Thread: Fwd: Re: SQL3 UNDER

Fwd: Re: SQL3 UNDER

From
"Robert B. Easter"
Date:

----------  Forwarded Message  ----------
Subject: Re: [HACKERS] SQL3 UNDER
Date: Tue, 23 May 2000 21:53:24 -0400
From: Robert B. Easter <reaster@comptechnews.com>



On Tue, 23 May 2000, Chris Bitmead wrote:
> "Robert B. Easter" wrote:
> > 
> > On Tue, 23 May 2000, Chris Bitmead wrote:
> > > Maybe it would help if you have two examples. One that only uses UNDER,
> > > and one that only uses INHERITS, and explain how one or the other can
> > > work differently.
> 
> Yes but how does a pure UNDER example actually work different to a pure
> INHERITS example? You've created various tables below (combining INHERIT
> and UNDER unfortunately), but how will the INHERITS hierarchies and
> UNDER hierarchies actually work differently in practice?
> 

I guess I've said most of what I can say about this idea.  Attached is another
GIF picture of the ideas, I suppose.  If I come up with a good example, I'll
post it.

I'm willing to admit my idea could be very flawed!  I'm hoping others in here
will find it worthy enough to try to find those flaws and examples on their own.

I've started posting this OO stuff to pgsql-hackers-oo@postgresql.org.  I'll
try to not post anymore oo stuff in pgsql-hackers (if there is even anything
else say about this).

Good luck,
Robert B. Easter





-------------------------------------------------------



-- 
Robert B. Easter
reaster@comptechnews.com

Re: Fwd: Re: SQL3 UNDER

From
Hannu Krosing
Date:
"Robert B. Easter" wrote:
> 
> I've started posting this OO stuff to pgsql-hackers-oo@postgresql.org.

Where can I subscribe to that list ?

I must have missed or lost the announcement possibly when I lost 3 days worth 
of mail sunday night ;(

> I'll try to not post anymore oo stuff in pgsql-hackers (if there is even 
> anything else say about this).

The OO-PostgreSQL discussion is not even near being over ...

-------------
Hannu


Re: Fwd: Re: SQL3 UNDER

From
"Robert B. Easter"
Date:
On Wed, 24 May 2000, Hannu Krosing wrote:
> "Robert B. Easter" wrote:
> > 
> > I've started posting this OO stuff to pgsql-hackers-oo@postgresql.org.
> 
> Where can I subscribe to that list ?

Just send a message to pgsql-hackers-oo-request@postgresql.org
with subject "subscribe"

> 
> 
> The OO-PostgreSQL discussion is not even near being over ...
> 

Well, I hope some one sees my point about INHERITS and UNDER
maybe being complementary.  UNDER is a single inheritance container/tree all
contained inside maximal supertable.  INHERITS provides multiple inheritance
and can provide links between tables in different containers/trees, subject to
some restrictions.  I think it deserves some looking at rather than just doing
away with INHERIT for just UNDER.  (again I can be wrong).  I guess its hard to
explain.  I still need to provide good examples.  I can best describe the
difference as UNDER creates circles within circles representing tables and
subtables.  INHERITS provides for circles/tables to overlap (to be cloned in a
sense) and allows it multiple overlapping/merging.  The INHERITS does it as it
is now that way, by merging same name attributes from two or more parents into a
single child.  INHERIT is like cells reproducing using one or n parents. 
UNDER is like a single cell making baby cells inside of itself.   :-)  hehe



Re: Fwd: Re: SQL3 UNDER

From
Chris Bitmead
Date:
"Robert B. Easter" wrote:
> Well, I hope some one sees my point about INHERITS and UNDER
> maybe being complementary.  UNDER is a single inheritance container/tree all
> contained inside maximal supertable.  INHERITS provides multiple inheritance
> and can provide links between tables in different containers/trees, subject to
> some restrictions.  I think it deserves some looking at rather than just doing
> away with INHERIT for just UNDER.  (again I can be wrong).  I guess its hard to
> explain.  I still need to provide good examples.  I can best describe the
> difference as UNDER creates circles within circles representing tables and
> subtables.  INHERITS provides for circles/tables to overlap (to be cloned in a
> sense) and allows it multiple overlapping/merging.  The INHERITS does it as it
> is now that way, by merging same name attributes from two or more parents into a
> single child.  INHERIT is like cells reproducing using one or n parents.
> UNDER is like a single cell making baby cells inside of itself.   :-)  hehe

Would you still be having these thoughts if you were looking at the
older SQL3 draft that included multiple inheritance UNDER? The newer
UNDER appears to be a subset, which I presume they adopted to get the
proposal out the door quicker. Personally I'd like to implement the
SQL3-1994 extensions as well, because they actually seemed well thought
out (I'm thinking particularly of the rename stuff).


Re: Fwd: Re: SQL3 UNDER

From
"Robert B. Easter"
Date:
On Wed, 24 May 2000, Chris Bitmead wrote:
> "Robert B. Easter" wrote:
> > Well, I hope some one sees my point about INHERITS and UNDER
> > maybe being complementary.  UNDER is a single inheritance container/tree all
> > contained inside maximal supertable.  INHERITS provides multiple inheritance
> > and can provide links between tables in different containers/trees, subject to
> > some restrictions.  I think it deserves some looking at rather than just doing
> > away with INHERIT for just UNDER.  (again I can be wrong).  I guess its hard to
> > explain.  I still need to provide good examples.  I can best describe the
> > difference as UNDER creates circles within circles representing tables and
> > subtables.  INHERITS provides for circles/tables to overlap (to be cloned in a
> > sense) and allows it multiple overlapping/merging.  The INHERITS does it as it
> > is now that way, by merging same name attributes from two or more parents into a
> > single child.  INHERIT is like cells reproducing using one or n parents.
> > UNDER is like a single cell making baby cells inside of itself.   :-)  hehe
> 
> Would you still be having these thoughts if you were looking at the
> older SQL3 draft that included multiple inheritance UNDER? The newer
> UNDER appears to be a subset, which I presume they adopted to get the
> proposal out the door quicker. Personally I'd like to implement the
> SQL3-1994 extensions as well, because they actually seemed well thought
> out (I'm thinking particularly of the rename stuff).

There are documents at

ftp://jerry.ece.umassd.edu/isowg3/dbl/BASEdocs/sql4hold/

that maybe we should look at.  It *might* represent what is planned for SQL4. 
It shows UNDER accepting multiple supertables like the 1994 draft.  However,
these documents are dated 1996 and probably don't really represent SQL4, which
might take many years still until its a standard.  By the time SQL4 comes out,
there's no telling what it will look like.

I'm thinking, it might be best just to implement UNDER as it stands in the
official standard for now.  Leave INHERIT the way it is (for the most part) and
implement UNDER separately.  Continue to use inherit if you need multiple
inheritance.  If you implement multiple inherit UNDER, it will create a user
base that will depend on that functionality.  Later, it will be difficult to
change that functionality if the official standard for UNDER does switch to
multiple inheritance but in a way that is incompatible with yours.  Its not a
good idea to second guess the future standard.  People already use INHERIT the
way it is and it can be used in combination with UNDER.
-- 
Robert B. Easter
reaster@comptechnews.com