[Bluej-discuss] language stuff

[Stephen Bloch]

Aryeh writes:

>... anything but infix requires an amazing amount of
>conceptual conversion between problem and code space.

Huh?  What real-world problem (other than parsing infix algebraic 
expressions) is inherently infix?  Infix notation was invented only a 
few hundred years ago; it's not exactly hardwired into our brains.

>That plus having to count parens basically rules out any lisp 
>variant in my mind.
>...
>the reason for using a mix of symbols vs. just one for grouping is
>breaks the counting up... something is very wrong if you have an
>expression that has more then 4 leading or closing parens in algol like
>langs because a[i] is not the same as a(i)... under your arg we should
>teach fully parened math only (because the only advantage of pre/postfix
>over infix for human readers {compilers have different reasons} is it is
>essencially fully parened without the need for parens)

Suit yourself.  I'd rather keep typing right-parens until everything 
matches than figure out the correct sequence of parentheses, 
brackets, curly-braces, and semicolons to close a complex (yet 
reasonable) expression like

myButton.addActionListener (
   new ActionListener () {
     public void actionPerformed(ActionEvent e) {
       if (showingValues) {
         myLabel.setText (Integer.toString(myButton.values()[i]));}}});


In the interest of fair comparison, here's a Scheme equivalent:

(add-action-handler my-button
   (lambda (e)
     (when showing-values
        (set-text my-label
           (number->string (vector-ref (values my-button) i))))))

>... there is also a fantastic language call APL or worse
>one where all the reserved words are white space

Slight correction.  In APL, most of the predefined functions (I don't 
know whether they're reserved words) are spelled using non-standard 
characters, but not whitespace.

>... this goes all the way back to the mid
>60's with the divide between Bell Labs and MIT on various aspects of
>Multics (MIT wanted theoretical purity [LISP] and Bell Labs wanted
>something praciticle [C/C++/Java])

I was about to point out that Multics was in fact written mostly in 
PL/1, as C, C++, and Java hadn't been invented yet.  But you're 
making an analogy, so never mind.

The dispute isn't "theoretical purity vs. something practical"; I 
think we're all in favor of something practical.  The question is 
"what approach is most practical and effective for teaching beginning 
programming?"  In a freshman programming class (or a non-majors' 
programming class), my main goal is not to get the students a 
programming job next month; it's to teach them habits of good 
programming.
In such a class, one of the real, practical hurdles is the time 
students spend memorizing syntax rules.  If I can lower that hurdle 
(e.g. by using a language with far simpler and more consistent 
syntax, or an IDE that shields them from some of the syntax rules), I 
have in a very practical way bought myself class time to teach more 
practical principles of programming: refactoring, design for re-use, 
polymorphism, data structures, etc.

For my own programming, I occasionally use C, C++, or Java, but in 
many cases I choose Scheme for its power, flexibility, and brevity. 
These are practical decisions, motivated by the desire to produce a 
working program as quickly as possible.

>There is conceptual ambiguity in pre/post fix because it is so
>unfamiliar to non-CS people....
>I was refering to arhithmatic (2 2 + 3 * would just baffle most
>beginners but 2+2*3 [assuming traditional human precendance] can be read
>instantly by anyone who is ready for college])

For the people on this list, 2+2*3 is familiar, comfortable, an old 
friend.  For many of my students, it's mysterious, scary, an old 
enemy.  I have had students disagree with one another in class over 
its correct value, since some of them never really grasped order of 
operations in high school.  For those students, using a different 
syntax allows them to start over with a clean slate.  I have seen 
light bulbs go on over the heads of students who had never understood 
before what this "order of operations" stuff meant.

That said, when I teach CS0 for non-majors, I don't introduce 
arithmetic at all until a month into the course, after they've 
learned to write functions in a disciplined way on graphical objects. 
By that time they're used to the prefix notation, and using the same 
syntax rule for arithmetic is no big deal.
-- 
Stephen Bloch
sbloch at adelphi.edu