GregTomkins
Active Member
IMO, approaches of trying to find problems based on XREF or looking at source can be non-productive because there are lots of cases where CAN-DO or USE-INDEX causes whole index reads ... or some other heinous thing ... which don't actually matter, because the tables involved never have more than a handful of records in them. Or, the problem is in code that hardly ever runs, or runs in a non-interactive way where the difference between a 1-second operation and a 1-minute operation is often irrelevant.
Of course, we should code properly regardless, but in the real world, with 1000's of .P's and 10's of programmers over 10's of years, it's more useful (IMHO) to concentrate on statistics and real-world measurements, eg. logging of the performance characteristics of your application using VST's and such.
Eg. in our case, and I expect this is typical, about 5% of our code handles 95% of our users' activity. So, we concentrate on identifying that 5% and fixing them. At least, that's the theory
Of course, we should code properly regardless, but in the real world, with 1000's of .P's and 10's of programmers over 10's of years, it's more useful (IMHO) to concentrate on statistics and real-world measurements, eg. logging of the performance characteristics of your application using VST's and such.
Eg. in our case, and I expect this is typical, about 5% of our code handles 95% of our users' activity. So, we concentrate on identifying that 5% and fixing them. At least, that's the theory