Re: CL pgm and pointer variables

Steve Richter wrote:

Eric,

Pointers are nothing to be afraid of. ( regardless of what Hans says ) And
why not use them when they allow your code to be more direct, easier to
read, more to the "point"?

Rarely the case in practice.

There are problems, maybe better said as "potential sources of errors", when
using pointers.  The first PSE is that pointers are typeless, that is the
compiler does not know whether the pointer points to a packed decimal, a
character string or data structure. A good start in dealing with this
problem is to allow a pointer to be declared with a type attribute:
  d pName       s           *    Like( OrdHdr.CustName )
  d pOrdHdr     s           *    likeds( OrdHdr )

As a seasoned C programmer, I know how often the C compiler
complains to *me* about pointer type mismatches. It's great to get
these diagnostics since it means that the error won't have to be
debugged at run-time. As you said, since RPG pointers are
"type-less", the RPG programmer has to wait until run-time to
discover pointer type problems (or worse, NOT discover the problems!).

Enhancements to the RPG language would be nice, but that still
doesn't mean a programmer has to use them.

An additional way to reduce type mixup errors is to permit data pointers to
be used in rpg.  Data pointers were built into the original s/38 by some
genius who was surely banished from Rochester long ago. A data pointer
contains both a pointer to something and also the attributes of the
something.  This elimiates the type mixup and makes for even cleaner code.
 d Name        s          30a
 d AnotherName s          30a
 d Amount      s           7p 2
 d dpSomething s            *     dataptr
   /free
     dpSomething = %AddrAndAttr( Name ) ;
     AnotherName = dpSomething ; // copies Name to AnotherName
   /end-free

You want data pointers? Program in MI. But in practice, over the
past 20 years, I can only think of one place where I've ever seen
data pointers put to use.

The 2nd source of error is not with the pointer itself, but the requirement
to free the memory that has been allocated as the program runs. That is you
must Dealloc what has been %alloc'd.

Or worse, using a pointer *after* the memory has been DEALLOC'ed.
(If you don't DEALLOC, the memory will still be freed anyways when
the AG ends.)

The current RPG ILE is wide open for this kind of error.  To deal with it
you either go the java route and handle all alloc and dealloc internally. Or
go the C++ route and implement real good scoping with constructor and
destructor code that runs automatically on scope entry and exit (
procedure/program entry and exit ).

Alternatively, you can just save yourself all the trouble of
pointers and use alternatives.

BTW, ever notice how pointers are *not* a feature of many modern
programming languages? Not just Java, but also Perl, Python, Ruby, etc.

(Oh, to be clear, pointers can be useful when using system API's.
But that falls into the realm of systems programming, for which C is
a more approriate language.)

Cheers!  Hans