glib/docs/reference/gobject/tut_howto.xml

<?xml version='1.0' encoding="UTF-8"?>
<!DOCTYPE part PUBLIC "-//OASIS//DTD DocBook XML V4.5//EN" 
               "http://www.oasis-open.org/docbook/xml/4.5/docbookx.dtd" [
]>
<part label="IV">
  <title>Tutorial</title>
  <partintro>
    <para>
      This chapter tries to answer the real-life questions of users and presents
      the most common use cases in order from most likely to least
      likely.
    </para>
  </partintro>

<chapter id="howto-gobject">
  <title>How to define and implement a new GObject</title>
  
  <para>
    This chapter focuses on the implementation of a subtype of GObject, for
    example to create a custom class hierarchy, or to subclass a GTK widget.
  </para>

  <para>
    Throughout the chapter, a running example of a file viewer program is used,
    which has a <type>ViewerFile</type> class to represent a single file being
    viewed, and various derived classes for different types of files with
    special functionality, such as audio files. The example application also
    supports editing files (for example, to tweak a photo being viewed), using
    a <type>ViewerEditable</type> interface.
  </para>

  <sect1 id="howto-gobject-header">
    <title>Boilerplate header code</title>
    
    <para>
      The first step before writing the code for your GObject is to write the
      type's header which contains the needed type, function and macro
      definitions. Each of these elements is nothing but a convention which
      is followed by almost all users of GObject, and has been refined over
      multiple years of experience developing GObject-based code. If you are
      writing a library, it is particularly important for you to adhere closely
      to these conventions; users of your library will assume that you have.
      Even if not writing a library, it will help other people who want to work
      on your project.
    </para>

    <para>
      Pick a name convention for your headers and source code and stick to it:
      <itemizedlist>
        <listitem><para>use a dash to separate the prefix from the typename:
        <filename>viewer-file.h</filename> and <filename>viewer-file.c</filename>
        (this is the convention used by Nautilus and most GNOME libraries).</para></listitem>
        <listitem><para>use an underscore to separate the prefix from the
        typename: <filename>viewer_file.h</filename> and
        <filename>viewer_file.c</filename>.</para></listitem>
        <listitem><para>Do not separate the prefix from the typename:
        <filename>viewerfile.h</filename> and <filename>viewerfile.c</filename>.
        (this is the convention used by GTK)</para></listitem>
      </itemizedlist>
      Some people like the first two solutions better: it makes reading file
      names easier for those with poor eyesight.
    </para>

    <para>
      The basic conventions for any header which exposes a GType are described
      in <xref linkend="gtype-conventions"/>.
    </para>

    <para>
      If you want to declare a type named ‘file’ in namespace ‘viewer’, name the
      type instance <function>ViewerFile</function> and its class
      <function>ViewerFileClass</function> (names are case sensitive). The
      recommended method of declaring a type differs based on whether the type
      is final or derivable.
    </para>

    <para>
      Final types cannot be subclassed further, and should be the default choice
      for new types — changing a final type to be derivable is always a change
      that will be compatible with existing uses of the code, but the converse
      will often cause problems. Final types are declared using
      <link linkend="G-DECLARE-FINAL-TYPE:CAPS"><function>G_DECLARE_FINAL_TYPE</function></link>,
      and require a structure to hold the instance data to be declared in the
      source code (not the header file).

<informalexample><programlisting>
/*
 * Copyright/Licensing information.
 */

/* inclusion guard */
#ifndef __VIEWER_FILE_H__
#define __VIEWER_FILE_H__

#include &lt;glib-object.h&gt;
/*
 * Potentially, include other headers on which this header depends.
 */

G_BEGIN_DECLS

/*
 * Type declaration.
 */
#define VIEWER_TYPE_FILE viewer_file_get_type ()
G_DECLARE_FINAL_TYPE (ViewerFile, viewer_file, VIEWER, FILE, GObject)

/*
 * Method definitions.
 */
ViewerFile *viewer_file_new (void);

G_END_DECLS

#endif /* __VIEWER_FILE_H__ */
</programlisting></informalexample>
    </para>

    <para>
      Derivable types <emphasis>can</emphasis> be subclassed further, and their class and
      instance structures form part of the public API which must not be changed
      if API stability is cared about. They are declared using
      <link linkend="G-DECLARE-DERIVABLE-TYPE:CAPS"><function>G_DECLARE_DERIVABLE_TYPE</function></link>:
<informalexample><programlisting>
/*
 * Copyright/Licensing information.
 */

/* inclusion guard */
#ifndef __VIEWER_FILE_H__
#define __VIEWER_FILE_H__

#include &lt;glib-object.h&gt;
/*
 * Potentially, include other headers on which this header depends.
 */

G_BEGIN_DECLS

/*
 * Type declaration.
 */
#define VIEWER_TYPE_FILE viewer_file_get_type ()
G_DECLARE_DERIVABLE_TYPE (ViewerFile, viewer_file, VIEWER, FILE, GObject)

struct _ViewerFileClass
{
  GObjectClass parent_class;

  /* Class virtual function fields. */
  void (* open) (ViewerFile  *file,
                 GError     **error);

  /* Padding to allow adding up to 12 new virtual functions without
   * breaking ABI. */
  gpointer padding[12];
};

/*
 * Method definitions.
 */
ViewerFile *viewer_file_new (void);

G_END_DECLS

#endif /* __VIEWER_FILE_H__ */
</programlisting></informalexample>
    </para>

    <para>
      The convention for header includes is to add the minimum number of
      <function>#include</function> directives to the top of your headers needed
      to compile that header. This
      allows client code to simply <function>#include "viewer-file.h"</function>,
      without needing to know the prerequisites for
      <filename>viewer-file.h</filename>.
    </para>
  </sect1>

  <sect1 id="howto-gobject-code">
    <title>Boilerplate code</title>

    <para>
      In your code, the first step is to <function>#include</function> the
      needed headers:
<informalexample><programlisting>
/*
 * Copyright information
 */

#include "viewer-file.h"

/* Private structure definition. */
typedef struct {
  gchar *filename;
  /* stuff */
} ViewerFilePrivate;

/* 
 * forward definitions
 */
</programlisting></informalexample>
    </para>

    <para>
      If the class is being declared as final using
      <function>G_DECLARE_FINAL_TYPE</function>, its instance structure should
      be defined in the C file:
<informalexample><programlisting>
struct _ViewerFile
{
  GObject parent_instance;

  /* Other members, including private data. */
};
</programlisting></informalexample>
    </para>

    <para>
      Call the <function>G_DEFINE_TYPE</function> macro (or
      <function>G_DEFINE_TYPE_WITH_PRIVATE</function> if your class needs
      private data — final types do <emphasis>not</emphasis> need private data)
      using the name
      of the type, the prefix of the functions and the parent GType to
      reduce the amount of boilerplate needed. This macro will:

      <itemizedlist>
        <listitem><simpara>implement the <function>viewer_file_get_type</function>
        function</simpara></listitem>
        <listitem><simpara>define a parent class pointer accessible from
        the whole .c file</simpara></listitem>
        <listitem><simpara>add private instance data to the type (if using
        <function>G_DEFINE_TYPE_WITH_PRIVATE</function>)</simpara></listitem>
      </itemizedlist>
    </para>

    <para>
      If the class has been declared as final using
      <function>G_DECLARE_FINAL_TYPE</function> (see
      <xref linkend="howto-gobject-header"/>), private data should be placed in
      the instance structure, <type>ViewerFile</type>, and
      <function>G_DEFINE_TYPE</function> should be used instead of
      <function>G_DEFINE_TYPE_WITH_PRIVATE</function>. The instance structure
      for a final class is not exposed publicly, and is not embedded in the
      instance structures of any derived classes (because the class is final);
      so its size can vary without causing incompatibilities for code which uses
      the class. Conversely, private data for derivable classes
      <emphasis>must</emphasis> be included in a private structure, and
      <function>G_DEFINE_TYPE_WITH_PRIVATE</function> must be used.

<informalexample><programlisting>
G_DEFINE_TYPE (ViewerFile, viewer_file, G_TYPE_OBJECT)
</programlisting></informalexample>
or
<informalexample><programlisting>
G_DEFINE_TYPE_WITH_PRIVATE (ViewerFile, viewer_file, G_TYPE_OBJECT)
</programlisting></informalexample>
    </para>

    <para>
      It is also possible to use the
      <function>G_DEFINE_TYPE_WITH_CODE</function> macro to control the
      <function>get_type</function> function implementation — for instance, to
      add a call to the <function>G_IMPLEMENT_INTERFACE</function> macro to
      implement an interface.
    </para>
  </sect1>

  <sect1 id="howto-gobject-construction">
    <title>Object construction</title>

    <para>
      People often get confused when trying to construct their GObjects because of the
      sheer number of different ways to hook into the objects's construction process: it is
      difficult to figure which is the <emphasis>correct</emphasis>, recommended way.
    </para>

    <para>
      <xref linkend="gobject-construction-table"/> shows what user-provided functions
      are invoked during object instantiation and in which order they are invoked.
      A user looking for the equivalent of the simple C++ constructor function should use
      the <function>instance_init</function> method. It will be invoked after
      all the parents’ <function>instance_init</function>
      functions have been invoked. It cannot take arbitrary construction parameters 
      (as in C++) but if your object needs arbitrary parameters to complete initialization,
      you can use construction properties.
    </para>

    <para>
      Construction properties will be set only after all
      <function>instance_init</function> functions have run.
      No object reference will be returned to the client of <function><link linkend="g-object-new">g_object_new</link></function>
      until all the construction properties have been set.
    </para>

    <para>
      It is important to note that object construction cannot <emphasis>ever</emphasis>
      fail. If you require a fallible GObject construction, you can use the
      <link linkend="GInitable"><type>GInitable</type></link> and
      <link linkend="GAsyncInitable"><type>GAsyncInitable</type></link>
      interfaces provided by the GIO library.
    </para>

    <para>
      You should write the following code first:
<informalexample><programlisting>
G_DEFINE_TYPE_WITH_PRIVATE (ViewerFile, viewer_file, G_TYPE_OBJECT)

static void
viewer_file_class_init (ViewerFileClass *klass)
{
}

static void
viewer_file_init (ViewerFile *self)
{
  ViewerFilePrivate *priv = viewer_file_get_instance_private (self);

  /* initialize all public and private members to reasonable default values.
   * They are all automatically initialized to 0 to begin with. */
}
</programlisting></informalexample>
    </para>

    <para>
      If you need special construction properties (with
      <link linkend="G-PARAM-CONSTRUCT-ONLY:CAPS"><function>G_PARAM_CONSTRUCT_ONLY</function></link>
      set), install the properties in
      the <function>class_init()</function> function, override the <function>set_property()</function>
      and <function>get_property()</function> methods of the GObject class,
      and implement them as described by <xref linkend="gobject-properties"/>.
    </para>

    <para>
      Property IDs must start from 1, as 0 is reserved for internal use by
      GObject.
<informalexample><programlisting>
enum
{
  PROP_FILENAME = 1,
  PROP_ZOOM_LEVEL,
  N_PROPERTIES
};

static GParamSpec *obj_properties[N_PROPERTIES] = { NULL, };

static void
viewer_file_class_init (ViewerFileClass *klass)
{
  GObjectClass *object_class = G_OBJECT_CLASS (klass);

  object_class->set_property = viewer_file_set_property;
  object_class->get_property = viewer_file_get_property;

  obj_properties[PROP_FILENAME] =
    g_param_spec_string ("filename",
                         "Filename",
                         "Name of the file to load and display from.",
                         NULL  /* default value */,
                         G_PARAM_CONSTRUCT_ONLY | G_PARAM_READWRITE);

  obj_properties[PROP_ZOOM_LEVEL] =
    g_param_spec_uint ("zoom-level",
                       "Zoom level",
                       "Zoom level to view the file at.",
                       0  /* minimum value */,
                       10 /* maximum value */,
                       2  /* default value */,
                       G_PARAM_READWRITE);

  g_object_class_install_properties (object_class,
                                     N_PROPERTIES,
                                     obj_properties);
}
</programlisting></informalexample>
      If you need this, make sure you can build and run code similar to the
      code shown above. Also, make sure your construct properties can be set
      without side effects during construction.
    </para>

    <para>
      Some people sometimes need to complete the initialization of an instance
      of a type only after the properties passed to the constructors have been
      set. This is possible through the use of the <function>constructor()</function>
      class method as described in <xref linkend="gobject-instantiation"/> or,
      more simply, using the <function>constructed()</function> class method.
      Note that the <function>constructed()</function>
      virtual function will only be invoked after the properties marked as
      <function>G_PARAM_CONSTRUCT_ONLY</function> or
      <function>G_PARAM_CONSTRUCT</function> have been consumed, but
      before the regular properties passed to <function>g_object_new()</function>
      have been set.
    </para>
  </sect1>

  <sect1 id="howto-gobject-destruction">
    <title>Object destruction</title>

    <para>
      Again, it is often difficult to figure out which mechanism to use to
      hook into the object's destruction process: when the last
      <function><link linkend="g-object-unref">g_object_unref</link></function>
      function call is made, a lot of things happen as described in
      <xref linkend="gobject-destruction-table"/>.
    </para>

    <para>
      The destruction process of your object is in two phases: dispose and
      finalize. This split is necessary to handle
      potential cycles due to the nature of the reference counting mechanism
      used by GObject, as well as dealing with temporary revival of
      instances in case of signal emission during the destruction sequence.
      See <xref linkend="gobject-memory-cycles"/> for more information.
<informalexample><programlisting>
struct _ViewerFilePrivate
{
  gchar *filename;
  guint zoom_level;

  GInputStream *input_stream;
};

G_DEFINE_TYPE_WITH_PRIVATE (ViewerFile, viewer_file, G_TYPE_OBJECT)

static void
viewer_file_dispose (GObject *gobject)
{
  ViewerFilePrivate *priv = viewer_file_get_instance_private (VIEWER_FILE (gobject));

  /* In dispose(), you are supposed to free all types referenced from this
   * object which might themselves hold a reference to self. Generally,
   * the most simple solution is to unref all members on which you own a 
   * reference.
   */

  /* dispose() might be called multiple times, so we must guard against
   * calling g_object_unref() on an invalid GObject by setting the member
   * NULL; g_clear_object() does this for us.
   */
  g_clear_object (&amp;priv->input_stream);

  /* Always chain up to the parent class; there is no need to check if
   * the parent class implements the dispose() virtual function: it is
   * always guaranteed to do so
   */
  G_OBJECT_CLASS (viewer_file_parent_class)->dispose (gobject);
}

static void
viewer_file_finalize (GObject *gobject)
{
  ViewerFilePrivate *priv = viewer_file_get_instance_private (VIEWER_FILE (gobject));

  g_free (priv->filename);

  /* Always chain up to the parent class; as with dispose(), finalize()
   * is guaranteed to exist on the parent's class virtual function table
   */
  G_OBJECT_CLASS (viewer_file_parent_class)->finalize (gobject);
}

static void
viewer_file_class_init (ViewerFileClass *klass)
{
  GObjectClass *object_class = G_OBJECT_CLASS (klass);

  object_class->dispose = viewer_file_dispose;
  object_class->finalize = viewer_file_finalize;
}

static void
viewer_file_init (ViewerFile *self);
{
  ViewerFilePrivate *priv = viewer_file_get_instance_private (self);

  priv->input_stream = g_object_new (VIEWER_TYPE_INPUT_STREAM, NULL);
  priv->filename = /* would be set as a property */;
}
</programlisting></informalexample>
    </para>

    <para>
      It is possible that object methods might be invoked after dispose is
      run and before finalize runs. GObject does not consider this to be a
      program error: you must gracefully detect this and neither crash nor
      warn the user, by having a disposed instance revert to an inert state.
    </para>
  </sect1>

  <sect1 id="howto-gobject-methods">
    <title>Object methods</title>

    <para>
      Just as with C++, there are many different ways to define object
      methods and extend them: the following list and sections draw on
      C++ vocabulary. (Readers are expected to know basic C++ concepts.
      Those who have not had to write C++ code recently can refer to e.g.
      <ulink url="http://www.cplusplus.com/doc/tutorial/"/> to refresh
      their memories.)
      <itemizedlist>
        <listitem><para>
            non-virtual public methods,
          </para></listitem>
        <listitem><para>
            virtual public methods and
          </para></listitem>
        <listitem><para>
            virtual private methods
          </para></listitem>
      </itemizedlist>
    </para>

    <sect2 id="non-virtual-public-methods">
      <title>Non-virtual public methods</title>

      <para>
        These are the simplest, providing a simple method which
        acts on the object. Provide a function
        prototype in the header and an implementation of that prototype
        in the source file.
<informalexample><programlisting>
/* declaration in the header. */
void viewer_file_open (ViewerFile  *self,
                       GError     **error);

/* implementation in the source file */
void
viewer_file_open (ViewerFile  *self,
                  GError     **error)
{
  g_return_if_fail (VIEWER_IS_FILE (self));
  g_return_if_fail (error == NULL || *error == NULL);

  /* do stuff here. */
}
</programlisting></informalexample>
      </para>
    </sect2>

    <sect2 id="virtual-public-methods">
      <title>Virtual public methods</title>

      <para>
        This is the preferred way to create GObjects with overridable methods:
        <itemizedlist>
          <listitem><para>
            Define the common method and its virtual function in the
            class structure in the public header
          </para></listitem>
          <listitem><para>
            Define the common method in the header file and implement it in the
            source file
          </para></listitem>
          <listitem><para>
            Implement a base version of the virtual function in the source
            file and initialize the virtual function pointer to this
            implementation in the object’s <function>class_init</function>
            function; or leave it as <constant>NULL</constant> for a ‘pure
            virtual’ method which must be overridden by derived classes
          </para></listitem>
          <listitem><para>
            Re-implement the virtual function in each derived class which needs
            to override it
          </para></listitem>
        </itemizedlist>
      </para>
      <para>
        Note that virtual functions can only be defined if the class is
        derivable, declared using
        <link linkend="G-DECLARE-DERIVABLE-TYPE:CAPS"><function>G_DECLARE_DERIVABLE_TYPE</function></link>
        so the class structure can be defined.
<informalexample><programlisting>
/* declaration in viewer-file.h. */
#define VIEWER_TYPE_FILE viewer_file_get_type ()
G_DECLARE_DERIVABLE_TYPE (ViewerFile, viewer_file, VIEWER, FILE, GObject)

struct _ViewerFileClass
{
  GObjectClass parent_class;

  /* stuff */
  void (*open) (ViewerFile  *self,
                GError     **error);

  /* Padding to allow adding up to 12 new virtual functions without
   * breaking ABI. */
  gpointer padding[12];
};

void viewer_file_open (ViewerFile  *self,
                       GError     **error);

/* implementation in viewer-file.c */
void
viewer_file_open (ViewerFile  *self,
                  GError     **error)
{
  ViewerFileClass *klass;

  g_return_if_fail (VIEWER_IS_FILE (self));
  g_return_if_fail (error == NULL || *error == NULL);

  klass = VIEWER_FILE_GET_CLASS (self);
  g_return_if_fail (klass->open != NULL);

  klass->open (self, error);
}
</programlisting></informalexample>
        The code above simply redirects the <function>open</function> call
        to the relevant virtual function.
      </para>

      <para>
        It is possible to provide a default
        implementation for this class method in the object's
        <function>class_init</function> function: initialize the
        <function>klass-&gt;open</function> field to a pointer to the
        actual implementation.
        By default, class methods that are not inherited are initialized to
        <function>NULL</function>, and thus are to be considered "pure virtual".
<informalexample><programlisting>
static void
viewer_file_real_close (ViewerFile  *self,
                        GError     **error)
{
  /* Default implementation for the virtual method. */
}

static void
viewer_file_class_init (ViewerFileClass *klass)
{
  /* this is not necessary, except for demonstration purposes.
   *
   * pure virtual method: mandates implementation in children.
   */
  klass->open = NULL;

  /* merely virtual method. */
  klass->close = viewer_file_real_close;
}

void
viewer_file_open (ViewerFile  *self,
                  GError     **error)
{
  ViewerFileClass *klass;

  g_return_if_fail (VIEWER_IS_FILE (self));
  g_return_if_fail (error == NULL || *error == NULL);

  klass = VIEWER_FILE_GET_CLASS (self);

  /* if the method is purely virtual, then it is a good idea to
   * check that it has been overridden before calling it, and,
   * depending on the intent of the class, either ignore it silently
   * or warn the user.
   */
  g_return_if_fail (klass->open != NULL);
  klass->open (self, error);
}

void
viewer_file_close (ViewerFile  *self,
                   GError     **error)
{
  ViewerFileClass *klass;

  g_return_if_fail (VIEWER_IS_FILE (self));
  g_return_if_fail (error == NULL || *error == NULL);

  klass = VIEWER_FILE_GET_CLASS (self);
  if (klass->close != NULL)
    klass->close (self, error);
}
</programlisting></informalexample>
      </para>
    </sect2>

    <sect2 id="virtual-private-methods">
      <title>Virtual private Methods</title>

      <para>
        These are very similar to <link linkend="virtual-public-methods">virtual
        public methods</link>. They just don't
        have a public function to call directly. The header
        file contains only a declaration of the virtual function:
<informalexample><programlisting>
/* declaration in viewer-file.h. */
struct _ViewerFileClass
{
  GObjectClass parent;

  /* Public virtual method as before. */
  void     (*open)           (ViewerFile  *self,
                              GError     **error);

  /* Private helper function to work out whether the file can be loaded via
   * memory mapped I/O, or whether it has to be read as a stream. */
  gboolean (*can_memory_map) (ViewerFile *self);

  /* Padding to allow adding up to 12 new virtual functions without
   * breaking ABI. */
  gpointer padding[12];
};

void viewer_file_open (ViewerFile *self, GError **error);
</programlisting></informalexample>
        These virtual functions are often used to delegate part of the job
        to child classes:
<informalexample><programlisting>
/* this accessor function is static: it is not exported outside of this file. */
static gboolean 
viewer_file_can_memory_map (ViewerFile *self)
{
  return VIEWER_FILE_GET_CLASS (self)->can_memory_map (self);
}

void
viewer_file_open (ViewerFile  *self,
                  GError     **error)
{
  g_return_if_fail (VIEWER_IS_FILE (self));
  g_return_if_fail (error == NULL || *error == NULL);

  /*
   * Try to load the file using memory mapped I/O, if the implementation of the
   * class determines that is possible using its private virtual method.
   */
  if (viewer_file_can_memory_map (self))
    {
      /* Load the file using memory mapped I/O. */
    }
  else
    {
      /* Fall back to trying to load the file using streaming I/O… */
    }
}
</programlisting></informalexample>
      </para>

      <para>
        Again, it is possible to provide a default implementation for this
        private virtual function:
<informalexample><programlisting>
static gboolean
viewer_file_real_can_memory_map (ViewerFile *self)
{
  /* As an example, always return false. Or, potentially return true if the
   * file is local. */
  return FALSE;
}

static void
viewer_file_class_init (ViewerFileClass *klass)
{
  /* non-pure virtual method; does not have to be implemented in children. */
  klass->can_memory_map = viewer_file_real_can_memory_map;
}
</programlisting></informalexample>
      </para>

      <para>
        Derived classes can then override the method with code such as:
<informalexample><programlisting>
static void
viewer_audio_file_class_init (ViewerAudioFileClass *klass)
{
  ViewerFileClass *file_class = VIEWER_FILE_CLASS (klass);

  /* implement pure virtual function. */
  file_class->can_memory_map = viewer_audio_file_can_memory_map;
}
</programlisting></informalexample>
      </para>
    </sect2>
  </sect1>

  <sect1 id="howto-gobject-chainup">
    <title>Chaining up</title>
    
    <para>Chaining up is often loosely defined by the following set of
    conditions:
      <itemizedlist>
        <listitem><para>Parent class A defines a public virtual method named <function>foo</function> and 
        provides a default implementation.</para></listitem>
        <listitem><para>Child class B re-implements method <function>foo</function>.</para></listitem>
        <listitem><para>B’s implementation of <function>foo</function> calls (‘chains up to’) its parent class A’s implementation of <function>foo</function>.</para></listitem>
      </itemizedlist>
      There are various uses of this idiom:
      <itemizedlist>
        <listitem><para>You need to extend the behaviour of a class without modifying its code. You create
          a subclass to inherit its implementation, re-implement a public virtual method to modify the behaviour
          and chain up to ensure that the previous behaviour is not really modified, just extended.
          </para></listitem>
        <listitem><para>You need to implement the
          <ulink url="http://en.wikipedia.org/wiki/Chain-of-responsibility_pattern">Chain
          Of Responsibility pattern</ulink>: each object of the inheritance
          tree chains up to its parent (typically, at the beginning or the end of the method) to ensure that
          each handler is run in turn.</para></listitem>
      </itemizedlist>
    </para>

    <para>
      To explicitly chain up to the implementation of the virtual method in the parent class, 
      you first need a handle to the original parent class structure. This pointer can then be used to 
      access the original virtual function pointer and invoke it directly.
      <footnote>
        <para>
          The <emphasis>original</emphasis> adjective used in this sentence is not innocuous. To fully 
          understand its meaning, recall how class structures are initialized: for each object type,
          the class structure associated with this object is created by first copying the class structure of its
          parent type (a simple <function>memcpy</function>) and then by invoking the <function>class_init</function> callback on
          the resulting class structure. Since the <function>class_init</function> callback is responsible for overwriting the class structure
          with the user re-implementations of the class methods, the modified copy of the parent class
          structure stored in the derived instance cannot be used. A copy of the class structure of an instance of the parent
          class is needed.
        </para>
      </footnote>
    </para>
    
    <para>
      Use the <function>parent_class</function> pointer created and initialized
      by the
      <link linkend="G-DEFINE-TYPE:CAPS"><function>G_DEFINE_TYPE</function></link>
      family of macros, for instance:
<informalexample><programlisting>
static void
b_method_to_call (B *obj, gint some_param)
{
  /* do stuff before chain up */

  /* call the method_to_call() virtual function on the
   * parent of BClass, AClass.
   *
   * remember the explicit cast to AClass*
   */
  A_CLASS (b_parent_class)->method_to_call (obj, some_param);

  /* do stuff after chain up */
}
</programlisting></informalexample>
  </para>

  </sect1>

</chapter>
<!-- End Howto GObject -->

<chapter id="howto-interface">
  <title>How to define and implement interfaces</title>

  <sect1 id="howto-interface-define">
    <title>Defining interfaces</title>
  
  <para>
    The theory behind how GObject interfaces work is given in
    <xref linkend="gtype-non-instantiatable-non-classed"/>; this section covers how to
    define and implement an interface.
  </para>

  <para>
    The first step is to get the header right. This interface
    defines three methods:
<informalexample><programlisting>
/*
 * Copyright/Licensing information.
 */

#ifndef __VIEWER_EDITABLE_H__
#define __VIEWER_EDITABLE_H__

#include &lt;glib-object.h&gt;

G_BEGIN_DECLS

#define VIEWER_TYPE_EDITABLE viewer_editable_get_type ()
G_DECLARE_INTERFACE (ViewerEditable, viewer_editable, VIEWER, EDITABLE, GObject)

struct _ViewerEditableInterface
{
  GTypeInterface parent_iface;

  void (*save) (ViewerEditable  *self,
                GError         **error);
  void (*undo) (ViewerEditable  *self,
                guint            n_steps);
  void (*redo) (ViewerEditable  *self,
                guint            n_steps);
};

void viewer_editable_save (ViewerEditable  *self,
                           GError         **error);
void viewer_editable_undo (ViewerEditable  *self,
                           guint            n_steps);
void viewer_editable_redo (ViewerEditable  *self,
                           guint            n_steps);

G_END_DECLS

#endif /* __VIEWER_EDITABLE_H__ */
</programlisting></informalexample>
    This code is the same as the code for a normal <link linkend="GType"><type>GType</type></link>
    which derives from a <link linkend="GObject"><type>GObject</type></link> except for a few details:
    <itemizedlist>
      <listitem><para>
        The <function>_GET_CLASS</function> function is called
        <function>_GET_IFACE</function> (and is defined by
        <link linkend="G-DECLARE-INTERFACE:CAPS"><function>G_DECLARE_INTERFACE</function></link>).
      </para></listitem>
      <listitem><para>
        The instance type, <type>ViewerEditable</type>, is not fully defined: it is
        used merely as an abstract type which represents an instance of
        whatever object which implements the interface.
      </para></listitem>
      <listitem><para>
        The parent of the <type>ViewerEditableInterface</type> is
        <type>GTypeInterface</type>, not <type>GObjectClass</type>.
      </para></listitem>
    </itemizedlist>
  </para>

  <para>
    The implementation of the <type>ViewerEditable</type> type itself is trivial:
    <itemizedlist>
      <listitem><para><function><link linkend="G-DEFINE-INTERFACE:CAPS">G_DEFINE_INTERFACE</link></function>
       creates a <function>viewer_editable_get_type</function> function which registers the
       type in the type system. The third argument is used to define a
       <link linkend="howto-interface-prerequisite">prerequisite interface</link>
       (which we'll talk about more later). Just pass <code>0</code> for this
       argument when an interface has no prerequisite.
       </para></listitem>
      <listitem><para><function>viewer_editable_default_init</function> is expected
      to register the interface's signals if there are any (we will see a bit
      later how to use them).</para></listitem>
      <listitem><para>The interface methods <function>viewer_editable_save</function>,
      <function>viewer_editable_undo</function> and <function>viewer_editable_redo</function> dereference the interface
      structure to access its associated interface function and call it.
      </para></listitem>
    </itemizedlist>
<informalexample><programlisting>
G_DEFINE_INTERFACE (ViewerEditable, viewer_editable, G_TYPE_OBJECT)

static void
viewer_editable_default_init (ViewerEditableInterface *iface)
{
    /* add properties and signals to the interface here */
}

void
viewer_editable_save (ViewerEditable  *self,
                      GError         **error)
{
  ViewerEditableInterface *iface;

  g_return_if_fail (VIEWER_IS_EDITABLE (self));
  g_return_if_fail (error == NULL || *error == NULL);

  iface = VIEWER_EDITABLE_GET_IFACE (self);
  g_return_if_fail (iface->save != NULL);
  iface->save (self, error);
}

void
viewer_editable_undo (ViewerEditable *self,
                      guint           n_steps)
{
  ViewerEditableInterface *iface;

  g_return_if_fail (VIEWER_IS_EDITABLE (self));

  iface = VIEWER_EDITABLE_GET_IFACE (self);
  g_return_if_fail (iface->undo != NULL);
  iface->undo (self, n_steps);
}

void
viewer_editable_redo (ViewerEditable *self,
                      guint           n_steps)
{
  ViewerEditableInterface *iface;

  g_return_if_fail (VIEWER_IS_EDITABLE (self));

  iface = VIEWER_EDITABLE_GET_IFACE (self);
  g_return_if_fail (iface->redo != NULL);
  iface->redo (self, n_steps);
}
</programlisting></informalexample>
    </para>
  </sect1>
  
  <sect1 id="howto-interface-implement">
    <title>Implementing interfaces</title>
  
    <para>
      Once the interface is defined, implementing it is rather trivial.
    </para>
  
    <para>
      The first step is to define a normal final GObject class exactly as in
      <xref linkend="howto-gobject-header"/>.
    </para>
  
    <para>
      The second step is to implement <type>ViewerFile</type> by defining
      it using
      <function><link linkend="G-DEFINE-TYPE-WITH-CODE:CAPS">G_DEFINE_TYPE_WITH_CODE</link></function>
      and
      <function><link linkend="G-IMPLEMENT-INTERFACE:CAPS">G_IMPLEMENT_INTERFACE</link></function>
      instead of
      <function><link linkend="G-DEFINE-TYPE:CAPS">G_DEFINE_TYPE</link></function>:
<informalexample><programlisting>
static void viewer_file_editable_interface_init (ViewerEditableInterface *iface);

G_DEFINE_TYPE_WITH_CODE (ViewerFile, viewer_file, G_TYPE_OBJECT,
                         G_IMPLEMENT_INTERFACE (VIEWER_TYPE_EDITABLE,
                                                viewer_file_editable_interface_init))
</programlisting></informalexample>
      This definition is very much like all the similar functions seen
      previously. The only interface-specific code present here is the use of
      <function><link linkend="G-IMPLEMENT-INTERFACE:CAPS">G_IMPLEMENT_INTERFACE</link></function>.
    </para>

    <note><para>Classes can implement multiple interfaces by using multiple calls to
    <function><link linkend="G-IMPLEMENT-INTERFACE:CAPS">G_IMPLEMENT_INTERFACE</link></function>
    inside the call to
    <function><link linkend="G-DEFINE-TYPE-WITH-CODE:CAPS">G_DEFINE_TYPE_WITH_CODE</link></function>
    </para></note>
  
    <para>
      <function>viewer_file_editable_interface_init</function>, the interface
      initialization function: inside it every virtual method of the interface
      must be assigned to its implementation:
<informalexample><programlisting>
static void
viewer_file_editable_save (ViewerFile  *self,
                           GError     **error)
{
  g_print ("File implementation of editable interface save method: %s.\n",
           self->filename);
}

static void
viewer_file_editable_undo (ViewerFile *self,
                           guint       n_steps)
{
  g_print ("File implementation of editable interface undo method: %s.\n",
           self->filename);
}

static void
viewer_file_editable_redo (ViewerFile *self,
                           guint       n_steps)
{
  g_print ("File implementation of editable interface redo method: %s.\n",
           self->filename);
}

static void
viewer_file_editable_interface_init (ViewerEditableInterface *iface)
{
  iface->save = viewer_file_editable_save;
  iface->undo = viewer_file_editable_undo;
  iface->redo = viewer_file_editable_redo;
}

static void
viewer_file_init (ViewerFile *self)
{
  /* Instance variable initialisation code. */
}
</programlisting></informalexample>
    </para>
    <para>
      If the object is not of final type, e.g. was declared using
      <function><link linkend="G-DECLARE-DERIVABLE-TYPE:CAPS">G_DECLARE_DERIVABLE_TYPE</link></function>
      then
      <function><link linkend="G-ADD-PRIVATE:CAPS">G_ADD_PRIVATE</link></function>
      macro should be added. The private structure should be declared exactly
      as for a normal derivable object, see <xref linkend="howto-gobject-code"/>.
<informalexample><programlisting>
G_DEFINE_TYPE_WITH_CODE (ViewerFile, viewer_file, G_TYPE_OBJECT,
                         G_ADD_PRIVATE (ViewerFile)
                         G_IMPLEMENT_INTERFACE (VIEWER_TYPE_EDITABLE,
                                                viewer_file_editable_interface_init))
</programlisting></informalexample>
    </para>
  </sect1>
  
  <sect1 id="howto-interface-prerequisite">
    <title>Interface definition prerequisites</title>
  
    <para>
      To specify that an interface requires the presence of other interfaces
      when implemented, GObject introduces the concept of
      <emphasis>prerequisites</emphasis>: it is possible to associate
      a list of prerequisite types to an interface. For example, if
      object A wishes to implement interface I1, and if interface I1 has a
      prerequisite on interface I2, A has to implement both I1 and I2.
    </para>
  
    <para>
      The mechanism described above is, in practice, very similar to
      Java's interface I1 extends interface I2. The example below shows
      the GObject equivalent:
<informalexample><programlisting>
/* Make the ViewerEditableLossy interface require ViewerEditable interface. */
G_DEFINE_INTERFACE (ViewerEditableLossy, viewer_editable_lossy, VIEWER_TYPE_EDITABLE)
</programlisting></informalexample>
      In the <function><link linkend="G-DEFINE-INTERFACE:CAPS">G_DEFINE_INTERFACE</link></function>
      call above, the third parameter defines the prerequisite type. This
      is the GType of either an interface or a class. In this case
      the <type>ViewerEditable</type> interface is a prerequisite of
      <type>ViewerEditableLossy</type>. The code
      below shows how an implementation can implement both interfaces and
      register their implementations:
<informalexample><programlisting>
static void
viewer_file_editable_lossy_compress (ViewerEditableLossy *editable)
{
  ViewerFile *self = VIEWER_FILE (editable);

  g_print ("File implementation of lossy editable interface compress method: %s.\n",
           self->filename);
}

static void
viewer_file_editable_lossy_interface_init (ViewerEditableLossyInterface *iface)
{
  iface->compress = viewer_file_editable_lossy_compress;
}

static void
viewer_file_editable_save (ViewerEditable  *editable,
                           GError         **error)
{
  ViewerFile *self = VIEWER_FILE (editable);

  g_print ("File implementation of editable interface save method: %s.\n",
           self->filename);
}

static void
viewer_file_editable_undo (ViewerEditable *editable,
                           guint           n_steps)
{
  ViewerFile *self = VIEWER_FILE (editable);

  g_print ("File implementation of editable interface undo method: %s.\n",
           self->filename);
}

static void
viewer_file_editable_redo (ViewerEditable *editable,
                           guint           n_steps)
{
  ViewerFile *self = VIEWER_FILE (editable);

  g_print ("File implementation of editable interface redo method: %s.\n",
           self->filename);
}

static void
viewer_file_editable_interface_init (ViewerEditableInterface *iface)
{
  iface->save = viewer_file_editable_save;
  iface->undo = viewer_file_editable_undo;
  iface->redo = viewer_file_editable_redo;
}

static void
viewer_file_class_init (ViewerFileClass *klass)
{
  /* Nothing here. */
}

static void
viewer_file_init (ViewerFile *self)
{
  /* Instance variable initialisation code. */
}

G_DEFINE_TYPE_WITH_CODE (ViewerFile, viewer_file, G_TYPE_OBJECT,
                         G_IMPLEMENT_INTERFACE (VIEWER_TYPE_EDITABLE,
                                                viewer_file_editable_interface_init)
                         G_IMPLEMENT_INTERFACE (VIEWER_TYPE_EDITABLE_LOSSY,
                                                viewer_file_editable_lossy_interface_init))
</programlisting></informalexample>
      It is very important to notice that the order in which interface
      implementations are added to the main object is not random:
      <function><link linkend="g-type-add-interface-static">g_type_add_interface_static</link></function>,
      which is called by
      <function><link linkend="G-DEFINE-INTERFACE:CAPS">G_IMPLEMENT_INTERFACE</link></function>,
      must be invoked first on the interfaces which have no prerequisites and then on
      the others.
    </para>
  </sect1>
  
  <sect1 id="howto-interface-properties">
    <title>Interface properties</title>
  
    <para>
      GObject interfaces can also have
      properties. Declaration of the interface properties is similar to
      declaring the properties of ordinary GObject types as explained in
      <xref linkend="gobject-properties"/>, except that
      <function><link linkend="g-object-interface-install-property">g_object_interface_install_property</link></function>
      is used to declare the properties instead of
      <function><link linkend="g-object-class-install-property">g_object_class_install_property</link></function>.
    </para>
  
    <para>
      To include a property named 'autosave-frequency' of type <type>gdouble</type> in the 
      <type>ViewerEditable</type> interface example code above, we only need to
      add one call in <function>viewer_editable_default_init</function> as shown
      below:
<informalexample><programlisting>
static void
viewer_editable_default_init (ViewerEditableInterface *iface)
{
  g_object_interface_install_property (iface,
                                       g_param_spec_double ("autosave-frequency",
                                                            "Autosave frequency",
                                                            "Frequency (in per-seconds) to autosave backups of the editable content at. "
                                                            "Or zero to disable autosaves.",
                                                            0.0,  /* minimum */
                                                            G_MAXDOUBLE,  /* maximum */
                                                            0.0,  /* default */
                                                            G_PARAM_READWRITE | G_PARAM_STATIC_STRINGS));
}
</programlisting></informalexample>
    </para>
  
    <para>
      One point worth noting is that the declared property wasn't assigned an 
      integer ID. The reason being that integer IDs of properties are used
      only inside the <function>get_property</function> and
      <function>set_property</function> virtual methods. Since interfaces
      declare but do not <emphasis>implement</emphasis> properties, there is no
      need to assign integer IDs to them.
    </para>
    
    <para>
      An implementation declares and defines its properties in the usual
      way as explained in <xref linkend="gobject-properties"/>, except for one
      small change: it can declare the properties of the interface it
      implements using <function><link linkend="g-object-class-override-property">g_object_class_override_property</link></function>
      instead of <function><link linkend="g-object-class-install-property">g_object_class_install_property</link></function>.
      The following code snippet shows the modifications needed in the
      <type>ViewerFile</type> declaration and implementation above:
<informalexample><programlisting>
struct _ViewerFile
{
  GObject parent_instance;

  gdouble autosave_frequency;
};

enum
{
  PROP_AUTOSAVE_FREQUENCY = 1,
  N_PROPERTIES
};

static void
viewer_file_set_property (GObject      *object,
                          guint         prop_id,
                          const GValue *value,
                          GParamSpec   *pspec)
{
  ViewerFile *file = VIEWER_FILE (object);

  switch (prop_id)
    {
    case PROP_AUTOSAVE_FREQUENCY:
      file->autosave_frequency = g_value_get_double (value);
      break;

    default:
      G_OBJECT_WARN_INVALID_PROPERTY_ID (object, prop_id, pspec);
      break;
    }
}

static void
viewer_file_get_property (GObject    *object,
                          guint       prop_id,
                          GValue     *value,
                          GParamSpec *pspec)
{
  ViewerFile *file = VIEWER_FILE (object);

  switch (prop_id)
    {
    case PROP_AUTOSAVE_FREQUENCY:
      g_value_set_double (value, file->autosave_frequency);
      break;

    default:
      G_OBJECT_WARN_INVALID_PROPERTY_ID (object, prop_id, pspec);
      break;
    }
}

static void
viewer_file_class_init (ViewerFileClass *klass)
{
  GObjectClass *object_class = G_OBJECT_CLASS (klass);

  object_class->set_property = viewer_file_set_property;
  object_class->get_property = viewer_file_get_property;

  g_object_class_override_property (object_class, PROP_AUTOSAVE_FREQUENCY, "autosave-frequency");
}
</programlisting></informalexample>
    </para>
  
  </sect1>

  <sect1 id="howto-interface-override">
    <title>Overriding interface methods</title>

    <para>
      If a base class already implements an interface and a derived
      class needs to implement the same interface but needs to override certain
      methods, you must reimplement the interface and set only the interface
      methods which need overriding.
    </para>

    <para>
      In this example, <type>ViewerAudioFile</type> is derived from
      <type>ViewerFile</type>. Both implement the <type>ViewerEditable</type>
      interface. <type>ViewerAudioFile</type> only implements one method of the
      <type>ViewerEditable</type> interface and uses the base class implementation of
      the other.
<informalexample><programlisting>
static void
viewer_audio_file_editable_save (ViewerEditable  *editable,
                                 GError         **error)
{
  ViewerAudioFile *self = VIEWER_AUDIO_FILE (editable);

  g_print ("Audio file implementation of editable interface save method.\n");
}

static void
viewer_audio_file_editable_interface_init (ViewerEditableInterface *iface)
{
  /* Override the implementation of save(). */
  iface->save = viewer_audio_file_editable_save;

  /*
   * Leave iface->undo and ->redo alone, they are already set to the
   * base class implementation.
   */
}

G_DEFINE_TYPE_WITH_CODE (ViewerAudioFile, viewer_audio_file, VIEWER_TYPE_FILE,
                         G_IMPLEMENT_INTERFACE (VIEWER_TYPE_EDITABLE,
                                                viewer_audio_file_editable_interface_init))

static void
viewer_audio_file_class_init (ViewerAudioFileClass *klass)
{
  /* Nothing here. */
}

static void
viewer_audio_file_init (ViewerAudioFile *self)
{
  /* Nothing here. */
}
</programlisting></informalexample>
    </para>

    <para>
      To access the base class interface implementation use
      <function><link linkend="g-type-interface-peek-parent">g_type_interface_peek_parent</link></function>
      from within an interface's <function>default_init</function> function.
    </para>

    <para>
      To call the base class implementation of an interface
      method from a derived class where than interface method has been
      overridden, stash away the pointer returned from
      <function><link linkend="g-type-interface-peek-parent">g_type_interface_peek_parent</link></function>
      in a global variable.
    </para>

    <para>
      In this example <type>ViewerAudioFile</type> overrides the
      <function>save</function> interface method. In its overridden method
      it calls the base class implementation of the same interface method.
<informalexample><programlisting>
static ViewerEditableInterface *viewer_editable_parent_interface = NULL;

static void
viewer_audio_file_editable_save (ViewerEditable  *editable,
                                 GError         **error)
{
  ViewerAudioFile *self = VIEWER_AUDIO_FILE (editable);

  g_print ("Audio file implementation of editable interface save method.\n");

  /* Now call the base implementation */
  viewer_editable_parent_interface->save (editable, error);
}

static void
viewer_audio_file_editable_interface_init (ViewerEditableInterface *iface)
{
  viewer_editable_parent_interface = g_type_interface_peek_parent (iface);

  iface->save = viewer_audio_file_editable_save;
}

G_DEFINE_TYPE_WITH_CODE (ViewerAudioFile, viewer_audio_file, VIEWER_TYPE_FILE,
                         G_IMPLEMENT_INTERFACE (VIEWER_TYPE_EDITABLE,
                                                viewer_audio_file_editable_interface_init))

static void
viewer_audio_file_class_init (ViewerAudioFileClass *klass)
{
  /* Nothing here. */
}

static void
viewer_audio_file_init (ViewerAudioFile *self)
{
  /* Nothing here. */
}
</programlisting></informalexample>
    </para>

  </sect1>

</chapter>
<!-- End Howto Interfaces -->

<chapter id="howto-signals">
  <title>How to create and use signals</title>

  <para>
    The signal system in GType is pretty complex and
    flexible: it is possible for its users to connect at runtime any
    number of callbacks (implemented in any language for which a binding
    exists)
    <footnote>
      <para>A Python callback can be connected to any signal on any
      C-based GObject, and vice versa, assuming that the Python object
      inherits from GObject.</para>
    </footnote>
    to any signal and to stop the emission of any signal at any 
    state of the signal emission process. This flexibility makes it
    possible to use GSignal for much more than just emitting signals to
    multiple clients.
  </para>

  <sect1 id="howto-simple-signals">
    <title>Simple use of signals</title>

    <para>
      The most basic use of signals is to implement event
      notification. For example, given a <type>ViewerFile</type> object with
      a <function>write</function> method, a signal could be emitted whenever
      the file is changed using that method.
      The code below shows how the user can connect a callback to the
      "changed" signal.
<informalexample><programlisting>
file = g_object_new (VIEWER_FILE_TYPE, NULL);

g_signal_connect (file, "changed", (GCallback) changed_event, NULL);

viewer_file_write (file, buffer, strlen (buffer));
</programlisting></informalexample>
    </para>
    
    <para>
      The <type>ViewerFile</type> signal is registered in the
      <function>class_init</function> function:
<informalexample><programlisting>
file_signals[CHANGED] = 
  g_signal_newv ("changed",
                 G_TYPE_FROM_CLASS (object_class),
                 G_SIGNAL_RUN_LAST | G_SIGNAL_NO_RECURSE | G_SIGNAL_NO_HOOKS,
                 NULL /* closure */,
                 NULL /* accumulator */,
                 NULL /* accumulator data */,
                 NULL /* C marshaller */,
                 G_TYPE_NONE /* return_type */,
                 0     /* n_params */,
                 NULL  /* param_types */);
</programlisting></informalexample>
      and the signal is emitted in <function>viewer_file_write</function>:
<informalexample><programlisting>
void
viewer_file_write (ViewerFile   *self,
                   const guint8 *buffer,
                   gsize         size)
{
  g_return_if_fail (VIEWER_IS_FILE (self));
  g_return_if_fail (buffer != NULL || size == 0);

  /* First write data. */

  /* Then, notify user of data written. */
  g_signal_emit (self, file_signals[CHANGED], 0 /* details */);
}
</programlisting></informalexample>
      As shown above, the details parameter can safely be set to zero if no
      detail needs to be conveyed. For a discussion of what it can be used for,
      see <xref linkend="signal-detail"/>
    </para>

    <para>
      The C signal marshaller should always be <literal>NULL</literal>, in which
      case the best marshaller for the given closure type will be chosen by
      GLib. This may be an internal marshaller specific to the closure type, or
      <function>g_cclosure_marshal_generic</function>, which implements generic
      conversion of arrays of parameters to C callback invocations. GLib used to
      require the user to write or generate a type-specific marshaller and pass
      that, but that has been deprecated in favour of automatic selection of
      marshallers.
    </para>

    <para>
      Note that <function>g_cclosure_marshal_generic</function> is slower than
      non-generic marshallers, so should be avoided for performance critical
      code. However, performance critical code should rarely be using signals
      anyway, as emitting a signal blocks on emitting it to all listeners, which
      has potentially unbounded cost.
    </para>
  </sect1>
</chapter>
</part>