[Swig-user] tcl export std vector

[lorubenet <lor...@go...>]

Hi swig-users!

I am using swig for quite some time now to export C++ to tcl and python. It is very good thanks!
I have never been able to solve the following problem though:
We want to export a method which returns an std::vector<double>. Swig works this out fine but returns a native List in tcl and a tuple in python. We want it to return a wrapped std::vector instead. Is there a way, like a flag or something, where I change that? 

I just found out today (which gave me hope) that in python this is possible by adding:
%{
#define SWIG_PYTHON_EXTRA_NATIVE_CONTAINERS
%}
in the i file. I tested that successfully. 

I cannot find something similar for tcl. Checking the cxx code generated by swig for python seems clear why this "#define" does the job. I also checked the cxx code generated for tcl but all seems hardcoded there. The code gets first a "std::vector<double> result" and then it copies it into a TclList:
    for (unsigned int i=0; i<(&result)->size(); i++) {
      Tcl_ListObjAppendElement(interp, (Tcl_GetObjResult(interp)), \
        Tcl_NewDoubleObj(((std::vector< double > &)result)[i]));
    }


I have been further playing with this (since the mailing list rejected my email at first). I did this change in the std_vector.i file in the tcl directory of the swig installation:
        %typemap(out) vector< T > {
        %#ifdef SWIG_TCL_EXTRA_NATIVE_CONTAINERS
            std::vector< T > * pRes = (std::vector< T > *)new std::vector< T >($1);
            Tcl_SetObjResult(interp, SWIG_NewInstanceObj( %as_voidptr(pRes), $&1_descriptor,0));
        %#else
            for (unsigned int i=0; i<$1.size(); i++) {
                Tcl_ListObjAppendElement(interp, $result, \
                                         CONVERT_TO((($1_type &)$1)[i]));
            }
        %#endif
        }


Adding then the ifdefs and the code within the SWIG_TCL_EXTRA_NATIVE_CONTAINERS. This seems to work, but I have no clue whether it makes sense to do it this way, nor whether it will fail in some other place... 

Thanks in advance,
Ruben