Re: More on RPG style

[Rory Hewitt <rory.hewitt@xxxxxxxxx>]

Dennis,

Some people would point out that using LEAVE violates a programming rule
(since it's really just a GOTO in disguise). I wouldn't necessarily agree
with them, though. Rules, schmules.

Here's my equivalent code:

http://code.midrange.com/125e58e898.html

As you can see, almost no difference - it still uses a single entry and exit
point, no more if/else logic. As far as "control flowing to the end of the
stream in the event of an exception", is that a good idea anyway? Surely it
should be made clear in the code that when an exception occurs, it should be
'handled' immediately - in my case, this means going straight to a dedicated
subroutine. In your case however, the user needs to scan down (possibly
several pages of) the source to find the ENDDO... Easier with consistent
indentation and WDSc, but not as easy as searching for 'RETURN'.

However, if one were to add another DO- or FOR-loop outside the main one (or
indeed, if any of your IF clauses was changedto a DO- or a FOR-loop), then
your code would need some changes, to ensure that the LEAVE exits both
loops, whereas my code would continue to work. I know, I know, it's easy to
think up situations where existing code would fail if someone made a
'flow-related' change without thoroughly seeing the ramifications, but I *
think* that my code is *slightly* more 'bulletproof' to future idiots in
that regard.

I guess I don't see where my approach is any more ugly than yours, but
that's the thing about code - one man's beauty is another man's ugliness :)
As long as we both have a consistent strategy for error-handling, then we're
both good.

Rory

On Tue, Mar 2, 2010 at 10:23 AM, Dennis Lovelady <iseries@xxxxxxxxxxxx>wrote:

> > I make a point of (generally, not always) having a subroutine called
> > return.
> > In my code, I will code an 'exsr return' instead of simply the RETURN
> > opcode, even though the 'return' subroutine may contain no more than
> > the
> > RETURN op-code.
>
> > Why?
>
> Here's an alternative for you to consider:
> http://code.midrange.com/85da238f9e.html   Maybe it will shed some light
> on
> an alternative method that doesn't violate any programming rules, doesn't
> add a bunch of new if/else logic, still allows control to flow to the end
> of
> the stream in the event of exception, et cetera, but with only one exit
> point.
>
> Please forgive the stray /free at the top.  It's a concept, not a working
> program.
>
> I think (and I seem to be alone here) that the effort is very well worth
> it!
> And with this particular approach, the effort is minimal.  To me (with the
> afore-mentioned 'alone' caveat) those other approaches ARE ugly next to
> this.
>
> Dennis Lovelady
> http://www.linkedin.com/in/dennislovelady