RE: Indicators

[CWilt@xxxxxxxxxxxx]

Michael,

At the risk of starting a religious war, the only thing you're missing is
"the point".

What you've written is perfectly acceptable RPG II.  It is not acceptable
RPGIV.  Conditioning indicators are simply a big no-no because they are
easily missed during maintenance.  The explicit IF is quickly seen.
Likewise using a named indicator instead of numbered.  

The point isn't that RPGIV takes less lines of code to do a given
"something".  In fact, RPG IV will take more lines initially.  The benefit
comes when you start to reuse procedures.  For example:

c                   if            not (DateIsOk(PRMSDT:PRMFMT) and
DateIsOk(PRMEDT:PRMFMT))
c                   eval          PRMERR = *ON
c                   endif


p DateIsOk        b 
d DateIsOk        pi              n
d  pDate                         8s 0     value
d  pFmt                          5a       value

c                   select
c                   when          pFmt = cDATE_FORMAT_ISO
c     *ISO          test (D)                pDate
c                   when          pFmt = cDATE_FORMAT_USA
c     *USE          test (D)                pDate
c                   endsl

c                   return        not %error
p DateIsOk        e


The point is to write the code to validate a date once, sick it in a service
program and use it anyplace you need to validate a date.

Which is quicker and easier to understand, my 3 lines of code or your 6.
Additionally, you'll need to keep writing 6 lines of code, while I'll only
need to write 3.

HTH,
Charles



> -----Original Message-----
> From: Michael Naughton [mailto:michael_naughton@xxxxxxxxxxxx]
> Sent: Wednesday, August 18, 2004 11:50 AM
> To: RPG programming on the AS400 / iSeries
> Subject: Re: Indicators
> 
> 
> I've tried to stop using indicators, too, but sometimes they just seem
> really handy. The other day, I came across a situation where I just
> couldn't see _not_ using them. I wanted to test two 8,0 
> fields to see if
> they held valid date values, based on either *ISO or *USA 
> format, and if
> either one of them was bad I wanted to set an error flag.
> 
> My code ended up looking like this:
> 
>  *                  Test passed date & set error flag if bad          
> C                   Select                                            
> C                   When          (PRMFMT = '*YYMD')                  
> C     *ISO          TEST (D)                PRMSDT                 99 
> C  N99*ISO          TEST (D)                PRMEDT                 99 
> C                   When          (PRMFMT = '*MDYY')                  
> C     *USA          TEST (D)                PRMSDT                 99 
> C  N99*USA          TEST (D)                PRMEDT                 99 
> C                   EndSL                                             
> C   99              Eval      PRMERR = *ON      
> 
> The shortest non-indicator version I could come up with was this:
> 
>  *                  Test passed date & set error flag if bad
> C                   Select                                  
> C                   When          (PRMFMT = '*YYMD')        
> C     *ISO          TEST (DE)               PRMSDT          
> C                   If             NOT %Error               
> C     *ISO          TEST (DE)               PRMEDT          
> C                   Endif                                   
> C                   When          (PRMFMT = '*MDYY')        
> C     *USA          TEST (DE)               PRMSDT          
> C                   If             NOT %Error               
> C     *USA          TEST (DE)               PRMEDT          
> C                   Endif                                   
> C                   EndSL                                   
> C                   If             %Error                   
> C                   Eval      PRMERR = *ON                  
> C                   Endif
> 
> which is a difference of 16 lines instead of 9. Is there a 
> better way to
> do it? Or is there some other reason why avoiding indicators 
> in this case
> is "better"? I fully agree that in many (most?) situations there are
> better ways, but "most" is not "all", and I'm not sure that indicators
> won't remain useful in the future. . . .
> 
> Am I missing something?
> 
> (PS: don't bother telling me that when I switch to /free I 
> won't be able
> to use indicators. I know that. If I'm forced to switch, I'll give up
> indicators, but IMHO that's not a good reason. So far, I find 
> the benefits
> of /free underwhelming, and I'm a guy who has done a fair amount of
> programming in Visual Basic and DIBOL -- both very free-format.)
> 
> Ron writes:
> >Yes I stopped using numbered indicators a while ago, 
> 
> 
> Mike Naughton
> Senior Programmer/Analyst
> Judd Wire, Inc.
> 124 Turnpike Road
> Turners Falls, MA  01376
> 413-863-4357 x444
> mnaughton@xxxxxxxxxxxx
>