Spline's type can only be deserialized as CatmullRom

[xiaojiao-zsw]

When Spline is serialized, type will be serialized to SplineType.CatmullRom by default, regardless of the current type value.

jmonkeyengine/jme3-core/src/main/java/com/jme3/math/Spline.java

Lines 465 to 468 in 1f31c14

	public void write(JmeExporter ex) throws IOException {
	OutputCapsule oc = ex.getCapsule(this);
	oc.writeSavableArrayList((ArrayList) controlPoints, "controlPoints", null);
	oc.write(type, "type", SplineType.CatmullRom);

Also, when deserializing, type reads the key name as “pathSplineType” instead of “type”.

jmonkeyengine/jme3-core/src/main/java/com/jme3/math/Spline.java

Lines 497 to 509 in 1f31c14

	public void read(JmeImporter im) throws IOException {
	InputCapsule in = im.getCapsule(this);
	controlPoints = in.readSavableArrayList("controlPoints", new ArrayList<>());
	/* Empty List as default, prevents null pointers */
	float list[] = in.readFloatArray("segmentsLength", null);
	if (list != null) {
	segmentsLength = new ArrayList<>(list.length);
	for (int i = 0; i < list.length; i++) {
	segmentsLength.add(list[i]);
	}
	}
	type = in.readEnum("pathSplineType", SplineType.class, SplineType.CatmullRom);

[stephengold]

Good catch!
Thank you, @xiaojiao-zsw, for documenting this issue.
I'll write some tests and try to get this corrected in the next release.

[stephengold]

On closer inspection, the Spline.write() code might be correct:
The value written by OutputCapsule.write() is the 1st argument. The 3rd argument to OutputCapsule.write() is a default value that is never written.

However, the mismatch of names ("pathSplineType" versus "type") is clearly a bug.

[stephengold]

Testing uncovered another issue with spline serialization: in the case of an NURBS spline, the knot list is a non-null List<Float>. Line 483

jmonkeyengine/jme3-core/src/main/java/com/jme3/math/Spline.java

Line 483 in 98c0f9d

oc.writeSavableArrayList((ArrayList<Float>) knots, "knots", null);

causes a ClassCastException to be thrown in BinaryOutputCapsule.writeSavableArrayList() because Float cannot be cast to Savable.

[stephengold]

I submitted PR #2368 to solve the bugs, which apparently date back to at least 2014:

https://github.com/jMonkeyEngine/jmonkeyengine/blob/a16857c8f4e7aadf3ca5fcb79bff5c8f16eb5481/jme3-core/src/main/java/com/jme3/math/Spline.java

In other words, they aren't recent regressions. It appears they were missed due to inadequate test coverage.

[xiaojiao-zsw]

On closer inspection, the Spline.write() code might be correct: The value written by OutputCapsule.write() is the 1st argument. The 3rd argument to OutputCapsule.write() is a default value that is never written.

However, the mismatch of names ("pathSplineType" versus "type") is clearly a bug.

Oh, that was my mistake, it's now clear that this is entirely due to the mismatch of names causing type to only be loaded as CatmullRom when deserializing, thanks for that!

[xiaojiao-zsw]

On closer inspection, the Spline.write() code might be correct: The value written by OutputCapsule.write() is the 1st argument. The 3rd argument to OutputCapsule.write() is a default value that is never written.
However, the mismatch of names ("pathSplineType" versus "type") is clearly a bug.

Oh, that was my mistake, it's now clear that this is entirely due to the mismatch of names causing type to only be loaded as CatmullRom when deserializing, thanks for that!

Now I totally think “Spline's type can only be deserialized as CatmullRom” is the correct title for this issue.

[stephengold]

I changed the title as @xiaojiao-zsw suggested.

Possibly related: I'm puzzled as to why Spline can't be deeply cloned. I've opened a new issue (#2370) for this.