On 2009-12-22 14:24:41 +0100, Andrei Alexandrescu
<[email protected]> said:
Daniel de Kok wrote:
...in pure FP. Of course, another problem is that some algorithms are
inherently imperative. For instance, I find calculating the edit
distance in a performant manner ugly in functional languages.
So then I guess I can be given a break about FP ueber alles.
Hey, it's not me arguing for that. I think every paradigm has its place...
And I am not in for parotting the 'we absolutely need pure FP for a
multicore world' stanza. While full purity certainly helps for
lock-free programming, it reduces the world to one where there are no
embarassingly parallel programs. In natural language parsing we get
(nearly) linear speedup without purity, with single-thread processes
(just split a corpus in N parts, and run a parser for each part). Or if
you do vector/matrix calculations on a GPU it's also a mutable world
(with the exception of texture memory).
Of course, there are good counter-examples where purity helps. But it's
not a black-white world. Not that I really have to say that on a D list
;-).
-- Daniel