• Matt Miller (6/16/2008)


    Having done some amount of testing on this - while this scenario is certainly convenient in some scenarios, it has a tendency to perform on par with CSQ's, so it will suffer rather severely on large datasets. It unfortunately just seems to be a cleaner way to write a Correlated sub-query, but with the same kinds of perf challenges the CSQ's had (it forces row-by-row evaluation, the logical reads tend to get out of control, etc...)

    Can someone comment on the veracity of Matt's claims?

    Or perhaps Matt, do you have evidence (tests we can run, etc.)?

    Thanks, 😀

    Paul DB