On Thu, Jul 6, 2017 at 2:49 PM, Simon Slavin <slav...@bigfraud.org> wrote:
> On 6 Jul 2017, at 12:32pm, Dominique Devienne <ddevie...@gmail.com> wrote: > > Actually not that much apparently. > > No Simon, I didn't have an index on code_type. > > In fact keys is itself a view, and cote_type a case expression. > > The index you created is not helpful. It is not useful for what you’re > doing. > You what evidence do you get that from? The new plan uses the index though, as I previously indicated, so I'm surprised you'd say that. It's a group-by query, so despite using the index, all rowids for the only 4 different "index entries" must still be counted, and that's still definitely longer to do that than full scanning the table once. > I suggest you remove the VIEW layer from your layered query, then > reconsider what indexes might be helpful. We can advise you on indexes if > you post the resulting query. I don't see how that would make the query any faster. The plan already shows it scan the "real" table, not the intermediate view (w/o the index that is). --DD _______________________________________________ sqlite-users mailing list sqlite-users@mailinglists.sqlite.org http://mailinglists.sqlite.org/cgi-bin/mailman/listinfo/sqlite-users
