Well, then you're handcuffing it when an index would be better but is in nowhere near rowid order.
-----Original Message----- From: sqlite-users [mailto:[email protected]] On Behalf Of Bart Smissaert Sent: Thursday, January 12, 2017 3:53 PM To: General Discussion of SQLite Database Subject: Re: [sqlite] Why this query plan? How about a pragma_sort_asc_on_rowid_or_primary_integer_key_for_unordered_selects ? RBS On 12 Jan 2017 20:35, "Richard Hipp" <[email protected]> wrote: > On 1/12/17, David Raymond <[email protected]> wrote: > > > > In the same vane I assume DRH's random ordering would be only random by > page > > of results. If you have 100+ million records in a table then keeping > track > > of which ones you've randomly picked so far would cripple systems with > the > > tracking requirements and with the slowdown of skipping all over the > file. > > Shuffling the order is one thing, killing performance is another. > > > > The idea is that as each new database connection is opened, the > reversed_unordered_selects pragma > (https://www.sqlite.org/pragma.html#pragma_reverse_unordered_selects) > would be enabled or disabled at random. That means that results might > be backwards from one run to the next, but within the same run they > would always be the same. That is not really "random" but I think it > should be sufficient to find instances of omitted ORDER BY clauses, at > least for the case where the developers test their application more > than once or twice. > -- > D. Richard Hipp > [email protected] > _______________________________________________ > sqlite-users mailing list > [email protected] > http://mailinglists.sqlite.org/cgi-bin/mailman/listinfo/sqlite-users > _______________________________________________ sqlite-users mailing list [email protected] http://mailinglists.sqlite.org/cgi-bin/mailman/listinfo/sqlite-users _______________________________________________ sqlite-users mailing list [email protected] http://mailinglists.sqlite.org/cgi-bin/mailman/listinfo/sqlite-users
