Quote:
Please don't take this the wrong way but that code is not template metaprogramming it's just ad hoc use of templates & macros. In general metaprogramming should be about reducing repetitive boilerplate code not increasing it and making code more declarative. The code is not flexible it looks very brittle to me.
Lastly no matter what solution you come up using purely templates for vertex declarators it will never be flexible enough, for instance you can't have custom vertex formats in a text file that can be read into the program it must be all in header-style files. If you want/need such flexiablity/extensibility you must scarifice static typing & some type safety.
I guess you're right about it not being true meta-programming, but it works and is more than flexible enough for my purposes. Specifying the internals of a vertex format through text files might be a bit too ambitious for me at the moment, and if I have to add new fields to the vertex data I'll just have to change the header then re-compile my libs and tools I guess, this isn't likely to happen very often though so it should be ok. I can't foresee any other problems and I'm pretty happy with what I've got (no doubt a couple of months down the line I'll discover some inherent fault and change my mind but nevermind) but of course suggestions and constructive criticism are always welcome.
I would like to find a more elegant solution than having to write up template classes for each vertex data field (see below) but short of writing a macro I can't really think of a better way to do it.
template <> struct VertexDataSize<VERTEX_FORMAT_POS>
{
static const unsigned ms_size = sizeof(float) * 3;
};
