From: MIGUEL A. B. L. <mb...@fe...> - 2001-11-20 20:11:19
|
Mensaje citado por: Ulrich Eckhardt <doo...@kn...>: > > > > What do you think is the best way to represent Name? > > > A string should be ok, another overload should take some kind of small > binary=20 > identifier (enum or so). The string would be when loading the Object > from a=20 > file, the binary identifier when it's a perception coming over the > network. No, I think that Brian is talking about how to structure the objects. Not about the std::string. So we have now active/animal/reptile/.... We need a new almost definitive structure. That structure SHOULD show how the RP interacts with the Objects. =20 > > Also, if you have any questions, please ask them! Any > > comments, I beg you to share them! > > > I wonder if we need such high complexity. As I said above, I see only > two overloads (binary and string) for the input that creates an Object.= =20 No need for this, read above. I agree, we don't need such a complex system at all. We can restrict by n= ow to=20 a much simpler version. > About the structure of the Oject itself, I also had the same thought=20 > (controller+attributes) but I see a big problem in that: either you > provide a=20 > generalized interface to the attributes (std::map<string, string> ) or > the=20 > controller must be matched with the actual dataset that it operates > with.=20 > The problem is that the generalized way probably takes lots of time and= =20 > memory ... at least something like map<enum, GenericAttrib*> should be > used instead. The <enum, GenericAttrib> approach is not as good as the other because it= break=20 the original idea of having a flexible, easy to improve system. I think that we should/could keep using string,string until it start to b= e a=20 big problem and them just replace the strings with pointers where the rea= l=20 string is located. > However, before we try to build that, I would like to see the need for > it. In=20 > the first run, I would rather use only a few predefined though limited=20 > Object-types and see how the game goes, and if we really need more=20 > flexibility, we can add that later. I think that we need that system, actual one is really limited and not us= eful=20 at all. The one the Brian proposed is great, although can overload the CPU, but w= ell,=20 every 18 months CPU double their power :-) > As a personal note, I must say that I once wrote The Perfect Flexible=20 > System(tm), it got me some interesting insights but was never really in > use=20 > or useful, just cost my employer a few thousands .... Hehehe Here you have a bit of UNICODE: x=B71000 =80, x>0 Regards, Miguel |