RE: Difference between Select and Omit in LF

["Reeve" <news@xxxxxxxxxx>]

Access path sharing is labeled "implicit access path sharing" and it's no
for both, which means each access path is unique to each file (the system
owns access paths).

The index size is quite different, and I don't know enough about DB2 to know
if an EVI is at work (I don't think you can create an EVI with DDS, however)
or the binary radix tree works differently for selects and omits.

There are two more possibilities: your testing procedure is flawed (signing
off and signing on doesn't necessarily purge data from memory); this means
you'd have to run the test one way, re-impl, and run it the other way,
re-impl, etc., several times, probably, to rule out any unknown/external
influences on the testing oricedure.  

The other possibility is the nature of your data: if you have a lot of
"A"'s, "omit" might work faster.  I don't know if the "omit" logic is in the
access path itself or if the test's performed by DB2 when you retrieve
records.

-reeve 
 
Regards,
Reeve Fritchman
Ayers Rock Software LLC
4915 S. West Shore Boulevard
Tampa, Florida 33611-3329
(813) 831-8574 (voice)
(813) 832-6391 (fax)