On Thu, 17 Mar 2005, Marcin 'Qrczak' Kowalczyk wrote:
> It's not the fault of the mapping function but of the stack being
> non-extensible. A user-written recursion can blow it too. Functional
> programming is supposed to encourage recursion, and a non-tail-recursive
> 'map' is much more readable than alternatives.

Interesting approach. Do you have any information as to how big the performance
hit is?

I've never used SML/NJ except for a few toy programs, but I recall that it
puts activation records on the Gc'ed heap (correct me if I'm wrong
someone) so that call/cc is more efficient, so stack overflows shouldn't
be a problem there either. Could you comment on why you chose extensible stacks
rather than the SMLNJ approach for Kogut.

> My implementation of my language Kogut has extensible stack.
> And transparent bignums when appropriate. Yes, it's slower,
> but correctness is more important.

Hard to disagree when you put it that way, but there you seem to be posing
a false dichotomy. With enough work, C code can be made safe.

What I think you intend is that you'd rather it be easy to write safe code
than that it be asy to write fast code, in the language. I wouldn't mind
that, as long as it  isn't ridiculously hard or impossible to do the latter in
the language.

-- Brian