Re: What predicates the following relation represents

From: mAsterdam <>
Date: Thu, 01 Apr 2004 10:28:33 +0200
Message-ID: <406bd2b2$0$567$>

Mikito Harakiri wrote:

> "mAsterdam" wrote

>>>>Key is fine, but what about the fact
>>>><PART=Nuts,  SOLD=0>
>>>>Is it the same as the absence of the tuple?

>>In other words: what fact is conveyed? What does it *mean* ?

> I don't understand. Do you request more details? More formality?

Not just any details. Specifically what does the fact "<PART=Nuts, SOLD=0>" tell us. The only formality I am looking for here is disambiguity - we should know which facts are conveyed. Stick to "don't lie in your database".

> OK, more formally, when do we find convenient to consider
> |--A
> |--SALE(Nuts, 0)
> as equivalent to
> |--A
> |--!exists x>0: SALE(Nuts, x)
> where A is some arbitrary set of clauses?

I'll (reluctantly) join the speculation a little bit to explain, I will try to conform to the guesswork Eric has done:   |--SALE(Nuts, 0)
may mean something like: Mr FooBar sold 0 nuts today. |--!exists x>0: SALE(Nuts, x)
may mean something like: Mr FooBar did not sell nuts today.

Do they convey the same fact? Not by itself. Maybe they do, but we don't have what we need to decide it. Maybe we could decide if we knew wether mr FooBar tried to sell nuts today.

This is not in the information we were provided with, so even more speculation is called for. This does not seem very useful. Most of the guessing would be unnecessay if there would have been something to go on at the start: the predicate.

> where A is some arbitrary set of clauses? (There might be not much substance
> in that kind of questions within logic realm, but I want to understand what
> aggregation is and its connection to logic)>

>>>Not necessarily - the above looks like a derived relation, rather than a
>>>base one. If it were a base one, given common-sense interpretations of
>>>the attribute names, I'd say the presence of the tuple was meaningless.
>>>If a derived one, then it has meaning - no nuts were sold.
>>>It's hard to say
>>>without understanding the requirements, but is certainly suspicious.
>>Requirements is much to broad. Meaning. Just meaning. Necessary and

> If you refer me to Model theory, let me think it over.

Not that I am aware of. Which model theory?

>>>>My humble interpretation is that the SALES table with attributes PART
>>>>and SOLD is still physical model. This relation on logical level is
>>>>select PART, sum(SOLD) from SALES
>>>>group by PART
>>>Agreed, that's probably the way it should be, but the SQL above isn't
>>>equivalent to the data you gave earlier, or nuts wouldn't appear twice.
>>>Perhaps there's some external predicate that might make the relation
>>>valid - but I doubt it. It's most likely simply poor design.
>>To achieve conceptual integrity one needs concepts.

> Be careful using the same abstract word twice in the sentence.

Where is the danger in this one? Received on Thu Apr 01 2004 - 10:28:33 CEST

Original text of this message