"Jose" <jmalv04@gmail.com> wrote
I think it would also be useful to have RTL's basic distance example shown in RML and then benchmark that with 10k+ points
Well, I guess it would be quite difficult to come up with exact locations of 10k+ cities (unless somebody already have this data). That's why my, I agree, quite silly example, was initially designed to mimique this problem (http://lists.boost.org/Archives/boost/2005/10/94890.php). All that needs to be done is to complicate the condition, so that such things as merge or nested loops join are not possible. I think something like "abs(c1 - c2) < 1" should do fine. I already suggested this before. This example has to show the strongest point of RTL. It's quite time-consumming to go through every possible pair evaluating the condition (selection over cross-product). With RTL it's possible to do it only once, create an index on top of this expression, and then all the subsequent queries would execute orders of magnitude faster. The updates to the table would not require to fully rebuild the index -- it would be updated incrementally. Regards, Arkadiy