Re: a union is always a join!
Date: Tue, 17 Mar 2009 20:11:10 GMT
Message-ID: <ytTvl.17237$Db2.8414_at_edtnps83>
Walter Mitty wrote:
> "Brian Selzer" <brian_at_selzer-software.com> wrote in message
> news:c4Fvl.26433$ZP4.2289_at_nlpi067.nbdc.sbc.com...
>
>> As an exercise, try to write a set-based update trigger (a statement >> trigger in Oracle) on a table with only a composite natural key. You'll >> find that when a column in the key may be the target of an update, that >> you can't just join deleted to inserted (old to new in Oracle) to >> determine exactly what is different. (That's one of the main reasons >> Oracle and many other implementations support row triggers.) >> >>
>
> The idea behind entity keys is that they are immutable. When analyzing the
> universe of discourse into entities, it's important to discover a reliable
> identifier for each entity. In practice, it's sometimes necessary to
> synthesize identifiers, because the identifiers in common use are
> unsuitable. An example might be the use of common names for people at the
> time when information systems were first being computerized.
> ...
Walter, please don't fall for a mystic trap/trick question. It is pointless to respond to a puzzle that mentions 'natural' keys which are a myth like the Emperor's New Clothes, some people don't see them and some people dream they see them. Any answer you give will be rejected in a picayune way for not using 'natural' keys, those being recognizable only by mystics and not the rest of us. The keys that are chosen by a design are all that matters. This means that 'synthetic' keys is a pointless term as well. The chosen keys will be as 'common' as the application needs them to be. There are some old posts from Bob B about this pointing out that a useful key will become 'familiar' very quickly, ie., immediately, even if it wasn't before an app goes into play.
Another trap here is Oracle - what's it got to do with db theory? I've never seen any Oracle documentation of its features/ops that used only algebra or calculus for definitions. What I've seen pretends that casual explanation is as good as a formal definition.
I think it might be more useful to ask whether there is a difference/delta set or relation that obeys all the constraints of the target variable. Received on Tue Mar 17 2009 - 21:11:10 CET