"ocaml_beginners"::[] Asserting polymorphism with annotations

Is there anyway to force the compiler to check that a function is
as polymorphic as its annotation?

I've large function that ocaml infers a less general type than I
expect, so I'd like to annotate some subexpressions until I can see
where the extra unification is happening.

E.g. say there's a function I expect to be polymorphic with a type
like 'a -> 'b -> bool, but in the body of the function I have an
expression which forces 'a='b. Can somehow I stop 'a and 'b from
being unified? Can I force a compile time error for something like "let
f (x:'a) (y:'b) : bool = x = y" instead of inferring "f: 'a -> 'a ->
bool"?

Messages in this topic (1) Reply (via web post) | Start a new topic

.

On Saturday 16 June 2007 17:59:24 ethan.aubin wrote:
> E.g. say there's a function I expect to be polymorphic with a type
>; like 'a -> 'b -> bool, but in the body of the function I have an
> expression which forces 'a='b. Can somehow I stop 'a and 'b from
>; being unified? Can I force a compile time error for something like "let
> f (x:'a) (y:'b) : bool = x = y" instead of inferring "f: 'a -> 'a ->
> bool"?

I don't believe you can. Your best bet is probably to break the large function
into many smaller functions, which is better ML style anyway.

--
Dr Jon D Harrop, Flying Frog Consultancy Ltd.
OCaml for Scientists
http://www.ffconsultancy.com/products/ocaml_for_scientists/?e

Messages in this topic (0) Reply (via web post) | Start a new topic

.

Zitat von Jon Harrop < jon%40ffconsultancy.com">jonffconsultancy.com&gt;:

&gt; On Saturday 16 June 2007 17:59:24 ethan.aubin wrote:
&gt; > E.g. say there's a function I expect to be polymorphic with a type
>; > like 'a -> 'b -> bool, but in the body of the function I have an
> > expression which forces 'a='b. Can somehow I stop 'a and 'b from
>; > being unified? Can I force a compile time error for something like "let
> > f (x:'a) (y:'b) : bool = x = y" instead of inferring "f: 'a -> 'a ->
&gt; > bool"?
>
> I don't believe you can.
[...]

And I see no reason why it should be possible or necessary...

...otherwise you could use Perl:

====================
oliversiouxsie ~ $ cat useperl.pl
#!/usr/bin/perl

print "use perl!n&quot; if "7&quot; == 7;
oliversiouxsie ~ $ perl useperl.pl
use perl!
oliversiouxsie ~ $
====================

Ciao,
Oliver

Messages in this topic (0) Reply (via web post) | Start a new topic

.

On Sat, 16 Jun 2007, "ethan.aubin&quot; < ethan.aubin%2Byahoo%40gmail.com">ethan.aubin+yahoogmail.com> wrote:
&gt;
> E.g. say there's a function I expect to be polymorphic with a type
>; like 'a -> 'b -> bool, but in the body of the function I have an
> expression which forces 'a='b. Can somehow I stop 'a and 'b from
>; being unified? Can I force a compile time error for something like "let
> f (x:'a) (y:'b) : bool = x = y" instead of inferring "f: 'a -> 'a ->
&gt; bool"?

No. If it was possible, then you could write [f 1 "1&quot;]. What should
OCaml do when you reach [1 = "1&quot;] ? Return true or false? According
to what you expect, define you type and the equality on it. You can
also pass an 'equality' (sic) function with type [eq: 'a -> 'b ->
bool] and the inferred type will be the one you want.

ChriS

Messages in this topic (0) Reply (via web post) | Start a new topic

.