Thread: BUG #2203: Group by, suggested "feature"

The following bug has been logged online:

Bug reference:      2203
Logged by:          Robert Hostetter
Email address:      robert@newcovsoft.com
PostgreSQL version: 8.1
Operating system:   CentOS 4.1
Description:        Group by, suggested "feature"
Details:

I'm sure this will be considered a missing feature and not a bug, I know I
would.  We are converting from MySQL to PostgreSQL, which is a huge project.
 But one minor irritation we've noticed is the group by handling in
PostgreSQL.  If you group by a primary key for a table, you still have to
include all the other columns you SELECT from that row, even though its
guaranteed uniuqe because its a primary key.  I understand requiring the
other columns when you aren't guaranteed uniqueness, but in cases of a group
by on a unique column you shouldn't need to include the rows, its
repetitive, and makes porting harder for no reason.  Also on a similar note,
you can't group by any column you don't select.

Robert Hostetter wrote:
> If you group by a primary key for a table,
> you still have to include all the other columns you SELECT from that
> row, even though its guaranteed uniuqe because its a primary key.

Yes, this feature request is known as supporting functional
dependencies.  We're aware of it.

> pAlso on a similar note, you can't group by any
> column you don't select.

Sure you can.

--
Peter Eisentraut
http://developer.postgresql.org/~petere/