Re: Target for numeric operation too small...hello, Toronto?

["David Morris" <David.Morris@xxxxxxxxxxxxx>]

Hans,

No sequence is available in SQL but you can write your own UDF. One
technique There is one on IBM's porting Web page that is pretty much
a clone of Oracle sequences, or you can write a simpler version that
uses a data area. In the one I wrote, I did check for overflow and it
wasn't too bad.

David Morris

>>> boldt@ca.ibm.com 12/06/01 12:55PM >>>
Tom wrote:
>Well:
>
>D PBcontrol     E DS                  ExtName(PBSCONTROL) Prefix(da_)
>D                                     DtaAra(PBACONTROL)
>...
>In *Lock PBcontrol;
>da_Sequence = da_Sequence + 1;
>Out PBcontrol;
>pa_Sequence = da_Sequence;
>...
>
>Where Sequence is a 7 digit number that used to, and still should,
just
roll
>over at hival.  (This is just one example of many sequence numbers)

Oh I fully realize there may be one or two reasonable applications
for numeric truncation.  And this may well be one of them.  But is
it worth all the bugs that may be masked?  As others have pointed
out, do you really want to risk seeing $2.34 on your paycheque when
it should be $1002.34?  Or leave out high-order digits on customer
bills?  (Hmmm...)  In this case, coding a MONITOR group around the
EVAL is a small price to pay, and doesn't even cost anything in
terms of run-time performance.

And if we provided an option to mask the overflow exception, how
many people would just blindly code the option without fully
analyzing the real problem?

(I was going to suggest using a SEQUENCE in SQL for this kind of
functionality, but I'm not sure SQL on OS/400 has that particular
extension.)

Hans Boldt, ILE RPG Development, IBM Toronto Lab, boldt@ca.ibm.com