Thread: Subselects lack functionality

Kuba Ober (winnie@hoth.amu.edu.pl) reports a bug with a severity of 3
The lower the number the more severe it is.

Short Description
Subselects lack functionality

Long Description
1.
Subselects don't allow one to use tuple set operators like UNION, INTERSECT, EXCEPT. It forces one to select the
resultsinto a temp table. 
2.
It would be gr8 if single column subselects would allow themselves to be treated the same way as lists do as right-side
argumentsof IN operator. 
This would allow e.g. much more powerful DELETE statement, like
DELETE FROM table WHERE id IN (SELECT ...)
3.
Subselects cannot be used as arguments to aggregation functions - I presume this is very weird and not expected at all,
butI had such an idea. 

I don't know if it is expected behaviour (standards?), this should be explained in the docs (or am I searching in the
wrongplace?) 


Sample Code
select (select id from table1 except select table1.id where expression) as id;

(assuming that subselect would result just one tuple)

No file was uploaded with this report

pgsql-bugs@postgresql.org wrote:
> 2.
> It would be gr8 if single column subselects would allow themselves to be treated the same way as lists do as
right-sidearguments of IN operator. 
> This would allow e.g. much more powerful DELETE statement, like
> DELETE FROM table WHERE id IN (SELECT ...)

This, AFAIK, works.  I am using just that construct in UPDATE -- I would
assume DELETE would work...

Lessee...This works, as I just tested it on a client's database (inside
a transaction, with a rollback at the end, of course).  You say
something like

DELETE from users WHERE uid IN (SELECT uid FROM users WHERE username ~*
'testing')

to delete (of course, I know that that is a contrived example, but a
working one).

More complex is

DELETE from personal_data WHERE uid IN (SELECT uid FROM users WHERE
username ~* 'testing')

-- which works just fine on a client's database (again, inside a
transaction so I could easily roll back the delete :-)).

--
Lamar Owen
WGCR Internet Radio
1 Peter 4:11

pgsql-bugs@postgresql.org writes:
> Subselects don't allow one to use tuple set operators like UNION,
> INTERSECT, EXCEPT.

Yeah, we know.  The existing implementation of UNION etc. is a horrid
kluge that only works at the top level of a SELECT (and not even very
well there).  Fixing this will require a redesign of querytrees, which
is currently planned for 7.2.

> It would be gr8 if single column subselects would allow themselves to
> be treated the same way as lists do as right-side arguments of IN
> operator.

As Lamar pointed out, this works now.

> Subselects cannot be used as arguments to aggregation functions

You have that backwards.  You don't do the select as an argument to
the aggregate, you use it as a context for the aggregate, eg

     ... (SELECT max(foo) FROM bar WHERE baz) ...

            regards, tom lane