[clean-list] discussing issues of uniqueness typing

Wolfgang Jeltsch wolfgang at jeltsch.net
Tue Jun 28 16:50:01 MEST 2005


Am Dienstag, 28. Juni 2005 09:03 schrieben Sie:
> Dear Wolfgang,
>
> First of all, in our system uniqueness does not always imply unique
> references: If a function delivers a unique result, this means that
> evaluating such a function will not increase the number of references to
> the closure.

Dear Sjaak,

do you think of situations like

    let x = fwritec char file in (x,x)

where x is referenced multiple times?  Are there other situations in which a 
closure marked as unique can be referenced more than once?

> Your remark concerning the subtyping relation
>
> 	t1 ->x t2 <= t1 ->* t2
>
> is right: it can be allowed. At the time we implemented the uniqueness type
> system we did not see the advantages of having this relation. But, for
> orthogonally reasons, it would have been better to allow it.

Ah, that's good.  Currently, I write my diploma thesis (similar to master's 
thesis) where I deal with different approaches for integrating functional and 
imperative programming concepts.  I'd like to allow the above subtype 
relation in my presentation of uniqueness typing if that's possible because I 
don't want to have unnecessary restrictions in the type system.

> Regards,
>
> Sjaak

Best wishes,
Wolfgang


More information about the clean-list mailing list