Re: Most-common value docs in PG 12

On Thu, Sep 26, 2019 at 04:20:59PM -0400, Bruce Momjian wrote:
>On Thu, Sep 26, 2019 at 05:17:55PM -0300, Alvaro Herrera wrote:
>> On 2019-Sep-26, Bruce Momjian wrote:
>>
>> > On Wed, Sep 11, 2019 at 06:15:22PM -0300, Alvaro Herrera wrote:
>> > > On 2019-Aug-30, Bruce Momjian wrote:
>> > >
>> > > > OK, how is this patch?  I didn't mention psql since I think everyone
>> > > > expects psql to show all information about tables and indexes.
>> > >
>> > > Why would you change perform.sgml?  It seems unnecessary; the commands
>> > > shown work fine.
>> >
>> > I realize they work fine, but the ordering in the examples not matching
>> > the defined order suggests that ordering matters, but it does not.
>>
>> Well, I mean exactly the other way around: the fact that the orders
>> don't match illustrates that the order is not important.  And that is
>> reinforced by the explanation indicating explicitly that it does not
>> matter:
>
>Uh, people normally list things in defined order, so you would usually
>not list them in non-defined order unless there is a purpose.  Doing
>that just to illustrate the order doesn't matter seems odd.
>

Well, that assumes there is a definition, and I don't think the zipcodes
table is defined anywhere. So how do you know in what order are those
columns defined?

Now, maybe the table should be defined somewhere in perform.sgml - I
don't recall why exactly I chose not to do that, maybe because there is
no universal definition (one country uses text, another number, ...).

I do however agree that had there been such definition, it's probably
natural to list columns in the same order. We know the order is not
important, the proposed patch states that explicitly, but this just
feels natural.

regards

-- 
Tomas Vondra                  http://www.2ndQuadrant.com
PostgreSQL Development, 24x7 Support, Remote DBA, Training & Services