On Fri, 2005-09-09 at 16:54 -0400, Ned Batchelder wrote:
> I can't follow this thread (NULLs make my head hurt), but it looks like Dr.
> Hipp has already taken action. Yesterday he made two changes to the source:
>
The changes yesterday brought SQLite's SUM function into compliance
with the (broken) SQL standard. My plan is to leave things that
way. I'll likely introduce some non-standard functions (perhaps
named "_sum" or "sqlite_sum") that work in a useful way - which is
to say the way that Martin Engelschalk and several others have
suggested:
sum(<empty-set>) == 0
sum(<anything-containing-null>) == NULL
sum(<numerics-without-nulls>) == the correct sum
--
D. Richard Hipp <[EMAIL PROTECTED]>
