Re: [Axiom-developer] Convert and Coerce

[Ralf Hemmecke]

First, there is a difference of Axiom and Aldor here.

If you say

  a::B

than that means EXACTLY the same as

  coerce(a)@B

Aldor treats :: as syntactic sugar and nothing more.

That also says if you have a function

  convert: A -> B

but no function

  coerce: A -> B

then the Aldor compiler will complain about a::B.

Axiom is different (and I guess it is the same for SPAD) in that it 
translates a::B into either coerce(a)@B or convert(a)@B. (Hello Axiom 
gurus, correct me if I am wrong.)

But anyway, inside library code NOTHING will be converted or coerced 
automatically.


On 03/28/2006 01:55 AM, Page, Bill wrote:
> On Monday, March 27, 2006 6:36 PM Antoine Hersen wrote

> I do not think that in general conversion implies any loss of
> information. (I am not sure how to define "loss of information".)
> But conversion is often a partial function i.e. only defined for
> some members of the domain. For example:
>
>   2.0::Integer
>
> If we only know
>
>   x:Float
>
> it is not "safe" to write:
>
>   x::Integer
>
> since
>
>   2.1::Integer
>
> will fail. Such conversions will not be done automatically.

Actually, to be a bit more exact...

2.1::Integer

currently fails, but it only fails since there is not function

  coerce: Float -> Integer

that does not fail for that input.

But read the thesis of Doye

http://portal.axiom-developer.org/forum/public/973207537422/view?searchterm=Doye

that says why you should not add a function definition like

coerce(f: Float): Integer == truncate f

Ralf