Re: Date and McGoveran comments on view updating 'problem'

On Dec 11, 1:28 pm, vadim..._at_gmail.com wrote:
> I fail to derive it with any assumptions, except the obvious, like
> demanding that D acted only on SP (or S). Therefore, there might be
> some truth to people questioning the validity of deletion from join
> view...

OK, here is what RL says about deletion from join view. In order to be able to represent deletion from the join view as a composition of that deletion into the base relations we need to prove the following:

(x ^ y) ^ R00 = z ^ R00 &
(z ^ (x ^ y)) = z
-> (x ^ (z' v (x ^ R00))) ^ (y ^ (z' v (y ^ R00))) = (x ^ y) ^ z'.

(For those who didn't follow all the messages, x and y are the renamed S and SP relations, and z is the delta).

If we plug in this sentence into QBQL (see the announcement in the other thread), we get the following counter example:

z = {<y=a,x=1,>,}
y = {<x=1,>,}
x = {<y=a,>,<y=b,>,}

(My apologies for using x,y,z names for relation variable names on the left and same variables as attribute names within set definitions on the right!)

On reflection, the result is quite obvious: deleting the <y=a,x=1> tuple from the join view would leave nonempty result {<y=b,x=1>} while deleting projection of the delta onto base results would produce the empty relation... Received on Fri Dec 12 2008 - 04:11:02 CET