Re: Adding pointers to RPG language.

[boldt@xxxxxxxxxx]

Simon wrote:
>I don't think you should regret adding pointer support to RPG IV.  Giving
>us pointers has meant that RPG can now do many things previously
>impossible or only posible with difficulty and truly ugly data structures.

As I said in my previous note (to Jim), perhaps "regret" was too
strong a word.  It's probably just a reflection of my frustration
in seeing pointers used in places where there are better
alternatives.

>Four major benefits from having pointers are:
>          1/ Based storage which makes processing user spaces easier and
>faster than the alternative and gives us access to dynamic storage thus
>more flexible array processing

This is the "dynamic array" issue.  As I've said before, handling
undetermined quantities of data is probably better handled in the
database, and not within program code.

(I like how undetermined quantities of data are handled in other
languages like Python, Rexx, Perl, and Ruby.  Varying sized lists
and hash tables are fundamental data structures in these languages.
But then, these have the advantage of being interpreted, not
compiled languages.)

>          2/ The ability to invoke C-runtime functions including those
which
>expect char** data types

This is not an unreasonable use of pointers, in my opinion.  The
C-runtime library should be considered a normal part of the RPG
programmers tool kit.

>          3/ Cleaner code -- there is a valid aphorism regarding coding
that
>states "Get the data structures correct and the code falls into place".
>It manifests itself in general as:
>                    complex data strucures = simpler code
>                    simpler data structures = complex code

I agree with the statement "Get the data structures correct and
the code falls into place".  I've said that many times myself.

I don't necessarily agree with your other.  Often complex data
structures can result in complex code.  The key point is getting
the data structures "right", which is different that making the
data structures complex.  I know from experience that there are
lots of complicated ways to design your data structures.  But
there are "good" ways to design the data structures that yield
both simple data structures and simple code.

For the average RPG programmer, I believe the key issue is
getting a proper database design.

Complexity seems to happen over time as functionality is added
to an application, and the developers cannot afford to
redesign existing data structures.

>          4/ Function pointers which give us dynamic call-back ability and
>therefore the ability to allow procedures to be enhanced or modifed by the
>caller -- if they've been designed correctly.

Function pointers are a different issue from the current
discussion that involves basing pointers.  If you're looking
at a lot of "call-back" functions in your application, you
probably should be using an object-oriented language instead.

>Just because certain programmers misuse or misunderstand pointers is no
>reason for you to regret providing them.  Carpenters cut themselves with
>chisels every day but that's no reason to remove chisels from the toolbox.
>You cannot protect the fools from themselves and if you try then you
>simply irritate the rest of us.

And I've got the scars to prove that point!  ;-)

>I can think of only three peculiarities involved in providing pointer
>support:

Only three?  ;-)

>          1/ Not allowing pointer arithmetic in the first release
>(thankfully that has been corrected but it was obviously a misguided
>attempt to protect RPG programmers from themselves)

Pointer arithmetic was indeed added quite a few releases
ago.

>          2/ Automatically declaring a pointer used in based storage (and
>I'm in two minds about that -- I frequently use that ability to advantage
>but like others get bitten by misspelling the pointer name which in
>reality is no worse than misspelling a field name)

Yes, we probably should have required explicit pointer
declaration.  I often code an initial value on a pointer
definition anyways when coding pointers.

>          3/ Not allowing BASED(%ADDR(somefield)) (which is likely to be
>fixed sometime in the future)

Not in the immediate future.  But who knows what will be
decided on in the next few years?

>The RPG compiler team is doing a good job on the whole.  There are
>certainly peculiar things done and I would certainly do some things
>differently but mostly everything you do improves the language (regardless
>of what Bob Cozzi would have us believe).
>
>I would not want to see typed pointers in RPG but I do want to see
>user-defined types (even COBOL has that ability!).  That's long overdue
>and I would spend my entire $100 on that single enhancement.  Then we can
>move on to overloaded procedures, he says hopefully.  I'd far rather see
>new features in the language than new ways of doing the same old stuff
>(e.g., %CHAIN, %READ, jeez give me a break :)

Cheers!  Hans

Hans Boldt, ILE RPG Development, IBM Toronto Lab, boldt@ca.ibm.com