Thread: Inheritance: Performance & Indexes

Just curious:
 create table a (x int4 primary key); create table a1 () inherits (a);
 [...]
 delete from a where x = 1000;

I take it that this is equivalent to:
 delete from only a where x = 1000; delete from only a1 where x = 1000;

In particular, for performance reasons I should try to limit the
delete to a1 wherever possible?

Also, it looks like a1 doesn't inherit the primary key index.  Is that
intentional or a bug?

Finally, is it possible to specify that x values be unique across
both a *and* a1, without triggers?

-- 
Itai Zukerman  <http://www.math-hat.com/~zukerman/>

On 18 Jun 2001, Itai Zukerman wrote:

> Just curious:
> 
>   create table a (x int4 primary key);
>   create table a1 () inherits (a);
> 
>   [...]
> 
>   delete from a where x = 1000;
> 
> I take it that this is equivalent to:
> 
>   delete from only a where x = 1000;
>   delete from only a1 where x = 1000;
> 
Yes.

> In particular, for performance reasons I should try to limit the
> delete to a1 wherever possible?
Well, you should be aware what'll actually happen when you do
updates/deletes on parent table. If you mean to delete only from parent,
use 'only'.

> Also, it looks like a1 doesn't inherit the primary key index.  Is that
> intentional or a bug?
Its a problem of current implementation (aka bug)
> 
> Finally, is it possible to specify that x values be unique across
> both a *and* a1, without triggers?
Not currently, but there are talks about revamping of inheritance system,
see archives

-alex