[Fortress-interest] exporting overloaded functions from an API

[Jan-Willem Maessen]

This message seems to have been eaten by a mailer glitch.

Begin forwarded message:

> From: Jan-Willem Maessen <JanWillem.Maessen@sun.com>
> Date: January 22, 2007 1:29:50 PM EST
> To: Jan-Willem Maessen <Janwillem.Maessen@sun.com>
> Cc: Mike Atkinson <mike@ladyshot.demon.co.uk>, Fortress- 
> interest@www.experimentalstuff.com
> Subject: Re: [Fortress-interest] exporting overloaded functions  
> from an API
>
>
> On Jan 21, 2007, at 10:14 PM, Jan-Willem Maessen wrote:
>
>>
>> On Jan 21, 2007, at 12:59 PM, Mike Atkinson wrote:
>>
>>> component Collections
>>> export {MyFn}
>>>
>>> MyFn(a:ZZ):Boolean = false
>>> MyFn(b:Boolean):Boolean = true
>>>
>>> end
>>>
>>>
>>> do both function definitions get exported? Is it possible to  
>>> export only  one?
>>
>> Actually, that depends on what you've declared in the api MyFn.   
>> The export list lists apis, not functions.  So if you wrote:
>>
>> api MyFn
>>
>> MyFn(a:ZZ):Boolean
>>
>> end
>>
>> then only the first of the two definitions will be exported.
>
>
> I should clarify at this point that this is looking ahead rather.   
> If you read the 1.0alpha spec (Section 22.2 in 1.0alpha, middle of  
> p164, says Victor), you'll quickly learn that we are required to  
> export the entire function, as in the MyBigFn api:
>
>> api MyBigFn
>>
>> MyFn(a:ZZ):Boolean
>> MyFn(b:Boolean):Boolean
>>
>> end
>
> This avoids some problems which crop up should your api slice your  
> multiple-inheritance trait hierarchy in a different direction than  
> your overloaded function declarations choose to.  My hope is that  
> we can address those problems and eventually permit narrower apis  
> like the MyFn api above.
>
> That said, the interpreter doesn't implement components and apis  
> yet; that's been a big focus of recent attention.
>
> -Jan
>
>