>>>>>If user does something, your code reacts to it and assigns the value to this property. You have a choice of bindevent() to the things user may do, or having an _assign method for this property etc. There are ways to avoid having the code in .refresh() doing any checking.
>>>>
>>>>I think I understand what you are saying (although I have not used bindevent() yet). But I am not convinced that having a code in the Refresh() method (the way I did it so far) is going to be a problem in the long run.
>>>
>>>In the long run, it would be a problem only if it runs too long, too often :). Test on slow machines.
>>
>>I used to try to design everything to the lowest common denominator (e.g. slowest computer). And I often ended up having two much code and too much maintenance. And a few customers who still used slow machines didn't really appreciate what I was doing for them.
>>No more Mr. Nice Guy.
>
>Someone had the tagline "Premature optimization is the source of all evil (Donald Knuth)".
>
>>Now I design everything for minimum code maintenance and if someone is having a problem I suggest that they spend some money and buy faster machines. Or add memory. Then they appreciate what I am saying :).
>
>Been through that cycle too, and we're actually on the same side: I've found refresh issues hard to track, easy to forget the reasons (why does the code do this? which problem was this solving?) and thus a drag on maintenance. The simpler the refresh, the less work later.

I never said that I know all the answers and my approach is the best. I am trying it. And if it does not work, or creates some 'side effects', I will go back to see how to make it better. So far I created this code in only ONE method of my look-up control class that I use in many, many forms/windows of the application. But this class is not using 'my' code because I have a property BIND2PK set to .F. by default. And when I do want to bind this control to a PK (in some forms of the application) I need to set this property to .T. and set 6 other properties to some values (e.g. PK field name, Pk field alias, and so on). And then 'my' code in Refresh works. As far as maintenance, I cannot see how I make it simpler. One method and about 10 lines of code; that is all. Performance is another story, and I will "watch" how it works. The good thing about VFP is that 10 lines of code are executed so fast (even on a not-so fast machine) that one never notices.