Midrange News for the IBM i Community


Posted by: Bob Cozzi
Rogue Programmer
Cozzi Productions, Inc.
Chicagoland
Test for numeric overflow
has no ratings.
Published: 01 Nov 2012
Revised: 23 Jan 2013 - 1707 days ago
Last viewed on: 25 Sep 2017 (2856 views) 

Using IBM i? Need to create Excel, CSV, HTML, JSON, PDF, SPOOL reports? Learn more about the fastest and least expensive tool for the job: SQL iQuery.

Test for numeric overflow Published by: Bob Cozzi on 01 Nov 2012 view comments(3)

Wow! I can't believe the number of numeric overflow operations found in legacy code that is converted to /Free syntax.

This is when you have a target variable that is (for example) defined as 7p2 and the code moves a value to that field that is larger than 5-digits, such as: 99999.00 + 1.00

So much legacy code.... So I'm using the ChkDecOvr()  (check for decimal overflow) subprocedure to first validate if it will fit and then if it won't take an alternate course of action; rather than just blow up. Or I could wrap the code in MONITOR/on-error groups.

 

/free
    if chkDecOvr( sales * qty : %len(ordTot) : %decPos(ordTot));
       // Value too big to fit in target variable!
    else;
       ordTot = sales * qty; // We're good.
    endif;
/end-free

 

Anyway, we've added ChkDecOvr() to the COZTOOLS runtime.

 

Return to midrangenews.com home page.
Sort Ascend | Descend

COMMENTS

(Sign in to Post a Comment)
Posted by: neilrh
Premium member *
Jackson, MI
Comment on: Test for numeric overflow
Posted: 4 years 10 months 24 days 4 hours 2 minutes ago
Edited: Fri, 02 Nov, 2012 at 09:04:18 (1789 days ago)

Yup, just this last week, 2 programs, 2 calcs in each... 

// Calc percentage
monitor;                                
  eval(h) Result = (Value1 * 100) / Value2;
on-error;                               
  if Value1 < 0 and Value2 < 0 or       
     Value1 > 0 and Value2 > 0;         
    Result = 100;                          
  else;                                 
    Result = -100;                         
  endif;                                
endmon;                                 

 But they're breeding like orange barrels on a highway.

Posted by: TFisher
Premium member *
Comment on: Test for numeric overflow
Posted: 4 years 10 months 24 days 3 hours 39 minutes ago

I use the MONITOR approach for the best possible performance.

Posted by: DaleB
Premium member *
Reading, PA
Comment on: Test for numeric overflow
Posted: 4 years 10 months 24 days 3 hours 27 minutes ago

I've also used the Monitor variant, but the chkDecOvr() looks interesting.

When converting existing code, I needed to emulate what the fixed format would have done so the program behaved the same way. Early on I think I did something with %REM(x:SomePowerOf10), but that gets messy when you have non-zero decimal places. A much simpler alternative is to use a DS, OVERLAYing so the decimal points align. In some cases you can do this with Packed, but Zoned is a no-brainer.