thanks for the reply and patience simon.
i understand the dynamics of compilation and i don't think my question is regarding that.
i think my primary question is that if tab2 does not have a histogram, then why cannot tab1's range/frequency weights and range density serve as a basis for evaluating tab2 ?
so if the query is looking for the the key value 1 (from your sample data) from tab2, why not simply use the frequency weight from tab1?
or for values 1 and 3, as per your sample data it could still use the frequency weights of the two. or if one of the values sought is part of a range cell, then could it not use the range density?
instead of using 380 rows from total density, why not use 700 rows from the frequency weights?