I uploaded some REXX code to
http://code.midrange.com/bdaa9cf4d1.html if anyone cares to refer to
it. It's a simple proc that I sent to IBM to find out why it wasn't
working.
I use REXX for a lot of utility work. In this case, the proc was a
variant of one that I use to test basic SQL code. I have a CL
wrapper that runs the proc with the QREXX API, while setting the
*EXECSQL command environment. Then it calls the QREXQ API to
retrieve a return value from the REXX external queue.
Usually all I need to do is modify the SQL in the middles of the
proc and call my CL. The returned value gets checked by the CL to
see how well the SQL I'm checking worked.
Just a general utility function for myself...
In this case, I was checking a formula that returned the date of the
end of the month after current month. In the past, I had usually
used a SELECT from SYSDUMMY1; but this time I tried to use VALUES
INTO, just because I hadn't used it in REXX before and because the
SQL docs said VALUES INTO was valid in REXX.
But it simply would not return anything. No error, but no value.
So, I started fiddling with it, trying every variation I could think
of. No value; no error. The uploaded proc is a little more messy
than it should be due to the variety of changes, but the intent
should be clear enough, and surely IBM support could work with it
with me answering any questions.
Eventually I ended up with the simple formula that's in the uploaded
code -- CHAR(DATE(CURRENT_DATE - 15 DAYS)) -- any formula that might
work.
The code has both the VALUES INTO and the SELECT from SYSDUMMY1
methods. That's just to send to IBM to show one working and the
other not working.
After a few days being reviewed by them, the apparent resolution
came back:
VALUES INTO isn't working in REXX. Further, it apparently never has
worked. Further, it's not judged to be a trivial fix. And since not
many people use REXX, the current decision is to change the
documentation, if possible for 6.1 (on-line, apparently) but
definitely for [next release].
I said that my immediate reaction was that many people don't use
REXX because IBM has seemed to abandon it. And that I'd think about
it before giving a final reply.
The REXX SQL ability (on i) is unique. It's really quite handy,
especially on systems that don't have the SQL Dev Kit. And the
interpreted nature of REXX makes it (on i) uniquely well suited for
a lot of rapid testing of concepts.
Personally, I'm irritated at the thought that the IBM default
position is "This looks hard. Let's not do it." Bluntly paraphrased,
but maybe still accurate.
But I can also see how decisions on resources have to be made. Is it
worth a developer being allocated if it takes away some feature of
[next release] that the majority of the community would rather have?
So, I figured I'd just rant about it here in case anyone else wanted
to say anything. I'm wavering between letting it go on one hand and
demanding "Fix it!" on the other.
Tom Liotta
midrange.com
