>(No cheating by looking at what I say afterwards< g >).

Didn't cheat, and guessed correctly < g >

The first and last tests were idenctical (for all practical purposes). Reason being at run-time they are identical, the VFP compiler has replaced all instance of SHORT_STRING with "AAAAA", so no performance gain or loss by using a constant.

>However, if you replace the defined constant with REPLICATE("A", 5), the first instance will drop to the slowest because it's being re-evaluated on each pass.

Right, the compiler has replaced all instances of SHORT_STRING with REPLICATE("A", 5). Resulting in a decrease in performance by using a constant.

In order to gain performance the compiler would have to evaluate REPLICATE("A", 5) prior to compling, which it doesn't do.

Just for S&G, here is your test using only constants. What suprised me is this gives different results for test 2 and 3 than just the straight code.

#DEFINE SHORT_STRING       "AAAAA"
#DEFINE FOR_BEGIN 	        FOR lni = 1 TO 100000
#DEFINE FOR_END 		NEXT
#DEFINE STOP_CLOCK 		lnstop = SECONDS()
#DEFINE DISP_CLOCK 		? lnstop - lnstart
#DEFINE SHORT_VAR 		lcshort = "AAAAA"
#DEFINE START_CLOCK 	        lnstart = SECONDS()
#DEFINE TEST_1 			lcfoo = SHORT_STRING
#DEFINE TEST_2 			lcfoo = lcshort
#DEFINE TEST_3 			lcfoo = "AAAAA"

SHORT_VAR
START_CLOCK
FOR_BEGIN
	TEST_1
FOR_END
STOP_CLOCK
DISP_CLOCK
FOR_BEGIN
	TEST_2
FOR_END
STOP_CLOCK
DISP_CLOCK
FOR_BEGIN
	TEST_3
FOR_END
STOP_CLOCK
DISP_CLOCK