Re: pgbench \for or similar loop

On Wed, Apr 20, 2011 at 10:46 AM, Merlin Moncure <mmoncure(at)gmail(dot)com> wrote:
> On Wed, Apr 20, 2011 at 9:14 AM, Alvaro Herrera
> <alvherre(at)commandprompt(dot)com> wrote:
>> Excerpts from David Fetter's message of mié abr 20 10:54:56 -0300 2011:
>>> On Wed, Apr 20, 2011 at 08:05:07AM +0200, Pavel Stehule wrote:
>>> > Hello
>>> >
>>> > I played with psql extensions two years ago - it can do it
>>>
>>> It's interesting, but it doesn't solve the fundamental problem, which
>>> is to allow every client, not just psql, to do this.
>>
>> Why is this problem fundamental?
>
> I happen to like your idea, even if we had stored procedures...they
> have a lot of overlap but so what?.  We have server side \copy and
> client side COPY -- both are useful.  Likewise, (getting back to the
> original point of the thread), bechmarking via client scripting and
> via procedure are also both useful.  Nobody will gripe if psql gets
> more features like this -- some people really want to do this on the
> client side and there are valid reasons to do that, say, to intermix
> client local shell commands between sql lines.

Yep, I agree. However, I think it's completely reasonable, as I said
upthread, to ask people not to implement \for as a loop over an
integer range in pgbench without answering questions like:

1. What happens if someone wants the other kind of for loop, that
iterates until a condition is met?

2. Are we going to get a different and incompatible implementation in psql?

--
Robert Haas
EnterpriseDB: http://www.enterprisedb.com
The Enterprise PostgreSQL Company