Al,

>I now have a theory:
>
>ALINES() is fast, and it's a relatively new function in VFP.
>AT(), SUBSTR() etc. are slow, these are old functions, around since FoxBASE+ days.

Not true. at() is just a thin wrapper around a strstr() family function.

The performance of code degrades because of using the 3rd parameter to the at() ala:

for i = 1 to 10
   j = at( x, y, i )
endfor

which is an O(n²) algorithm. It makes the at() function internally run several times, the next iteration of the loop forces it into redoing the all same string seeks that it already found in the iteration before.

This sort of code construct:

for i = 1 to 10
   j = at( x, y )
   y = substr( y, j+1)
endfor

reduces the seek time by removing the repeated seeks, at the expense of moving lots of string memory around and some memory management in deallocating the used memory of y as it shortens. This also is O(n²) performance.

>I wonder if the "old" functions are still mired in the 16-bit world and are losing performance with 16/32 bit conversions ("thunking") and/or hitting limitations at 64K chunks? The moral - for fast VFP performance, use new functions??

No, at(x,y) in VFP is linear in performance O(n) where n length of string for n anywhere from a few thousand characters upto 16mb.

The alines() sort of code is O(n) because you only need to make 2 linear passes through the string, once to put it into the array and once to iterate the array.