Re: Does Codd's view of a relational database differ from that ofDate&Darwin?[M.Gittens]

Alexandr Savinov wrote:
>
> I still do not want to write joins.

If you really don't want to do joins, and you're willing to give up expressive power to achieve this, simply limit yourself to querying one table at a time.

But your desire to avoid joins is misplaced. What would you say to someone who wanted to avoid projection? What would you say to a programmer who wanted to avoid iteration?

> You propose to implelent it at the user interface level
> but I would prefer to have more support from the database. For example,
> the database should know about alternative paths, which are represented
> in a way different from explicit joins. One approach consists in
> specifying an intermediate table in the path. In this case the qurey
> might look as follows:
>
> get all houses related to 'Smith' via HouseEnsurance

select * from Houses natural join HouseInsurance natural join People where Name = 'Smith';

> Nice format, is not it? Would not you like to have such a facility?

I *already* have such a facility; it's just syntax on natural join.

> So the question is do we really need to have such a freedom which allows
> us (makes it easy) to produce meaningless results?

We certainly need not to throw out certain queries just because we don't see today how they're useful. (Your use of the word "meaningless" is incorrect; all RM queries are meaningful, you just might not need that particular meaning at this time.)

> For example, it will include not only facts but also relationships.

Relationships are facts.

Marshall Received on Fri Jun 17 2005 - 17:12:26 CEST