Extending Python with Numerical Recipes

Abstraction: Tutorial for wrapping C++ Numerical Recipes code as callable Python modules

Key points:

• Uses nr3python.h header to bridge Python and C++ types: NRpyArgs unpacks Python arguments, NRpyObject wraps return values, NRpyTuple returns multiple values
• Vectors and matrices pass by reference (no copy); changes in C++ are immediately visible in Python numpy arrays
• Six examples: scalar variables, vector/matrix/string/dict, encapsulating C++ functions, passing Python functions to C++, persistent OO objects (NRpyCONSTRUCTOR/NRpyCONNECT macros), and sharing objects by name between namespaces
• Compilation produces .so (Linux) or .pyd (Windows) shared libraries; first-time compile requires locating Python.h and numpy/arrayobject.h include paths
• Key motivation: C++ executes orders of magnitude faster than Python for numerical functions, while Python controls the workflow

Connections: Numerical Recipes · Python · Python C Extension · Numerical Computing

Source: http://numerical.recipes/nr3_python_tutorial.html