Re: 4 the FAQ: Are Commercial DBMS Truly Relational?

From: Marshall Spight <mspight_at_dnai.com>
Date: Sun, 10 Oct 2004 15:43:25 GMT
Message-ID: <xMcad.156343$wV.96061@attbi_s54>

"Laconic2" <laconic2_at_comcast.net> wrote in message news:Wv2dnSILNblTtPTcRVn-gg_at_comcast.com...
>
> "Marshall Spight" <mspight_at_dnai.com> wrote in message
> news:Yo2ad.220257$D%.89507_at_attbi_s51...
>
> > Right, but although you certainly need an operation like this,
> > SQL's outer join is not the only way to get it. I rather think
> > that GROUP/UNGROUP, and result columns of relation type,
> > are a better approach, in part because NULLs are not required.
>
> I'm not familiar with UNGROUP. Could you say some more, or give me a
> pointer?

Relation 1:

(1,a)
(2,a)
(2,b)

Relation 2:

(a)
(b)
(c)

R1 group R2:

({1,2}, a)
({2}, b)
({}, c)

This is essentially a right outer join, but there are no nulls in the result.

I believe TTM has such an operator. I should probably go back and read about it.

Marshall Received on Sun Oct 10 2004 - 10:43:25 CDT