Having these serialVersionUID fields screws up my script which converts the
float code to double code. Meh. I guess I can improve it so that it preserves
the double versions... yaks!
The correct way to mirror an angle in [-PI, PI] around the x-axis is simply to
negate it, which doesn't seem ceremonious enough to require a utility method.
The old mirrorAngle mirrors around the "y-axis" (PI/2 or -PI/2 as appropriate)
which is necessary for other code (particularly angularDifference) to work
correctly. The new mirrorAngleOrigin mirrors around the "x-axis" (zero).
- nixed all transforms except identity and affine; the intermediate transforms
were more trouble than they were worth
- fixed bugs in AffineTransform.translate/scaleX/scaleY/rotate; aiya!
- replaced Transform.clone with Transform.copy; deprecated clone
- rewrote transform test in Scala and using Java AffineTransform as a
reference.
XY provides a common super-interface for IPoint and IVector for APIs that don't
care if you have a point or vector, and just want it's x/y coordinates.
Returning a vector when subtracting two points was clever and all, but turns
out to be annoying. More often, you're just adjusting a point and don't want to
switch to vector-baesd math. If we want an easy way to obtain a vector that is
the difference of two points we can add Vectors.subtract() or a new
constructor.
Compilers are presumably smart enough to throw away the bulk of that
automagically, but some were storing the return value of a method call
of some object that implemented an interface whic his a lot harder for
it to realize won't have side effects.
I've tried to make a useful distinction between Point and Vector, but sometimes
you just want to use a Vector as a Point. I'm not going to add Point3 and try
to push this distinction into the third dimension, so I'll accommodate using
Vector as Point in 2D, and we'll just use Vector3 as a 3D point when needed.
Michael Bayne