Re: CHAIN Versus SETLL and READ When Data Needed

On 2/28/07, Dunfee, Randy <rdunfee@xxxxxxxxxxxxx> wrote:

I would prefer that the LEAVE opcode had never been added.  IMHO, it is
EVEN WORSE than a GOTO because the destination is implicit.

People have deeply felt (and not necessarily rational) feelings on this subject.

All I have to say about LEAVE (and ITER) today is what I had to say about
GOTO twenty years ago; if I ever need it I will use it, but I have never needed
it.

I believe that when I see a conditioned DOU loop or a DOW loop that I
have the right to expect that the ONLY way out of this loop is to meet
the stated condition, and I can therefore skip desk-checking the "loop"
code if that particular condition is not of interest to me.  Is that
REALLY too much to ask?

Straight from the Gospel according to Dijkstra (Go To Stament Considered
Harmful) (http://www.acm.org/classics/oct95/).

If the code is to be self-documenting, then all of the Do arguments need to
be 'invariant' (i.e. the loop *can not* be exited until the stated
conditions are
meant).

While it is possible to write non-spaghetti (structured) code with
LEAVE/ITER/GOTO, it is very easy to fall into bad habits.

But the people who use GOTOs (and its 'structured' bastard cousins) can be
very defensive; claiming it is a bad practice is a good way to start
an argument.

Bottom line is, if the code works, it works. I, personally, find code with
GOTO/LEAVE/ITERs hard to maintain, and really wish free-form RPG did
not support LEAVE/ITER (similarly to the way it doesn't support GOTO)..

Chris

P.S. This discussion reminds me of my freshman programming class in
college (rough 1979); Dr. Perlman (our instructor) said "If you use a GOTO,
I will break your fingers, and then I will fail you."

--
chris@xxxxxxxxx           | Nothing is true; All is permitted
www.pando.org             |   Hasan i Sabbah