Re: histrionics

[Robert Dewar <robert_dewar@my-deja.com>]

In article <rtiq3si91iv24@corp.supernews.com>,
  "Vladimir Olensky" <vladimir_olensky@yahoo.com> wrote

> May be this is because writing complex programs in Prolog
> requires very high level of abstract thinking to be able to
> express complex problems using short logic predicates
> sentences.

Actually I disagree strongly with your criticisms of Prolog.
On the contrary, I think it is MUCH easier to think at the
prolog level, than to think in the low level sequential
manner that conventional programming requires (as I mentioned
the sad thing in Prolog, is that once you use cut, which in
practice you do need to in real programs, you have to go back
to much lower level thinking).

> It is always more easy to describe something in many words
> than in couple of  words:-)

No, again I disagree, the power of very high level languages
is directly related to the fact that you want to be able
to describe things at a much higher level, and of course
the descriptions are shorter. If I can say (this is SETL,
without the nice character set :-)

primes = {n in 2 .. 100 |
           notexists j in 2 .. n - 1 |
             n mod j = 0};

that's a WHOLE lot easier to understand than a seive programmed
in a conventional procedural language. Vladimir, you can supply
the corresponding prolog program if you like :-) I prefer the
statement in SETL, since it seems close to the mathematical
definition of primes, both syntactically and semantically

> Another thing is that it may be difficult for many people to
> grasp= the idea of backtracking

Actually the nice thing about angelic non-determinism is
precisely that you DON'T need to think about it, the right
thing happens. Backtracking is not a fundamental part of
the semantics of Prolog at a high level, rather it is an
implementation mechanism for acheiving a correct pattern
match as required by angelic ND. Of course, once again,
if you worry about cut, then suddenly the backtracking
becomes non-transparent. A very similar situation happens
in SNOBOL-4 pattern matching, which at best can be considered
to be fully declarative, but which provides lots of cut-like
low level junk that requires you to know the procedural
algorithm that is used.

> and recursion

Surely you are not saying recursion is hard to understand. Sure
I know that beginning students may have trouble, but surely
anyone who considers themselves a competent professional
programmer has got past this long ago, and is completely
comfortable with recursion ...

> Iterative algorithms implemented using conventional languages
> may have much more lines of code but be much more easy
> understandable.

This is the kind of criticism you expect from someone who
basically does not like Prolog :-) I would strongly disagree
an elegant program written in Prolog is far far easier to
understand than the extended gobbledygook that passes for
low level procedural programming :-)

So, it's interesting, Dewar and Olensky have almost
diametrically opposed views of Prolog.

Dewar likes very high level languages in general, and finds
examples of programs written in these languages compelling,
and for examples, that feeling applies to Prolog. Elegant
Prolog examples, even quite complex ones are quite
beautiful, easy to understand, concise, and clear.

Where Dewar gets off the Prolog admiration bus is where people
claim that it is an appropriate and practical language for
writing large scale applications, including expert systems.

Olensky on the other hand is rather suspicious of Prolog, thinks
it is hard to understand because of backtracking, logic etc, but
nevertheless thinks it is the language of choice for large
complex expert systems.

Interesting ....




Sent via Deja.com http://www.deja.com/
Share what you know. Learn what you don't.