Re: Multivariate relation (Was topological databases)

Tegiri

Thank you very much for answering my question, both concisely, and (separately) completely. Reminds me of someone.

> On Wednesday, 28 January 2015 15:41:01 UTC+11, Tegiri Nenashi wrote:

> Getting back to your question, the term "multivariate relation" is synonym for "relation with named attributes", which emphasizes named perspective.

Perfect.

(So it is not what I suspected, it is not n-ary relations.)

> Section 3.2 from the Alice book

Oh God ! A well-known abortion. An intended textbook, available free. Unfortunately cited in many papers. Which damages the credibility of, and declares the watery foundation of, said papers. I take it, it is beloved of mathematicians.

> Unnamed perspective is something that has been used in mathematics for a long long time. Please note that Tarski's algebra of binary relations also uses unnamed perspective. It seems that only after Codd's invention of relational model the named perspective reigned supreme, so that database practitioners might not be even aware of relations with unnamed/positional attributes.

Confirmed. I do know about it, but I readily agree that most practitioners do not.

> In those rare cases where SQL uses positional notation, it is more of misnomer.

Meaning, in SQL, not mathematics, it is a misnomer.

> Getting back to your question, the term "multivariate relation" is synonym for "relation with named attributes", which emphasizes named perspective.

By "emphasises", do you mean "implements". The unnamed perspective is not implemented.

> Apparently, a person who originally designed SQL "order by" clause was not familiar with continuous functions.

I think not. He understood completely, that in the practical world, the physical universe, the unnamed perspective is irrelevant. Or, the unnamed/posititional perspective would be a nightmare for the developer, and for an SQL implementation, ridiculous in a practical universe.

And then it follows, that the positional in SQL is the position number of the [named] attributes (in the SELECT list).

> Compare
>
> select name from employees order by 1
>
> with
>
> select name from employees order by 1.000001

Well, only the commercial SQLs (not "rare") handle things like that, and handle them correctly. The second example has a value distinction that is meaningless, irrelevant, to the operation. This might give a better understanding:

SELECT /column_name_list/
____FROM { /table/ | /view/ | /derived_table/ } ____ORDER BY { /column_name/ | /column_no } [, { /column_name/ | /column_no } ] ...

Where /view/ and /derived_table/ are merely SELECT statements.

SQL is a practical-only data sublanguage.

It is not reasonable to expect theoretical closure from it. We do have full transitive closure (if one implemented that in the tables, not by magic); join transitive closure (by magic, fully automatic); etc, that is relevant in the physical universe.

If the TTM groupies collected half a brain amongst the lot of them, they would simply write an interpreter for relational algebra, and stick it on top of SQL. I said so six years ago; someone else said the same thing three years ago. Deaf ears. But for the mathematical papers (un-scientific) that support, and re-inforce, the monolith, the Tower of Babel, the everything-in-one-language. In these days of components and libraries.

To anyone who says that SQL is "not relational" or that something in a Relational Database can't be done in SQL, I have always said: give me the problem, and I will give you the fully Relational solution, free of charge. While practitioners and developers take up my offer frequently, and leave satisfied, mathematicians never take it up. They prefer to hang on to the notion, un-evidenced, that "SQL is broken, it is not relational".

Have you seen the O.2.SQL abortion ? The creatures the OO boys come up with these days, sheeesh. I do all that in plain SQL, on a pure RDb, with standards and simple rules for implementation. No fuss, no mess, no fanfare. That one is not free.

Last question. Why then, call the unnamed perspective "multi-variate relations" ?

Cheers
Derek Received on Wed Jan 28 2015 - 07:01:11 CET