🚀 go-pugleaf

Jan Steinman wrote:
> Don't get me wrong, I'm one of the biggest ST fans around, but I
wouldn't
> suggest it for serious number-crunching, unless you're doing infinite
> precision math, in which case it does okay -- not fabulous, but not bad.


How serious is serious Jan? If I'm trying to predict the weather, I agree
with you. However, the pure OOP capabilities of Smalltalk make it the best
number-crunching language I've ever come across -- except FORTRAN, of
course, but I had to give that up in the interests of programming in the
long (i.e., still using my code 10 years from now). In Smalltalk, the
Number class and the potential for extending it are absolute mental
hard-ons.

Once again, for most of the applications developed, what really matters is
not the speed of the executable, its the speed to develop, modify, and the
ability to head off the wall of complexity. (Okay, speed matters when the
run-time is pokey. It also matters when your spreadsheet is being sold to
tens of millions of people.)

Finally, I took my old Fortran matrix inversion code and converted it to
Smalltalk. I then ran a test problem from my old Numerical Methods Book
(Carnahan, Luther Wilkes). They have a worked out long-hand  example where
the values in the original matrix are Fractions and the inverted matrix
are also Fractions. I ran this on Smalltalk and actually got back
Fractions! Big deal? Yes, perhaps the most intensely profound programming
result I've ever experienced. Sorry if this sounds wierd, but spend your
life in Numerical Methods and Matrices and forever having to deal with
roundoff and you would understand.