Gergely Buday wrote:
do I understand correctly that only the following built-in types admit equality?
class eq
val eq : t ::: Type -> eq t -> t -> t -> bool
val ne : t ::: Type -> eq t -> t -> t -> bool
val eq_int : eq int
val eq_float : eq float
val eq_string : eq string
val eq_char : eq char
val eq_bool : eq bool
val eq_time : eq time
val mkEq : t ::: Type -> (t -> t -> bool) -> eq t
You are free to add equality witnesses for any type at all. You may
even shadow the built-in witnesses in local contexts. (Though I
wouldn't want to make a strong statement about which witness would be
chosen by automatic resolution in such a case. ;])
And, what about xml? Is it not possible to compare two xml fragments?
No [eq] instance for the [xml] family is provided in the standard
library. You can see that [xml] _does_ have a [show] instance, which
provides an easy path to equality testing, but this operation (and a few
others) frightens me in its capacity to break abstraction, by, for
instance, revealing implementation choices for abstract types. I may
make the compiler by default warn about uses of such scary functionality.
And, the only other types that have equality are those that I create
an equality function with mkEq for?
Right.
_______________________________________________
Ur mailing list
Ur@impredicative.com
http://www.impredicative.com/cgi-bin/mailman/listinfo/ur
