Principal of view equality?

Thread view:

Enable EMail Alerts

Start New Thread

Thread rating:

paul c - 04 Jun 2008 03:53 GMT

I was just looking at Codd's RM 2 book again (the rather short chapter
on views from acm.org) and it seemed to me that what he wrote took it as
essential that a view must always equal the expression that defines the
view. If so, does this in effect constitute a kind of indirect
constraint on any base relations involved?

(Assuming that all relations a dbms stores are to be considered 'true')

Reply to this Message

Brian Selzer - 04 Jun 2008 05:35 GMT

>I was just looking at Codd's RM 2 book again (the rather short chapter on
>views from acm.org) and it seemed to me that what he wrote took it as
>essential that a view must always equal the expression that defines the
>view. If so, does this in effect constitute a kind of indirect constraint
>on any base relations involved?

If you issue a query, does this in effect constitute a kind of indirect
constraint on any base relations involved?

From P --> Q and ~P, you can't conclude ~Q, but
from P --> Q and ~Q, you can conclude ~P. This is applicable here because
that something appears in a base relation doesn't necessarily mean that it
will appear in a view derived from the base relation, so from the absence of
that something in the view you can't conclude that it doesn't appear in the
base relation. On the other hand, from the absence of that something in the
base relations, you should be able to conclude that it shouldn't appear in
the view.

Reply to this Message

Bob Badour - 04 Jun 2008 17:00 GMT

> I was just looking at Codd's RM 2 book again (the rather short chapter
> on views from acm.org) and it seemed to me that what he wrote took it as
> essential that a view must always equal the expression that defines the
> view. If so, does this in effect constitute a kind of indirect
> constraint on any base relations involved?

No. It does, however, specify constraints on the value of the view.
Tuples can exist in the base relations that have no effect on the value
of the view.

Reply to this Message

Evan Keel - 09 Jun 2008 19:04 GMT

> > I was just looking at Codd's RM 2 book again (the rather short chapter
> > on views from acm.org) and it seemed to me that what he wrote took it as
[quoted text clipped - 5 lines]
> Tuples can exist in the base relations that have no effect on the value
> of the view.

I thought that "base relations" were also views. What am I missing?

Reply to this Message

Bob Badour - 09 Jun 2008 20:53 GMT

>>>I was just looking at Codd's RM 2 book again (the rather short chapter
>>>on views from acm.org) and it seemed to me that what he wrote took it as
[quoted text clipped - 7 lines]
>
> I thought that "base relations" were also views. What am I missing?

A base relation in database design is kind of like ground in electrical
design.

The base relations may not be stored as is. One might derive the value
of the base relations from all sorts of different physical structures
arranged and distributed all sorts of ways.

One generally achieves physical data independence by deriving the base
relations from physical structures. One generally achieves logical data
independence by deriving application views from the base relations. One
generally describes the database constraints in terms of the base relations.

Regardless, the expression that defines a view constrains the view and
nothing else. A constraint declared on a view constrains the entire
database.

Reply to this Message

Brian Selzer - 10 Jun 2008 04:40 GMT

>> > I was just looking at Codd's RM 2 book again (the rather short chapter
>> > on views from acm.org) and it seemed to me that what he wrote took it
[quoted text clipped - 8 lines]
>
> I thought that "base relations" were also views. What am I missing?

A base relation is not a derived relation.

Reply to this Message

Evan Keel - 10 Jun 2008 14:49 GMT

> >> > I was just looking at Codd's RM 2 book again (the rather short chapter
> >> > on views from acm.org) and it seemed to me that what he wrote took it
[quoted text clipped - 10 lines]
>
> A base relation is not a derived relation.

Yes, it is.

Evan

Reply to this Message

Brian Selzer - 10 Jun 2008 18:21 GMT

>> >> > I was just looking at Codd's RM 2 book again (the rather short
> chapter
[quoted text clipped - 16 lines]
>>
> Yes, it is.

That's news to me. According to Codd (pages 17-18 of his 1990 book):

Those relations, or R-tables, that are internally represented by stored
data in some implementation-defined way are called the /base relations/ or
/base R-tables/. All R-tables other than base R-tables are called /derived
relations/ or, synonymously, /derived R-tables/. An example of a derived
relation is a /view/. A view is a virtual R-table defined in terms of other
R-tables, and is represented by its defining expression only.

Reply to this Message