At 08:45 -0700 1999/04/19, Mark P Jones wrote:
...
>Your observations are correct, and this was an intentional design feature of
>Gofer, which has carried over into Hugs.
In private communications with Pablo Azero <[EMAIL PROTECTED]>, we were able
to figure out some of this stuff.
>There is a simple way to release
>the
>storage for fib, which is to reload fib.hs (or any other file, or even
>nothing
>using :load with no arguments).
So this one I noticed.
>..note that, if you
>want to write a version of fib that doesn't hold on to the partially
>computed list beyond a single use of that list, then you should rewrite your
>definition as:
>
> fib :: () -> [Integer]
> fib () = fibs where fibs = 1 : 1 : zipWith (+) fibs (tail fibs)
>
>And compare the behavior of this definition with the following, seemingly
>modest variation:
>
> fib :: () -> [Integer]
> fib () = 1 : 1 : zipWith (+) (fib ()) (tail (fib ()))
>From what I could see, Hugs views the top level definitions, and its
partial computations, as a root set for the GC (garbage collector). My
guess is that has to do something with not having to do partial
computations over again, would they be needed.
>This has been a controversial point in the design, and some folks will tell
>you that I made the wrong choice. I'm sure that I've written about my
>rationale for this behaviour in the Gofer user manual or implementation
>report, so I won't go into it again here.
So my guess is that one wants to partially collect a partial top level
computation, one needs figure out a good strategy for doing that.
Hans Aberg
* Email: Hans Aberg <mailto:[EMAIL PROTECTED]>
* Home Page: <http://www.matematik.su.se/~haberg/>
* AMS member listing: <http://www.ams.org/cml/>
