4. Defining Dynamic Objects

By default, Jazz visual components are static, or still. That is, they have a persistent spatial position within the Jazz global coordinate system. Unless a visual component is redefined, or the transform of one of its ancestor nodes is modified, it stays in the same place. Of course, a camera can change its view so an object may to move on the screen, but that is really just changing the viewpoint into the scenegraph. The object has not moved within the scene.

However, there are many times when an application may want to define objects that are more dynamic. Perhaps they should be animated and move on their own, or should respond to some activity in the system. An object might change something about itself based on the current camera view. For instance, an object might behave normally, but never be allowed to be rendered below a minimum size. Or, it might pan normally with other objects, but when the camera changes magnification, the object wouldn't be magnified with the other objects, but instead would stay the same size. This might be useful for a label on top of an image. Or, an application might want to define a fisheye view where objects change their size and position based on the current camera view.

One obvious approach to creating dynamic objects would be to build the dynamic behavior into the object. So, if you want a label that doesn't shrink below 4 pixels, then you could define an object whose paint method would check the current magnification, and render itself bigger when the overall magnification was smaller. This approach involves a few details, but the big problem is that every object must be modified to implement the desired behavior. If you want to change behaviors, you must change all your objects! Also, there is no way to change the behavior at run-time with this approach.

For completeness, the details involved in implementing a dynamic object this way involve two primary things. 1) Since the object may be different depending on the camera viewpoint, the bounds computation must also be changed to reflect this; 2) Since the object is dynamic, Jazz must be informed of this so it doesn't use some efficiency mechanisms to pre-compute and cache certain things, such as the object's position. This is done by setting the volatileBounds property of the visual component using ZVisualComponent.setVolatileBounds()

A better approach to defining dynamic objects is to define the dynamic behavior completely separately from the visual components themselves. This allows the behavior to be written independently of specific object types, and allows behaviors to be dynamically associated with objects.

Jazz supports a general-purpose mechanism for creating dynamic behavior in a way that decouples the behavior definition from the object itself. The approach is to define a new node type that encapsulated this behavior, and then insert this node above whatever sub-tree you want to apply this behavior to.

Jazz uses this kind of special node itself in a few place. Perhaps the simplest kind just stores some information related task at hand. It doesn't define a dynamic behavior, but it does store some application-specific data in a way that doesn't modify the node. Jazz uses this approach to define ZAnchorGroup which stores the destination of a spatial hyperlink associated with a particular node.

Jazz also uses nodes to define some more complex dynamic behaviors. One is ZSelectionGroup which represents a selection rectangle around a visual component. It expands the bounds of the object to represent the extra size of the selection handle it draws. It is also dynamic because it draws the selection handle with constant width. This means that as the camera view is zoomed in and out, the selection handle is always one pixel wide.

A more interesting example is ZConstraintGroup. The constraint node changes the transform based on the current camera view so that the child below it gets rendered with a different transform (resulting in arbitrary scaling, translation, rotation, or shear). The first use of this constraint is to make various kinds of "sticky" objects.

ZStickyGroup makes objects "sticky" with respect to a camera. That is, as the camera changes its viewpoint, and most objects pan and zoom, the sticky objects don't move. It is as if they are stuck to the camera's lens. The sticky node works by applying the inverse of current camera view transform, effectively removing the camera's view changes from the rendering pipeline. This node has an alternative behavior called "sticky Z". In this case, the child objects pan from side to side normally as the camera view pans, but when the camera view changes magnification, the sticky Z child doesn't. This is designed for for labels which should stay with the relevant portion of the scene they are labelling, but shouldn't change size since a big or small label doesn't provide any useful function.

4.1 Fisheye Views

Custom nodes can even be used to create certain kinds of fisheye views in Jazz. A fisheye view is a general term to describe visualization systems that distort the content to show the focus area in more detail while showing the global context in less detail. By putting a fisheye decorator above all the components that you want to be distorted, we can create a fisheye view.

The following very simple fisheye view decorator changes the magnification of the component below the decorator so that objects that are closer to the center of the screen are bigger, and objects that are further from the center of the screen are smaller. To use this code, simply create a fisheye decorator for each object that you want to be affected by the fisheye view. A simple way to try this out is to modify the SquiggleEventHandler in the HiNote demo so that whenever you create a squiggle, it adds the fisheye decorator. That is, in SquiggleEventHandler.mousePressed(), change the polyline creation code to look like this:

	    polyline = new ZPolyline(pt);
	    ZVisualLeaf leaf = new ZVisualLeaf(polyline);
	    ZFisheyeGroup fisheye = new ZFisheyeGroup(camera, leaf);

Here's the code for the sample fisheye decorator:
import java.awt.geom.*;

import edu.umd.cs.jazz.*;
import edu.umd.cs.jazz.util.*;
import edu.umd.cs.jazz.component.*;

public class ZFisheyeGroup extends ZConstraintGroup {
    private AffineTransform constraintTransform = new AffineTransform();

     * Constructs a new fisheye group with a specified camera.
     * @param camera The camera the component is related to.
    public ZFisheyeGroup(ZCamera camera) {
     * Constructs a new fisheye group with a specified camera that 
     * decorates the specified child.
     * @param camera The camera the component is related to.
     * @param child The child that should go directly below this group.
    public ZFisheyeGroup(ZCamera camera, ZNode child) {

     * Computes the fisheye constraint that makes children larger in the 
     * center of the screen as the camera view changes.
     * @return the affine transform the defines the constraint.
    public AffineTransform computeTransform() {

	if (camera != null) {
	    Point2D viewCtr = camera.getViewBounds().getCenter2D();
	    double mag = camera.getMagnification();

				// We want to get the child's global bounds, but we 
                                // can't do it directly by calling node.getGlobalBounds() 
                                // because that would result in an infinitely recursive 
                                // call to this.  So, instead, we manually compute it.
	    ZBounds childBounds = new ZBounds();
	    ZNode[] childrenRef = getChildren();
	    for (int i=0; i<getNumChildren(); i++) {

	    Point2D childCtr = childBounds.getCenter2D();
	    double dist = viewCtr.distance(childCtr) / mag;
	    double s = 100.0f / dist;
	    constraintTransform.translate(childCtr.getX(), childCtr.getY());
	    constraintTransform.scale(s, s);
	    constraintTransform.translate(- childCtr.getX(), - childCtr.getY());

	return constraintTransform;