Re: is pivoted phones view updateable?

NENASHI, Tegiri wrote:
> "Vadim Tropashko" <vadimtro_invalid_at_yahoo.com> wrote in
> news:1163183196.730989.262270_at_m73g2000cwd.googlegroups.com:
>
>
> > This view is exactly the opposite of oldcontactview. While
> > oldcontactview defines what data must be in the old table if we know
> > the content of the new table, the newcontactview defines new table
> > data if we know the old one. Yes, this view is definitely useful,
> > especially if we want to migrate the data from old table to new one,
> > but I have more advanced application in mind. I reserve new name for
> > it - let's call it an inverse view, while the original oldcontactview
> > would simply be a direct view. Formally, an inverse view is such a
> > view that applying it after direct view will restore the content of
> > the base table that the direct view uses in its declaration.
>
> Let r1 = {'x' = x|x = 1..100}., v1 = {r1|x = 1..10}. One can not recover
> r1 from v1, it means v1 does not have the inverse. It means that a
> random view is not invertible in general.

Random view is not invertible. This is not surprising at all, given that a system of equation in classic algebra is rarely invertible too.

In your example, what is 'x'? Have you meant

r1(x)

a relation with one argument x, such that domain of x is 1..100? This is a constraint:

r1 union {(x=1),(x=2),...,(x=100)} = r1.

Then, what is v1? Perhaps

v1 = r1 join a1

where a1={(x=1),(x=2),...,(x=10)} ?

I agree that this example is not invertible. But switch the roles of 10 and 100 and it becomes updateable! Received on Thu Nov 16 2006 - 15:43:21 CST