Hacker News Reader

Comment by maximusprime | Hacker News Reader

Or you could just use java...

The title says only fast languages are interesting, then rules out using Java because it's not lispy enough :/

Either you want speed, or you want your favorite syntax and high level stuff. You can't have both.

FWIW My favorite syntax happens to be c/java/assembly type stuff. I'd hate to be one of the developers who hates those, must put you at a big disadvantage - as shown in the original post.

View on HN

Why not? Apparently lush gets it right: C-like speed on numerics with a Lisp-ish syntax this guy likes. Yes, C-style stacks are going to beat Lisp-style closures; but there is no reason why a "fancy" language can't do fast matrix operations, which is 99.9% of numerical code anyway.

Even Scheme has (rather un-Lisp-ish) vectors in addition to the standard linked lists.

> C-style stacks are going to beat Lisp-style closures

It has been a long time since the "λ the ultimate..." series of papers gave the lie to this myth.

I once made the joke that a Java web app usually catches up with a Django one in about six months. The Java app is faster, but, the Django one stars serving users three months before the Java enters live beta.

And if you are serious about number crushing, my friends tell me you should use FORTRAN, not Java.

Here's the summation problem in Fortran, quite succinct, IMO:

    real, dimension(30000000) :: a

    call random_number(a)

    print *, sum(a)

That's 30 million doubles, takes 0.04s on my machine.

Except if you're like 99% of engineers whose favorite syntax is C/C++-ish and are happily using C, C++, C#, Java or maybe Python.

Reply on news.ycombinator.com