Theory of Signals and Callbacks


In version 2.0, the signal system has been moved from GTK to GLib, therefore the functions and types explained in this section have a "g_" prefix rather than a "gtk_" prefix. We won't go into details about the extensions which the GLib 2.0 signal system has relative to the GTK 1.2 signal system.

Before we look in detail at helloworld, we'll discuss signals and callbacks. GTK is an event driven toolkit, which means it will sleep in gtk_main() until an event occurs and control is passed to the appropriate function.

This passing of control is done using the idea of "signals". (Note that these signals are not the same as the Unix system signals, and are not implemented using them, although the terminology is almost identical.) When an event occurs, such as the press of a mouse button, the appropriate signal will be "emitted" by the widget that was pressed. This is how GTK does most of its useful work. There are signals that all widgets inherit, such as "destroy", and there are signals that are widget specific, such as "toggled" on a toggle button.

To make a button perform an action, we set up a signal handler to catch these signals and call the appropriate function. This is done by using a function such as:

gulong g_signal_connect( gpointer      *object,
                         const gchar   *name,
                         GCallback     func,
                         gpointer      func_data );

where the first argument is the widget which will be emitting the signal, and the second the name of the signal you wish to catch. The third is the function you wish to be called when it is caught, and the fourth, the data you wish to have passed to this function.

The function specified in the third argument is called a "callback function", and should generally be of the form

void callback_func( GtkWidget *widget,
                    gpointer   callback_data );

where the first argument will be a pointer to the widget that emitted the signal, and the second a pointer to the data given as the last argument to the g_signal_connect() function as shown above.

Note that the above form for a signal callback function declaration is only a general guide, as some widget specific signals generate different calling parameters.

Another call used in the helloworld example, is:

gulong g_signal_connect_swapped( gpointer     *object,
                                 const gchar  *name,
                                 GCallback    func,
                                 gpointer     *slot_object );

g_signal_connect_swapped() is the same as g_signal_connect() except that the callback function only uses one argument, a pointer to a GTK object. So when using this function to connect signals, the callback should be of the form

void callback_func( GtkObject *object );

where the object is usually a widget. We usually don't setup callbacks for g_signal_connect_swapped() however. They are usually used to call a GTK function that accepts a single widget or object as an argument, as is the case in our helloworld example.

The purpose of having two functions to connect signals is simply to allow the callbacks to have a different number of arguments. Many functions in the GTK library accept only a single GtkWidget pointer as an argument, so you want to use the g_signal_connect_swapped() for these, whereas for your functions, you may need to have additional data supplied to the callbacks.