Re: is pivoted phones view updateable?

NENASHI, Tegiri wrote:
> Very well, I understand the notation now. But I am disappointed. Your
> rules, one can better write them in logical notation:
>
> 1. X : predicate that defines the relation
> 2. X --> A : X implies B; the constraint

Perhaps. Although X greater than A in lattice terms, and the lattice in question is not a boolean algebra, remember?

> 3. A --> B : A implies B; how B is connected to A

Again, A greater than B in lattice terms.

> 4. Y = X and B : predicate that defines the view.
>
> By modus ponens (X, X-->A, A-->B|= B)

By transitivity of the lattice order

X>=B

which is the same as

X /\ B = X

> one immediately sees that B is
> true and X = Y
>
> Can you give a more interesting concrete example of the invertible view
> ?

This example is obvious informally as well. We had an unitary relation constrained in {1..10}, we made a view with a "relaxed" constraint {1..100}. Clearly it is the same as the base relation!

The phones example (the one where there were two tables and no NULLs)

X = (Y /\ A) \/ (Z /\ B) -- aka ConsolidatedPhones

is less trivial. Next, there goes an example with join

Y = X join A

subject of constraint

project_y (X(x,y)) is subset of project_y (A(y,z))

which I admittedly failed to work out formally yet (although, I know what an inverse view is).

And, finally, there is a join of relation with function where the joint column is projected away. I'm not even sure if relational lattice is applicable, as it seems that we need a difference operator... Received on Sat Nov 18 2006 - 00:13:22 CET