Financial Modelling in Python

P1: JYS

app02 JWBK378-Fletcher April 24, 2009 8:20 Printer: Yet to come

212 Financial Modelling in Python

B.4 PYTHON OPERATORS

Boost.Python makes it extremely easy to wrap C++ operator-powered classes. A simple

example should suffice. Consider the class:

class Vector{ /*...*/};

Vector operator+(Vector const&, float);

Vector operator+(float, Vector const&);

Vector operator-(Vector const&, float);

Vector operator-(float, Vector const&);

Vector& operator+=(Vector&, float);

Vector& operator-=(Vector&, float);

bool operator<(Vector const&, Vector const&);

The class and operators can be mapped to Python by writing:

class <Vector>(‘‘Vector’’)

.def(self + float() )

.def(float() + self )

.def(self - float() )

.def(float() - self )

.def(self += float())

.def(self -= float())

.def(self < self)

;

B.5 FUNCTIONS

In C++ it is common to come across functions with arguments and return types that are pointers

or references. The problem with such primitive types is that we don’t know the owner of the

pointer or referenced object. Although most C++ programmers now use smart pointers with

clear ownership semantics, nevertheless there exists a lot of older C++ code with raw pointers.

So Boost.Python has to be able to deal with them. The main issue to solve is the problem of

dangling pointers and references. Let’s consider the following simple C++ function:

X& f(Y& y, Z* z)

{

y.z = z;

return y.x;

}

The above function binds the lifetime ...