From: René J. <rvj...@xs...> - 2008-02-04 22:12:42
|
I think the cleanest way is the best - cleanest in the sense that there need to be no unnecessary exceptions to rules. I did not know about .STATIC_REQUIRES and even less about .PUBLIC_ROUTINES. I will probably never need them. If oodialog depends on them it must be doing something wrong - I can say this easily because I do not use it. Still, when the usage of these constructs is limited to oodialog, I would vote to adapt it into not needing them. The notion of "super public" does not seem to be very useful, and not having it will do away with potential name clashes. I think that as long as the removal of these coincides with a level of oodialog that runs with it, there would be no problem. The :REQUIRES cache seems to be very sensible both from a performance and also from an class identity point of view. I would want to make one remark about a "super public" construct - if it brings back something that some of the OS/2 users did expect from ooRexx, it might not be that useless - but why then not implement it like it was in those days. anyway, my 2 eurocents. best regards, Rene. On 4 feb 2008, at 19:28, Rick McGuire wrote: > Shoot, one thing I forgot to mention in the first posting. If we go > the route of a "super public" scope for methods and routines, then > we need to define what happens if there is a name collision. I > think I'd favor issuing an error if there is already an entity with > a given name defined at the "super public" scope. > > Rick > > On Feb 4, 2008 1:25 PM, Rick McGuire <obj...@gm...> wrote: > Currently, I'm redoing all of the code that manages ::REQUIRES > processing so that multiple programs/packages that have a ::REQUIRES > for the same file will use a shared reference to the package rather |