Re: More on view updates and inverse views
Date: Thu, 10 Sep 2009 23:31:24 GMT
Mr. Scott wrote:
> "Vadim Tropashko" <vadimtro_at_gmail.com> wrote in message
>> On Sep 8, 9:35 am, "Walter Mitty" <wami..._at_verizon.net> wrote: >>> This makes the projection view >>> "invertable" even though without the default value it would not be >>> invertable. >> And what is default value in relational terms?
> I think that depends on how a default constraint is expressed logically.
> For a table (x,y,z) that neither allows nulls nor has default constraints,
> the predicate could be Pxyz. If the table were to have a default constraint
> for y, what would the logical expression for the table become? Isn't it
> (Pxyz xor Pxdz) or (Pxyz and (y = d))?
> If it is, then I think a view over (x,z) should be updatable.
My attitude is that if it can't be expressed in an algebra it shouldn't be implemented. Not saying there isn't a way, but I don't know how to express defaults as a constraint nor algebraically. Seems to me a union is always involved, perhaps an exclusive one as you suggest, but the main thing is that I don't know how to do it without declaring a base relvar with attributes x and z, (so it wouldn't be a view). I'd be happy to be corrected! Received on Thu Sep 10 2009 - 18:31:24 CDT