Nicolas Neuss wrote on Tue, Jan 14, 2003 at 01:38:12PM +0100: > > Martin Cracauer <[EMAIL PROTECTED]> writes: > > > I don't use CLOS and as mentioned do not recommend it when teaching a > > Java or C++ programmer Lisp. You must be sure they get that this is a > > very high-level tool before you allow them to use it. > > On the other hand, if those people do not see CLOS I assume that they will > stay with their languages. For my application, I have used CLOS and, of > course, have run into the performance problem you mentioned. But > nevertheless, I do not regret it. Instead, I'm thinking about CLOSifying > some remaining low-level structs. Perhaps, this is possible with Gerd's > PCL changes...
In my opinion - if the person you are teaching is so infected by the OO virus that he insists on using classes, inheritance and virtual functions (whatever you name it) for every problem he sees, then there is zero hope whatsoever that he can benefit from learning Lisp. The only thing you will do by trying this is create another person claiming "Lisp is slow". A full dylan-style OO system with sealing would be another matter, but CLOS is about a factor of 100 too overkilly for what these people throw at it. Last time I looked you could easily get a 50 to 100 times (not %) performance hit from using CLOS for small objects, compared to tight code. Interestingly enough, all the OO idiots (there are clever OO people, here I just mean the subportion who have the hammer/nail view) are now hoocked at Java or similar languages. But most of the hard-working C++ guys now use generic programming, not OO. I have a very hard time bashing our C++ guys because they totally dismissed the big inheritance-oriented OO style and do their actual work with templates, around minimal encapsulation. Martin
