Search for disctint values using the ORM

On Wednesday 07 April 2010, you wrote:
> Public bug reported:
>
> Answers for Question #106692 indicate that for the moment we can't do a
> "SELECT DISTINCT" using the ORM. There are cases where this would be
> useful (like searching the order ids in order lines), so I propose we add
> a new parameter "distinct=False" to the search() method of th OSVs. When
> set to True, it would return distinct tuples only.
> It seems easy to add in osv using a SELECT DISTINCT statement.
> It may be a bit more tricky in osv_memory, and it will run slower of
> course, and none sane will use it there I suppose, so we should just throw
> a "not implemented" exception I guess. Lionel
>

Aren't the orm.search() and orm.read() results already distinct, by definition?
Those functions use the id column as a key, and thus never expected to return
duplicate results.

Or am I missing something?

A Dissabte, 10 de juliol de 2010, xrg va escriure:
> On Wednesday 07 April 2010, you wrote:
> > Public bug reported:
> >
> > Answers for Question #106692 indicate that for the moment we can't do a
> >
> > "SELECT DISTINCT" using the ORM. There are cases where this would be
> > useful (like searching the order ids in order lines), so I propose we
> > add a new parameter "distinct=False" to the search() method of th OSVs.
> > When set to True, it would return distinct tuples only.
> >
> > It seems easy to add in osv using a SELECT DISTINCT statement.
> > It may be a bit more tricky in osv_memory, and it will run slower of
> >
> > course, and none sane will use it there I suppose, so we should just
> > throw a "not implemented" exception I guess. Lionel
>
> Aren't the orm.search() and orm.read() results already distinct, by
> definition? Those functions use the id column as a key, and thus never
> expected to return duplicate results.
>
> Or am I missing something?

Well, the thing is that search() function doesn't receive a list of fields so
it's not possible to execute a SELECT DISTINCT in search function. IMHO read()
function should accept both a list of IDS and a domain. In the latter case
it'd be possible what the bug report asks for.

--
Albert Cervera i Areny
http://www.NaN-tic.com
OpenERP Partners
Mòbil: +34 669 40 40 18

http://twitter.com/albertnan
http://albert-nan.blogspot.com

xrg & Albert Cervera i Areny : search() returns ids so yes it's probably already distinct. The feature is missing in read(), not search() - I must have been distracted.

xrg: it's quite common for search to return several identical tuples when different ids are passed. One example is: search for the order id in sale order lines.

Maybe distinct should be the default but that requires a lot of testing, I wouldn't recommend it for v6.0.

Albert Cervera i Areny
> IMHO read() function should accept both a list of IDS and a domain.
Yes I think Tryton made it this way too, please file a distinct bug report.

Lionel.

Adding a way to select distinct values seems like an interesting improvement post-v6, along with search_read and the likes..

Thanks for the suggestion

Has this been implemented already?