Re: SELECT block enhancement?

[mlazarus <mlazarus@xxxxxxxxxxxx>]

Joe,

 That's onother good example of where enhancements can be made to the 
SELECT block.  Once the conditions and fallbacks start piling up that IF 
statement can get pretty nasty.

 -mark

On 1/29/2019 8:47 AM, Joe Pluta wrote:
> Here's another wrinkle, Mark, that's very similar.  I run into this 
> all the time.  The basic design pattern is this: I am checking a 
> record for some condition.  If the condition is met, great.  If not, 
> then I have to check a different file.  Very similar to yours.  I'd 
> like to see something like this:
>
> CHAIN (X:Y:Z) ORDERLINE;
>
> SELECT;
>
>     WHEN OL_QTYREM = 0;
>         STATUS = '90';
>
>     WHEN OL_LINETYPE = 'C';
>         STATUS = '80'';
>
>     // It didn't match either condition, so now I have to check the 
> header
>     NOSELECT;
>         CHAIN (X:Y) ORDERHEADER;
>
>     WHEN OH_DIVISION = 'USA';
>         STATUS = '60';
>
>     OTHER;
>         STATUS = '40';
>
> ENDSL;
>
> In this pattern, I always have to chain to the order line, but I don't 
> always have to chain to the header.  Depending on how often the lines 
> match one of the first conditions, this could significantly reduce my 
> I/O.  So the compiler would see the "NOSELECT" clause and would 
> execute that code prior to moving ahead to the next WHEN statement.
>
> The is the kind of code that ends up in an IF/ENDIF block now, which 
> isn't necessarily bad, but it does start indenting:
>
> IF OL_QTYREM = 0;
>     STATUS = '90';
> ELSEIF OL_LINETYPE = 'C';
>     STATUS = '80'';
> ELSE;
>     CHAIN (X:Y) ORDERHEADER;
>     IF OH_DIVISION = 'USA';
>         STATUS = '60';
>     ELSE;
>        STATUS = '40';
>     ENDIF;
> ENDIF;
>
>
> On 1/28/2019 11:38 PM, mlazarus wrote:
> > I'm throwing this out there to see what people think of the idea.
> >
> > Has anyone run into the need to have a "NEXT" or "CONTINUE" or 
> > "NOBREAK" opcode in a SELECT block?  It would move on to the next 
> > WHEN condition instead of exiting the entire SELECT block.
> >
> > GotIt = *Off ;
> >
> > SELECT ;
> > WHEN X = 1 ;
> > CHAIN FILE1 ;
> > IF %Found( FILE1 ) And DeleteFlag1 = *Blank ;
> >    GotIt = *On ;
> > ELSE ;
> >    NOBREAK ;
> > ENDIF ;
> >
> > WHEN Not GotIt ;
> > CHAIN FILE2 ;
> > IF %Found( FILE2 ) And DeleteFlag2 = *Blank ;
> >    GotIt = *On ;
> > ELSE ;
> >    NOBREAK ;
> > ENDIF ;
> >
> > WHEN Not GotIt ;
> > CHAIN FILE3 ;
> > IF %Found( FILE3 ) And DeleteFlag3 = *Blank ;
> >    GotIt = *On ;
> >
> > ENDSL ;
> >
> >   This is a simple example, but it demonstrates a desire to have a 
> > "cascading" set of conditions, using the existing SELECT structure, 
> > but allowing the next set of condition(s) to be executed, instead if 
> > needing to repeat the previous conditions or adding an exit value for 
> > nested conditions.
> >
> >   We've all found ways around it, but I think it would be a nice 
> > addition.  It's basically the opposite of C's break opcode.
> >
> >  -mark