Re: uniqueness typing

[Fergus Henderson]

On 12-Mar-2002, Dana Harrington <[EMAIL PROTECTED]> wrote:
> 
> I believe Mercury borrowed their uniqueness type (mode) system from 
> Clean.

Nope.  The support for unique modes in Mercury was developed before
we were aware of Clean.  They achieve similar aims, but it's a case
of convergent evolution, because we were both trying to solve the same
problem, rather than borrowing.  (This is in contrast to e.g. Mercury's
support for type classes, which was pretty much directly lifted from
Haskell.)

In fact Clean's support for uniqueness is better than Mercury's.
In particular, Clean supports uniqueness polymorphism,
whereas Mercury only supports overloading on uniqueness.

-- 
Fergus Henderson <[EMAIL PROTECTED]>  |  "I have always known that the pursuit
The University of Melbourne |  of excellence is a lethal habit"
WWW:   | -- the last words of T. S. Garp.
___
Haskell mailing list
[EMAIL PROTECTED]
http://www.haskell.org/mailman/listinfo/haskell

Re: uniqueness typing

[Dana Harrington]

Andre W B Furtado wrote:
> I found the following text when visiting the Clean (a functional language)
> site:
> 
> "Clean is the only functional language in the world which has a special type
> system, uniqueness typing. It enables to update function arguments
> destructively retaining the purity of the language."

I believe Mercury borrowed their uniqueness type (mode) system from 
Clean. But since Mercury is functional/logical then I guess its still 
true that Clean is the only functional language with a uniqueness type 
system.

> Then I have some questions:
> 1. Can anyone explain what does "update function arguments destructively"
> mean?

Uniquely typed values are guaranteed to be referenced at most once under 
program evaluation, these values can be modified in-place. As an 
example, suppose I have an array and I want to modify an element.  In 
general I need to create a whole new copy of the array in order to make 
the modification since there may be other places in the program that 
require access to the old array.  However, if I know that the old array 
has no other references (because the array has a unique type) then I do 
not need to make a copy, I can just modify the original array.  This has 
obvious efficiency benefits.

> 2. If this is a desirable thing, why does Haskell not implement it?

- Uniqueness types, of course, require replacing the type system, this 
is a non-trivial task,
- Destructive updates can already be accomplished with compiler 
supported libraries using monads,
- Uniqueness types invalidate some program transformations.

On the other hand I think uniqueness types are quite useful. 
Particularly as they allow much more flexible interaction between 
mutable data-structures than monadic approaches.  I don't know of any 
concrete reasons a uniqueness type system couldn't be added to Haskell.

This seems like a fine time to insert a plug for my Master's thesis, 
which describes a denotational semantics of uniqueness types:
http://www.cpsc.ucalgary.ca/~danaha/uniqueness-types.ps

Dana

___
Haskell mailing list
[EMAIL PROTECTED]
http://www.haskell.org/mailman/listinfo/haskell

uniqueness typing

[Andre W B Furtado]

I found the following text when visiting the Clean (a functional language)
site:

"Clean is the only functional language in the world which has a special type
system, uniqueness typing. It enables to update function arguments
destructively retaining the purity of the language."

Then I have some questions:
1. Can anyone explain what does "update function arguments destructively"
mean?
2. If this is a desirable thing, why does Haskell not implement it?

Thanks,
-- Andre

___
Haskell mailing list
[EMAIL PROTECTED]
http://www.haskell.org/mailman/listinfo/haskell