Re: is pivoted phones view updateable?

NENASHI, Tegiri wrote:
> "Vadim Tropashko" <vadimtro_invalid_at_yahoo.com> wrote in
> news:1163523276.819401.288540_at_m7g2000cwm.googlegroups.com:
>
> > NENASHI, Tegiri wrote:
> >> The view is not correct: suppose (2) has not the FAX. One goes to
> >> lose (2) in the view oldcontactview.
> >
> > The view (2) is supposed to reflect the data in the table
> >
> > ID VOICE FAX
> > 1 4150000000 4081111111
> > 2 80012345672 6501234567
> >
> > which doesn't have NULLs. Perhaps this should be stated as an explicit
> > constraint?
>
> Pardon me but you did not understand well what I have wrote:

indeed

> Suppose that the row with id = 2 of the newcontact table has not the fax
> number. You go to lose the personne whose id = 2 in the view
> oldcontactview. What you do in this situation ?

This has been vaguely mentioned in the other thread. There are implicit constraints in a view that must not allow such updates.

Consider the function

y = 2*x

on the domain of integers. Scalar value "x" is analogous to a a relational variable, and scalar value "y" corresponds to a view. I can change the state of x to any integer value. I cannot change the value of y to anything, because there is an implicit constraint that the range of y is that of even numbers.

> > Regardless, the example in the beginning of the thread is even more
> > simple (and NULL immune). Although, it seems to require a whole lot of
> > algebraic manipulations to invert it.
>
> I do not understand all the notation.

This is why it is kept to the minimum?-) OK, clearly you know what /\ and \/ is. Next, 00 is an empty relation with empty attribute set. Finally, 01 is the relation with empty attribute set and one tuple (it is the bottom element in the lattice). Did I miss anything else? Received on Wed Nov 15 2006 - 02:08:54 CET