Re: Unlimited size result sets

> 100Meg return variables? Show me that in the "real world".

Hi Bob!

I think that pretty much everyone on the list agrees that returning a
several hundred thousamd record result set is not really workable.  The only
reason I posted in the first place is that I believe that other RPG
programmers will be asked to do the same thing I am being asked to do.  As
we begin using the web, we green screen people will be pooh-poohed by the
web people.  'You are good at green screen but this is a new world; a new
paradigm.  You need to change your ways and step into the new era.'  And in
truth, there probably is a lot I could learn if I shed my inhibitions and
give things a whirl.

In this particular case, I am being asked to produce an unlimited size
result set.  Because 'no other OS limits the result set.'  So, if RPG can't
handle that, then RPG is weak, and by extension, so is the iSeries.  The
whole point is that Java/web people who are familiar with other database
systems think nothing of making a request like that, and if it turns out
that I, an RPG programmer can't fulfill that request, then I am going to be
seen as holding the company back.  Along with the iSeries.

If, however, I can fairly easily do what she asks, and it turns out to have
been an unworkable solution, then myself, RPG and the iSeries are not going
to be seen as the weak link here.  And in fact, my assertion that this is
not an optimum design choice will be proved true.  Which means that at least
_some_ of those old green screen ideas still hold water in the new era.

Now, it is quite possible that I am overreacting; I've certainly done that
before.  But I feel like I am on trial, having to prove myself and my
iSeries all the time.  So far, I've been holding my own, but it looks like
RPG alone has a hard limit on the size of the result set which can be
returned to a stored procedure.  The SQL procedure calling RPG UDFs is
working great.  It was easy to implement, and it was a drop-in replacement
for the existing stored procedure.  So our Java person got to try it out on
Friday.  On her PC, running Apache/Tomcat on Windows XP with 256MB of RAM,
she got about 85,000 records in 6 minutes before getting a 500 error on
Apache.  By comparison, I was able to print (well, spool) roughly 11 million
records in 11 minutes.

I am trying really really hard to keep this from becoming an 'us vs. them'
situation, and by remaining flexible and delivering requested changes
quickly, I think I am able to demonstrate to management that a co-operative
Java-RPG project is going to work.  Because the alternative is Java-only,
touching the database directly via JDBC.

I hope I have not muddied the waters.
  --buck