Re: [boost] how to do from_python in v2?

From: David Abrahams (dave_at_[hidden])
Date: 2002-12-18 12:43:10

"HellcatV HellcatV" <hellcatv_at_[hidden]> writes:

> I've been struggling to port boost v1 to boost v2 and I haven't
> located a nice sort of documentation for it--

What's wrong with

http://cvs.sourceforge.net/cgi-bin/viewcvs.cgi/*checkout*/boost/boost/libs/python/doc/index.html?

??
We worked very hard on it!

> luckily i have almost finished figuring it out...and just need to
> figure out how to register from_python
>
> I have tried and failed to contact the mailing list 

What did you try to do?
Instructions are at http://www.python.org/sigs/c++-sig/

> so I'm hoping you can help me, Dave :-)
> or point me to where I cna find infos

See above.

> so I'm trying to get boost v2 to work with my new project at LBL
>
> my test code appears to be 100% working ( to_python converted to that
> big macro,  

Huh? That big macro?

> sloppy casts of "member" functions that included something
> that casted to the true member function has been removed)

I'm still confused, but it sound like you're saying things have
improved.

> but I get a type error when trying to call a function with a tuple
> that really takes a Vector.

> struct Vector {
> float i; float j; float k;
> };
>
> This is because I have not figured out how to properly define a
> from_python operator in boost v2... I have scoured google and all the
> docs and have only encountered half-debates on what convention to use
> and no solid rules--and I cannot appear to make anything I have tried
> work.

http://cvs.sourceforge.net/cgi-bin/viewcvs.cgi/*checkout*/boost/boost/libs/python/doc/v2/faq.html#question2

Ralf also works at Berkeley Labs, so you have the expert on-site.

> can you point me to some documentation on which functioN I should make
> that takes in a PyObject * and return a Vector  so that I can call
> functions that take a Vector as an argument?
>
> my old from_python method is as follows:
> inline Vector from_python(PyObject *p,boost::python::type<Vector>) {
>          Vector vec(0,0,0);
>          PyArg_ParseTuple(p,"fff",&vec.i,&vec.j,&vec.k);
>          return vec;
> }
> error checking is not required :-)
>
> if anyone can help me get this function to work in v2 I believe the
> rest of my code will magically work (because
>
> the only functions currently failing involve taking IN
> vectors...returning them is accomplished with
> BOOST_PYTHON_TO_PYTHON_BY_VALUE(QVector,
> boost::python::to_python_value<boost::python::tuple> ()(bo\
> ost::python::make_tuple((double)x.i,(double)x.j,(double)x.k)));

Ahh, that macro is an implementation detail; I'm not sure that you
should be using it.  I probably should be #undef ing it at the end of
the header.

    static void* QVector_convertible(PyObject* p)
    {
        return PyTuple_Check(p);
    }
    
    static void QVector_construct(PyObject* source, rvalue_from_python_stage1_data* data)
    {
        void* const storage = ((converter::rvalue_from_python_storage<QVector>*)data)->storage.bytes;
        new (storage) QVector();
        // Fill in QVector values from source tuple here
        // details left to reader.
        data->convertible = storage;
    }

    ...
    boost::python::converter::registry::insert(QVector_convertible, QVector_construct, type_id<QVector>());

> is the above calling syntax going to be portable to newer versions of
> boost? or am I going to have to scour through millions of header files
> for version 1.30?

Scouring headers is never a good idea.  I recommend reading the docs,
then asking questions.

> anyhow thanks for a great library....

You're welcome!

> unfortunately only boost v1 compiles on my macintosh with gcc-3.1
> it appears there is a preprocessing error with gcc-3.1 :-( and I'm not
> sure what to do about it...
>
> it says: (postprocessed)
>
>   template < typename 0 = mpl::void_ , typename B1 = mpl::void_ ,
> typename B2 = mpl::void_ , typena\
> me B3 = mpl::void_ , typename B4 = mpl::void_ , typename B5 =
> mpl::void_ , typename B6 = mpl::void_\
> , typename B7 = mpl::void_ , typename B8 = mpl::void_ , typename B9 =
> mpl::void_ >
>   struct bases : detail::type_list< B0 , B1 , B2 , B3 , B4 , B5 , B6 ,
> B7 , B8 , B9 >::type
>   {};
>
> where I think it means B0 instead of just plain ole  0....

Ouch!  Yes, that's right.

> I think it's a gcc-3.1 b0rken sign but I don't know how to get around
> it :-/

GCC-3.2.1?

-- 
                       David Abrahams
   dave_at_[hidden] * http://www.boost-consulting.com
Boost support, enhancements, training, and commercial distribution