<html> <head> <title>SWIG:Examples:tcl:class</title> </head> <body bgcolor="#ffffff"> <tt>SWIG/Examples/tcl/class/</tt> <hr> <H2>Wrapping a simple C++ class</H2> <tt>$Header: /cvs/projects/SWIG/Examples/tcl/class/index.html,v 1.1.4.1 2001/08/30 04:18:46 beazley Exp $</tt><br> <p> This example illustrates the most primitive form of C++ class wrapping performed by SWIG. In this case, C++ classes are simply transformed into a collection of C-style functions that provide access to class members. <h2>The C++ Code</h2> Suppose you have some C++ classes described by the following (and admittedly lame) header file: <blockquote> <pre> /* File : example.h */ class Shape { public: Shape() { nshapes++; } virtual ~Shape() { nshapes--; }; double x, y; void move(double dx, double dy); virtual double area() = 0; virtual double perimeter() = 0; static int nshapes; }; class Circle : public Shape { private: double radius; public: Circle(double r) : radius(r) { }; virtual double area(); virtual double perimeter(); }; class Square : public Shape { private: double width; public: Square(double w) : width(w) { }; virtual double area(); virtual double perimeter(); }; </pre> </blockquote> <h2>The SWIG interface</h2> A simple SWIG interface for this can be built by simply grabbing the header file like this: <blockquote> <pre> /* File : example.i */ %module example %{ #include "example.h" %} /* Let's just grab the original header file here */ %include "example.h" </pre> </blockquote> Note: when creating a C++ extension, you must run SWIG with the <tt>-c++</tt> option like this: <blockquote> <pre> % swig -c++ -tcl example.i </pre> </blockquote> <h2>Some sample Tcl scripts</h2> SWIG performs two forms of C++ wrapping-- a low level interface and a high level widget-like interface. <ul> <li> Click <a href="example1.tcl">here</a> to see a script that calls the C++ functions using the low-level interface. <li> Click <a href="example2.tcl">here</a> to see a the same script written with the high-level interface. </ul> <h2>Key points</h2> <ul> <li>The low-level C++ interface works like this: <p> <ul> <li>To create a new object, you call a constructor like this: <blockquote> <pre> set c [new_Circle 10.0] </pre> </blockquote> <p> <li>To access member data, a pair of accessor functions are used. For example: <blockquote> <pre> Shape_x_set $c 15 ;# Set member data set x [Shape_x_get $c] ;# Get member data </pre> </blockquote> Note: when accessing member data, the name of the base class must be used such as <tt>Shape_x_get</tt> <p> <li>To invoke a member function, you simply do this <blockquote> <pre> puts "The area is [Shape_area $c]" </pre> </blockquote> <p> <li>Type checking knows about the inheritance structure of C++. For example: <blockquote> <pre> Shape_area $c # Works (c is a Shape) Circle_area $c # Works (c is a Circle) Square_area $c # Fails (c is definitely not a Square) </pre> </blockquote> <p> <li>To invoke a destructor, simply do this <blockquote> <pre> delete_Shape $c # Deletes a shape </pre> </blockquote> <p> <li>Static member variables are wrapped as C global variables. For example: <blockquote> <pre> set n $Shape_nshapes # Get a static data member set Shapes_nshapes 13 # Set a static data member </pre> </blockquote> </ul> <p> <li>The high-level interface works like a Tk widget <p> <ul> <li>To create a new object, you call a constructor like this: <blockquote> <pre> Circle c 10 # c becomes a name for the Circle object </pre> </blockquote> <p> <li>To access member data, use cget and configure methods. For example: <blockquote> <pre> c configure -x 15 ;# Set member data set x [c cget -x] ;# Get member data </pre> </blockquote> <p> <li>To invoke a member function, you simply do this <blockquote> <pre> puts "The area is [c area]" </pre> </blockquote> <p> <li>To invoke a destructor, simply destroy the object name like this: <blockquote> <pre> rename c "" # c goes away </pre> </blockquote> <p> <li>Static member variables are wrapped as C global variables. For example: <blockquote> <pre> set n $Shape_nshapes # Get a static data member set Shapes_nshapes 13 # Set a static data member </pre> </blockquote> </ul> </ul> <h2>General Comments</h2> <ul> <li>The low-level function interface is much faster than the high-level interface. In fact, all the higher level interface does is call functions in the low-level interface. <p> <li>SWIG *does* know how to properly perform upcasting of objects in an inheritance hierarchy (including multiple inheritance). Therefore it is perfectly safe to pass an object of a derived class to any function involving a base class. <p> <li>A wide variety of C++ features are not currently supported by SWIG. Here is the short and incomplete list: <p> <ul> <li>Overloaded methods and functions. SWIG wrappers don't know how to resolve name conflicts so you must give an alternative name to any overloaded method name using the %name directive like this: <blockquote> <pre> void foo(int a); %name(foo2) void foo(double a, double b); </pre> </blockquote> <p> <li>Overloaded operators. Not supported at all. The only workaround for this is to write a helper function. For example: <blockquote> <pre> %inline %{ Vector *vector_add(Vector *a, Vector *b) { ... whatever ... } %} </pre> </blockquote> <p> <li>Namespaces. Not supported at all. Won't be supported until SWIG2.0 (if at all). </ul> <hr> </body> </html>