On Dec 14, 3:15 pm, "[EMAIL PROTECTED]" <[EMAIL PROTECTED]> wrote: > On Dec 14, 2:48 pm, "Chris Mellon" <[EMAIL PROTECTED]> wrote: > > > > > On Dec 14, 2007 2:09 PM, [EMAIL PROTECTED] <[EMAIL PROTECTED]> wrote: > > > > On Dec 11, 10:34 pm, "Terry Reedy" <[EMAIL PROTECTED]> wrote: > > > > "Ron Provost" <[EMAIL PROTECTED]> wrote in message > > > > >news:[EMAIL PROTECTED] > > > > But here's my problem, most of my coworkers, when they see my apps and > > > > learn that they are written in Python ask questions like, "Why would you > > > > write that in a scripting language?" Whenever I hear a comment like > > > > that I > > > > can feel myself boiling inside. > > > > =================== > > > > > I don't blame you. Python is an full-fledged algorithm/programming > > > > language that was designed to *also* be used a scripting language. > > > > When you buy a new car, which is more important, the styling > > > or what's under the hood? > > > <snip> > > > > That's for the whole sequence, not a single term! > > > > 25.67 MINUTES compared to 17.65 HOURS! > > > > So, sure, some languages compile to .exe programs. > > > > In the trade, we call that "polishing a turd". > > > While I agree with the sentiment, surely this can't be a good example > > of F#s general performance? > > Of course. They example was deliberately chosen to make > F#'s BigInt library look as bad as possible. > > > I would expect .NET code to smoke stock > > (non-Psycoed) Python in this benchmark. > > Probably. It just so happens that the example chosen > is typical of _my_ number theory research and I was > toying with the idea of converting my Collatz Conjecture > function library to F#. > > And that would take a lot of work since F# doesn't have > anywhere near the functionality of gmpy. There isn't > even a BigInt.mod function for cryin' out loud, let alone > stuff like GCD or modular inverse, critical to my research. > > Sure, I could write my own Extended Euclidean Algorithm, > but now that I now that F# is just a Volkswagen with a > fiberglass shell that looks like a Ferrari, there isn't > much point, eh? > > And don't even get me started about the stupid stuff, > like having TWO different type of Big Rationals. One's > more efficient than the other, they say. So why two? > The less efficient one has functions not implemented > in the other. And F# is strongly typed, so you can't > use BigRational methods with BigNum types. And they > have all sorts of conversion methods for to/from ints, > strings, bigints, etc. Guess which conversions they > don't have? BigRational <-> BigNum, of course. > > MAybe I'll check it out again in the future after it > has gone through several revisions.
Since F# is a caml derivative, you may want to look at OCaml. Though, I've never used OCaml for any kind of heavy number-crunching, so maybe it has even worse library support; can't say. Ps. The two types are probably for boxed and unboxed versions, q.v. this article, about half- way down: http://msdn.microsoft.com/msdnmag/issues/1200/dotnet/ Regards, Jordan -- http://mail.python.org/mailman/listinfo/python-list
