Does it make sense to remove slots ?

[Massimo Dentico]

Todd Fleming wrote:


> Massimo Dentico wrote:
> 
>>    http://www.dbdebunk.com/page/page/626998.htm
> 
> 
> Nice article. Like the rest of their site, I agree with some points he 
> makes and disagree with others. (I agree with articles on his site that 
> say that SQL NULL is incompatible with the relational model). Here are 
> some choice quotes:
> 
> "In THE THIRD MANIFESTO, Hugh Darwen and I demonstrate clearly--and I 
> think conclusively -- that if your inheritance model involves "pointers 
> or references," then it's logically impossible to deal properly with the 
> idea that a circle is an ellipse!  In fact, we strongly suspect that 
> it's this fact (the fact, that is, that pointers and a good model of 
> inheritance are fundamentally incompatible) that's the root of the problem."

Now I remember why they forbid pointers in their model;
assume that (informal for brevity):

Ellipse is a type, defined early.
Ellipse has a method ScaleXBy(s), which scale the X axis by the value s.

type Circle: subtype(Ellipse) // Circle is a type, a subtype of Ellipse

var e: Ellipse // e is a variable of type Ellipse
var c: Circle  // c is a variable of type Circle

c.Init(10)     // c is a Circle of radius 10

You can usually write:

e <- c         // assign c to e

Now, if this means "the value of c is copied in e"
there are no problems: the variable e now contains
a value of type Circle which is a proper subtype
of Ellipse.

But if it means "e now points to the same value
pointed by c" you are in trouble as soon as you
write this:

e.ScaleXBy(2)

Now e is pointing to a value of type Ellipse,
so this statement is correct...

... but c is pointing to a value of type Ellipse
also, which is NOT correct!

So this statement is at the same time correct
and incorrect: an absurdity.


Going back to your example, which is different;
in that case they call your "squish" (my "ScaleXBy")
a *mutator* and they wrote:

    http://www.dbdebunk.com/page/page/626958.htm

    Liskov and Wing tacitly seem to support unconditional
    inheritance of mutators, but they never discuss--in fact,
    they don't even mention--the logical absurdities that
    are necessary consequences of such unconditional
    inheritance.

So, as I understood it, in their model such mutator
is NOT inherited.

But if you wrote "squish" in a way such that it returns
an Ellipse, without updating the object, then it is
an *observer* and so there are no problems in inheriting
it.

Again, this is *my* understanding and I could be wrong.


Regards.

--
Massimo Dentico