Philippa Cowderoy <flippa@[EMAIL PROTECTED]
> writes:
> On Sat, 9 Feb 2008, Stephen J. Bevan wrote:
>
>> George Neuner <gneuner2/@[EMAIL PROTECTED]
> writes:
>> > However, Dylan is still a Scheme is still a Lisp.
>>
>> Aren't all functional languages a Lisp then?
>
> Dylan really did start out as a more conventional lisp, and then the
> syntax changed. There are plenty of functional languages that didn't
> descend from lisp though - notably the descendants of ISWIM. Naturally
the
> lambda calculus predates both.
Naturally. Now back to my question which is if Dylan is a Lisp thus
MPS has been used in a Lisp, aren't all functional languages
effectively a Lisp when it comes to whether MPS can be used in them or
not?
>> Or is there something
>> different about some functional languages (Haskell?) which makes their
>> GC requirements different?
>>
>
> As functional languages go, lisps take pure functional programming the
> least seriously - having lots of definitely-immutable data around does
> change the game a little, especially in the presence of concurrency.
A pure functional language using graph reduction as an implemenation
strategy is using mutation as part of graph reduction and thus I'd
argue as far as a GC is concerned that language looks more like a Lisp
than (S)ML. That was true of Glasgow Haskell circa 1993 when Sansom
published his FPLCA paper, I've no idea if GHC has changed is
implementation in the meantime so as to invalidate that conclusion.
So, except for some pure languages whose implementation does not use
any form of update (i.e. not all known Hasekell implementations) then
functional languages have mutable structures whether they are explicit
in the language or not. If you are supporting some kind of
incremental collection then you need a software/hardware read/write
barrier. Those four values provide quite a bit of design space to
chose from. I'd argue that the program in the language has just as
much effect as the language itself. That is, I may write SML or
O'Caml with little use of references but the next person, say coming
over from Python, may think nothing of using references because that's
what they are used to. Same language, same implementation,
dramatically different barrier profile. Thus whichever style I assume
is the most common, there are always going to be programs for which my
GC assumptions are less than ideal.
|
