Re: RA with MV attributes
From: Bob Badour <bbadour_at_pei.sympatico.ca>
Date: Sat, 20 Jan 2007 14:53:20 GMT
Message-ID: <A1qsh.3113$1x.53267_at_ursa-nb00s0.nbnet.nb.ca>
>
> Certainly, in the platonic sense. It is also useful to consider
> when types and values are reified, and when they cease to
> be so. In some systems, types may be reified at compile
> time but not at runtime at all; this is how C works. Values
> of course are not reified until runtime.
>
>
> Sure. Or perhaps we may be speaking of system that does not have
> subtyping. Out of curiosity, what does Alphora do? I vaguely recall
> reading that they abandoned subtyping at some point.
Date: Sat, 20 Jan 2007 14:53:20 GMT
Message-ID: <A1qsh.3113$1x.53267_at_ursa-nb00s0.nbnet.nb.ca>
Marshall wrote:
> On Jan 17, 5:43 am, Bob Badour <bbad..._at_pei.sympatico.ca> wrote:
>
>>Marshall wrote: >> >>>Types may disappear (or not) at runtime; values do not exist at >>>compile time. If we remove this distinction, we remove valuable >>>capabilities from our language. >> >>I don't follow that at all. Types and values just are.
>
> Certainly, in the platonic sense. It is also useful to consider
> when types and values are reified, and when they cease to
> be so. In some systems, types may be reified at compile
> time but not at runtime at all; this is how C works. Values
> of course are not reified until runtime.
>
>>>Consider the idea of join compatibility. This is violated if we attempt >>>to join two relations each having an attribute of a given name but a >>>different type in each relation. Now, whatever your feelings >>>about how this should be handled, warning, error, whatever, we have >>>to consider the question of when and whether this situation can be >>>detected. >> >>I assume by "different type" you mean neither type shares a subtype >>other than the universal subtype.
>
> Sure. Or perhaps we may be speaking of system that does not have
> subtyping. Out of curiosity, what does Alphora do? I vaguely recall
> reading that they abandoned subtyping at some point.
Indeed. I wrote Bryn Rhodes at Alphora just recently to ask about this very issue. Apparently, they replaced subtyping with a system of implicit conversions, and they hope to reintroduce subtyping later (probably via inheritance by the sound of it.)
I can see how one might use such a mechanism as a limited replacement to subtyping. If one does not understand subtyping, though, I wonder how one is to recognize when it is appropriate (and more importantly inappropriate) to have an implicit conversion.
It seems to me the only safe implicit conversions are conversions from subtypes to supertypes. Received on Sat Jan 20 2007 - 15:53:20 CET