Re: Software landmines (loops)

[rgardner@nyx.net (Ray Gardner)]

ell@access.digex.net (Ell) wrote:
> rgardner@nyx.net (Ray Gardner) wrote:
> 
> >.... Of course my main point was 
> >that the founder(s) of SP did limit the control structures to 
> >single-entry / single-exit structures, and that still stands.
> 
> And your proof is what?
> 
> Elliott

Ell, the "proof" is in my post of 5 Sept 1998, more-or-less 
directed to you personally, which was the result of more than a 
couple of hours of research for *your* benefit.  Not that I 
expected any thanks.  But since you seem to have missed it, I'll 
waste a little bandwidth and repost.  (I could just give you a 
pointer to DejaNews, but you seem to never look up any cites 
people give you.  Consider that you asked for a quote from a 
founder, RCM gives you a direct quote from *the* primary source, 
and you complain that it's out of print and "RCM knows it".  Did 
you expect a primary source from 25 years ago to be in print?  As 
another poster mentioned, that's what libraries are for.  I don't 
know if you live near any major metropolitan areas, but a quick 
check shows two copies on the shelf in Denver and one in Boulder 
in libraries where they can be browsed by anyone.)  So once more: 


[Repost of my message <905028596.876380@iris.nyx.net>]

Ell wrote:
 > >>For example, there's a pernicious myth that exiting (or returning) from
 > >>the middle of a loop is bad, and that the only proper way to write a
 > >>loop is to state the termination condition explicitly, as a predicate
 > >>appearing at the top of the loop.
 > 
 > >Yes.  That pernicious myth is called "structured programming".
 > 
 > Please cite even just a single structured programming text, or founder
 > where what you say here was asserted.

And later Ell wrote:
 > The assertion that "single entry, single exit" is a required, or even
 > desirable, maxim of structured programming is a myth.
 > 
 > No one (including RCM) can show that this maxim is in fact a coding
 > heuristic put forward by any of the founders of the structured
 > paradigm. [Check past posts in this thread.]

RCM had responded with a direct quote from Dijkstra, but Ell 
rejected it without understanding it.  And rejected it again 
after RCM explained it.  

To recap, in Dijkstra's Notes on Structured Programming in the 
classic _Structured Programming_ (1972), EWD referred to the 
single-entry/single-exit property of the flowcharts of the if-
then-else, case-of, while-do, and repeat-until control 
structures.  Ell apparently thought EWD was talking about how 
flowcharts should be represented on a page.  EWD was actually 
using the flowcharts as a graphical aid to make a point about 
program topology. 

To support this interpretation, I'll just quote this from the 
same source, p. 20:

"... restricting ourselves to the three mentioned types of 
decomposition lead to flowcharts of a restricted topology compared 
with the flowcharts one can make when arrows can be drawn from 
any block leading into any other."

EWD then goes on at length to explain why such a topology is 
desireable.

Ell, you might consider this: in the entire monograph EWD 
never mentions goto, exit, break, return or any other change in 
the flow of control outside of the selection and repetition 
statements mentioned above.  And he certainly had no such other 
modes of flow control in mind when he wrote his "GOTO considered 
harmful" letter; he was accustomed to Algol, which has no 
exit/break/return or similar statements.  (Also consider that 
goto-less programming in Pascal has the same restriction, since 
Pascal lacks exit/break/return etc.)

Ell, remember that you asked for a cite from "even a single 
structured programming text, or founder" to support single-
entry/single-exit.  Okay.  Hard to say who you can consider "a" 
founder other than "the" founder, EWD.  But consider Niklaus 
Wirth: "For the intellectual manageability, it is crucial that 
the constituent operations at each level of abstraction are 
connected according to sufficiently simple, well understood 
_program schemas_, and that each operation is described as a 
piece of program with _one starting point_ and a _single 
terminating point_."  [Emphases were italics in the original.]  
"On the Composition of Well-Structured Programs", N. Wirth, in 
ACM Computing Surveys, December 1974 (Special Issue on 
Programming), p.249. 

How about this:  "A _proper program_ is a program with a control 
structure that (1.) has a single entry line and a single exit 
line, and (2.) for each node, has a path through that node from 
the entry line to the exit line."  _Structured Programming: 
Theory and Practice, R.C. Linger, H.D. Mills, B.I. Witt, Addison-
Wesley, 1979, p. 94.  (Ell has already commented on a similar use 
of the word "proper".  Ell, you should understand that the word 
is being used here in a technical fashion as part of a 
*definition*, similar to the use of "well-formed formula" in 
mathematical logic.  And, anticipating that you'll claim that 
this has nothing to do with structured programming per se, I'll 
point out that the entire book is about programming via the use 
of "proper programs" as defined above.) 

Since you mentioned structured COBOL, I'll mention this:

"GO TO statements are not allowed.  The rationale here is to 
force programmers to think ahead and use _only_ 1-in, 1-out 
control structures." ... "STOP RUN may only occur as the last 
statement of the main procedure of a program.  EXIT PROGRAM may 
only occur as the last paragraph of the main procedure of a 
subprogram.  The rationale here is to make the logical exit of a 
program identical to the lexical end of its main procedure." 
_COBOL With Style (Programming Proverbs)_, L.J. Chmura, H.F. 
Ledgard, Hayden, 1976.  A book on structured programming in 
COBOL.

Many other writers have discussed structured programming in terms 
of single-entry/single-exit control structures, often without 
using those terms explicitly.  But what else could they mean when 
they develop a programming style that uses only sequence, 
if-then-else, while-do, repeat-until, and sometimes case-of, 
without ever mentioning or using goto, return, break, exit, 
continue, or any other discontinuity of control?

For an even heavier dose of Dijkstra, try his _A Discipline of 
Programming_ (Prentice-Hall, 1976) where he develops a fairly 
complete theory of proof-oriented _nondeterministic_ procedural 
programming semantics using only assignment and "guarded 
commands" for conditional and repetitive (if ... fi and do ... 
od) statements.  He never explicitly mentions "single-
entry/single-exit", but it's clear from his exposition that 
there is no other form of flow control in his system.

You mentioned (twice) page 72 of Knuth's _Literate Programming_ 
(not _The Art of ... as you've twice mis-cited it, and the 
publisher is CSLI, not CLSC or CLCS as you've twice misnamed 
them).  This page is part of a reprint of Knuth's classic paper 
"Structured Programming with go to Statements", originally in ACM 
Computing Surveys, December 1974.  You might consider what Knuth 
said on p. 18 of your book: "It is impossible to read the recent 
book _Structured Programming_ [Dahl (not Dahle), Dijkstra, and 
Hoare] without having it change your life."  And then take RCM's 
advice and read the book.  Perhaps that's what it'll take to get 
you to believe that Dijkstra was advocating exactly what RCM said 
he was.  (You won't do it because you might find RCM is right.)  

BTW, while it's true that Dijkstra considered abstraction to be 
central to his notion of structured programming, page 72 of 
Knuth's book nowhere says "It was Dijkstra who thinks of 
*abstraction* when someone mentions 'structured programming'."   
It does quote Dijkstra as saying that when he's dealing with one 
level of abstraction, it occupies 90% of his attention while the 
rest is kept in mind in the background.  Yes, abstraction is key 
to Dijkstra's notion of structured programming.  But so is the 
use of the limited set of control structures mentioned above, and 
so is his oft-mentioned "separation of concerns". 

I've been doing this long enough (30 years) to remember when this 
stuff was new, and I have many of the original books and papers 
in my library.  I also had the personal experience of bringing 
structured programming into a COBOL spaghetti shop in 1974.   So 
I do know a little about this stuff.  I'm not saying structured 
programming is only about control structures.  And I'm not saying 
that no one has tried "enhancing" the original idea of structured 
programming with additional control structures.  Wulf, Zahn, 
Dahl, Knuth and others have suggested other forms of loops.  But 
the original idea, as expounded by Dijkstra, Mills, Wirth, 
Ledgard, and others does indeed restrict you to single-
entry/single-exit control structures, and they don't exit loops 
in the middle.