There is a difference as to whether a line executes in 0.0009 or 0.0006. For example, my test processed 10 records, but in the real world a user will be importing data consisting of an average of 300,000 to 500,000 records. The test took 5-6 seconds. Do the math for 500,000 records...its too long.

In my test, the profiler showes how many times each line was run (as you know). Those lines that executed just once and took 0.000 to do so, I am not concerned about. Some lines executed 10 times at an average of 0.000. Well, when the whole process took 5 seconds, I must look at what a real user is going to be faced with. Some lines for the 10 record test executed a total of 800 times (it is parsing through the data) at an average of 0.000, again do the math for 500,000 records using "real" time. These lines of code do take "real" time to execute.

You're right in that if a line of code executes once, there is no difference between it taking 0.0006 or 0.0009. But, if this line executes 40,000,000 times, then I need to be concerned about the time that it takes relative to the execution time of other lines or routines. For example, in using my own debug code, I show that for the 3 main functions of this routine, the first took 20 percent of the time, the second took 60 percent and the third took 20 percent of the time. But, using the profiler, almost all lines show that they took 0.000 time to execute. I am having a hard time going in and changing a line of code to try to optimize it when the orginal line took 0.000 and the change also took 0.000 to run (in profiler time).

My suggestion for the FoxPro team at Microsoft is to replace their development machines with lets say even the bottom line 300's that are almost extinct on the shelves today. And to run the profiler against some code. This way they too can hopefully see that reporting that almost all lines of code in a routine take 0.000 to execute makes the profiler tool useless.

Thanks
Dean