[m-dev.] functional dependencies syntax

Ralph Becket rafe at cs.mu.OZ.AU
Tue Apr 5 17:24:17 AEST 2005


Mark Brown, Tuesday,  5 April 2005:
> 
> There has been some discussion in the office, which I shall summarise here.
> Two main suggestions were made, both of which I am inclined to agree with.
> 
> Suggestion 1:
> 
> Use '->'/2 as the top-level operator for a functional dependency, and use
> ','/2 as the operator to construct sets of type variables in the domain
> and range of the FD.
> 
> Some example FDs with this syntax:
> 
> 	(A -> B)
> 	(A, B, C -> D)
> 	(A, B -> C, D, E)
> 
> The drawback of this is that there is no obvious way to specify an empty
> domain or range.

I find it hard to imagine what such a thing would even mean.

I like this suggestion.

> Suggestion 2:
> 
> Rather than introduce a new keyword to precede the FDs (I used "given" in
> my earlier example), make '->'/2 into a special builtin typeclass with the
> semantics of FDs and place any FDs in the set of constraints on your
> typeclass declaration.
> 
> For example:
> 
> 	:- typeclass container(C, E) <= (C -> E) where ...
> 	:- typeclass foo(A, B) <= (bar(A, B), (A -> B)) where ...
> 	:- typeclass baz(A, B, C) <= (quux(A), (A, B -> C), quuux(B, C)).
> 
> The FDs would be able to appear anywhere in the set of constraints.

This is also a good suggestion.

> An alternative is that FDs must be at the start of the list of constraints.
> Another alternative is that they must be at the end of the list.

These seem needlessly restrictive to me.

> 	- This would preclude the use of '->'/2 as the name of a user
> 	defined typeclass, unless we provide some way of escaping the
> 	special meaning of this name in typeclass declarations.

This is not a significant loss.

> 	- There is a slight lack of orthogonality here because, unlike
> 	ordinary constraints, FDs can only appear on typeclass declarations.
> 	They can't appear on instances, preds, funcs, or existentially
> 	quantified types.

FDs are a property of the type class, not of those other things.  So
this is fine.

-- Ralph
--------------------------------------------------------------------------
mercury-developers mailing list
Post messages to:       mercury-developers at cs.mu.oz.au
Administrative Queries: owner-mercury-developers at cs.mu.oz.au
Subscriptions:          mercury-developers-request at cs.mu.oz.au
--------------------------------------------------------------------------



More information about the developers mailing list