Thread: [OT] why not keeping the original column name in catalog after a drop?

[OT] why not keeping the original column name in catalog after a drop?

From
Luca Ferrari
Date:
Hi all,
when you drop a column on a table the pg_attribute is updated and the
name of the column is changed with an almost fixed identifier that
reports only the original column position:

/** Change the column name to something that isn't likely to conflict*/       snprintf(newattname, sizeof(newattname),
             "........pg.dropped.%d........", attnum);       namestrcpy(&(attStruct->attname), newattname);
 

I'm wondering what is the problem in placing the old/original name
after the "pg.dropped" prefix. I know that the tuple in pg_attribute
is temporary, but what are the possible conflicts the comment talks
about?

Thanks,
Luca



Re: [OT] why not keeping the original column name in catalog after a drop?

From
Pavan Deolasee
Date:
On Wed, Nov 13, 2013 at 1:22 PM, Luca Ferrari <fluca1978@infinito.it> wrote:


I'm wondering what is the problem in placing the old/original name
after the "pg.dropped" prefix. I know that the tuple in pg_attribute
is temporary, but what are the possible conflicts the comment talks
about?


May be when a column with the same name is added and again dropped ? Of course, we can have the attribute number and column name both to avoid conflict.

Thanks,
Pavan 

--
Pavan Deolasee
http://www.linkedin.com/in/pavandeolasee

Re: [OT] why not keeping the original column name in catalog after a drop?

From
Andres Freund
Date:
Hi,

On 2013-11-13 08:52:27 +0100, Luca Ferrari wrote:
> when you drop a column on a table the pg_attribute is updated and the
> name of the column is changed with an almost fixed identifier that
> reports only the original column position:
> 
> /*
>  * Change the column name to something that isn't likely to conflict
>  */
>         snprintf(newattname, sizeof(newattname),
>                  "........pg.dropped.%d........", attnum);
>         namestrcpy(&(attStruct->attname), newattname);
> 
> I'm wondering what is the problem in placing the old/original name
> after the "pg.dropped" prefix. I know that the tuple in pg_attribute
> is temporary, but what are the possible conflicts the comment talks
> about?

The old name might not fit there, attribute names have a relatively low
maximum length (64 by default), so we cannot always fit the entire old
name there.

Also, think about:
CREATE TABLE foo(cola int);
ALTER TABLE foo DROP COLUMN cola;
ALTER TABLE foo ADD COLUMN cola;
ALTER TABLE foo DROP COLUMN cola; -- should not error out

I don't really see much need for anything better than the current
solution, why is the old name interesting?

Greetings,

Andres Freund

-- Andres Freund                       http://www.2ndQuadrant.com/PostgreSQL Development, 24x7 Support, Training &
Services



Re: [OT] why not keeping the original column name in catalog after a drop?

From
Luca Ferrari
Date:
On Wed, Nov 13, 2013 at 9:00 AM, Andres Freund <andres@2ndquadrant.com> wrote:

> The old name might not fit there, attribute names have a relatively low
> maximum length (64 by default), so we cannot always fit the entire old
> name there.


Thanks, I was guessing this.

>
> Also, think about:
> CREATE TABLE foo(cola int);
> ALTER TABLE foo DROP COLUMN cola;
> ALTER TABLE foo ADD COLUMN cola;
> ALTER TABLE foo DROP COLUMN cola; -- should not error out

Well, I was talking about appending the original column name, and
therefore the above should have been respectively  pg.dropped.1.cola.
and pg.dropped.2.cola. Of course the original name is not very much
interesting, I was just curios about the conflicts.

Luca