Re: 0.99999998 (was: Unknown SQL)

Sam Staton wrote:
>

 [snip]
>
> He had, as many beginners would, assumed a higher notion of equality
> than was supported. He assumed that 'Mr Fred Smith' = 'Mr Frederick
> Smith'. And they ARE equal on a higher level.
>
> Incidentally, the person who wrote the initial queries was a newcomer
> to SQL with little experience (I notice that people have been claiming
> that SQL is easy to use for newcomers). I am not particularly biassed
> to OO or relational, but I must admit that in the case of an OO world
> with enforced unique identifiers for each object, the initial
> programmer would have been forced to think twice before assuming that
> the names would match, instead of the GUI (OK, Access) deciding it for
> him based on the primary keys.

In normal OO you don't have object identifiers inherent in the data. The object is an entity of it's own and contains the data of some object in the application domain. I don't know of any standard OO way of ensuring uniqueness of the data. Unless you store all object of the same kind (class and subclasses) in a big container and search for matching data all the time.

> This comes down to the debate about the use of unique identifiers.
>
> - An OODBMS can enforce them (many do implicitly).

An OODBMS can't enforce this on it's own. How is it supposed to guess that names should be unique or what really identifies a person in a certain context?

> - The relational model can support them but it is up to the user to
> implement them (as you highlight, in most cases the user will do).
> However, they are not guaranteed.

Actually the DBA should put the correct uniqueness constraints. Then the RDMBS knows how to identify a person. And probably there will be some other kind of surrogate key which corresponds to OODBMSs' object identity.

Regards,
Heinz Received on Sat Jul 21 2001 - 18:29:57 CDT