A couple of us have been having an off-line conversation about the
possibilty of getting incorrect results as a result of using
query-flattening. Ingres does query flattening by default, but it can
be turned off using SET NOFLATTEN. The consequence of turning off
flattening will usually be vastly slower joins (2, 3, maybe 5 times
slower, or worse)--although there are certainly cases where
subquery joins will be faster.

I have to admit I've never knowingly encountered a case where flattening
gives incorrect results, but I've certainly seen the claim made more
that once. So now I am curious. Can flattening give incorrect results
in recognizable situations, or is this one of those superstitions that
has grown up around some ancient and long-since fixed bug(s)?

--
Roy

UK Ingres User Association Conference 2010 will be on Tuesday June 8 2010
Go to http://www.iua.org.uk/join to get on the mailing list.