Proposal: Relaxing restrictions in Curry

From: Michael Hanus <mh_at_informatik.uni-kiel.de>
Date: Tue, 02 Nov 2004 12:11:57 +0100

Dear Colleagues,

I'd like to propose to relax two restrictions in the definition
of Curry which seem to me unnecessarily strong.

The first one is related to the introduction of free variables.
The current language definition (Curry Report, C.3) requires
that in an expression of the form

  let x free in e

e must be of type "Success" (1).

On the one hand, this seems justified by the fact that existential
quantification is reasonable for constraints only. On the other hand,
this demands for some nasty transformations if one needs local
variables in a non-constraint scope. For instance, the (currently
illegal) definition

  f x = \y -> let z free in g x y z -- (2)

(where (g x y z) is not of type Success) can be transformed
in the valid Curry program

  f x = g' x
  g' x y = g x y z where z free

The transformation can be avoided by introducing a
"generate-free-variable" function (actually, this is the
solution which I often use):

  f x = \y -> g x y any
  any = z where z free

Both solutions are less readable than (2). Furthermore,
the alternatives demonstrate that restriction (1) is not
really necessary. In particular, if one defines a type
isomorphic to Success, the restriction becomes questionable.

Thus, I propose to drop the restriction (2).
This would also make the application of let and where
"more equivalent".


The second restriction concerns the sequential conjunction
of constraints, which is currently defined as

  (&>) :: Success -> Success -> Success
  c &> x | c = x

in the prelude.

We have a number of applications where we want to put constraints
during a functional evaluation. Until now, we have defined for this
purpose a "guard" function that takes a constraint and an expression as
input and is the identity on the second argument provided that the
constraint is satisfied. Thus, it is defined as

  guard :: Success -> a -> a
  guard c x | c = x

so that you can use it as in (guard (x=:=1) (2+x)).
Now, you see that the definition of the guard function is
identical to (&>) apart from the different types.
Since such a guard function is quite useful, I propose
to generalize the type of (&>) to

  (&>) :: Success -> a -> a

so that it is a general function to establish new constraints
during arbitrary computations.


Since both proposals have no influence on existing programs but
allow more valid Curry programs, I see no problem in them.
However, maybe somebody sees some problem?

Best regards,

Michael

_______________________________________________
curry mailing list
curry_at_lists.RWTH-Aachen.DE
http://MailMan.RWTH-Aachen.DE/mailman/listinfo/curry
Received on Tue Nov 02 2004 - 14:14:47 CET

This archive was generated by hypermail 2.3.0 : Mon Nov 11 2019 - 07:15:06 CET