>>>Hi Naomi,
>>>
>>>I believe the first is marginally faster. I could be wrong, but I think I remember that there's a lot more processing going on with an INSERT due to VARCHARMAPPING and other late changes.
>>>
>>>Shame I don't have source code access anymore :-)
>>
>>You mean scatter memvar is faster? Both tables have almost identical structures... Too bad I switched, now I have to switch back <g>
>>
>>My tests seems to confirm it, because I believe after my change the speed decreased. Though it's really hard to tell, because in addition I'm getting Invalid Seek Offset on about each third attempt, which really drives me nuts...
>
>Naomi - if you really want to figure out why your code is so slow, run it with Coverage Logging on and see where the bottlenecks. Much more efficient than this long thread.
>
>Tamar

That is, if I can get passed INVALID SEEK offset problem.

I just got it again on my first attempt to run the process on the new fresh set of data just re-indexed.