Re: Thinking about MINUS

On Jan 7, 4:57 pm, paul c <toledobythe..._at_oohay.ac> wrote:
> Marshall wrote:
> > On Jan 7, 8:48 am, Bob Badour <bbad..._at_pei.sympatico.ca> wrote:
> >>paul c wrote:
[lots of snippage]
>
> >>> eg., the
> >>>case when two operands share an attribute name that has different types.
>
> >>Wouldn't one end up with a resulting attribute defined as a union-type
> >>in a relation with no rows?
>
> > That's certainly one way to handle it. However I would be more inclined
> > to think of it as a type error.
>
> I don't believe that there is any such thing as a type error and say
> that they are sort-of outside what we are talking about. There are user
> errors, programmer errors and designer errors and maybe some others, but
> no such thing, in our possible understanding, as something like a
> universal error.

I didn't really understand that. However I will note that a type error is usually considered a programmer error.

> I think types stand on their own, unchangeable and
> constant and that it is in fact illogical and a fallacy to say that an
> agree-upon type is invalid, unless the details of its definition are
> somehow contradictory, that is, contradictory in a way that matters to
> the point.

Ah, but when we are joining, which is to say unifying, two relations, and the two relations each make different, *incompatible* claims about the type associated with a particular attribute name, then I would say that well meets the criteria for being "contradictory."

> When we define the qualities of a result, say its attributes, I think we
> are dismissing, out-of-hand and rightly, certain values and operational
> conditions. I think this is just as logical, maybe more logical if that
> makes any loose sense, than basing "valid procession" on the inputs.
>
> Marshall, one reason I like your posts is because you are good at coming
> up with examples. Grateful if you'd try to counter the above with some.
> Maybe my attitude is a useless one!

Thanks! However I'm not quite clued in on what you're talking about yet so I'm not in a good position to respond.

Marshall Received on Mon Jan 08 2007 - 09:45:03 CET