mutter/doc/cookbook/animations.xml

<!DOCTYPE chapter PUBLIC "-//OASIS//DTD DocBook XML V4.2//EN"
  "http://www.oasis-open.org/docbook/xml/4.2/docbookx.dtd">

<chapter id="animations" xmlns:xi="http://www.w3.org/2003/XInclude">
  <title>Animations</title>

  <epigraph>
    <attribution>Walt Disney</attribution>
    <para>Animation can explain whatever the mind of man can conceive.</para>
  </epigraph>

  <section id="animations-introduction">
    <title>Introduction</title>

    <para>Clutter actors have a variety of <emphasis>properties</emphasis>
    (position, size, rotation in 3D space, scale, opacity) which govern
    their visual appearance in the UI. They may also have
    <emphasis>constraints</emphasis> on how they are aligned
    and/or positioned relative to each other.</para>

    <para>The Clutter animation API provides a means of changing
    properties and constraints as a function of time: moving, scaling,
    rotating, changing opacity and colour, modifying postional
    constraints, etc.</para>

    <note><para>Clutter also makes it possible to animate non-visual
    properties if desired.</para></note>

    <section>
      <title>High level overview</title>

      <para>Here are the main concepts behind animation in Clutter:</para>

      <itemizedlist>
        <listitem>
          <para>An <emphasis>animation</emphasis> changes one or more
          properties of one or more actors over time: their rotation in
          a particular dimension (<varname>x</varname>, <varname>y</varname>,
          <varname>z</varname>), scale, size, opacity etc.</para>
        </listitem>
        <listitem>
          <para>An animation has an associated <emphasis>timeline</emphasis>.
          Think of this as analogous to the "thing" you're controlling when
          you watch a video on the internet: it's what you control with
          the play/pause button and what is measured by the bar
          showing how far through the video you are. As with the
          controls on a video player, you can play/pause/skip a Clutter
          timeline; you can also rewind it, loop it, and play it
          backwards.</para>
          <note>
            <para>If a timeline is reversed, the progress along the
            timeline is still measured the same way as it is in the forward
            direction: so if you start from the end of the timeline and run
            it backwards for 75% of its length, the progress is reported
            as 0.25 (i.e. 25% of the way from the start of the
            timeline).</para>
          </note>
        </listitem>
        <listitem>
          <para>The <emphasis>duration</emphasis> of a timeline
          (e.g. 500 milliseconds, 1 second, 10 seconds) specifies how
          long its animation will last. The timeline can be inspected
          to find out how much of it has elapsed, either as a value in
          milliseconds or as a fraction (between 0 and 1) of the total
          length of the timeline.</para>
        </listitem>
        <listitem>
          <para>An animation is divided into <emphasis>frames</emphasis>.
          The number of frames which make up the animation isn't
          constant: it depends on various factors, like how powerful
          your machine is, the state of the drivers for your hardware,
          and the load on he system. So you won't always get the same
          number of frames in an animation of a particular duration.</para>
        </listitem>
        <listitem>
          <para>The change to a property in an animation occurs over
          the course of the timeline: the start value of the property
          heads toward some target value. When it reaches the end of
          the timeline, the property should have reached the target
          value.</para>
        </listitem>
        <listitem>
          <para>Exactly how the property changes over the course of the
          timeline is governed by an <emphasis>alpha</emphasis>. This
          is the trickiest idea to explain, so it has its own section
          below.</para>
        </listitem>
      </itemizedlist>
    </section>

    <section id="animations-introduction-alphas">
      <title>Alphas</title>

      <para>An alpha is generated for each frame of the animation.
      The alpha varies between -1.0 and 2.0, and changes during the
      course of the animation's timeline; ideally, the value should
      start at 0.0 and reach 1.0 by the end of the timeline.</para>

      <para>The alpha for any given frame of the animation is determined
      by an <emphasis>alpha function</emphasis>. Usually, the alpha
      function will return a value based on progress along the timeline.
      However, the alpha function doesn't have to respect or pay
      attention to the timeline: it can be entirely random if desired.</para>

      <para>To work out the value of a property at a given frame
      somewhere along the timeline for a given alpha:</para>

      <orderedlist>
        <listitem>
          <para>Determine the difference between the start value and
          the target end value for the property.</para>
        </listitem>
        <listitem>
          <para>Multiply the difference by the alpha for the current
          frame.</para>
        </listitem>
        <listitem>
          <para>Add the result to the start value.</para>
        </listitem>
      </orderedlist>

      <para>The shape of the plot of the alpha function over time is
      called its <emphasis>easing mode</emphasis>. Clutter provides
      various modes ranging from <constant>CLUTTER_LINEAR</constant>
      (the alpha value is equal to progress along the timeline),
      to modes based on various polynomial and exponential functions,
      to modes providing elastic and bounce shapes. See the
      ClutterAlpha documentation for examples of the shapes produced
      by these functions. There is also a good interactive demo
      of the modes on
      <ulink url="http://www.robertpenner.com/easing/easing_demo.html">Robert Penner's site</ulink>.
      </para>

      <para>Most of the time, you can use the built-in Clutter easing
      modes to get the kind of animation effect you want. However,
      in some cases you may want to provide your own alpha function.
      Here's an example (based on the quintic ease in mode from
      <filename>clutter-alpha.c</filename>):</para>

      <informalexample>
        <programlisting>
<![CDATA[
static gdouble
_alpha_ease_in_sextic (ClutterAlpha *alpha,
                       gpointer      dummy G_GNUC_UNUSED)
 {
  ClutterTimeline *timeline = clutter_alpha_get_timeline (alpha);
  gdouble p = clutter_timeline_get_progress (timeline);

  return p * p * p * p * p * p;
}
]]>
        </programlisting>
      </informalexample>

      <para>An alpha function just has to have a specified method
      signature and return a <type>gdouble</type> value when called.
      As stated above, you'd typically base the return value on the
      timeline progress; the function above shows how you get the
      timeline associated with the alpha, so you can apply the alpha
      function to it.</para>
    </section>

    <section id="animations-introduction-api">
      <title>Clutter's animation API</title>

      <para>All of the animation approaches in Clutter use the same
      basic underpinnings (as explained above), but the API provides
      varying levels of abstraction and/or ease of use on top of those
      underpinnings.</para>

      <itemizedlist>
        <listitem>
          <para><emphasis>Implicit animations</emphasis> (created using
          <function>clutter_actor_animate()</function> and related
          functions) are useful where you want to apply
          a simple or one-off animation to an actor. They enable you
          to animate one or more properties using a single easing mode;
          however, you only specify the target values for the properties
          you're animating, not the start values.</para>
        </listitem>
        <listitem>
          <para><emphasis>ClutterAnimator</emphasis> provides support
          for declarative animations (defined using <type>ClutterScript</type>).
          You can animate multiple actors with this approach, and
          have more control over the easing modes used during an
          animation: while implicit animations only allow a single
          easing mode for all properties, <type>ClutterAnimator</type>
          supports <emphasis>multiple</emphasis> easing modes for
          <emphasis>each</emphasis> property; <emphasis>key frames</emphasis>
          are used to indicate where in the animation each easing mode
          should be applied.</para>
        </listitem>
        <listitem>
          <para><emphasis>ClutterState</emphasis> enables you to describe
          <emphasis>states</emphasis>: property values across one or
          more actors, plus the easing modes used to transition to
          those values. It can also be combined with <type>ClutterAnimator</type>
          for finer grained definition of transitions if desired.</para>
          <para>States are particularly useful if you need actors to
          animate between a known set of positions/sizes/opacities etc.
          during their lifecycles (e.g. animating a list of items in
          a menu, or for animations in a picture viewer where you
          click on thumbnails to display a full view of a photograph).</para>
        </listitem>
      </itemizedlist>

      <para>The recipes in this section show when and where it is
      appropriate to use each of these approaches.</para>

    </section>

  </section>

  <section id="animations-inversion">
    <title>Inverting Animations</title>

    <section>
      <title>Problem</title>

      <para>You want to have an animation exactly mirroring another one
      that you just played.</para>
    </section>

    <section>
      <title>Solution</title>

      <para>Reverse the direction of the <type>ClutterTimeline</type>
      associated with the animation.</para>

      <para>For example, here's how to invert an implicit
      animation which moves an actor along the <varname>x</varname>
      axis. The direction of the animation is inverted when the
      movement along the <varname>x</varname> axis is completed; it is
      also inverted if the mouse button is pressed on the actor.</para>

      <para>First, set up the animation:</para>

      <informalexample>
        <programlisting>
<![CDATA[
ClutterAnimation *animation;

/*
 * animate actor to x = 300.0;
 * the implicit animation functions return a ClutterAnimation
 * which we can use to invert the timeline
 */
animation = clutter_actor_animate (actor,
                                   CLUTTER_EASE_IN_OUT_CUBIC,
                                   2000,
                                   "x", 300.0,
                                   NULL);

/* callback for when the animation completes */
g_signal_connect (animation,
                  "completed",
                  G_CALLBACK (_animation_done_cb),
                  NULL);

/*
 * callback for when the mouse button is pressed on the actor;
 * note the animation is passed as user data, so we can
 * get at the timeline
 */
g_signal_connect (actor,
                  "button-press-event",
                  G_CALLBACK (_on_click_cb),
                  animation);
]]>
        </programlisting>
      </informalexample>

      <para>Next, add a function for inverting the timeline:</para>

      <informalexample>
        <programlisting>
<![CDATA[
static void
_invert_timeline (ClutterTimeline *timeline)
{
  ClutterTimelineDirection direction = clutter_timeline_get_direction (timeline);

  if (direction == CLUTTER_TIMELINE_FORWARD)
    direction = CLUTTER_TIMELINE_BACKWARD;
  else
    direction = CLUTTER_TIMELINE_FORWARD;

  clutter_timeline_set_direction (timeline, direction);
}
]]>
        </programlisting>
      </informalexample>

      <para>Then add a function which calls <function>_invert_timeline</function>
      when the animation completes. More importantly, the callback should
      stop emission of the "completed" signal by the animation. This
      prevents the <type>ClutterAnimation</type> underlying the implicit
      animation from being unreferenced; which in turn allows it to be
      inverted:</para>

      <informalexample>
        <programlisting>
<![CDATA[
static void
_animation_done_cb (ClutterAnimation *animation,
                    gpointer          user_data)
{
  /* stop the completed signal before the ClutterAnimation is unreferenced */
  g_signal_stop_emission_by_name (animation, "completed");

  /* invert the timeline associated with the animation */
  ClutterTimeline *timeline = clutter_animation_get_timeline (animation);
  _invert_timeline (timeline);
}
]]>
        </programlisting>
      </informalexample>

      <para>Finally, the click callback function uses the same
      <function>_invert_timeline</function> function if the animation
      is playing; but if the animation is stopped, it will
      start it instead:</para>

      <informalexample>
        <programlisting>
<![CDATA[
static void
_on_click_cb (ClutterActor *actor,
              ClutterEvent *event,
              gpointer      user_data)
{
  ClutterAnimation *animation = (ClutterAnimation *)user_data;

  ClutterTimeline *timeline = clutter_animation_get_timeline (animation);

  if (clutter_timeline_is_playing (timeline))
    {
      _invert_timeline (timeline);
    }
  else
    {
      clutter_timeline_start (timeline);
    }
}
]]>
        </programlisting>
      </informalexample>

    </section>

    <section>
      <title>Discussion</title>

      <para>If you are using <type>ClutterAnimator</type> rather than
      implicit animations, <function>clutter_animator_get_timeline()</function>
      enables you to get the underlying timeline; you could then use
      the techniques shown above to invert it.</para>

      <para><type>ClutterState</type> enables a different approach
      to "inverting" an animation: rather than having a single animation
      which you invert, you would define two or more
      <emphasis>keys</emphasis> for an actor (or set of actors) and
      transition between them.</para>

      <para>For the example above, you would define two keys:
      one for the actor's initial position; and a second for the actor
      at <code>x = 300.0</code>. You would also define the
      transition between them: 2000 milliseconds with a
      <constant>CLUTTER_EASE_IN_OUT_CUBIC</constant> easing mode.</para>

      <para>With the states defined, you would then use
      <function>clutter_state_set_state()</function> inside callbacks to
      animate the actor between the two <varname>x</varname> positions.
      Behind the scenes, <type>ClutterState</type> would handle the
      animations and timelines for you.</para>

    </section>

  </section>

  <section id="animations-fading">
    <title>Fading an actor out of or into view</title>

    <section>
      <title>Problem</title>

      <para>You want to animate an actor so that it fades out of or into
      view.</para>
    </section>

    <section>
      <title>Solution</title>

      <para>Animate the actor's opacity property.</para>

      <para>You can do this using any of the approaches provided
      by the animation API. Here's how to fade out an actor (until it's
      completely transparent) using implicit animations:</para>

      <informalexample>
        <programlisting>
<![CDATA[
/* fade out actor over 4000 milliseconds */
clutter_actor_animate (actor,
                       CLUTTER_EASE_OUT_CUBIC,
                       4000,
                       "opacity", 0,
                       NULL);
]]>
        </programlisting>
      </informalexample>

      <para>Here's an example of a rectangle fading out using this
      animation:</para>

      <inlinemediaobject>
        <videoobject>
          <videodata fileref="videos/animations-fading-out.ogv"/>
        </videoobject>
        <alt>
          <para>Video showing an actor fading out using implicit
          animations</para>
        </alt>
      </inlinemediaobject>

      <para><constant>CLUTTER_EASE_OUT_CUBIC</constant> is one of the
      Clutter easing modes; see
      <link linkend="animations-introduction-alphas">the introduction</link>
      for more details about what these are and how to choose one.</para>

      <para>Here's an example of the transitions you could use to
      fade an actor in and out using <type>ClutterState</type>:</para>

      <informalexample>
        <programlisting>
<![CDATA[
ClutterState *transitions = clutter_state_new ();

/* all transitions last for 2000 milliseconds */
clutter_state_set_duration (transitions, NULL, NULL, 2000);

/* transition from any state to "fade-out" state */
clutter_state_set (transitions,
                   NULL,        /* from state (NULL means "any") */
                   "fade-out",  /* to state */
                   actor, "opacity", CLUTTER_EASE_OUT_QUAD, 0,
                   NULL);

/* transition from any state to "fade-in" state */
clutter_state_set (transitions, NULL, "fade-in",
                   actor, "opacity", CLUTTER_EASE_OUT_QUAD, 255,
                   NULL);

/* put the actor into the "fade-out" state with no animation */
clutter_state_warp_to_state (transitions, "fade-out");
]]>
        </programlisting>
      </informalexample>

      <para>You would then trigger an animated state change as events
      occur in the application (e.g. mouse button clicks):</para>

      <informalexample>
        <programlisting>
<![CDATA[
clutter_state_set_state (transitions, "fade-in");
]]>
        </programlisting>
      </informalexample>

      <para>Here's an example of this animation fading in then out again:</para>

      <inlinemediaobject>
        <videoobject>
          <videodata fileref="videos/animations-fading-in-then-out.ogv"/>
        </videoobject>
        <alt>
          <para>Video showing an actor fading in then out using
          <type>ClutterState</type></para>
        </alt>
      </inlinemediaobject>

      <note>
        <para><type>ClutterState</type> is most useful where you
        need to animate an actor backwards and forwards between multiple
        states (e.g. fade an actor in and out of view). Where you just
        want to fade an actor in or out once,
        <function>clutter_actor_animate()</function> is adequate.</para>
      </note>

    </section>

    <section>
      <title>Discussion</title>

      <para>Reducing an actor's transparency to zero does not make it
      inactive: the actor will still be reactive even if it's not
      visible (responding to key events, mouse clicks etc.).
      To make it really "disappear", you could use
      <function>clutter_actor_hide()</function> once you'd made the actor
      fully transparent.</para>

    </section>

  </section>

  <section id="animations-rotating">
    <title>Rotating an actor</title>

    <section>
      <title>Problem</title>

      <para>You want to animate rotation of an actor. Some example cases
      where you might want to do this:</para>

      <itemizedlist>
        <listitem>
          <para>To rotate an image so it's the right way up for
          viewing.</para>
        </listitem>
        <listitem>
          <para>To make actors more or less prominent, rotating them
          towards or away from the view point.</para>
        </listitem>
        <listitem>
          <para>To turn an actor "around" and display different UI
          elements "behind" it.</para>
        </listitem>
      </itemizedlist>
    </section>

    <section>
      <title>Solution</title>

      <para>Animate one of the <varname>rotation-angle-(x|y|z)</varname>
      properties of the actor.</para>

      <para>The most "obvious" (and probably most commonly used) rotation is
      in the <emphasis>z axis</emphasis> (parallel
      to the 2D surface of the UI). The other rotation axes
      (<emphasis>x</emphasis> and <emphasis>y</emphasis>)
      are less obvious, as they rotate the actor in the depth dimension,
      "away from" or "towards" the view point.</para>

      <para>Examples of each type of rotation are given below. While the
      examples use <link linkend="animations-introduction-api">implicit
      animations</link>, it is also possible to use
      <type>ClutterAnimator</type> and <type>ClutterState</type> to animate
      rotations: see <link linkend="animations-rotating-example">the
      full example at the end of this recipe</link> for some
      <type>ClutterState</type> code.</para>

      <note>
        <para>I've added an inaccurate (but hopefully useful) metaphor to
        each rotation axis ("wheel", "letter box", "door"), to make it
        easier to remember the effect you get from animating in that axis
        (and when the rotation center is inside the actor).</para>
      </note>

      <para><emphasis>Rotating on the z axis</emphasis> ("wheel")</para>

      <informalexample>
        <programlisting>
<![CDATA[
clutter_actor_animate (actor,
                   CLUTTER_LINEAR,            /* easing mode */
                   1000,                      /* duration in milliseconds */
                   "rotation-angle-z", 90.0,  /* target rotation angle in degrees */
                   NULL);
]]>
        </programlisting>
      </informalexample>

      <para>The above code animating a texture:</para>

      <inlinemediaobject>
        <videoobject>
          <videodata fileref="videos/animations-rotating-z-90.ogv"/>
        </videoobject>
        <alt>
          <para>Video showing an actor rotating to 90 degrees on the
          z axis</para>
        </alt>
      </inlinemediaobject>

      <para>By default, the center of the rotation is derived from
      the anchor point of the actor; unless you've changed the anchor
      point, the default is the top-left corner of the actor. See the
      Discussion section below for more about setting the rotation center.</para>

      <note>
        <para>An animated rotation moves an actor <emphasis>to</emphasis>
        the specified rotation angle; it <emphasis>does not</emphasis>
        increment or decrement the actor's current rotation angle by
        the amount specified.</para>
      </note>

      <para><emphasis>Rotating on the <code>x</code> axis</emphasis>
      ("letter box")</para>

      <informalexample>
        <programlisting>
<![CDATA[
clutter_actor_animate (actor,
                   CLUTTER_LINEAR,
                   1000,
                   "rotation-angle-x", -45.0,
                   NULL);
]]>
        </programlisting>
      </informalexample>

      <para>The above code animating a texture:</para>

      <inlinemediaobject>
        <videoobject>
          <videodata fileref="videos/animations-rotating-x-minus-45.ogv"/>
        </videoobject>
        <alt>
          <para>Video showing an actor rotating to -45 degrees on the
          x axis</para>
        </alt>
      </inlinemediaobject>

      <para>Notice how the texture rotates away from the view point,
      and also how perspective effects are applied (as the actor is rotating
      "into" the depth dimension).</para>

      <para><emphasis>Rotating on the <code>y</code> axis</emphasis>
      ("door")</para>

      <informalexample>
        <programlisting>
<![CDATA[
clutter_actor_animate (actor,
                   CLUTTER_LINEAR,
                   1000,
                   "rotation-angle-y", 45.0,
                   NULL);
]]>
        </programlisting>
      </informalexample>

      <para>The above code animating a texture:</para>

      <inlinemediaobject>
        <videoobject>
          <videodata fileref="videos/animations-rotating-y-45.ogv"/>
        </videoobject>
        <alt>
          <para>Video showing an actor rotating to 45 degrees on the
          y axis</para>
        </alt>
      </inlinemediaobject>

      <para>Again, the rotation is into the depth dimension, so
      you get perspective effects.</para>

    </section>

    <section>
      <title>Discussion</title>

      <para>It can sometimes be difficult to predict exactly
      how a particular rotation animation will appear when applied.
      Often the only way to find out is to experiment. However,
      the sections below outline some of the most common factors which
      affect animated rotations, with the aim of minimising the
      experimentation you need to do.</para>

      <section>
        <title>Setting the rotation center for an animation</title>

        <para>The examples in the previous section used the default
        center of rotation for each axis. However, it is possible to
        change the rotation center for an axis, in turn changing
        the appearance of the animation.</para>

        <note>
          <para>Rotation center coordinates are relative to the
          actor's coordinates, not to the coordinates of the actor's
          container or the stage.</para>
        </note>

        <section>
          <title>Setting a rotation center inside an actor</title>

          <para>You can set the center for rotation on the x or y axes
          like this:</para>

          <informalexample>
            <programlisting>
<![CDATA[
/*
* only required for y axis rotation;
* here set to the mid point of the actor's y axis
*/
gfloat x_center = clutter_actor_get_height (actor) * 0.5;

/*
* only required for x axis rotation;
* here set to the mid point of the actor's x axis
*/
gfloat y_center = clutter_actor_get_width (actor) * 0.5;

/*
* depth for the rotation center: positive numbers
* are closer to the view point, negative ones
* are further away
*/
gfloat z_center = 0.0;

/* set rotation center */
clutter_actor_set_rotation (actor,
                          CLUTTER_X_AXIS,  /* or CLUTTER_Y_AXIS */
                          0.0,             /* set the rotation to this angle */
                          x_center,
                          y_center,
                          z_center);
]]>
            </programlisting>
          </informalexample>

          <para>Because z axis rotations are more common, Clutter
          provides some convenience functions to set the rotation
          center for this axis:</para>

          <informalexample>
            <programlisting>
<![CDATA[
clutter_actor_set_z_rotation_from_gravity (actor,
                                         0.0,
                                         CLUTTER_GRAVITY_CENTER);
]]>
            </programlisting>
          </informalexample>

          <para><constant>CLUTTER_GRAVITY_CENTER</constant> makes the
          center of the actor the rotation center for
          the z axis. See the <type>ClutterGravity</type> enumeration for
          acceptable values for this parameter.</para>

          <note>
            <para>Setting the rotation center for the z axis using gravity
            is recommended, as Clutter will automatically recompute the
            rotation center if the actor's size changes. For the x and y
            axes, you have to do this computation yourself if you
            want an actor's center of rotation to stay in the same place
            if it is resized.</para>
          </note>

          <para>Rotation on the x axis around an actor's center:</para>

          <inlinemediaobject>
            <videoobject>
              <videodata fileref="videos/animations-rotating-x-centered.ogv"/>
            </videoobject>
            <alt>
              <para>Video showing an actor rotating around its center
              on the x axis</para>
            </alt>
          </inlinemediaobject>

          <para>Rotation on the y axis around an actor's center:</para>

          <inlinemediaobject>
            <videoobject>
              <videodata fileref="videos/animations-rotating-y-centered.ogv"/>
            </videoobject>
            <alt>
              <para>Video showing an actor rotating around its center
              on the y axis</para>
            </alt>
          </inlinemediaobject>

          <para>Rotation on the z axis around an actor's center:</para>

          <inlinemediaobject>
            <videoobject>
              <videodata fileref="videos/animations-rotating-z-centered.ogv"/>
            </videoobject>
            <alt>
              <para>Video showing an actor rotating around its center
              on the z axis</para>
            </alt>
          </inlinemediaobject>

        </section>

        <section>
          <title>Setting the rotation center outside an actor</title>

          <para>Rather than rotating the actor around a point inside
          itself, the rotation center can be moved to a position
          outside the actor. (In the case of the z axis,
          any rotation center setting is outside the actor as its depth
          is 0.) When animated, the actor will describe an arc around the
          rotation center, as if it's swinging from an invisible thread.</para>

          <para>The same code as shown above can be used to set the
          rotation center: just set the rotation center coordinates to
          negative numbers (outside the actor). However, you can't use the
          gravity functions if the rotation center falls outside an actor.</para>

          <para>For example, here's a rotation to -180 degrees in the x
          axis, with the y rotation center set to -96 (the same as the height
          of the actor):</para>

          <inlinemediaobject>
            <videoobject>
              <videodata fileref="videos/animations-rotating-x-minus-180-with-y-minus-96.ogv"/>
            </videoobject>
            <alt>
              <para>Video showing an actor rotating to -180 degrees on
              the x axis with y rotation center set to -96</para>
            </alt>
          </inlinemediaobject>

          <para>Similarly, moving the z rotation center (for a rotation
          in the x or y axis) will cause the actor to swing "into" or "out
          of" the UI. Its final apparent size may be different, as it could
          reach a different depth in the UI by the end of the
          animation.</para>

          <para>For example, here's a rotation to -180 in the x axis,
          with the z rotation center set to -96 (the same as the height
          of the actor):</para>

          <inlinemediaobject>
            <videoobject>
              <videodata fileref="videos/animations-rotating-x-minus-180-with-z-minus-96.ogv"/>
            </videoobject>
            <alt>
              <para>Video showing an actor rotating to -180 degrees on
              the x axis with z rotation center set to -96</para>
            </alt>
          </inlinemediaobject>

          <para>The apparent final size of the actor is reduced, as it
          has rotated away from the view point.</para>

        </section>

      </section>

      <section id="animations-rotating-discussion-direction">
        <title>Direction of rotation</title>

        <para>The apparent direction of an animated rotation depends on
        two things:</para>

        <orderedlist>
          <listitem>
            <para>Whether the angle of rotation is positive or negative.</para>
          </listitem>
          <listitem>
            <para>The rotation of the container(s) the actor is inside.</para>
          </listitem>
        </orderedlist>

        <para>In the case of the sign of the rotation, here's what
        happens for each axis and rotation angle sign (positive or
        negative).</para>

        <informaltable>
          <thead>
            <tr>
              <th>Axis</th>
              <th>Sign of rotation angle</th>
              <th>Effect on actor</th>
            </tr>
          </thead>
          <tbody>
            <tr>
              <td>z</td>
              <td>+</td>
              <td>
                Clockwise spin about the <code>x,y</code> center of
                rotation.
              </td>
            </tr>
            <tr>
              <td>z</td>
              <td>-</td>
              <td>
                Anti-clockwise spin about the <code>x,y</code>
                center of rotation.
              </td>
            </tr>
            <tr>
              <td>x</td>
              <td>+</td>
              <td>
                The top swings away from the view point and the
                bottom swings towards it. If y rotation center == 0,
                the top is fixed; if y rotation center == the actor's
                height, the bottom is fixed.
              </td>
            </tr>
            <tr>
              <td>x</td>
              <td>-</td>
              <td>
                The bottom swings away from the view point and the
                top swings towards it. If y rotation center == 0,
                the top is fixed; if y rotation center == the actor's
                height, the bottom is fixed.
              </td>
            </tr>
            <tr>
              <td>y</td>
              <td>+</td>
              <td>
                The right-hand side swings away from the view point and
                the left-hand side swings towards it. When x rotation
                center == 0, the left-hand side if fixed; when x
                rotation center == the actor's width, the right-hand
                side is fixed.
              </td>
            </tr>
            <tr>
              <td>y</td>
              <td>-</td>
              <td>
                The right-hand side swings towards the view point and
                the left-hand side swings away from it. When x rotation
                center == 0, the left-hand side if fixed; when x
                rotation center == the actor's width, the right-hand
                side is fixed.
              </td>
            </tr>
          </tbody>
        </informaltable>

        <para>If an actor's container is rotated, this may affect the
        appearance of rotation animations applied to the actor. In
        particular, if an actor's container has been rotated
        by 180 degrees in one axis, the direction of that actor's
        rotation may appear reversed.</para>

        <para>For example, the video below shows an actor being animated
        to 90 degrees on the z axis, then back to 0 degrees;
        the actor's container is then rotated by 180 degrees in the y
        axis; then the same rotation 90 degree rotation is applied
        to the actor again. Note that the first time the animation
        is applied, the rotation is clockwise; but the second time (as
        the actor is effectively "reversed"), it is anti-clockwise.</para>

        <inlinemediaobject>
          <videoobject>
            <videodata fileref="videos/animations-rotating-container-reverses-direction.ogv"/>
          </videoobject>
          <alt>
            <para>Video showing how an actor's apparent rotation is
            affected by the rotation of its parent</para>
          </alt>
        </inlinemediaobject>

      </section>

      <section>
        <title>Apparent vs. actual rotation</title>

        <para>There is a difference between an actor's <emphasis>apparent</emphasis>
        rotation (how much an actor appears to be rotating, from the
        perspective of someone looking at the UI) and its
        <emphasis>actual</emphasis> rotation (how much that actor is
        really rotating).</para>

        <para>For example, if you rotate an actor and its container
        simultaneously, each by 90 degrees in the same direction, the
        actor will appear to have rotated by 180 degrees by the end
        of the animation. However, calling the
        <function>clutter_actor_get_rotation()</function> function
        for that axis on the actor still returns a rotation of 90
        degrees.</para>
      </section>

      <section>
        <title>Orientation of rotation axes</title>

        <para>The rotation axes remain fixed in the same place on
        the actor regardless of its rotation, even though from the viewer's
        perspective they may appear to move.</para>

        <para>For example, when rotation in the z axis is 0 degrees,
        the actor's x axis is horizontal (across the UI) from both the
        actor's and the viewer's perspective. However, if you rotate the
        actor by 90 degrees in the z axis, the x axis is now vertical from
        <emphasis>the viewer's</emphasis> perspective, but still horizontal
        across the actor from <emphasis>the actor's</emphasis>
        perspective.</para>
      </section>

    </section>

    <section>
      <title>Full example</title>

      <example id="animations-rotating-example">
        <title>Rotating an actor around x, y, and z axes using
        <type>ClutterState</type></title>
        <programlisting>
<xi:include href="examples/animations-rotating.c" parse="text">
  <xi:fallback>a code sample should be here... but isn't</xi:fallback>
</xi:include>
        </programlisting>
      </example>
    </section>

  </section>

  <section id="animations-complex">
    <title>Creating complex animations with
    <type>ClutterAnimator</type></title>

    <section>
      <title>Problem</title>

      <para>You want to create a complex animation involving one or more
      actors. The animation will consist of a sequence of transitions
      over multiple properties on each actor.</para>

      <para>An example might be moving several actors between points,
      with different types of movement for each part of the path, while
      transforming each actor (e.g. scaling or rotating it).</para>

    </section>

    <section>
      <title>Solution</title>

      <para>Use a <type>ClutterAnimator</type> to define the animation.</para>

      <para>Because there are many complex animations you
      <emphasis>could</emphasis> implement, the example below does
      this:</para>

      <inlinemediaobject>
        <videoobject>
          <videodata fileref="videos/animations-complex.ogv"/>
        </videoobject>
        <alt>
          <para>Video showing a complex animation of an actor
          using <type>ClutterAnimator</type></para>
        </alt>
      </inlinemediaobject>

      <para>Although this uses a single actor, the animation is complex
      enough to make it difficult to implement with implicit animations
      or <type>ClutterState</type> (see
      <link linkend="animations-complex-why-clutteranimator">the Discussion
      section</link> for reasons why).</para>

      <para>Here is a JSON definition of the stage, actors, and
      the <type>ClutterAnimator</type> for this
      animation:</para>

      <example id="animations-complex-example-1">
        <title>JSON definition of a complex animation using
        <type>ClutterAnimator</type></title>
        <programlisting>
<xi:include href="examples/animations-complex.json" parse="text">
  <xi:fallback>a code sample should be here... but isn't</xi:fallback>
</xi:include>
        </programlisting>
      </example>

      <note>
        <para>The core to understanding this example is understanding
        how to define keys for a <type>ClutterAnimator</type>. As
        this is an involved topic, further explanation
        is given in <link linkend="animations-complex-discussion-keys">the
        Discussion section</link>.</para>
      </note>

      <para>The program for loading this JSON definition from a file
      is as follows:</para>

      <example id="animations-complex-example-2">
        <title>Simple program for loading a JSON script;
        any key press starts the animation</title>
        <programlisting>
<xi:include href="examples/animations-complex.c" parse="text">
  <xi:fallback>a code sample should be here... but isn't</xi:fallback>
</xi:include>
        </programlisting>
      </example>

      <note>
        <para>It is also possible to use the <type>ClutterAnimator</type>
        C API to define keys for an animation, but this will
        typically be much more verbose than the JSON equivalent.</para>

        <para>One other advantage of JSON is that it is much simpler
        to tweak and test an animation, as you don't have to recompile
        the application each time you edit it (you just load
        the new JSON file).</para>
      </note>

    </section>

    <section id="animations-complex-discussion">
      <title>Discussion</title>

      <para>You can think of <type>ClutterAnimator</type>
      as a way to give directions to actors. For example,
      you could give a real (human) actor a direction like "move
      downstage; when you get there, stop and
      rotate 90 degrees to your right". In code,
      this might equate to a transition in the <varname>x</varname>
      and <varname>y</varname> properties of the actor, followed by a
      rotation in one axis.</para>

      <note>
        <para><type>ClutterAnimator</type> can give
        "directions" to any type of GObject, but we concentrate
        on animating <type>ClutterActors</type> in this section.</para>
      </note>

      <para>Each direction like this has an implicit
      timeline, spanning the length of time the direction should
      take to fulfil (you set the length of the timeline through
      the <varname>duration</varname> property of the
      <type>ClutterAnimator</type>). But within that timeline, you may
      change the proportion of time spent on each action: "move
      downstage quickly, then slowly rotate 90 degrees
      to your right". The direction is the same, but we've
      specified how much of the timeline should be devoted to each
      action.</para>

      <para>In <type>ClutterAnimator</type>, this concept is
      captured by <emphasis>key frames</emphasis>. A
      key frame represents a point somewhere along the timeline,
      with one or more target property values for one or more actors.
      A <type>ClutterAnimator</type> manages the transitions
      between property values for each object, ensuring that
      the target values are reached when the associated key frame
      is reached.</para>

      <para>To change the amount of time a transition
      should take, you change the percentage of the timeline
      between key frames. Using our real stage directions as an
      example, you might define the key frames like this:</para>

      <itemizedlist>
        <listitem>
          <para><emphasis>0.2 (after 20% of the timeline):</emphasis>
          arrive downstage</para>
        </listitem>
        <listitem>
          <para><emphasis>1.0 (by the end of the timeline):</emphasis>
          achieve a 90 degree rotation to the right</para>
        </listitem>
      </itemizedlist>

      <para>See
      <link linkend="animations-complex-discussion-keys">this
      section</link> for more details about keys and key frames.</para>

      <para>Finally, a direction might be further refined with
      a description of the kind of movement to use:
      rather than saying "move downstage quickly, then
      slowly rotate 90 degrees to your right" a director could say:
      "start off slowly, but build up to a run;
      run downstage quickly; then stop and start rotating
      slowly to your right, gradually speeding up, turn a little more, then slow
      down gradually; you should end up rotated 90 degrees to your right"
      (this granularity of description is closer to what you might
      see in dance notation like
      <ulink href="http://en.wikipedia.org/wiki/Labanotation">Laban</ulink>;
      though of course you can't animate human opacity, scale, dimensions
      etc...).</para>

      <para><type>ClutterAnimator</type> gives you this level of
      granularity. Each transition to a property value between
      key frames can have a separate <emphasis>easing mode</emphasis>:
      for example, starting off slowly and building to a constant
      speed equates to an "ease in" mode; starting slowly, speeding
      up, maintaining a constant speed, then gradually slowing down
      equates to "ease in and ease out".</para>

      <para>To summarise: creating a complex animation means deciding:</para>

      <itemizedlist>
        <listitem>
          <para>Which properties need to change on which actors?</para>
        </listitem>
        <listitem>
          <para>What target value should each property transition to?</para>
        </listitem>
        <listitem>
          <para>How quickly (by which key frame) should the property
          reach the target value?</para>
        </listitem>
        <listitem>
          <para>What "shape" (easing mode) should the change to
          the target value follow?</para>
        </listitem>
      </itemizedlist>

      <section id="animations-complex-discussion-keys">
        <title>Understanding keys and key frames</title>

        <para>A <type>ClutterAnimator</type> maintains a list of
        <varname>properties</varname> objects, each being a unique pair
        of <varname>object</varname> (an object to be animated) +
        <varname>name</varname> (name of the property
        to be animated on that object).</para>

        <para>Each <varname>properties</varname> object in turn has a
        list of keys, with each key having three elements:</para>

        <itemizedlist>
          <listitem>
            <para>The <emphasis>key frame</emphasis>, expressed as a fraction
            (between 0.0 and 1.0) of the duration of the animation. At this
            point, the named property should reach a target value.</para>
          </listitem>
          <listitem>
            <para>The <emphasis>easing mode</emphasis> to use to transition
            the property to that value.</para>
          </listitem>
          <listitem>
            <para>The <emphasis>target value</emphasis> the property
            should transition to.</para>
          </listitem>
        </itemizedlist>

        <para>For example:</para>

        <informalexample>
          <programlisting>
{
  "object" : "rectangle",
  "name" : "x",
  "ease-in" : true,
  "keys" : [
    [ 0.0, "linear", 0.0 ],
    [ 0.1, "easeInCubic", 150.0 ],
    [ 0.8, "linear", 150.0 ],
    [ 1.0, "easeInCubic", 0.0 ]
  ]
}
          </programlisting>
        </informalexample>

        <para>defines a sequence of transitions for the <varname>x</varname>
        property (position on the x axis) of the <code>rectangle</code>
        object, as follows:</para>

        <orderedlist>
          <listitem>
            <para><emphasis>[ 0.0, "linear", 0.0 ]</emphasis>:
            At the start of the animation, <code>x</code> should be
            0.0; <code>linear</code> is used as the easing mode, as there
            is no transition here.</para>
          </listitem>

          <listitem>
            <para><emphasis>[ 0.1, "easeInCubic", 150.0 ]</emphasis>:
            By 10% of the way through the animation,
            <code>x</code> should reach a value of <code>150.0</code>.
            This moves the rectangle horizontally across the stage.</para>

            <para>The <code>easeInCubic</code> easing mode means that
            the transition to the new value starts slow and speeds up.
            This makes the movement look more "natural".</para>
          </listitem>

          <listitem>
            <para><emphasis>[ 0.8, "linear", 150.0 ]</emphasis>:
            From 10% of the way through the animation to 80%
            of the way through, the <code>x</code> value remains at
            <code>150.0</code>. This makes the rectangle stay still
            on the x axis throughout this period.</para>

            <para>It's important to specify interim key frames if
            in a later key frame you intend to change the value again
            (as is done for the <code>x</code> value here). Otherwise
            you can get premature transitions to a value over longer
            periods than you intended. By specifying the interim
            key frames where the value remains constant, you ensure
            that it doesn't change before you want it to.</para>
          </listitem>

          <listitem>
            <para><emphasis>[ 1.0, "easeInCubic", 0.0 ]</emphasis>:
            From 80% of the way through the animation to the end,
            the <code>x</code> value should transition back to
            <code>0.0</code>. This moves the actor back to its
            starting position on the x axis. Again, an <code>easeInCubic</code>
            easing mode is used to make the transition appear more natural.</para>
          </listitem>
        </orderedlist>

        <para>There are two more properties you can set for each
        object/property pair:</para>

        <orderedlist>
          <listitem>
            <para>Set <varname>ease-in</varname> to <code>true</code> to
            animate to the target value at the first key frame. If
            <varname>ease-in</varname> is false, the animation will
            "jump" to the target value instead (if the target value is
            different from the current value).</para>
          </listitem>

          <listitem>
            <para>Set <varname>interpolation</varname> to either
            <code>"linear"</code> (the default) or <code>"cubic"</code>.
            This sets how <type>ClutterAnimator</type> transitions between
            key frames; in effect, it further modulates any easing modes
            set on individual keys: if set to <code>"cubic"</code>, you
            get a slightly more natural and gentle transition between
            key frames than you do if set to <code>"linear"</code>.</para>
          </listitem>
        </orderedlist>

      </section>

      <section id="animations-complex-why-clutteranimator">
        <title>Why <type>ClutterAnimator</type>?</title>

        <para>Why use <type>ClutterAnimator</type> and not the other
        <link linkend="animations-introduction-api">Clutter animation
        approaches</link> for complex animations?</para>

        <itemizedlist>
          <listitem>
            <para><emphasis>Implicit animations</emphasis> can animate
            properties on a single actor; however, you can only specify a
            single transition for each property. Also, it's not possible
            to describe complex movement along a path in a single implicit
            animation: you would have to chain several animations together
            to do that.</para>

            <para>To animate multiple actors, you'd also need multiple
            implicit animations, one for each actor. These animations would
            also need to be synchronized (for example, by sharing a
            single timeline).</para>

            <para>So it would be possible, but more difficult than
            an implementation using <type>ClutterAnimator</type>.</para>
          </listitem>

          <listitem>
            <para><emphasis><type>ClutterState</type></emphasis> can
            be used for complex animations: each state can describe
            transitions for multiple actors and multiple properties.
            However, to make continuous movement (as in the example),
            you would need to write a state for each movement between a
            pair of points; then add a callback so that when each state
            is reached, the animation moves onto the next state. This
            adds some code (a handler for the <code>completed</code>
            signal emitted by the <type>ClutterState</type> to set
            the next state). This could work OK for a few states,
            but doesn't scale as well as <type>ClutterAnimator</type>
            if you have many transitions.</para>

            <note>
              <para><type>ClutterState</type> and
              <type>ClutterAnimator</type> are not mutually exclusive. If
              you generally need to transition between several known states
              (e.g. hiding/revealing menus which stay in the same place,
              moving between two UI layouts), but want to create a
              complex animation between states, you can use
              <type>ClutterAnimators</type> to define the transitions: see
              the documentation for
              <function>clutter_state_set_animator()</function> for
              details.</para>
            </note>

          </listitem>
        </itemizedlist>

        <para><type>ClutterAnimator</type> is a good fit for complex
        animations, and probably the best fit for the most complex:
        it is the simplest way to encode a sequence of transitions
        for a list of object/property pairs which can be treated
        as a single animation. This is largely because
        <type>ClutterAnimator</type> is effectively managing the
        chaining together of the individual transitions into a whole.</para>

        <para>One other feature of <type>ClutterAnimator</type> which
        isn't demonstrated here is how it enables transitions to overlap.
        For example, let's say you wanted an actor
        to move along a complex path (e.g. described by five pairs of
        x,y coordinates); but during that movement, you
        wanted the actor to continuously transition to a scale of
        4.0 on both the x and y axes.</para>

        <para>To achieve this with <type>ClutterState</type>, you would
        need to set up five transitions (one to move to each pair of
        x,y coordinates); plus a callback to chain the state transitions
        together; and within each transition, you'd have to figure out a
        percentage of the scaling to apply, so that the actor
        was at a scale of 4.0 on reaching the final state.</para>

        <para>With <type>ClutterAnimator</type>, you can treat the
        movement between the coordinates and the scaling separately
        within the same animation, but overlap their key frames. This
        makes coding overlapping animations of different properties
        much more straightforward. See
        <link linkend="animations-complex-example-3">this JSON
        definition</link> for an example of how to do this.</para>

      </section>

    </section>

    <section>
      <title>Full example</title>

      <example id="animations-complex-example-3">
        <title>Running multiple transition sequences with
        different key frames in parallel using
        <type>ClutterAnimator</type></title>

        <note>
          <para>This JSON file can be loaded with the same code
          as used for <link linkend="animations-complex-example-2">this
          example</link>, by passing the JSON file name on the command line:</para>

          <screen>
            <prompt>$</prompt> <command>./animations-complex animations-complex-overlapping.json</command>
          </screen>
        </note>

        <programlisting>
<xi:include href="examples/animations-complex-overlapping.json" parse="text">
  <xi:fallback>a code sample should be here... but isn't</xi:fallback>
</xi:include>
        </programlisting>
      </example>
    </section>

  </section>

  <section id="animations-reuse">
    <title>Reusing a complex animation on different actors</title>

    <section id="animations-reuse-problem">
      <title>Problem</title>

      <para>You want to apply the same complex animation to several
      different actors.</para>

    </section>

    <section id="animations-reuse-solution">
      <title>Solution</title>

      <para>Instead of animating each actor separately, create a
      <emphasis>rig</emphasis>: an empty container with an associated
      animation, which will be animated in lieu of
      animating the actor directly. Do this as follows:</para>

      <orderedlist>
        <listitem>
          <para>Initialise the stage and actors, including those
          to be animated.</para>
        </listitem>
        <listitem>
          <para>Define a <type>ClutterContainer</type> and a
          <type>ClutterAnimator</type> animation to animate it.</para>
        </listitem>
        <listitem>
          <para>When you need to animate an actor:</para>

          <orderedlist>
            <listitem>
              <para>Create an instance of the rig and its animator.</para>
            </listitem>
            <listitem>
              <para>Reparent the actor to the rig.</para>
            </listitem>
            <listitem>
              <para>Run the rig's animation.</para>
            </listitem>
          </orderedlist>
        </listitem>
      </orderedlist>

      <para>For this solution, we're using
      <ulink url="http://json.org/">JSON</ulink> to define the
      animation and the user interface elements. For more
      details about this approach, see
      <link linkend="script-introduction">the chapter
      on <type>ClutterScript</type></link>.</para>

      <para>Here's an extract of the JSON definition for the stage and
      one of five rectangles placed at its left edge (the full definition
      is in <link linkend="animations-reuse-example-1">the
      appendix</link>):</para>

      <informalexample>
        <programlisting>
<![CDATA[
[
  {
    "type" : "ClutterStage",
    "id" : "stage",

    ... stage properties, signal handlers etc. ...

    "children" : [
      {
        "type" : "ClutterRectangle",
        "id" : "rect1",
        "color" : "white",
        "width" : 50,
        "height" : 50,
        "y" : 50,
        "reactive" : true,
        "signals" : [
          { "name" : "button-press-event", "handler" : "foo_button_pressed_cb" }
        ]
      },

      ... more children defined here ...
    ]
  }
]
]]>
        </programlisting>
      </informalexample>

      <para>The key point to note is how a signal handler is defined
      for the <code>button-press-event</code>, so that the
      <function>foo_button_pressed_cb()</function> function will trigger
      the animation when a (mouse) button is pressed on each rectangle.</para>

      <para>The second JSON definition includes the rig
      (an empty <type>ClutterGroup</type>) and a
      <type>ClutterAnimator</type> to animate it. The animation moves the
      container across the stage and scales it to twice its original
      size. (This is the <link linkend="animations-reuse-example-2">same
      code as in the appendix</link>):</para>

      <informalexample>
        <programlisting>
<xi:include href="examples/animations-reuse-animation.json" parse="text">
  <xi:fallback>a code sample should be here... but isn't</xi:fallback>
</xi:include>
        </programlisting>
      </informalexample>

      <para>The remaining parts of the application code load
      the user interface definition, setting up the stage and rectangles;
      and define the callback. The full code is
      <link linkend="animations-reuse-example-3">in the appendix</link>,
      but below is the most important part, the callback function:</para>

      <informalexample>
        <programlisting>
<![CDATA[
gboolean
foo_button_pressed_cb (ClutterActor *actor,
                       ClutterEvent *event,
                       gpointer      user_data)
{
  ClutterScript *ui = CLUTTER_SCRIPT (user_data);
  ClutterStage *stage = CLUTTER_STAGE (clutter_script_get_object (ui, "stage"));

  ClutterScript *script;
  ClutterActor *rig;
  ClutterAnimator *animator;

  /* load the rig and its animator from a JSON file */
  script = clutter_script_new ();

  /* use a function defined statically in this source file to load the JSON */
  load_script_from_file (script, ANIMATION_FILE);

  clutter_script_get_objects (script,
                              "rig", &rig,
                              "animator", &animator,
                              NULL);

  /* remove the button press handler from the rectangle */
  g_signal_handlers_disconnect_matched (actor,
                                        G_SIGNAL_MATCH_FUNC,
                                        0,
                                        0,
                                        NULL,
                                        foo_button_pressed_cb,
                                        NULL);

  /* add a callback to clean up the script when the rig is destroyed */
  g_object_set_data_full (G_OBJECT (rig), "script", script, g_object_unref);

  /* add the rig to the stage */
  clutter_container_add_actor (CLUTTER_CONTAINER (stage), rig);

  /* place the rig at the same coordinates on the stage as the rectangle */
  clutter_actor_set_position (rig,
                              clutter_actor_get_x (actor),
                              clutter_actor_get_y (actor));

  /* put the rectangle into the top-left corner of the rig */
  clutter_actor_reparent (actor, rig);

  clutter_actor_set_position (actor, 0, 0);

  /* animate the rig */
  clutter_animator_start (animator);

  return TRUE;
}
]]>
        </programlisting>
      </informalexample>

      <para>The code creates a new rig and associated animation
      at the point when the rectangle is clicked. It then positions the
      rig at the same coordinates as the rectangle, reparents
      the rectangle to the rig, and starts the rig's animation.</para>

      <note>
        <para>The signal handler has to be declared non-static and
        you must use <code>-export-dynamic</code> as an option to the
        compiler, otherwise the function isn't visible to
        <type>ClutterScript</type> (as outlined
        <link linkend="script-signals">in this recipe</link>).</para>
      </note>

      <para>This is what the animation looks like:</para>

      <inlinemediaobject>
        <videoobject>
          <videodata fileref="videos/animations-reuse.ogv"/>
        </videoobject>
        <alt>
          <para>Video of a simple reusable animation</para>
        </alt>
      </inlinemediaobject>

    </section>

    <section id="animations-reuse-discussion">
      <title>Discussion</title>

      <para>The above solution reparents an actor to be animated
      into a rig (an empty placeholder). The rig is a container
      which acts as a temporary parent for the actor we
      <emphasis>really</emphasis> want to animate. By animating the rig,
      it appears as though the actor inside it is being animated (but
      <link linkend="animations-reuse-discussion-rig-not-actor">see
      these caveats</link>). This means the same animation can be
      easily applied to different actors: create an
      instance of the rig, reparent an actor to it, then
      run the rig's animation. This is simpler than creating
      a separate animation for each actor individually, or
      reusing a single <type>ClutterAnimator</type> on different
      actors (see
      <link linkend="animations-reuse-discussion-one-or-many">this
      section</link>).</para>

      <para>Using JSON enhances the animation's reusability (it's even
      potentially reusable in another application), makes the code
      simpler (an animation can be loaded directly from the script),
      and makes refactoring easier (the animation can be modified
      without recompiling the application code). However, it also puts
      some minor limitations on the animation's reusability; namely, you
      can only set absolute property values in a JSON animation
      definition. This makes JSON less useful in cases where
      you need to animate properties relative to their starting
      values: for example, "move 50 pixels along the x axis" or
      "rotate by 10 degrees more on the z axis". (This type of animation
      is probably less portable anyway.) In such cases, the programmable
      API may be a better option: see the <type>ClutterAnimator</type>
      documentation for examples.</para>

      <section id="animations-reuse-discussion-one-or-many">
        <title>One animation vs. many</title>

        <para>In the sample code, a new instance of the rig and its
        animation are created for each actor. One side effect of this
        is that all of the actors can animate simultaneously with the
        "same" animation. If you don't want this behaviour, but still
        want to use a rig approach, you could create a single instance
        of the rig and its animation. Then, you could reparent each actor
        to it in turn.</para>

        <para>To ensure that the rig only animates one actor (or group
        of actors) at a time, you could track whether the rig is
        currently animating (e.g. by examining the animation's
        timeline with <function>clutter_animator_get_timeline()</function>).
        Then, if the animation is running, prevent any other actor
        from being reparented to the rig.</para>

        <para>Note that you would also need to "reset" the rig each time the
        animation completed (move it back to the right start values for
        its properties), ready to animate the next actor.</para>
      </section>

      <section id="animations-reuse-discussion-rig-not-actor">
        <title>Caveats about animating a rig instead of an actor</title>

        <para>There are a few issues to be aware of in cases
        where you animate a rig with contained actors, rather than
        animating the actor directly:</para>

        <itemizedlist>
          <listitem>
            <para>Animating a rig doesn't <emphasis>always</emphasis>
            produce the same visual effect as animating an actor directly.
            For example, compare the following cases:</para>

            <itemizedlist>
              <listitem>
                <para>You rotate an actor by 180 degrees in the
                <code>y</code> axis, then by 90 degrees in the
                <code>z</code> axis. The actor appears to rotate in
                a <emphasis>clockwise</emphasis> direction.</para>
              </listitem>
              <listitem>
                <para>You rotate the parent container of an actor
                by 180 degrees in the <code>y</code> axis; then rotate
                the actor by 90 degrees in the <code>z</code> axis.
                The actor appears to rotate in an
                <emphasis>anti-clockwise</emphasis> direction. By
                rotating the container, the "back" of the
                actor faces the view point, so the actor's movement
                appears reversed. See
                <link linkend="animations-rotating-discussion-direction">this
                recipe</link> for more details.</para>
              </listitem>
            </itemizedlist>

            <para>There may be other situations where you get similar
            discrepancies.</para>
          </listitem>

          <listitem>
            <para>Animating a rig doesn't change an actor's properties,
            but animating the actor does.</para>

            <para>When you animate a container rather than the actor
            directly, the reported properties of the actor may not
            reflect its visual appearance. For example, if you apply
            a scale animation to a container, the final scale of
            actors inside it (as returned by
            <function>clutter_actor_get_scale()</function>) will not
            reflect the scaling applied to their container; whereas
            directly animating the actors would cause their scale
            properties to change.</para>
          </listitem>

          <listitem>
            <para>Reparenting an actor to a rig can cause the actor
            to "jump" to the rig's position, unless you align the
            actor to the rig first.</para>

            <para>Note that in the sample code, the position of the actor
            (<code>x</code>, <code>y</code> coordinates) is copied to
            the rig before the reparenting happens. The actor is then
            reparented to the rig, and positioned in the rig's
            top-left corner. So the actor appears to be in the same
            position, but is now actually inside a rig at the actor's old
            position.</para>

            <para>Why bother to do this? Because the rig has a default
            position of <code>0,0</code> (top-left of <emphasis>its</emphasis>
            container, the stage). If you reparent the actor to the rig,
            without first copying the actor's position to the rig, the
            actor appears to "jump" to the rig's position.</para>
          </listitem>
        </itemizedlist>
      </section>

    </section>

    <section id="animations-reuse-examples">
      <title>Full example</title>

      <note>
        <para>The three separate code examples in this section
        constitute a single application which implements the above
        solution.</para>
      </note>

      <example id="animations-reuse-example-1">
        <title><type>ClutterScript</type> JSON defining several
        rectangles with signal handlers</title>
        <programlisting>
<xi:include href="examples/animations-reuse-ui.json" parse="text">
  <xi:fallback>a code sample should be here... but isn't</xi:fallback>
</xi:include>
        </programlisting>
      </example>

      <example id="animations-reuse-example-2">
        <title><type>ClutterScript</type> JSON describing a "rig"
        and a <type>ClutterAnimator</type> animation</title>
        <programlisting>
<xi:include href="examples/animations-reuse-animation.json" parse="text">
  <xi:fallback>a code sample should be here... but isn't</xi:fallback>
</xi:include>
        </programlisting>
      </example>

      <example id="animations-reuse-example-3">
        <title>Loading <type>ClutterScript</type> from JSON files
        in response to events in a user interface</title>
        <programlisting>
<xi:include href="examples/animations-reuse.c" parse="text">
  <xi:fallback>a code sample should be here... but isn't</xi:fallback>
</xi:include>
        </programlisting>
      </example>

    </section>

  </section>

  <section id="animations-moving">
    <title>Moving actors</title>

    <section>
      <title>Problem</title>

      <para>You want to animate the movement of one or more actors.
      For example:</para>

      <itemizedlist>
        <listitem>
          <para>To move user interface elements in response to user input
          (e.g. keyboard control of a character in a game).</para>
        </listitem>
        <listitem>
          <para>To move a group of actors "off stage" to make way
          for another group of actors (e.g. paging through
          thumbnails in a photo viewer).</para>
        </listitem>
        <listitem>
          <para>To move an actor to a different position in the
          interface (e.g. moving an icon for a trashed file into
          a wastebin).</para>
        </listitem>
      </itemizedlist>

    </section>

    <section>
      <title>Solutions</title>

      <para>Animate the actors movement on one or more axes
      (<varname>x</varname>, <varname>y</varname>,
      <varname>z/depth</varname>) using one or more of the approaches
      available in the Clutter API (implicit animations,
      <type>ClutterState</type>, <type>ClutterAnimator</type>).</para>

      <section>
        <title>Solution 1: Implicit animations</title>

        <para>This works well for simple movement of a single actor to
        a single set of coordinates. Here is an example of how to animate
        movement of a <type>ClutterActor</type> <varname>actor</varname>
        to position <code>100.0</code> on <varname>x</varname> axis:</para>

        <informalexample>
          <programlisting>
clutter_actor_animate (actor, CLUTTER_LINEAR, 500,
                       "x", 100.0,
                       NULL);
          </programlisting>
        </informalexample>

        <para>See <link linkend="animations-moving-example-1">this
        example</link> which demonstrates movement in each axis,
        in response to (mouse) button presses.</para>

      </section>

      <section>
        <title>Solution 2: <type>ClutterState</type></title>

        <para>This suits simple, repeated movement of one or more actors
        between sets of coordinates. Here is an example of how to
        create two states for a <type>ClutterState</type> instance to
        move two actors, <varname>actor1</varname> and
        <varname>actor2</varname>:</para>

        <informalexample>
          <programlisting>
ClutterState *transitions = clutter_state_new ();

/* all state transitions take 250ms */
clutter_state_set_duration (transitions, NULL, NULL, 250);

/* create a state called move-down which moves both actors to y = 200.0 */
clutter_state_set (transitions, NULL, "move-down",
                   actor1, "y", CLUTTER_EASE_OUT_CUBIC, 200.0,
                   actor2, "y", CLUTTER_EASE_OUT_CUBIC, 200.0,
                   NULL);

/* create a state called move-up which moves both actors to y = 0.0 */
clutter_state_set (transitions, NULL, "move-up",
                   actor1, "y", CLUTTER_EASE_OUT_CUBIC, 0.0,
                   actor2, "y", CLUTTER_EASE_OUT_CUBIC, 0.0,
                   NULL);

/* move the actors by setting the state */
clutter_state_set (transitions, "move-down");
          </programlisting>
        </informalexample>

        <para>This <link linkend="animations-moving-example-2">full
        example</link> shows how to move and simultaneously
        scale two actors. When a button is pressed on one actor, it is
        moved and scaled to occupy the right-hand side of the stage;
        the other actor is simultaneously moved back to the left-hand
        side of the stage and scaled down.</para>

      </section>

      <section>
        <title>Solution 3: <type>ClutterAnimator</type></title>

        <para>This is a good way to implement complex movement of
        one or more actors between sets of coordinates.</para>

        <informalexample>
          <programlisting>
ClutterAnimator *animator = clutter_animator_new ();

/* the animation takes 500ms */
clutter_animator_set_duration (animator, 500);

/* at the start of the animation, actor should be at 0.0,0.0;
 * half-way through, at 100.0,100.0;
 * by the end, actor should be at 150.0,200.0;
 * note that you can set different easing modes for each
 * part of the animation and for each property at each key
 */
clutter_animator_set (animator,

                      /* keys for the start of the animation */
                      actor, "x", CLUTTER_LINEAR, 0.0, 0.0,
                      actor, "y", CLUTTER_LINEAR, 0.0, 0.0,

                      /* keys for half-way through the animation */
                      actor, "x", CLUTTER_EASE_OUT_CUBIC, 0.5, 100.0,
                      actor, "y", CLUTTER_EASE_IN_CUBIC, 0.5, 100.0,

                      /* keys for the end of the animation */
                      actor, "x", CLUTTER_EASE_OUT_EXPO, 1.0, 150.0,
                      actor, "y", CLUTTER_EASE_OUT_CUBIC, 1.0, 200.0,

                      NULL);

/* run the animation */
clutter_animator_start (animator);
          </programlisting>
        </informalexample>

        <para>The <link linkend="animations-moving-example-3">full
        example</link> demonstrates how <type>ClutterAnimator</type>
        can be used to programmatically animate multiple actors: in this
        case, to simultaneously move three actors to random positions
        along the <varname>x</varname> axis. Synchronising the
        movement of three actors simultaneously using implicit
        animations would be possible but awkward;
        <type>ClutterState</type> might be another option,
        but it wasn't really designed for this case: there are no persistent
        states to transition between, as the actor positions are
        generated on each key press.</para>

        <note>
          <para>If you want to apply the same movement to a group of
          actors, rather than different movements for each actor,
          it's often better to put the actors into a container
          of some kind and move that instead of moving the actors
          individually.</para>
        </note>

      </section>

    </section>

    <section>
      <title>Discussion</title>

      <section>
        <title>Movement can take an actor "outside" its container</title>

        <para>Actor movement in the <varname>x</varname> and
        <varname>y</varname> axes is relative to the actor's parent
        container. There is nothing to stop you animating an actor
        until it falls outside the bounds of its container. This
        could result in the actor moving "off" the interface; though it's
        worth remembering that the actor is not unparented or destroyed
        if this happens.</para>

        <para>To ensure that an actor remains visible, its position
        should remain within the visible area of the container. In practice,
        this means either anywhere in the container, if no clip area
        has been set; or within the container's clip area, if set.</para>

      </section>

      <section>
        <title>Anchor points can affect movement</title>

        <para>An actor's anchor point is defined as an <code>x,y</code>
        coordinate relative to the top-left of the actor. The default
        anchor point for an actor is in its top-left
        corner. However, it is possible to set this to some other
        coordinate, relative to the actor's top-left corner,
        using the <function>clutter_anchor_set_anchor_point()</function>
        function.</para>

        <para>For example:</para>

        <informalexample>
          <programlisting>
/* set the actor's size to 100px x 100px */
clutter_actor_set_size (actor, 100, 100);

/* set an anchor point half-way along the top of the actor */
clutter_actor_set_anchor_point (actor, 50.0, 0.0);
          </programlisting>
        </informalexample>

        <note>
          <para>In GL terms, the anchor point of an actor is the equivalent
          of applying an additional transformation of <code>-x, -y</code> to
          the actor's modelview. If the anchor point is <code>0, 0</code>,
          i.e. the top-left corner, then the transformation will leave the
          actor in the same place.</para>
          <para>It is important to note that the anchor point will affect the
          position in which an actor is painted, but will not change the
          position or size that its parent allocated for it.</para>
          <para>Finally, the anchor point will affect the other transformations
          that can be applied to an actor: scaling and rotating.</para>
        </note>

        <para>A positive anchor point within the width/height bounds of the
        actor is inside the actor. An anchor point outside these bounds
        is outside the actor. You can also set a negative
        <varname>x</varname> or <varname>y</varname> value for
        the anchor point, which will again place the point outside
        the actor's bounds.</para>

        <para>This is important with respect to moving an actor, because
        you are actually moving the anchor point and "dragging" the
        actor along with it.</para>

        <para>For example: you have an actor with width 50px, and you
        set its <varname>anchor-x</varname> property to <code>25.0</code>.
        If you move that actor on the <varname>x</varname> axis, you are
        effectively moving a point half-way across the top of the
        actor along the <varname>x</varname> axis (which in turn moves the
        actor).</para>

        <para>Similarly, you could set the same actor's
        <varname>anchor-x</varname> to <code>-25.0</code>. If you then
        moved the actor along the <varname>x</varname> axis, you would
        effectively be moving the point 25px left of the top of the actor
        along that axis.</para>

        <para>The video below demonstrates the effect on movement of shifting
        the anchor point on the <varname>x</varname> axis. The
        <emphasis>red</emphasis> rectangle has <varname>anchor-x</varname>
        set to <code>25.0</code>; the <emphasis>green</emphasis> rectangle has
        <varname>anchor-x</varname> set to <code>0.0</code> (the default); the
        <emphasis>blue</emphasis> rectangle has <varname>anchor-x</varname>
        set to <code>-25.0</code>.</para>

        <inlinemediaobject>
          <videoobject>
            <videodata fileref="videos/animations-moving-anchors.ogv"/>
          </videoobject>
          <alt>
            <para>Video showing the effect of anchor point on movement</para>
          </alt>
        </inlinemediaobject>

        <para> A <type>ClutterAnimator</type> is
        used to move each of the rectangles to <code>x = 225.0</code>.
        Although the three rectangles move to the same position on the
        <varname>x</varname> axis, it's actually the anchor points
        which are at the same position. These all align on the
        <varname>x</varname> axis with the left-hand edge of the green
        rectangle.</para>

      </section>

      <section>
        <title>Actors can move in the <varname>z</varname> axis</title>

        <para>The examples so far have shown how to move actors in
        the <varname>x</varname> and <varname>y</varname> axes; but it
        is also possible to move actors in the <varname>z</varname>
        axis (i.e. move them closer or further away from the view point).
        This lets you move actors under/over each other.</para>

        <para>To move an actor in the <varname>z</varname> axis, animate
        its <varname>depth</varname> property. Animating to a negative
        depth moves the actor away from the view point; animating to a
        positive depth moves the actor towards the view point.</para>

        <para>Changing the depth of an actor also causes perspective
        effects: the actor gets smaller and converges on the center
        of the stage as it gets further from the view point, and
        gets larger and diverges from the center of the stage as it gets
        closer. This results in an apparent (but not actual) change in
        the <code>x,y</code> position and scale of the actor.</para>

        <note>
          <para>Animating the depth of an actor is slightly different
          from animating its x and y coordinates, as depth is relative
          to the whole stage, not just the parent container of the
          actor. This means that perspective effects are with
          respect to the whole stage: so as an actor's depth
          moves below <code>0.0</code>, it converges on the center
          of the stage, and may even apparently move outside its
          container (if the container stays at the same depth).</para>
        </note>

        <para>The video below demonstrates the effect of animating
        the depth of four actors to a value of <code>-15000.0</code>.
        Note how the actors converge on the center of the stage,
        as well as appearing to change position and scale; also note
        that they appear to move outside the bounds of their parent containers
        (the four yellow <type>ClutterBoxes</type>).</para>

        <inlinemediaobject>
          <videoobject>
            <videodata fileref="videos/animations-moving-depth.ogv"/>
          </videoobject>
          <alt>
            <para>Video showing perspective effects when animating
            actor depth</para>
          </alt>
        </inlinemediaobject>

      </section>

      <section>
        <title>Movement is affected by constraints</title>

        <para>An actor can have its x,y position constrained by
        the position of other actors through <type>ClutterBindConstraints</type>.
        This can affect movement in two ways:</para>

        <orderedlist>
          <listitem>
            <para>If an actor has its <varname>x</varname> and/or
            <varname>y</varname> properties
            bound or aligned to another actor's, you can't animate
            those properties.</para>

            <para>In effect this means that the bound actor can't be
            moved on a bound axis directly, but can only be moved by
            animating the constraint's properties.</para>
          </listitem>

          <listitem>
            <para>If you move an actor which has other actors bound to
            it, the bound actors will also move. For example, if
            the actor has several other actors whose <varname>x</varname>
            properties are bound to its <varname>x</varname> property,
            moving the actor on the <varname>x</varname> axis will also
            move the bound actors on that axis.</para>

            <para>Similarly, if some actor is the source for
            alignment constraints on other actors, moving the source
            will cause those other actors to move, so that they remain in
            alignment with it.</para>
          </listitem>
        </orderedlist>

        <para>For example, consider two actors bound by constraints
        as follows:</para>

        <informalexample>
          <programlisting>
/* the source actor for the constraint */
ClutterActor *source;

/* the actor bound by the constraint */
ClutterActor *target;

/* a constraint to be added to target */
ClutterConstraint *constraint;

/* ...initialize actors etc... */

/* create a constraint for binding the x position of some actor to the
 * x position of source
 */
constraint = clutter_bind_constraint_new (source, CLUTTER_BIND_X, 0.0);

/* add the constraint to target with a name */
clutter_actor_add_constraint_with_name (target, "bind-x", constraint);
          </programlisting>
        </informalexample>

        <para>Animating <varname>source</varname> on the <varname>x</varname>
        axis also animates <varname>target</varname> on the same axis:</para>

        <informalexample>
          <programlisting>
clutter_actor_animate (source, CLUTTER_LINEAR, 500,
                       "x", 250.0,
                       NULL);
          </programlisting>
        </informalexample>

        <para>...while this has no effect, as it would violate
        <varname>constraint</varname> (it's best not to animate
        <varname>target's</varname> <varname>x</varname> property
        directly):</para>

        <informalexample>
          <programlisting>
clutter_actor_animate (target, CLUTTER_LINEAR, 500,
                       "x", 250.0,
                       NULL);
          </programlisting>
        </informalexample>

        <para>But the constraint's properties can be animated, to change
        how <varname>source</varname> and <varname>target</varname>
        are bound; which in turn moves <varname>target</varname>:</para>

        <informalexample>
          <programlisting>
clutter_actor_animate (target, CLUTTER_LINEAR, 500,
                       "@constraints.bind-x.offset", 250.0,
                       NULL);
          </programlisting>
        </informalexample>

        <para>Note the <code>@constraints.&lt;constraint name&gt;.&lt;constraint property&gt;</code>
        syntax (which is why we needed to use
        <function>clutter_actor_add_constraint_with_name()</function>,
        so that the constraint can be accessed through the actor).
        We are still animating <varname>target</varname>, but really
        we're indirectly animating a property of one of its constraints.</para>

        <para>Another alternative would be to directly animate
        the constraint's properties through <type>ClutterState</type>
        or <type>ClutterAnimator</type>, rather than using
        pseudo-properties on the actor animation:</para>

        <informalexample>
          <programlisting>
ClutterAnimator *animator = clutter_animator_new ();
clutter_animator_set_duration (animator, 500);

clutter_animator_set (animator,
                      constraint, "offset", CLUTTER_LINEAR, 0.0, 0.0,
                      constraint, "offset", CLUTTER_LINEAR, 1.0, 250.0,
                      NULL);

clutter_animator_start (animator);
          </programlisting>
        </informalexample>

        <para>This could be useful if you need to animate
        multiple constraints between multiple values simultaneously.</para>

      </section>

    </section>

    <section id="animations-moving-examples">
      <title>Full examples</title>

      <example id="animations-moving-example-1">
        <title>Simple movement using implicit animations</title>
        <programlisting>
<xi:include href="examples/animations-moving-implicit.c" parse="text">
  <xi:fallback>a code sample should be here... but isn't</xi:fallback>
</xi:include>
        </programlisting>
      </example>

      <example id="animations-moving-example-2">
        <title>Using <type>ClutterState</type> to repeatedly move
        (and scale) two actors</title>
        <programlisting>
<xi:include href="examples/animations-moving-state.c" parse="text">
  <xi:fallback>a code sample should be here... but isn't</xi:fallback>
</xi:include>
        </programlisting>
      </example>

      <example id="animations-moving-example-3">
        <title>Using <type>ClutterAnimator</type> to randomly move
        three actors along the <varname>x</varname> axis</title>
        <programlisting>
<xi:include href="examples/animations-moving-animator.c" parse="text">
  <xi:fallback>a code sample should be here... but isn't</xi:fallback>
</xi:include>
        </programlisting>
      </example>

    </section>

  </section>

  <section id="animations-looping">
    <title>Looping an animation</title>

    <section>
      <title>Problem</title>

      <para>You want to loop an animation so it plays multiple times.</para>
    </section>

    <section>
      <title>Solutions</title>

      <para>Each <link linkend="animations-introduction-api">animation
      approach</link> can be used to create a looping animation, as
      described in the following sections.</para>

      <para>The animation implemented in each case is a simple repeated
      movement of a rectangle from the right (<code>x = 150.0</code>)
      to the left (<code>x = 50.0</code>) of the stage, and back again,
      looped; like this (just a few iterations):</para>

      <inlinemediaobject>
        <videoobject>
          <videodata fileref="videos/animations-looping.ogv"/>
        </videoobject>
        <alt>
          <para>Video showing simple looped movement of an actor</para>
        </alt>
      </inlinemediaobject>

      <section id="animations-looping-solutions-implicit">
        <title>Solution 1: looping an implicit animation</title>

        <para>Implicit animations, started using
        <function>clutter_actor_animate()</function>, can be looped via
        their associated <type>ClutterTimeline</type>.</para>

        <para>Create a <type>ClutterTimeline</type> which is
        set to loop:</para>

        <informalexample>
          <programlisting>
ClutterTimeline *timeline = clutter_timeline_new (1000);
clutter_timeline_set_loop (timeline, TRUE);
          </programlisting>
        </informalexample>

        <para>Use this timeline when starting an implicit animation on an
        actor; in this case, to animate the actor's <varname>x</varname>
        coordinate from its initial value to <code>50.0</code>:</para>

        <informalexample>
          <programlisting>
/* assume <varname>actor</varname> is a <type>ClutterActor</type> instance */

/* actor's initial x value is 150.0 */
clutter_actor_set_x (actor, 150.0);

/* animate the actor (starting the timeline is implicit) */
clutter_actor_animate_with_timeline (actor,
                                     CLUTTER_LINEAR,
                                     timeline,
                                     "x", 50.0,
                                     NULL);
          </programlisting>
        </informalexample>

        <para>One further technique is to swop the timeline's
        direction to create a "closed loop" animation (one which returns
        to its origin at the end of each iteration). See
        <link linkend="animations-looping-discussion-closed-loop">this
        section</link> for details.</para>

        <para><link linkend="animations-looping-example-1">The full
        code example</link> shows how to run an implicit animation on
        a loop.</para>
      </section>

      <section>
        <title>Solution 2: looping with <type>ClutterAnimator</type></title>

        <para>A <type>ClutterAnimator</type> animation can also be looped
        via its <type>ClutterTimeline</type>. However, as
        <type>ClutterAnimator</type> enables more complex animations,
        you don't have to manually invert the timeline at the
        end of each iteration. Instead, you can animate
        an actor's properties back to their initial values
        at the end of each iteration of the loop.</para>

        <para>Creating the timeline and setting it to loop is the same
        as for implicit animations:</para>

        <informalexample>
          <programlisting>
ClutterTimeline *timeline = clutter_timeline_new (2000);
clutter_timeline_set_loop (timeline, TRUE);
          </programlisting>
        </informalexample>

        <para>Note that the timeline is twice the length of the one for
        the implicit animation: this is because, unlike the implicit
        animation, the movement from right to left and back again
        is a <emphasis>single</emphasis> animation. By contrast, in the
        implicit animation, the timeline runs forward, for the right to
        left movement; and then backwards, for the left to right
        movement. So rather than a 1000ms timeline running twice (once
        forward,  once backward for the implicit animation),
        we have a 2000ms timeline running once (for
        <type>ClutterAnimator</type>).</para>

        <para>Next, create a <type>ClutterAnimator</type> which animates
        the actor from right to left, then left to right:</para>

        <informalexample>
          <programlisting>
/* assume <varname>actor</varname> is a <type>ClutterActor</type> instance */
ClutterAnimator *animator = clutter_animator_new ();

/* use the looping timeline as the timeline for the animator */
clutter_animator_set_timeline (animator, timeline);

/* set positions for the actor at various points through the animation:
 * at progress 0.0, x = 150.0 (right of the stage)
 * at progress 0.5, x = 50.0 (left of the stage)
 * at progress 1.0, x = 150.0 again (back to the right)
 */
clutter_animator_set (animator,
                      actor, "x", CLUTTER_LINEAR, 0.0, 150.0,
                      actor, "x", CLUTTER_LINEAR, 0.5, 50.0,
                      actor, "x", CLUTTER_LINEAR, 1.0, 150.0,
                      NULL);
          </programlisting>
        </informalexample>

        <para>Finally, start the animation:</para>

        <informalexample>
          <programlisting>
clutter_animator_start (animator);
          </programlisting>
        </informalexample>

        <para>See <link linkend="animations-looping-example-2">the full
        example</link> for more details.</para>
      </section>

      <section>
        <title>Solution 3: looping with <type>ClutterState</type></title>

        <para>You can loop <type>ClutterState</type> animations by
        creating a cycle of states which
        <ulink url="http://en.wikipedia.org/wiki/Ouroboros">"swallows
        its own tail"</ulink>: i.e. goes from a start state, through
        intermediate state(s), back to the start state, then again
        through the intermediate states(s), back to the start state,
        etc., ad infinitum.</para>

        <para>For the animation we're implementing, there are two states
        the actor transitions between:</para>

        <orderedlist>
          <listitem>
            <para>The actor's <varname>x</varname> value
            is <code>150.0</code> (the start/end state, on the right
            of the stage).</para>
          </listitem>
          <listitem>
            <para>The actor's <varname>x</varname> value is
            <code>50.0</code> (the intermediate state, on the left
            of the stage).</para>
          </listitem>
        </orderedlist>

        <para>Here is how to add those states to a
        <type>ClutterState</type> instance:</para>

        <informalexample>
          <programlisting>
ClutterState *transitions = clutter_state_new ();

/* the duration for a transition from any state to any other is 1 second */
clutter_state_set_duration (transitions, NULL, NULL, 1000);

clutter_state_set (transitions, NULL, "right",
                   actor, "x", CLUTTER_LINEAR, 150.0,
                   NULL);

clutter_state_set (transitions, NULL, "left",
                   actor, "x", CLUTTER_LINEAR, 50.0,
                   NULL);
          </programlisting>
        </informalexample>

        <para>You also need a handler to move the <type>ClutterState</type>
        to its next state, called each time a state transition
        is completed:</para>

        <informalexample>
          <programlisting>
/* handler to move the <type>ClutterState</type> to its next state */
static void
next_state (ClutterState *transitions,
            gpointer      user_data)
{
  const gchar *state = clutter_state_get_state (transitions);

  if (g_strcmp0 (state, "right") == 0)
    clutter_state_set_state (transitions, "left");
  else
    clutter_state_set_state (transitions, "right");
}
          </programlisting>
        </informalexample>

        <para>Then connect the <type>ClutterState's</type>
        <code>completed</code> signal to the handler, so that each time
        a state is reached, the transition to the next state begins:</para>

        <informalexample>
          <programlisting>
/* connect the <type>ClutterState</type> <code>completed</code> signal to the handler */
g_signal_connect (transitions,
                  "completed",
                  G_CALLBACK (next_state),
                  NULL);
          </programlisting>
        </informalexample>

        <para>Finally, put the <type>ClutterState</type> into the start
        state to begin the animation:</para>

        <informalexample>
          <programlisting>
clutter_state_warp_to_state (transitions, "right");
          </programlisting>
        </informalexample>

        <para>See <link linkend="animations-looping-example-3">the full
        example</link> for more details.</para>
      </section>

    </section>

    <section>
      <title>Discussion</title>

      <para>We use two different approaches to looping in the solutions:</para>

      <orderedlist>
        <listitem>
          <para>Setting the <type>ClutterTimeline</type> to loop
          (via <function>clutter_timeline_set_loop()</function>). This
          is the best approach where the timeline is explicit (for
          <type>ClutterAnimator</type> and implicit animations).</para>
        </listitem>
        <listitem>
          <para>Cycling through states in a <type>ClutterState</type>. In
          this case, the timeline is implicit and we don't need to
          manually control it: the loop is a consequence of cycling
          repeatedly through a series of states.</para>
        </listitem>
      </orderedlist>

      <para>The following sections cover some other aspects of looping
      animations.</para>

      <section>
        <title>Looping a fixed number of times</title>

        <para><type>ClutterTimeline</type> doesn't have any built-in
        functionality to support looping a certain number of times. But
        it is reasonably easy to count the number of iterations completed and
        stop the animation when some limit is reached.</para>

        <para>For example, you could use a static counter to keep track
        of the iteration count:</para>

        <informalexample>
          <programlisting>
static guint counter = 0;
          </programlisting>
        </informalexample>

        <para>Implement the looping behaviour as in the above solutions,
        but use a callback function to set/reset the counter each time
        the timeline completes. For example, for the
        <type>ClutterAnimator</type> solution, you would connect the
        <code>completed</code> signal of the timeline
        to a callback function:</para>

        <informalexample>
          <programlisting>
g_signal_connect (timeline,
                  "completed",
                  G_CALLBACK (timeline_completed_cb),
                  NULL);
          </programlisting>
        </informalexample>

        <para>And implement a callback function which resets the counter and
        stops the timeline if more than two iterations have been counted:</para>

        <informalexample>
          <programlisting>
static void
timeline_completed_cb (ClutterTimeline *timeline,
                       gpointer         user_data)
{
  counter++;

  if (counter &gt; 2)
    {
      counter = 0;
      clutter_timeline_stop (timeline);
    }
}
          </programlisting>
        </informalexample>

        <para>Note that it's simple to count iterations and
        control the timeline using <type>ClutterAnimator</type> or
        <type>ClutterState</type>, as the whole animation (right to left
        and back) is a discrete unit. Doing the same with implicit
        animations is possible (one forward + one backward run along the
        timeline is one iteration). But you will be really stretching the
        implicit animation API beyond its intended use cases.</para>
      </section>

      <section id="animations-looping-discussion-closed-loop">
        <title>Creating a "closed loop" with an implicit animation</title>

        <para>When using implicit animations, at the end of the timeline
        (before the next iteration of the loop), an actor's properties
        "jump" back to their initial values (as they were when the timeline
        started). For example, in the
        <link linkend="animations-looping-solutions-implicit">earlier
        solution</link>, the actor's initial <varname>x</varname> value was
        <code>150.0</code>; so the default behaviour on each iteration
        of the loop would be to animate the actor to <code>x = 50.0</code>
        then jump it immediately back to <code>x = 150.0</code>, before
        continuing the loop.</para>

        <para>To prevent this happening, you can create a "closed" loop:
        animate the actor's properties away from their initial values, then
        back again.</para>

        <para>This could be done manually, by creating two separate
        animations, one the inverse of the other, and chaining them together.</para>

        <para>However, a simpler solution is to run forward through the timeline
        once; then invert its direction when the end of timeline is reached.
        The animation continues, but in reverse. Once the backward iteration
        completes, set the timeline to run forward again. Keep changing the
        timeline's direction each time it completes. This
        is the approach used in <link linkend="animations-looping-example-1">the
        example</link>, which results in a smooth, repeated right to left,
        left to right motion.</para>

        <para>See <link linkend="animations-inversion">this
        recipe</link> for more details about inverting a timeline.</para>
      </section>

    </section>

    <section id="animations-looping-examples">
      <title>Full examples</title>

      <example id="animations-looping-example-1">
        <title>Looping an implicit animation</title>
        <programlisting>
<xi:include href="examples/animations-looping-implicit.c" parse="text">
  <xi:fallback>a code sample should be here... but isn't</xi:fallback>
</xi:include>
        </programlisting>
      </example>

      <example id="animations-looping-example-2">
        <title>Looping with <type>ClutterAnimator</type></title>
        <programlisting>
<xi:include href="examples/animations-looping-animator.c" parse="text">
  <xi:fallback>a code sample should be here... but isn't</xi:fallback>
</xi:include>
        </programlisting>
      </example>

      <example id="animations-looping-example-3">
        <title>Looping with <type>ClutterState</type></title>
        <programlisting>
<xi:include href="examples/animations-looping-state.c" parse="text">
  <xi:fallback>a code sample should be here... but isn't</xi:fallback>
</xi:include>
        </programlisting>
      </example>

    </section>

  </section>

  <section id="animations-scaling">
    <title>Animated scaling</title>

    <section>
      <title>Problem</title>

      <para>You want to animate scaling of an actor. Example use
      cases:</para>

      <itemizedlist>
        <listitem>
          <para>To animate zooming in/out of a texture in an
          image viewer application.</para>
        </listitem>
        <listitem>
          <para>To add an animated "bounce" effect (quick scale up
          followed by scale down) to a UI element
          to indicate it has received focus.</para>
        </listitem>
      </itemizedlist>
    </section>

    <section>
      <title>Solution</title>

      <para>Animate the actor's <varname>scale-x</varname> and
      <varname>scale-y</varname> properties to change the scaling on
      the <varname>x</varname> and <varname>y</varname> axes
      respectively.</para>

      <para>For example, to animate an actor to twice its current scale
      with implicit animations:</para>

      <informalexample>
        <programlisting>
<![CDATA[
gdouble scale_x;
gdouble scale_y;

/* get the actor's current scale */
clutter_actor_get_scale (actor, &scale_x, &scale_y);

/* animate to twice current scale on both axes */
clutter_actor_animate (actor, CLUTTER_LINEAR, 1000,
                       "scale-x", scale_x * 2,
                       "scale-y", scale_y * 2);
]]>
        </programlisting>
      </informalexample>

      <para>Alternatively, <type>ClutterAnimator</type> or
      <type>ClutterState</type> can be used to animate an actor's scale
      properties. See <link linkend="animations-scaling-example-1">this
      example</link> which uses <type>ClutterState</type> to animate
      scaling.</para>
    </section>

    <section>
      <title>Discussion</title>

      <para>Scaling an actor is done through its <varname>scale-x</varname>
      and <varname>scale-y</varname> properties, each of which takes
      a <code>double</code> value. A value of less than
      <code>1.0</code> for an axis scales an actor down on that axis,
      reducing its apparent size; values greater than <code>1.0</code>
      scale an actor up, increasing its apparent size.</para>

      <para>Why "apparent" size? Because scaling applies a transform
      to an actor which changes how it appears on the
      stage, without changing its "real" size. Similarly, scaling an
      actor may transform its position: it could appear to move to a
      different position within its container,
      although it is "really" at its original position. Run
      <link linkend="animations-scaling-example-1">the
      example</link> to see how size and position are
      transformed by scaling.</para>

      <para>It can be useful to know an actor's
      <emphasis>transformed</emphasis> position and size after scaling:
      for example, if you were implementing a reflowing layout manager
      which used scaling as part of its allocation algorithm.
      Here's an example of how to get these properties for an
      actor:</para>

      <informalexample>
        <programlisting>
<![CDATA[
gfloat transformed_x;
gfloat transformed_y;

gfloat transformed_width;
gfloat transformed_height;

clutter_actor_get_transformed_position (actor, &transformed_x, &transformed_y);
clutter_actor_get_transformed_size (actor, &transformed_width, &transformed_height);
]]>
        </programlisting>
      </informalexample>

      <para>Note that you can scale an actor on both axes by the same
      amount (uniform scaling), or by a different amount on each axis
      (differential scaling).</para>

      <para>Use <function>clutter_actor_is_scaled()</function> to determine
      whether scaling has been applied to an actor: this function returns
      <code>FALSE</code> if both <varname>scale-x</varname> and
      <varname>scale-y</varname> are <code>1.0</code>; otherwise, it
      returns <code>TRUE</code>.</para>

      <section>
        <title>Scaling vs. resizing</title>

        <para>Scaling changes the <emphasis>apparent</emphasis> size
        of an actor, while leaving its real size unchanged. By contrast,
        resizing changes the <emphasis>real</emphasis> size of the actor,
        by modifying its <varname>width</varname> and
        <varname>height</varname> properties.</para>

        <para>Resizing and scaling produce the same visual
        effect, as both make an actor appear to be larger or
        smaller. Therefore, for most purposes, they are interchangeable
        if you just want to change an actor's apparent size.</para>

        <para>So why would you scale an actor rather than resize it?</para>

        <itemizedlist>
          <listitem>
            <para>If you've scaled an actor, you can easily reset it
            to its original size, by setting its
            scale back to <code>1.0</code> on both axes. By contrast,
            to reset a resized actor to its original size,
            you would have to track the original size manually: the
            actor doesn't make its original size accessible.</para>
          </listitem>

          <listitem>
            <para>Scaling can easily change the apparent size
            of multiple actors inside a container. For example, say you
            wanted to shrink multiple actors inside a container
            to half their original size. There are two ways you
            could do this:</para>

            <orderedlist>
              <listitem>
                <para>The hard way would be to resize
                each actor individually. You couldn't just resize the container,
                as resizing a container doesn't resize its children: usually
                they will be clipped so that they are either partially or
                wholly hidden.</para>
              </listitem>

              <listitem>
                <para>The easy way would be to set the container's scale
                to half its initial value: the actors
                in the container would retain their original sizes, but would
                appear at half size.</para>
              </listitem>
            </orderedlist>
          </listitem>
        </itemizedlist>

      </section>

      <section>
        <title>Scaling, layouts and containers</title>

        <para>It is possible to scale actors inside containers. For
        example, if you were using a <type>ClutterBox</type>
        which has a <type>ClutterBoxLayout</type> layout manager,
        you could scale the children of that layout.</para>

        <para>However, you should remain aware that layout managers
        don't take account of the scale of their children, only their
        size. So if you scale up an actor inside a layout manager,
        it may overlap other actors in the layout: the size allocated
        by the layout manager doesn't increase as an actor's scale
        increases.</para>

        <para>Similarly, scaling an actor down doesn't reduce the space
        it will be allocated by a layout.</para>
      </section>

      <section>
        <title>Setting the scale center</title>

        <para>An actor's scale center is the point around which
        scaling occurs: when you scale the actor, it will "shrink"
        into (if scale &lt; 1.0) or "expand" out of (if scale &gt; 1.0)
        its scale center.</para>

        <para>You can change an actor's scale center using
        either gravity (a named position on the actor; for example, the
        middle of the top edge of the actor is
        <constant>CLUTTER_GRAVITY_NORTH</constant>); or
        x,y coordinates relative to the actor's anchor point (by default,
        the anchor point for an actor is at <code>0,0</code>).</para>

        <para>Setting scale gravity has the same consequences as
        setting both the <varname>scale-center-x</varname> and
        <varname>scale-center-y</varname> properties for an actor.
        For example, <constant>CLUTTER_GRAVITY_NORTH_EAST</constant>
        sets the scale center to <code>&lt;width of the actor&gt;, 0</code>,
        relative to the actor's anchor point (defaults to the top-right
        corner of the actor). However, the advantage of scale
        gravities is that they change with the actor: so if the
        actor is resized, you don't have to manually reset the scale
        center. This means that <constant>CLUTTER_GRAVITY_NORTH_EAST</constant>
        will always represent the top-right corner of the actor,
        regardless of how it is scaled or resized. The same is true
        of each of the other scale gravities.</para>

        <para>If you're animating an actor's scale but want a different
        scale center, set it before the animation begins. One way to
        do this is to leave the actor's scale unchanged, but with
        a different scale center:</para>

        <informalexample>
          <programlisting>
<![CDATA[
gdouble scale_x;
gdouble scale_y;

/* get the actor's current scale */
clutter_actor_get_scale (actor, &scale_x, &scale_y);

/* set scale center using x,y coordinates, leaving scale unchanged;
 * the actor's size here is assumed to be 200x200
 */
clutter_actor_set_scale_full (actor,
                              scale_x,
                              scale_y,
                              100.0,  /* center x */
                              100.0   /* center y */);

/* set scale center using gravity, leaving scale unchanged */
clutter_actor_set_scale_with_gravity (actor,
                                      scale_x,
                                      scale_y,
                                      CLUTTER_GRAVITY_CENTER);
]]>
          </programlisting>
        </informalexample>

        <para>Another approach is to set scale center properties
        via GObject, which doesn't require you to figure out the
        actor's scale first:</para>

        <informalexample>
          <programlisting>
<![CDATA[
/* set scale center using x,y coordinates */
g_object_set (actor,
              "scale-center-x", 100.0, /* center x */
              "scale-center-y", 100.0, /* center y */
              NULL);

/* set scale center using gravity */
g_object_set (actor,
              "scale-gravity", CLUTTER_GRAVITY_CENTER,
              NULL);
]]>
          </programlisting>
        </informalexample>

        <para>Once the scale center is set, you can animate the
        scaling as per usual.</para>

        <para>It is even possible to animate the
        <varname>scale-center-*</varname> properties, which can
        produce interesting, though slightly
        unpredictable, effects. It's usually better to change the
        scale center before the animation starts.</para>

        <para><link linkend="animations-scaling-example-1">The
        example</link> cycles through the available scale gravities,
        showing the effect on the animation of each of the scale
        centers.</para>

        <para>The <link linkend="animations-scaling-example-2">second
        example</link> shows how to combine scaling in and out on a
        texture, in response to mouse button presses. In this case,
        the scale gravity remains at <constant>CLUTTER_GRAVITY_NORTH_WEST</constant>
        (i.e. at the anchor point of the actor). However, the anchor
        point is moved to the coordinates of each double click on button 1
        (usually the left mouse button) or button 3 (usually the right
        mouse button); which in turn automatically moves the scale center
        before the texture is scaled. As a result, the texture
        "expands" or "contracts" around the clicked point,
        while the point remains still.</para>

        <warning>
          <para>One final caveat about scale centers: if an actor is
          already scaled, the scale center coordinates are relative to
          the <emphasis>real size</emphasis> of the actor, rather than
          its <emphasis>transformed</emphasis> size. This can result in
          a "jumping" effect if you change the scale center on
          a scaled actor.</para>

          <para>For example, you might set the scale gravity of an actor
          to <constant>CLUTTER_GRAVITY_WEST</constant>, then
          scale the actor to <code>0.5</code> on both axes. Later, you
          change the actor's scale gravity to
          <constant>CLUTTER_GRAVITY_EAST</constant>. The effect of this
          is to "jump" the actor to the right, so its right-hand edge
          is aligned with where it was at scale <code>1.0</code>.</para>

          <para>If this isn't desirable, you can just retain the scale
          center on a scaled actor, and only change it when the actor
          is unscaled.</para>
        </warning>
      </section>

    </section>

    <section id="animations-scaling-examples">
      <title>Full examples</title>

      <example id="animations-scaling-example-1">
        <title>Animated scaling of an actor using each of the
        scale gravities. Press any key to start the animation.</title>
        <programlisting>
<xi:include href="examples/animations-scaling.c" parse="text">
  <xi:fallback>a code sample should be here... but isn't</xi:fallback>
</xi:include>
        </programlisting>
      </example>

      <example id="animations-scaling-example-2">
        <title>Animated scaling (up and down) of a texture in response
        to button presses. Call with the path to an image as the
        first argument.</title>
        <programlisting>
<xi:include href="examples/animations-scaling-zoom.c" parse="text">
  <xi:fallback>a code sample should be here... but isn't</xi:fallback>
</xi:include>
        </programlisting>
      </example>
    </section>

  </section>

</chapter>