Hi Jim,
I've solved a similiar problem to yours, except for contour value
labels. Here's my controlChanged() method with a few comments:
.
.
.
public void controlChanged(ControlEvent e)
{
matrix = proj.getMatrix();
MouseBehaviorJ3D.unmake_matrix(rot_a, scale_a, trans_a, matrix);
double factor = first_scale/scale_a[0];
float f_scale = (float) ((scale_a[0] - first_scale)/scale_a[0]);
trans.getTransform(t3d);
Vector3f trans_vec
new Vector3f(f_scale*vertex[3*closest_vert+0],
f_scale*vertex[3*closest_vert+1],
f_scale*vertex[3*closest_vert+2]);
t3d.set((float)factor, trans_vec);
trans.setTransform(t3d);
}
-----
I use a little matrix algebra, and a little experimentation, to
compute the f_scale value, essentially I solve for translation
components required to place the label where I want it, at the
desired scale. It works very well, and under all combinations
of transformation although I can only prove it works for pure
scaling, and scaling with translation. The closest_vert is sort
of the anchor point for the label.
Hope this helps,
TomR
Bill Hibbard wrote:
>
> I wrote:
>
> > 2. Since you know that:
> >
> > Transform3D theTransform = new Transform3D();
> > theTransform.set( 1 / aScale[ 0 ], theTranslationVector );
> > theTransformGroup_.setTransform( theTransform );
> >
> > works for fixing the location and size, you might try:
> >
> > Transform3D theTransform = new Transform3D();
> > theTransform.set( 1 / aScale[ 0 ], theTranslationVector );
> > Transform3D scale = new Transform3D();
> > scale.set( 1 / aScale[ 0 ] );
> > theTransform.mul(scale);
> > theTransformGroup_.setTransform( theTransform );
>
> Clearly this should not re-apply the scale, but apply the
> inverse of the translation in 'scale'. That is:
>
> Transform3D theTransform = new Transform3D();
> theTransform.set( 1 / aScale[ 0 ], theTranslationVector );
> Transform3D translate = new Transform3D();
> Vector3d inverse = new Vector3d();
> inverse.x = -theTranslationVector.x;
> inverse.y = -theTranslationVector.y;
> inverse.z = -theTranslationVector.z;
> translate.set( inverse );
> theTransform.mul(translate);
> theTransformGroup_.setTransform( theTransform );
>
> > I may have some detail wrong, like the order of multiplying
> > matrices, but with a few experiments you may get it. When I
> > write such transform matrix code, I always have to experiment
> > to get it right.
>
> And of course, you can probably accomplish the same thing
> with your original code and experimenting with arithmetic
> operations on the x, y and z values in theTranslationVector.
>
> Good luck,
> Bill
> ----------------------------------------------------------
> Bill Hibbard, SSEC, 1225 W. Dayton St., Madison, WI 53706
> hibbard@xxxxxxxxxxxxxxxxx 608-263-4427 fax: 608-263-6738
> http://www.ssec.wisc.edu/~billh/vis.html
