On Sun, Nov 28, 2010 at 11:29 AM, Tom Lane <tgl@sss.pgh.pa.us> wrote:
> Marti Raudsepp <marti@juffo.org> writes:
>> This patch returns command tag "CREATE X" or "REPLACE X" for
>> LANGAUGE/VIEW/RULE/FUNCTION. This is done by passing completionTag to
>> from ProcessUtility to more functions, and adding a 'bool *didUpdate'
>> argument to some lower-level functions. I'm not sure if passing back
>> the status in a bool* is considered good style, but this way all the
>> functions look consistent.
>
> This is going to break clients that expect commands to return the same
> command tag as they have in the past. I doubt that whatever usefulness
> is gained will outweigh the compatibility problems.
You complained about this when we changed the SELECT tag for 9.0 to
include row-counts for CTAS etc. where it hadn't before. Have we
gotten any complaints about that change breaking clients?
I think more expessive command tags are in general a good thing. The
idea that this particular change would be useful primarily for humans
examining the psql output seems a bit weak to me, but I can easily see
it being useful for programs. Right now a program has no reason to
look at this command tag anyway; it'll always be the same.
--
Robert Haas
EnterpriseDB: http://www.enterprisedb.com
The Enterprise PostgreSQL Company