Giving a figure a transform context

TransformedFigureTutorial

This tutorial illustrates how to write a figure that contains its own transform context. On the FigureTutorial page, we showed how to make a custom figure, and how to transform the various 2D shapes in the paint() method. Here, we will use an AffineTransform to do the same thing. This technique is a little more work to figure out how to do, but it probably better if your figure has more than a couple of Shapes in it.

Transforms are a little tricky to get right, so the Diva Canvas provides a class, TransformContext, that you need to use to give a figure its own transform. Each instance of TransformContext contains a single AffineTransform, and a bunch of methods that deal with it.

The start of the CloudFigure class contains this code:

    private TransformContext _transformContext;
    private Rectangle2D _cachedBounds = null;
    private Shape _cachedShape = null;

    public CloudFigure (
            double x, double y,
            double width, double height ) {

        _transformContext = new TransformContext(this);
        AffineTransform at = _transformContext.getTransform();
        at.translate(x,y);
        at.scale(width/100, height/100);
        _transformContext.invalidateCache();

	....
    } 

Now, because the shape of this figure is fairly expensive to transform, the two instance variables _cachedBounds and _cachedShape store the bounds and shape for the current transform. If you look at the source code for this class, you will see that these are created and remembered in getBounds() and getShape(). In getShape(), for example, the internally-stored shape needs to be transformed into "external" coordinates:

    public Shape getShape () {
        if (_cachedShape == null) {
            AffineTransform at = _transformContext.getTransform();
            _cachedShape = at.createTransformedShape(_shape);
        }
        return _cachedShape;
    }

    public void transform (AffineTransform at) {
        repaint();
        _cachedShape = null;
        _cachedBounds = null;
        _transformContext.preConcatenate(at);
        repaint();
    }

The only other interesting thing about this class is the paint() method. Because paint() is called recursively down the tree of figures, the TransformContext class provides two methods that "stack" transform contexts as the tree is traversed. The paint() method calls push() and pop() before and after painting the figure's contents:

    public void paint (Graphics2D g) {
        _transformContext.push(g);

	....
        // Paint the big cloud
        AlphaComposite c = AlphaComposite.getInstance(
                AlphaComposite.SRC_OVER,0.5f);
        g.setComposite(c);
        g.setPaint(Color.magenta);
        g.fill(_shape);

	....
        _transformContext.pop(g);
    }