Alexander Gattin <xr...@umc.com.ua> writes: > On Fri, Nov 19, 2010 at 12:32:19PM +0100, Alain > Ketterlin wrote: >> Alexander Gattin <xr...@yandex.ru> writes: >> > It's better to select count(1) instead of >> > count(*). The latter may skip rows consisting >> > entirely of NULLs IIRC. >> >> Wrong: count(anyname) ignores NULL, whereas count(*) does not. > > I'm using count(1), which is a count over constant > non-NULL expression (1). It doesn't ignore NULLs > or duplicate rows, as my attempts with Oracle 10g, > 8g and sqlite3 show.
I'm not saying anything about count(1) (which works like count(*) as you explain, even though I see no point in using 1, since *'s hehavior is well-defined). What was wrong in your first message is that count(*) ignores NULL (granted, you've corrected your mistake in a later message, which I didn't see before posting). -- Alain. -- http://mail.python.org/mailman/listinfo/python-list
