Mailman 3 August 2002 - Cplusplus-sig

Documentation
by Achim Domma Sept. 25, 2002

Sept. 25, 2002

Hi Dave, I will have some time to contribute to boost.python, but I don't think that I'm able to implement functionality or write documentation. I thought it might be helpfull to prepare at least the html code for the docs. Do you think it would be usefull, if I create files like the attached one where somebody else could fill out the missing parts. So you would at least not waste time with copy and paste and don't have to care about layout. Achim

3 4

convertibility and "Pythonicity"
by David Abrahams Sept. 12, 2002

Sept. 12, 2002

[Is "Pythonicity" the right word?] I'm interested in getting some qualitative feedback about something I'm doing in Boost.Python. The questions are, 1. How well does this behavior match up with what Python users have probably come to expect? 2. (related, I hope!) How close is it to the intended design of Python? When wrapping a C++ function that expects a float argument, I thought it would be bizarre if people couldn't pass a Python int. Well, Python ints have a lovely __float__ function which can be used to convert them to floats. Following that idea to its "logical" conclusion led me to where I am today: when matching a formal argument corresponding to one of the built-in Python types, first use the corresponding conversion slot. That could lead to some surprising behaviors: char index(const char* s, int n); // wrapped using Boost.Python >>> index('foobar', 2) # ok 'o' >>> index(3.14, 1.2) # Wierd (floats have __str__) '.' >>> index([1, 3, 5], 0.0) # Super wierd (everything has __str__) '[' So I went back and tried some "obvious" test in Python 2.2.1: >>> 'foobar'[3.0] Traceback (most recent call last): File "<stdin>", line 1, in ? TypeError: sequence index must be integer Well, I had expected this to work, so I'm beginning to re-think my "liberal conversion" policy. It seems like Python itself isn't using these slots to do "implicit conversion". But then: >>> 'foobar'[3L] 'b' [The int/long unification I've heard about hasn't happened yet, has it?] Furthermore: >>> range(3.3, 10.3) [3, 4, 5, 6, 7, 8, 9] But: >>> range('1', '5') Traceback (most recent call last): File "<stdin>", line 1, in ? TypeError: an integer is required Now I note that strings don't have __int__, so I guess the int type handles int('42') itself using special knowledge about strings. I suppose that's to keep strings from seeming to be numbers, since the nb_int slot fills in the number_methods. And: >>> class zero(object): ... def __int__(self): return 0 ... >>> range(zero(), 5) [0, 1, 2, 3, 4] So, is there any general practice, (even if it's not universal)? Do Python functions usually tend to coerce their arguments into the types they're expecting? I'm guessing the answer is no... ----------------------------------------------------------- David Abrahams * Boost Consulting dave(a)boost-consulting.com * http://www.boost-consulting.com

3 2

Boost.Pyton v2 & gcc 2.95.3
by Peter Schoen Sept. 8, 2002

Sept. 8, 2002

Dear y'all, is it possible to compile Boost.Python v2 using the gcc 2.95.3 compiler (I am using SuSE 8)? I tried to find the answer in the archive, but didn't succeed. When I try to compile the lib with: "bjam -sTOOLS=gcc test", I get many template related erros. If others were able to compile the lib using 2.95.3, I can also post the error messages. Many thanks & very best wishes Peter

4 8

lvalue converter (and other) questions
by Ralf W. Grosse-Kunstleve Aug. 30, 2002

Aug. 30, 2002

I have questions (for David), but first the Context: I have made a new cvs module scitbx that contains an "array_family" incl. Python bindings. At the C++ level there is the type shared<ElementType> which is just like std::vector, but reference counted or IOW with shallow copy semantics. Then there is the type versa<ElementType, AccessorType> to cover multi-dimensional arrays (or arrays with any other method for accessing the elements). versa inherits<> from shared<>. At the C++ level I have two general accessor types (among others): grid<N> // number of dimensions N known at compile-time flex_grid<> // number of dimensions determined at run-time To minimize the memory footprint for the Python bindings, only one array type is visible in Python: C++: versa<ElementType, flex_grid<> > Python: flex.int, flex.double, etc. shared<T> and versa<T, grid<N> > are automatically mapped (two-way) to versa<ElementType, flex_grid<> >/flex.xxx: only the accessor is changed, the actual data are not copied or modified. The shared<->flex mappings are achieved with the code in this file: http://cvs.sourceforge.net/cgi-bin/viewcvs.cgi/cctbx/scitbx/include/scitbx/… First question: Fragment from the file referenced above: static void* convertible(PyObject* obj_ptr) { using namespace boost::python; object obj(borrowed(obj_ptr)); extract<flex_type const&> flex_proxy(obj); if (!flex_proxy.check()) return 0; //... } Am I using extract<> in the intended way? Is there a more direct way to just check the type? Something similar to PyInt_Check(), e.g. if (!boost::python::class_check<flex_type>(obj_ptr)) return 0; Would using this make sense as an optimization: extract<> is only called after this test? Second question: If I understand correctly, struct shared_from_flex in the reference code implements an rvalue converter only. However, I am also interested in a similar lvalue converter. Here is an example of what I am doing at the moment (btw: the C++ language /really/ needs template typedefs): static void push_back(versa<ElementType, flex_grid<> >& a, ElementType const& x) { // first check that "a" is a 0-based, 1-dimensional, unpadded array // and then convert "a" to a shared<ElementType> base_array_type b = flex_as_base_array(a); b.push_back(x); // make a new (1-dimensional) accessor for "a". a.resize(flex_grid<>(b.size())); } Could I do it like this if I had an lvalue converter? static void push_back(shared<ElementType>& a, ElementType const& x) { a.push_back(x); } This would require that the lvalue converter can do the a.resize() after push_back() returns. Is this possible (and where can I find an example for creating a lvalue converter)? Another question, related to the previous one: Since push_back() is .def'ed as a member function of a wrapped versa<T,A> there is an additional concern: the first argument is a shared<T>& instead of a versa<T,A>& . Will this work/is it safe? Thanks, Ralf __________________________________________________ Do You Yahoo!? Yahoo! Finance - Get real-time stock quotes http://finance.yahoo.com

2 4

"bpl_utils" reorganized
by Ralf W. Grosse-Kunstleve Aug. 30, 2002

Aug. 30, 2002

For those who are using the "bpl_utils": The name has changed, and the code has been reorganized. It is now part of the "scitbx" module (under the old "cctbx" cvs project). The container conversions can be found in this directory: http://cvs.sourceforge.net/cgi-bin/viewcvs.cgi/cctbx/scitbx/include/scitbx/… File name: container_conversions.h The regression test is here: http://cvs.sourceforge.net/cgi-bin/viewcvs.cgi/cctbx/scitbx/array_family/bo… File names: regression_test_module.cpp regression_test.py Example usage: scitbx::boost_python::container_conversions::from_python_sequence< std::vector<double>, scitbx::boost_python::container_conversions::variable_capacity_policy>(); Fragment from regression_test.py to illustrate how it works from Python: assert rt.std_vector((8,9,10)) == 27 assert rt.std_vector([7,9,10]) == 26 assert rt.std_vector(iter(xrange(10,13))) == 33 assert rt.std_vector((8,9,10,11)) == 38 Tested against the latest boost cvs. Ralf P.S.: To check out the scitbx module for convenient access to the files: cvs -z3 -d:pserver:anonymous@cvs.cctbx.sourceforge.net:/cvsroot/cctbx co scitbx boa:/net/cci/rwgk % boa:/net/cci/rwgk % boa:/net/cci/rwgk % cat zmsg For those who are using the "bpl_utils": The name has changed, and the code has been reorganized. It is now part of the "scitbx" module (under the old "cctbx" cvs project). The container conversions can be found in this directory: http://cvs.sourceforge.net/cgi-bin/viewcvs.cgi/cctbx/scitbx/include/scitbx/… File name: container_conversions.h The regression test is here: http://cvs.sourceforge.net/cgi-bin/viewcvs.cgi/cctbx/scitbx/array_family/bo… File names: regression_test_module.cpp regression_test.py Example usage: scitbx::boost_python::container_conversions::from_python_sequence< std::vector<double>, scitbx::boost_python::container_conversions::variable_capacity_policy>(); Fragment from regression_test.py to illustrate how it works from Python: assert rt.std_vector((8,9,10)) == 27 assert rt.std_vector([7,9,10]) == 26 assert rt.std_vector(iter(xrange(10,13))) == 33 assert rt.std_vector((8,9,10,11)) == 38 Tested against the latest boost cvs. Ralf P.S.: To check out the scitbx module for convenient access to the files: cvs -z3 -d:pserver:anonymous@cvs.cctbx.sourceforge.net:/cvsroot/cctbx co scitbx __________________________________________________ Do You Yahoo!? Yahoo! Finance - Get real-time stock quotes http://finance.yahoo.com

1 0

Boost.Pyton v2 & gcc 2.95.3
by Peter Schoen Aug. 30, 2002

Aug. 30, 2002

Hi! From: "David Abrahams" <dave(a)boost-consulting.com> > From: "Peter Schoen" <Peter.Schoen1(a)web.de> >> When I try to compile the lib with: "bjam -sTOOLS=gcc test", I get many >> template related erros. If others were able to compile the lib using >> 2.95.3, I can also post the error messages. >I see that the fix Joel posted (and checked into CVS) was also sitting on >my hard disk, waiting to be checked in. Everything *should* be fine now. Yes, that solved the problem, many thanks. -Peter-

1 0

tuple::size()
by Ralf W. Grosse-Kunstleve Aug. 30, 2002

Aug. 30, 2002

If I want to know the length of a tuple, list or dict, currently I have to #include a certain file with a long path. On top of that, if I am not in namespace boost::python, I have to type "boost::python::len(o)". Points: - Having to #include <boost/python/whatever.hpp> is cumbersome and brings more code into scope than I really need. - boost::python::len() is quite a bit of clutter. - .size() is what one expects to have in C++. - .size() could directly look at sq_length and thus generate less code. David, would you be very upset if I added .size() to tuple.hpp, list.hpp and dict.hpp? Ralf __________________________________________________ Do You Yahoo!? Yahoo! Finance - Get real-time stock quotes http://finance.yahoo.com

3 4

offer-invitation
by Peter Steenwinkel Aug. 30, 2002

Aug. 30, 2002

Helllo, First, let me talk one joke-parable: Once, a man was perishing in a river and wanting a help he had begun to cry the heavens – God! Save me! ... Continuation see http://www.geocities.com/makeincome2002

1 0

List of Boost.Python projects?
by David Abrahams Aug. 26, 2002

Aug. 26, 2002

If you have a project or product which uses Boost.Python, I would appreciate it if you'd send me a link for the Boost.Python documentation. Thanks, Dave ----------------------------------------------------------- David Abrahams * Boost Consulting dave(a)boost-consulting.com * http://www.boost-consulting.com

5 4

adding types to module A from modules B, C, etc.
by Ralf W. Grosse-Kunstleve Aug. 24, 2002

Aug. 24, 2002

I have made good progress with exposing a family of C++ array types to Python (of course via Boost.Python). At the Python level it works like this: from somewhere import flex d = flex.double() # one-dimensional d.append(3) c = flex.complex_double(flex.grid((10,20,30))) # three-dimensional # etc. for bool int, long, float and a few other types I also need a number of arrays with custom element types. E.g. this simple one: t = flex.triple() # one-dimensional array of triples t.append((1,2,3)) # append one triple To make this generally usable, I'd like to separate the array toolbox from the rest of my (crystallographic) application. At the same time I'd like to avoid this: from somewhere import standard_flex # with int, double, etc. from somewhere_else import custom_flex # with triple and other custom types In other words, I'd like to add additional types to a central module flex from a number of other modules: the modules that expose the custom types. Isn't this related to Dave Hawke's ideas? Could I do it right now in V2? If not, how difficult would it be do get there? Thanks, Ralf __________________________________________________ Do You Yahoo!? Yahoo! Finance - Get real-time stock quotes http://finance.yahoo.com

1 0