[prev in list] [next in list] [prev in thread] [next in thread] 

List:       e-lang
Subject:    Re: [e-lang] Joe-E
From:       "Rob Meijer" <capibara () xs4all ! nl>
Date:       2010-04-30 19:04:05
Message-ID: 0888b8462e5e659b99fa72feb5e089a5.squirrel () webmail ! xs4all ! nl
[Download RAW message or body]

On Fri, April 30, 2010 18:41, David Wagner wrote:
> Rob Meijer wrote:
>> wouldn't it be more logical for any 'container' templates to be made
>> honorarily transitively Powerless with respect to T ?  That is, a
>> smarter
>> validation rather than a richer library ?
>
> Can you explain more what you are proposing?  I don't follow yet...
>
> In Joe-E, collection classes like ImmutableArray and PowerlessArray are
> provided by the Joe-E library.  They are exempt from static verification,
> and instead are carefully manually vetted.
>
> We can't ever allow creation of an ImmutableArray that contains a
> non-Immutable elements, because other code relies upon the property
> that every object of type ImmutableArray<T> is guaranteed to be
> Immutable, regardless of the type T.  (Reminder: Immutable is a
> transitive property.)
>
> We generally can't reuse Java collection classes, because they generally
> aren't capability-safe and don't have the properties we need.

Ok, that basically disqualifies most of what I tried to suggest.

What I suggested was that if you have 3 classes:

* String          ( Powerless)
* PowerfullFoo    ( non powerless )
* PowerlessBar    ( Explicitly Powerless)

Than for any collection class for example the priority queue you could
transitively infer:

*  PriorityQueue<String>         ( Powerless )
*  PriorityQueue<PowerfullFoo>   ( non powerless )
*  PriorityQueue<PowerlessBar>   ( Powerless )

But if as you stated those containers like PriorityQueue aren't capability
safe, such inference would not be possible.

But given this major issue, and given the fact that a
CapabilitySafePriorityQueue  would be needed anyway if anyone needed a
priority queue, would it not still make more sense to look at the explicit
tagging for generics, for example introducing a 'CapabilitySafeContainer'
concept, for what these simple inference rules would make sense? That
would seem like a lot less work than having to rewrite capability safe
alternatives for all of the java ones?


_______________________________________________
e-lang mailing list
e-lang@mail.eros-os.org
http://www.eros-os.org/mailman/listinfo/e-lang
[prev in list] [next in list] [prev in thread] [next in thread]