Oracle FAQ Your Portal to the Oracle Knowledge Grid

Home -> Community -> Usenet -> comp.databases.theory -> Re: Example of expression bias?

Re: Example of expression bias?

From: Tony D <>
Date: 20 Jun 2006 17:04:58 -0700
Message-ID: <>

Well, here we are again ...

Cimode wrote:
> You are redefining RM.

No I'm not. You're about to try though. I would also point you towards a boxed out section, in bold type, two-thirds of the way down page 114 in the 7th edition of Date's Introduction to Database Systems. The preceding paragraph is also helpful.

> A relvar has associated properties among which
> a data type expressing how it draws values from a domain and restricts
> that extraction only to permissible values.

Mmmm. How about, a relvar is a variable which can indicate relation values. It has an associated set of constraints that define which relation values it is acceptable for the relvar to indicate. Clumsy I'll grant you.

> A data type in RM = (a domain1 to draw values from) + (restrictions
> implemented on domain1 --> domain constraint) + (operators that can be
> defined using that data type)

This may be one definition of a data type (not quite one I'd accept, as we've thrashed over elsewhere), but there is nothing particular to RM about this.

> An attribute does not belong to a data type. A data type is the set of
> rules that apply to attribute for saying whether or not values in
> tuples are permissible values in the system.

If an attribute doesn't belong (for want of a better word) to a data type, how do you specify what the acceptable values for it are ? Your second sentence moves in a single breath from attributes to tuples to the system as a whole.

> That's what I mean by saying you are at implementation level.

Actually no, that's a base level definition, valid in beginner's denotational semantics, of what variables *mean*. You *could* implement them that way, but not necessarily; as long as the net effect is provably the same you have produced an acceptable implementation of the definition. (You may care to do some reading on denotational semantics; google for Christopher Strachey, Dana Scott and/or Joe Stoy for starting points.)

That's another plus for FP; if you manage to define an acceptable set of denotational semantics, it doesn't take a lot of work to turn them into an executable specification. Maybe that's why Perl 6 is (according to Larry Wall, anyway) being trial implemented in Haskell. Off topic, but mildly interesting in context.

> If you are considering that data types are not defined in RM. That
> shows the level of ignorance about RM concepts your in.

And if you're thinking that what RM has to say about data types constitutes a complete, sound and usable definition of data types, you've got some reading to do.

> If you state that RM has said nothing about data types: It can only
> mean one thing -->you have never read a book about RM.

I've read a few; they've had little to say about data types, apart from they exist, you need them, and you need to be able to define your own. (Date moves on into how to define your own via Tutorial D; but since the starting point is that types are orthogonal to the relational model, this is "add-on" territory, not fundamental to RM itself.)

> I am tired with you.

Time for sleepy-byes, then.

> I have asked a question you did not answer.

And I told you I wasn't going to blow by blow our last exchange. (And here I am doing it again. Oscar, you have much to answer for !)

> Definition RM, through SQL perception limits the understanding of RM.

I'm not sure this is a sentence, and even if it is, I'm not sure what (if any) meaning it is intended to convey.

> That FP is a set of definition sufficiently clear. All I have seen so
> far in your definition is fuziness..

Clear, precise definitions of lambda calculus are readily available. In fact, the wikipedia entry is entirely acceptable, because the syntax and semantics of the lambda calculus are so straightforward. If you suspect wikipedia, then you can buy just about any FP book and it will explain it in no time (the operational semantics in Peyton Jones's "Implementation of Functional Programming Languages" takes a whole 9 pages (pp 14 - 23), including worked examples and quite a lot of explanatory text.)

> No. If the answers were convincing, I may have paid a closer look.
> But I confirm what I thought.

Now there's a surprise. Received on Tue Jun 20 2006 - 19:04:58 CDT

Original text of this message