> We I think option 1 - the dynamic bindings would first need to be done

Considering that a lot of languages are not dynamic, would it not be best to 
settle on one OSG-generic static way of doing this first?

In my work with SWIG I have found that it _could_ be very easy to generate 
full bindings for any language by parsing the include files.
If we modify the include files to be more parse-friendly (or to comply more 
fully with Swig or Boost) we would not have to maintain a secondary API via 
osgIntrospection and we could (in theory) generate bindings for Python, 
Ruby, C#, etc using one command on the prompt.

If one could generate bindings directly from the API headers that would be 
the least maintenance, it would not reduce speed and it would not limit the 
binding language to dynamic ones.
It also seems like the least work to get working intially. Using the time to 
dually maintain the API via osgIntrospection could be used to slightly 
modify the headers.
I've completed around 20% of the base API using Swig without having to 
modify any header file. So far, only nested structs and some constants have 
been a problem, but I easily mended it by writing ifdef sections, sometimes 
just adding some new typedef's at the end.

This is just a suggestion, whatever the community at large feel is the right 
way to go is my way too. Better that we all focus on one effort.

Best regards, Peter
_______________________________________________
osg-users mailing list
[email protected]
http://openscenegraph.net/mailman/listinfo/osg-users
http://www.openscenegraph.org/

Reply via email to