>Just one small thing, Scott...
>
>While I do think your test using VFP9 is valid (all done on same), there is the prospect that debugging code and/or testing assists are impacting the actual numbers.
>
>In particular I'd suspect that the slight penalty with alias prefixing might disappear in the final version.
>
>cheers
Hi Jim.
No, this is intrinsic.
The access more express than VFP happens on the first field of the active workarea withou the table prefix
( The prefixed one of table demands a job in more from part of the VM )
But m. issue it is not a speed issue:
m. without a active workarea it is slower
( for the same reason of the table prefix ):
CLEAR
CREATE CURSOR ZZ (AA I,A2 I,A3 I)
LOCAL Z
t1 = SECONDS()
FOR k=1 TO 10000000
=z
NEXT
? SECONDS()-T1
t1 = SECONDS()
FOR k=1 TO 10000000
=m.z
NEXT
? SECONDS()-T1
SELECT 0
t1 = SECONDS()
FOR k=1 TO 10000000
=z
NEXT
? SECONDS()-T1
t1 = SECONDS()
FOR k=1 TO 10000000
=m.z
NEXT
? SECONDS()-T1
This is a correctness aspect.
If you write a program that use a memvar,
with m. it is correct without m. it is mistaken,
it can works, but it remains mistaken.
it is like resolving a mathematics problem, with a mistaken method and obtaining exact results.
Fabio