Re: OO versus RDB

"Daniel Parker" <danielaparker_at_gmail.com> wrote in message news:1152186211.639700.279770_at_j8g2000cwa.googlegroups.com...

> H. S. Lahman wrote:

>> Similarly, functional programming represents the best
>> approach I know of to algorithmic processing when requirements are
>> nonvolatile.

> I doubt if you can justify that statement.

The "nonvolatile" bit, certainly.

> The classical matrix and
> integration algorithms, for example, are written in languages like
> FORTRAN, ALGOL, C and C++, and would be extremely difficult to port to
> functional languages, relying as they do on a sequence of mutations to
> data structures.

> If you do a search for work on functional languages
> and numerical algorithms, you won't find much, apart from experimental
> work.

A fellow undergrad in my year (this was 1988) implemented ray-tracing using FP.
The images he was rendering were not trivial, and the output was not much slower (allegedly) than other 'standard' implementations.

You will find that many maths algorithms are suitable for attack by FP.

But the problem IMHO is that the procedural/imperative mindset is so strong in us all (cultural perhaps ?? ) that the massive values in / massive values out
becomes difficult to comprehend (it was for me with the ray-tracing stuff) without the "mental crutch" of holding intermediate values/results that procedural/imperative allows.

Regards,
Steven Perryman Received on Thu Jul 06 2006 - 15:33:53 CEST