|From:||Brad Hilton <bhilton(at)vpop(dot)net>|
|Subject:||Re: switching from mysql|
|Views:||Raw Message | Whole Thread | Download mbox|
On Thu, 2001-11-15 at 22:16, Tom Lane wrote:
> Brad Hilton <bhilton(at)vpop(dot)net> writes:
> > I issue
> > select * from items GROUP BY id
> > I get:
> > "Attribute items.name must be GROUPed or used in an aggregate function"
> > It appears in MySQL if you group on a unique key, then you aren't
> > required to group on the rest of the fields you select from the table.
> If you know it's a unique key, why are you bothering to GROUP?
> If it's not a unique key, then there's no unique value for the other
> columns. Either way, I can see no sense to this query. The SQL92
> spec doesn't see any sense in it either.
The example I gave was a simple one to demonstrate my problem. Let me
give a better example. Let's say I have 3 tables: items, stores, and
store_items. 'store_items' maps items to stores. If I want to select
all items that are in at least one store, I currently do:
select items.* from items, store_items where items.id =
store_items.item_id group by items.id
Is there a better way to do this query?
> I believe MySQL thinks that this query means
> ... FROM (a CROSS JOIN b) LEFT JOIN c ON condition
> which unfortunately for them is not what the SQL spec says the
> syntax means. The spec's interpretation is
> ... FROM a CROSS JOIN (b LEFT JOIN c ON condition)
Thank you for your help!
|Next Message||Helge Bahmann||2001-11-16 17:26:10||Re: switching from mysql|
|Previous Message||Jason Earl||2001-11-16 16:44:13||Re: Perl and pgsql...|