Mark,

It's kind of hard to read exactly what you are trying to do in that code. You basically have an O(n²) algorithm there. And (1.5x10^6)² is quite a bit bigger than (5x10^5)². And it may be even worse then O(n²) depending on how the other two fields end up inside your looping structure.

The GO TOP is not rushmore optimizeable you should use a LOCATE instead. Changing ORDER etc might be better handled by using the same table under alternate aliases. There may even be a couple of SQL statements that can handle the operation.

Can you describe in words what your loop is doing? Maybe show a couple of records before and after processing?

>We use a SEEK, followed by a DO WHILE loop. Here's a psuedo-code description:
>

>SET ORDER TO nGroup
>
>DO WHILE nGroup = 0
> liRecno = RECNO()
> lcField1 = cField1
> SET ORDER TO cField1
> SEEK lcField1
> DO WHILE cField1 = lcField1
>  IF (RECNO() <> liRecno) .AND. (nGroup > 0)
>   *- This is a matching record, other than the one we're using..
>   lnGroup = nGroup   && Get the group ID from this one...
>   GO liRecno
>   REPLACE nGroup WITH lnGroup
>   EXIT
>  ENDIF
> ENDDO
>
> IF nGroup = 0
>  *- No match, just create a new group ID here...
> ENDIF
>
> SET ORDER TO nGroup
> GO TOP   && Go to the next un-matched record...
>ENDDO

>
>It basically assigns the current record to a record that matches on one field. There are actually 3 fields used to match by, I left them out for clarity.