Re: Date, Darwen, Pascal and the alternative to Nulls in the RM

From: paul c <toledobythesea_at_oohay.ac>
Date: Thu, 23 Mar 2006 20:56:15 GMT
Message-ID: <PXDUf.175407$H%4.146848_at_pd7tw2no>


David Cressey wrote:
> "falcon" <shahbazc_at_gmail.com> wrote in message
> news:1143134130.259128.181350_at_g10g2000cwb.googlegroups.com...
> ...

>>Notice the missing [222-22-2222, joe, null, C] entry since this tuple
>>doesn't evaulate to TRUE (due to the null), it can not be part of any
>>result set.
>>

>
>
> I'm with you up to this point. Notice that the tuple you flag as missing is
> missing because you preformed an INNER join. See below.
>
> ...

Although I haven't read the SQL ANSI/ISO Standard (hard to believe that something so big could possibly be coherent, even snippets of it seem to confirm this), I have read elsewhere that it doesn't mention the word 'relation' (which may be to its credit in a half-assed way). Don't see how SQL creations such as 'INNER join' could involve tuples.

pc

>>The missing entry for 'joe' in the last query might be very difficult
>>for someone coming from an sql background, working at a real company.
>>I understand that logical models should not worry about 'real-world'
>>matters.  As far as I can see, this problem of missing joe can be
>>mitigated by allowing something similar to outer joins when the
>>result-set does indeed include a null, such as follows:
>>[111-11-1111, jim, 21, A]
>>[222-22-2222, joe, null, C]
>>[333-33-3333, jack, 23, B]
>>

>
>
> Yes. Coming from and SQL background, my reaction would be "what the hell
> did they do with the outer join?" Once they provide an outer join (to
> satisfy information requirements), it will give you the answer you outlined
> above, null and all.
>
>
>>But then, this is not a relation.  In other words, the previous three
>>lines may appear on a user's computer screen, but they can no longer be
>>part any further relational operations.
>>

>
>
> Too bad. This is one of the differences between relational algebra and the
> relational data model. The model was closely based on the algebra, with due
> attention to detail. But the relational data model is a DATA model, and not
> a mathematical model.
>
>
>
Received on Thu Mar 23 2006 - 21:56:15 CET

Original text of this message