07 января 1987

        PROGRAM STRUCTURE
         By Brian Cavers
        _________________

I wish somebody would explain
what the difference is between a
procedure and a subroutine. No,
I don't mean the superficial
difference I mean the structural
difference. Let's look at a
definition:-.


A BASIC PROGRAM  IS  A  LIST  OF
INSTRUCTIONS WHICH  IS  FOLLOWED
IN  SEQUENCE  BY  THE  COMPUTER.
THERE IS A SPECIAL  INSTRUCTION,
FOLLOWED BY A LABEL,WHICH CAUSES
THE COMPUTER TO JUMP OUT OF  THE
SEQUENCE AND TO CONTINUE FROM  A
PLACE IN THE PROGRAM POINTED  TO
BY  THE  LABEL.  THE    COMPUTER
CONTINUES  TO  CARRY  OUT    THE
INSTRUCTIONS   FOLLOWING    THIS
POINT,  IN  SEQUENCE,  UNTIL  IT
MEETS  ANOTHER  SPECIAL  COMMAND
WHICH CAUSES IT TO JUMP BACK AND
CONTINUE  WITH   THE    ORIGINAL
SEQUENCE..
It seems to me that this could
equally be the definition of a
procedure or of a subroutine.
The "special instruction": PROC
with a label is no different to
GOSUB with a label and ENDPROC
is the exact equivalent of
RETURN. I realize that the
Spectrum doesn't support true
labels but they can be simulated
with variables and other
machines do provide labelled
GOSUBS.

I own an Atari ST with FAST
BASIC which provides both
procedures and subroutines. The
start of the subroutine is
marked with a label and the
finish with a RETURN. It is
called with GOSUB LABEL in the
.
usual way. The start of a
procedure is marked with DEF
PROC LABEL and the end with
ENDPROC in the usual way. It is
called from the main program
with PROC LABEL.

The manual actively discourages
GOSUBS as if they contribute to
bad programming and it is stated
that they were only included in
the BASIC to be compatible with
other BASICS! However, it does
not give any rational argument
as to why it is thought that
procedures are structurally
superior. After a week or two
programming using procedures I
discovered that FAST BASIC
executes subroutines FASTER than
procedures! I know what I shall
use in future!.
Actually this is what infuriates
me about what I call "structure"
fanatics. There is this smug,
superior claim, as if it is
self-evident, that subroutines
are obsolete and that one
"should" use procedures but I
HAVE NEVER YET SEEN ANY
EXPLANATION! THEY REALLY DO LOOK
LIKE IDENTICAL COMMANDS TO ME! I
CHALLENGE ANYONE TO PROVIDE A
RATIONAL EXPLANATION AS TO WHY
PROCEDURES ARE STRUCTURALLY
SUPERIOR TO SUBROUTINES.

The poor old GO TO takes even
more of a hammering from the
structural purists. I realize
that numerous jumps back and
forth can turn a program into
a debugger's nightmare but I.
cannot see what is wrong in
having several main routines
making up a program, each one
being jumped to as needed. For
example, in a database: INPUT
ROUTINE, SORT ROUTINE, PRINT
ROUTINE, DISPLAY ROUTINE etc.

Additionally, why is:

  10 LET A = 1
  20 LET A = A + 1
  30 PRINT A
  40 IF A <25 THEN GO TO 20

any less structurally sound
than:

  10 FOR A = 1 TO 25
  20 PRINT A
  30 NEXT A.
or:

  10 LET A = 1
  20 WHILE A <=25
  30 PRINT A
  40 LET A = A + 1
  50 WEND.

In every case what the program
is actually doing is jumping
back in a loop. The loop is the
structure - the labels: GO TO,
FOR NEXT, WHILE WEND, to me, are
cosmetic - irrelevant to the
structure of the program. (This
is not to say, of course, that
these commands don't all have
their uses in the right place -
they do but it's not
STRUCTURALLY RELEVANT in my
view!)..
Another thing! (I'm really
getting annoyed now - but then,
this is SHOUTBURST isn't it!).
In many BASICS, the computer
keeps track of the program's
position in FOR NEXT loops by
use of a stack, I believe. This
means it could cause problems if
you jump out of the middle of a
FOR NEXT loop without finishing
it properly. This has given rise
to the nonsense that it is
structurally inefficient to
escape FOR NEXT loops in that
way. On the Spectrum, the FOR
NEXT variable is just another
variable and can be abandoned at
any time and another loop
started using the same variable,
for example:-.
  10 FOR N = 1 TO 25
  20 GO TO 50
  30 NEXT N
  40 STOP
  50 FOR N = 1 TO 4
  60 NEXT N

NO PROBLEM TO THE SPECTRUM AND
NO STRUCTURAL PROBLEM EXCEPT IN
SOMEONE'S IMAGINATION!

To sum up, I think what make's
my blood boil most is not the
claim that certain features are
structurally superior but that
such claims never seem to be
accompanied by rational,
objective supporting evidence or
arguments. After all, it is not
up to me to support my position
- I'm not the one who's.
SUPPORTING anything (in this
particular topic). It's the
people who claim that certain
labels somehow CONSTITUTE
correct structure who need to
prove what they say is true! To
finish, I'd just like to
emphasise how important I
believe structure is in
programming - I take a great
deal of care when writing the
structure of my own programs.
However, I believe that
structure lies in the WAY THE
PROGRAMMER ARRANGES HIS
INSTRUCTIONS AND NOT IN
SUPERFICIAL LABELS!

Grrrrrrrrrrrrrrr!.