Re: [sqlite] Ticket 3128: DISTINCT -> GROUP BY can lead to performance degradation

2008-05-20 Thread Ralf Junker
>> This should NOT happen as SQLite usually rejects duplicate rowids with a >constraint error. > >For this table, the badly-named column "RowID" is actually a text field >that's allowed to be null, not a primary key. > >-- Describe NDXPARAMVALUES_LOCALIZEDSTRING >CREATE TABLE

Re: [sqlite] Ticket 3128: DISTINCT -> GROUP BY can lead to performance degradation

2008-05-20 Thread Bradley Town
On Tue, May 20, 2008 at 6:57 AM, Ralf Junker <[EMAIL PROTECTED]> wrote: > This should NOT happen as SQLite usually rejects duplicate rowids with a constraint error. For this table, the badly-named column "RowID" is actually a text field that's allowed to be null, not a primary key. -- Describe

Re: [sqlite] Ticket 3128: DISTINCT -> GROUP BY can lead to performance degradation

2008-05-20 Thread Ralf Junker
Correcting myself: This should NOT happen as SQLite usually rejects duplicate rowids with a constraint error. >This should happen as SQLite usually rejects duplicate rowids with a >constraint error. ___ sqlite-users mailing list

Re: [sqlite] Ticket 3128: DISTINCT -> GROUP BY can lead to performance degradation

2008-05-20 Thread Ralf Junker
I looked at the database attached to the ticked and noticed that the table contains NULL rowids as well duplicate rowids. This should happen as SQLite usually rejects duplicate rowids with a constraint error. When I run this query: select rowid, count() from ndxparamvalues_localizedstring

[sqlite] Ticket 3128: DISTINCT -> GROUP BY can lead to performance degradation

2008-05-20 Thread Bradley A. Town
I've only found one reference to slower queries with the DISTINCT/GROUP BY optimization that went in back in November for 3.5.3 and later. I would have expected more given the number of our queries causing problems. The problem as I wrote in ticket 3128