[Haskell-cafe] Re: questions about Arrows

[Ben]

Thanks for the prompt reply.  Some questions / comments below :

On Wed, Sep 1, 2010 at 12:33 AM, Maciej Piechotka <uzytkownik2 at gmail.com> wrote:

> rSum2 :: ArrowCircuit a => a Int Int
> rSum2 = proc x -> do
>    rec out <- delay 0 -< out + x
>    returnA -< out + x

Wow, that was simple.  I guess I never thought to do this because it
evaluates (out + x) twice, but one can always write

rSum3 :: ArrowCircuit a => a Int Int
rSum3 = proc x -> do
  rec let next = out + x
      out <- delay 0 -< next
  returnA -< next

I have a follow-up question which I'll ask in a new thread.

>> 3) One can define fix in terms of trace and trace in terms of fix.
>>
>> trace f x = fst $ fix (\(m, z) -> f (x, z))
>> fix f = trace (\(x, y) -> (f y, f y)) undefined
>>
>> Does this mean we can translate arbitrary recursive functions into
>> ArrowLoop equivalents?
>>
>
> Yes. In fact fix is used on functional languages that do not support
> recursion to have recursion (or so I heard)

In which case my question is, why is the primitive for Arrows based on
trace instead of fix?

Best regards, Ben