Re: Representing Student Activity Score

"Neo" <neo55592_at_hotmail.com> wrote in message news:4b45d3ad.0407221642.2ddcaed0_at_posting.google.com...
> > > Could you show how to explicity represent the following info in RM:
> > > john take test.
> > > john get 95.
> > > (john get 95) for (john take test).
> >
> > PERSON SCORE
> > John 50
> > Mary 75
>
> A program could derive each tuple's relationship to a person and score
> (ie tuple1's person is john, tuple1's score is 50. It doesn't encode
> the original info explicitly.

Because it is not necessary (or proper) to refer to the tuple by its (row) number. Sets are not ordered, so tuple numbers are meaningless. They can change. Besides, why refer to a tuple by its number rather than by the Primary Key? The PK uniqiely identifies the tuple. You don't need a unique identifier for the unique identifier. That is redundant. Boy, this sounds familiar, as I knew it would.

>
> > Or, if you want a sentence as a result;
> > SELECT person||' scored '||score||' in the exam.' FROM =test_scores;
> > John scored 50 in the exam.
> > Mary scored 75 in the exam.
>
> In the above, which is closer to desired, the SQL statement itself is
> providing some of the relationship between the data, not data in the
> db, unlike that shown at www.xdb2.com/Example/StudentActivityScore.asp

Ummm, dude, somewhere along the line, some bit of code is going to provide the relationship, unless you've invented a psychic computer interface. Anyway, in this case, it's very simple SQL providing the link because the underlying database engine (i.e., Oracle) handles all the details of physical storage by mapping this physical layer to a logical layer that SQL accesses. People ask the questions, people type the question in sql and optionally format the answer so it is more readable, and people read the answers. In your system, the programmer must know all possible questions that will be asked, and all possible relationships among the data- all before entering any data. Guess what- requirements change. A lot. Your system works well with known sample data under controlled conditions. Try it in the real world. Create a printed report requested ad hoc in 5 minutes or less. If you can do it, Gene may even forgo the $1000 you owe him (up to him, of course).

I suspect you have an interest in AI, and it looks like you are trying to connect an inference engine to an rdb, but you tripped over the db theory. Received on Fri Jul 23 2004 - 05:05:15 CEST