Re: Lazy Evaluation [Side-Effects in Functions]

[dewar@gnat.com (Robert Dewar)]

Darren New <dnew@san.rr.com> wrote in message news:<3BE8BBC2.A8563DBD@san.rr.com>...
> Robert Dewar wrote:
> > Of course there is no particular conflict between side
> > effects and call by name, as anyone who has programmed
> > in Algol-60 extensively is perfectly aware.
> 
> Uh, unless you do something like
>   Swap(A(I), I)
> (where A is an array, of course)

Well yes, we all know Algol-60! (or at least I hope any
serious student of programming languages does). But there
is no special problem here. In Algol-60, the effect will be canonical,
depending on the exact order of statements in the body. Sure, it might
not do what the programmer expects
in this case, but it will do EXACTLY what the programmer
programmed. No language can avoid this discrepancy in
all cases.

Are you arguing that in practice call-by-name is troublesome for this
reason? I think that's a bogus
argument. I certainly never found it to be the case
in my Algol-60 programming days, how about you?

Note that in Ada, simple out parameters generate similar
concerns, except that the result is *not* canonical, if
you do

   x (m.m)

when x has two out parameters, the order in which things
get assigned is undefined in Ada.