Many thanks Matthew for the clarification. Subtle :-)
Since datatype declarations declare both a type operator
and constructors, I think the explanation deserves to appear
in http://mlton.org/EqualityTypeVariable, last part.
The last two declarations declare identical type operators, but
not iden
I think that the reasoning is that the type function induced by a
datatype (which maps the type variables to the fresh type name) is
ignorant of equality type variables, but the type schemes induced by
the constructors do respect equality type variables.
Thus, with
datatype ('a, ''b) t = A of '
Hello,
Funnily (where my belief came from ?), the page about equality type
variables on mlton.org (http://mlton.org/EqualityTypeVariable)
says that they play no special role in either type bindings or
datatype bindings, contradicting the compiler ...
BB.
On 9/16/16 10:42 PM, Bernard Berthomie
David,
Yes, you are right; p19 applies to type functions (abbreviations),
not to type constructors introduced by datatype.
My understanding was that both obeyed the same rule, but I guess
it is not the case (can't find anything about it in the book ..).
mlton also rejects (C 1.0), by the way.
Bernard,
That refers to a "type function" not a datatype binding.
type ''a t = ''a * ''a
creates a type function and indeed Poly/ML ignores the equality
attribute in that case. A datatype binding creates a new "type name"
and that is dealt with elsewhere.
The fact that both mlton and hamlet b
> On 16 Sep 2016, at 20:59, Bernard Berthomieu
> wrote:
>
> ...
> I like the treatment of polyML :-), but I guess it is not standard:
>
> The "definition" (http://sml-family.org/sml97-defn.pdf) says, page 19:
> "... In particular, the equality attribute has no significance in a bound
>
Hello,
Besides any printing/elaboration problem with polyml 5.6:
In polyml 5.5:
> datatype ''a t = C of ''a;
datatype ''a t = C of ''a
> C 1.0;
Error-Type error in function application.
Function: C : ''a -> ''a t
Argument: 1.0 : real
Reason: Can't unify ''a to real (Requires equality t
Rob,
Actually what I was trying to find was the bit that says that the
equality attribute in a type variable used as a datatype parameter
should be honoured where it isn't when used as a type parameter.
> type ''a t = ''a * ''a;
type 'a t = 'a * 'a
> (1.0, 2.0): real t;
val it = (1.0, 2.0): re
David,
> On 16 Sep 2016, at 19:19, David Matthews
> wrote:
>
> I've checked hamlet and mlton and they both reject it so I think in this case
> Poly/ML is right and SML/NJ is wrong. I can't point to the bit of the
> definition that says that, though.
It's in Appendix C: "Of [the type names
On 16/09/2016 16:12, Tjark Weber wrote:
Hi,
Is the following behavior (of Poly/ML 5.6) intended?
> datatype ''a t = C of ''a;
datatype 'a t = C of ''b
I'll look at the printing here. There was a change in 5.5 that seems to
have broken it.
Moreover, Poly/ML rejects a following
> C 1.
Hi,
Is the following behavior (of Poly/ML 5.6) intended?
> datatype ''a t = C of ''a;
datatype 'a t = C of ''b
Moreover, Poly/ML rejects a following
> C 1.0;
while SML/NJ accepts it.
Best,
Tjark
___
polyml mailing list
polyml@inf.ed.ac.uk
ht
11 matches
Mail list logo
