2. I don't like introduction "safe" operators - now the basic operators 
> are doubled, and nobody without documentation will use @* operators.
> 
> It is not intuitive. I think is better to map this functionality to 
> basic operators +- * and implement it just for pairs (Multirange, 
> Multirange) and (Multirange, Range) if it is possible
> 
> It's same relation line Numeric X integer. There should not be 
> introduced new operators. If somebody need it for ranges, then he can 
> use cast to multirange, and can continue.
 > [snip]
> 3. There are not prepared casts -
> 
> postgres=# select int8range(10,15)::int8multirange;
> ERROR:  cannot cast type int8range to int8multirange
> LINE 1: select int8range(10,15)::int8multirange;
>                                 ^
> There should be some a) fully generic solution, or b) possibility to 
> build implicit cast when any multirange type is created.

Okay, I like the idea of just having `range + range` and `multirange + 
multirange`, then letting you cast between ranges and multiranges. The 
analogy to int/numeric seems strong. I guess if you cast a multirange 
with more than one element to a range it will raise an error. That will 
let me clean up the docs a lot too.

Thanks!

-- 
Paul              ~{:-)
pj@illuminatedcomputing.com