Re: Variable substitution in psql backtick expansion

On Mon, Nov 13, 2017 at 5:21 AM, Tom Lane <tgl@sss.pgh.pa.us> wrote:
> Pavel Stehule <pavel.stehule@gmail.com> writes:
>> [ psql-server-version-2.patch ]
>
> I think this patch should be rejected.  It adds no new functionality;
> you can get the string in question with "select version()".  Moreover,
> you've been able to do that for lo these many years.  Any application
> that tried to depend on this new way of getting the string would fail
> when working with an older server or older psql.  That does not seem
> like a good property for a version check.  Also, because the string
> isn't especially machine-friendly, it's not very clear to me what the
> use-case is for an application to use it at all, rather than the other
> version formats we already provide.

+1 for rejection as version() returns PG_VERSION_STR already. It is
also already possible to define a VERSION variable psqlrc simply with
that:
\set VERSION 'version();'

+-- check consistency of SERVER_VERSION
+-- which is transmitted as GUC "server_version_raw"
+SELECT :'SERVER_VERSION' = VERSION()
+   AND :'SERVER_VERSION' = current_setting('server_version_raw')
+   AND :'SERVER_VERSION' = :'VERSION'
+       AS "SERVER_VERSION is consistent";
Not much enthusiastic with this test when thinking about cross-upgrades.
-- 
Michael