Matthew Allum
mallum@openedhand.com
Emmanuele Bassi
ebassi@linux.intel.com
Creating Animations with Clutter With Clutter using hardware accelration for graphics rendering, complex and fast animations are possible. This chapter describes basic techniques and the utilities Clutter provides in aiding animation creation.
Basic Animations The most basic way to create animations with Clutter is via the use of g_timeout_add(). This enables a callback function to be called at a defined interval. The callback function can then modify actors visual properties as to produce an animation. Simple timeout example Implement a rotating actor using 360 "frames" struct RotationClosure { ClutterActor *actor; gdouble final_angle; gdouble current_angle; }; static gboolean rotate_actor (gpointer data) { struct RotationClosure *clos = data; clutter_actor_set_rotation (clos->actor, clos->current_angle, 0, 0, 0); /* add one degree */ clos->current_angle += 1.0 if (clos->current_angle == clos->final_angle) return FALSE; return TRUE; } static void rotate_actor_cleanup (gpointer data) { struct RotationClosure *clos = data; g_object_unref (clos->actor); g_free (clos); } ... struct RotationClosure *clos = NULL; clos = g_new (struct RotationClosure, 1); clos->actor = g_object_ref (an_actor); clos->final_angle = 360.0; clos->current_angle = 0; g_timeout_add_full (1000 / 360, /* 360 updates in one second */ rotate_actor, clos, rotate_actor_cleanup); Priorities %G_PRIORITY_DEFAULT should always be used as the timeouts priority (in case of g_timeout_add_full()) as not to intefere with Clutter's scheduling of repaints and input event handling.
Timelines Using g_timeout_add() to control an animation is complicated and does not work in concert with the rest of the operations Clutter must perform for each redraw cycle. For this reason, Clutter provides #ClutterTimeline, a class that allows scheduling animations with a definite duration. Timelines are advanced during the redraw cycle so that animations are ready to be performed at the right time. This also means that animations will not affect the event processing; it also means that if the animation is too complex it will be called with a longer delay, thus not blocking the whole UI. A Timeline is created with: clutter_timeline_new (duration_in_milliseconds); The duration of the timeline then be modifed via the #ClutterTimeline:duration property or by using clutter_timeline_set_duration(). A timeline is started via clutter_timeline_start() and its playback further manipulated by the clutter_timeline_pause(), clutter_timeline_stop(), clutter_timeline_rewind() and clutter_timeline_skip() functions. By attaching a handler to the timeline's #ClutterTimeline::new-frame signal a timeline can then be used to drive an animation by altering an actor's visual properties. The callback looks like: void on_new_frame (ClutterTimeline *timeline, gint elapsed_msecs, gpointer user_data) { } The elapsed_msecs parameter is set to the amount of time elapsed since the beginning of the timeline, and its value is always between 0 and the #ClutterTimeline:duration value. The function clutter_timeline_get_progress() can also be used to get a normalised value of the timeline's current position between 0 and 1. Timelines can also be played in reverse by setting the direction using clutter_timeline_set_direction(), and can also have a one-time delay set before they begin playing by using the function clutter_timeline_set_delay(). Timelines can also control a pyshical simulation; the clutter_timeline_get_delta() function allows retrieving the number of milliseconds elapsed since the previous callback to ensure the physics engine to be able to take the actual time elapsed between iterations into account. Using a Timeline to drive an animation Rewrite the example above with a #ClutterTimeline instead of using g_timeout_add() #include <clutter/clutter.h> static void on_new_frame (ClutterTimeline *timeline, gint elapsed_msecs, ClutterActor *actor) { gdouble angle = 360 * clutter_timeline_get_progress (timeline); clutter_actor_set_rotation (actor, CLUTTER_Z_AXIS, angle, clutter_actor_get_width (actor) / 2, clutter_actor_get_height (actor) / 2, 0); } ... ClutterTimeline *timeline; timeline = clutter_timeline_new (1000); /* one second */ clutter_timeline_set_loop (timeline, TRUE); g_signal_connect (timeline, "new-frame", G_CALLBACK (on_new_frame), actor); clutter_timeline_start (timeline); Multiple timelines can be sequenced in order by using a #ClutterScore. See the #ClutterScore documentation for more details on using this.
Behaviours With a large application containing many animations, the use of just timelines can become unwieldy and difficult to manage, with much code duplication in the #ClutterTimeline::new-frame handlers that can require over-complex code changes for minor animation modifications. To ease these problems the #ClutterAlpha and #ClutterBehaviour classes were created. #ClutterAlpha and #ClutterBehaviour attempt to generalise the #ClutterTimeline::new-frame function by defining common actions (or behaviours) that can be quickly modified, applied to multiple actors or mixed on a single actor. A #ClutterAlpha is a 'function of time' (and does not refer to the alpha channel of a RGBA color). It is created by referencing a source timeline and an "easing mode" whichproduces a value between -1.0 and 2.0 depending on the progress of the timeline. Clutter provides various easing modes, as described by the #ClutterAnimationMode enumeration. It is also possible to register new animation modes using the function clutter_alpha_register_func() or to provide a custom #ClutterAlphaFunc for a specific #ClutterAlpha instance. A #ClutterBehaviour is created with a #ClutterAlpha and a set of parameters for whatever the behaviour modifies in an actor. The value of a #ClutterAlpha during the animation is then mapped to a value for the behaviour parameters and then applied on the actors referenced by the #ClutterBehaviour. With the #ClutterAlpha's underlying timeline playing the produced value will change and the behaviour will animate an actor. A #ClutterBehaviour is effectively 'driven' by a supplied #ClutterAlpha and when then applied to an actor it will modify a visual property or feature of the actor dependant on the Alpha's value. For example, a path-based behaviour applied to an actor will alter its position along a #ClutterPath, depending on the current alpha value over time. The actual progress of the motion will depend on the chosen "easing mode". Multiple behaviours can of course be applied to an actor as well as a single behaviour being applied to multiple actors. The separation of timelines, alphas and behaviours allows for a single timeline to drive many behaviours each potentially using different alpha functions. Behaviour parameters can also be changed on the fly.
Effects of alpha functions on a path
The actors position between the path's end points directly correlates to the #ClutterAlpha's current alpha value driving the behaviour. With the #ClutterAlpha's animation mode set to %CLUTTER_LINEAR the actor will follow the path at a constant velocity, but when changing to %CLUTTER_EASE_SINE_IN_OUT the actor initially accelerates before quickly decelerating.
 The behaviours included in Clutter are: #ClutterBehaviourDepth Changes the depth of actors #ClutterBehaviourEllipse Moves actors along an elliptical path #ClutterBehaviourOpacity Changes the opacity of actors #ClutterBehaviourPath Moves actors along a path #ClutterBehaviourRotate Rotates actors along an axis #ClutterBehaviourScale Changes the scaling factors of actors Using a #ClutterBehaviour The following example demonstrates an ellipse behaviour in action. #include <clutter/clutter.h> int main (int argc, char *argv[]) { ClutterTimeline *timeline; ClutterBehaviour *behave; ClutterAlpha *alpha; ClutterActor *stage, *actor; GdkPixbuf *pixbuf; clutter_init (&argc, &argv); stage = clutter_stage_get_default (); actor = clutter_texture_new_from_file ("ohpowers.png, NULL); clutter_container_add_actor (CLUTTER_CONTAINER (stage), actor); /* set up the animation to be 4 seconds long */ timeline = clutter_timeline_new (4000); clutter_timeline_set_loop (timeline, TRUE); /* set up a sinusoidal easing mode to power the behaviour; the * alpha will take a reference on the timeline so we can safely * release the reference we hold */ alpha = clutter_alpha_new_full (timeline, CLUTTER_EASE_SINE_IN_OUT); g_object_unref (timeline); /* the behaviour will own the alpha by sinking its floating * reference (if needed) */ behave = clutter_behaviour_ellipse_new (alpha, 200, /* center x */ 200, /* center y */ 400, /* width */ 300, /* height */ CLUTTER_ROTATE_CW, /* direction */ 0.0, /* initial angle */ 360.0); /* final angle */ clutter_behaviour_apply (behave, actor); clutter_actor_show_all (stage); clutter_timeline_start (timeline); clutter_main(); /* clean up; behaviours are top-level objects */ g_object_unref (behave); return 0; } The parameters of a #ClutterBehaviour can be changed whilst a animation is running. There can be many #ClutterAlpha's attached to a single timeline. There can be many behaviours for a #ClutterAlpha. There can be many behaviours applied to an actor. A #ClutterScore can be used to chain many behaviours together. Combining behaviours that effect the same actor properties (i.e two separate paths) will cause unexpected results. The values will not be merged in any way with only the last applied behaviour taking precedence. Tips for implementing a new behaviour can be found here.
Implicit Animations Using behaviours for simple animations of a single actor may be too complicated, in terms of memory management and bookkeeping of the object instances. For this reason, Clutter also provides a simple animation API for implicit animations using properties of an actor: clutter_actor_animate(). The clutter_actor_animate() family of functions will create and use an implicit #ClutterAnimation instance, which will then handle the animation of one or more #ClutterActor properties between a range of values. Using clutter_actor_animate() The following example demonstrates how to use the clutter_actor_animate() method to tween an actor between the current position and a new set of coordinates. The animation takes 200 milliseconds to complete and uses a linear progression. clutter_actor_animate (actor, CLUTTER_LINEAR, 200 "x", 200, "y", 200, NULL); The clutter_actor_animate() method returns a #ClutterAnimation instance that can be used to start, stop and modify the animation while it's running. The #ClutterAnimation::completed signal will be emitted when the animation has been completed. When the animation is complete it will be automatically unreferenced, and disposed if nothing else is holding a reference on it. Animating inside an event handler The following example demonstrates how to animate an actor inside the signal handler for a button press event. If the user presses the button on a new position while the animation is running, the animation will be restarted with the new final values updated. static gboolean on_button_press (ClutterActor *actor, ClutterEvent *event, gpointer user_data) { gfloat event_x, event_y; clutter_event_get_coords (event, &event_x, &event_y); clutter_actor_animate (actor, CLUTTER_EASE_SINE_OUT, 500, "x", event_x, "y", event_y, NULL); return TRUE; } Calling clutter_actor_animate() multiple times on an actor which is being animated will cause the animation to be updated with the new values. If you need to chain up multiple animations created using clutter_actor_animate() you should connect to the #ClutterAnimation::completed signal using g_signal_connect_after() to have the guarantee that the current #ClutterAnimation has been detached from the actor. The documentation for clutter_actor_animate() has further examples.
Conclusion Clutter provides a number of utility classes to aid animations and complex animations can be produced by combining the various features provided. Of course animations can becreated outside of the Clutter animation framework, as the framework is not expected to cover every kind of possible animation scenario. The animation functionality in Clutter is primarily suited to building animations with a set or finite running time - i.e transitions between states. For animations involving variable input (such as touchscreen handling) physical simulations may be more suited.