Faces area in a group ?

Matt666

De rien mon initiateur de Ruby

todd burch

Personally, I'd call that a bug in the #area method if it is not taking the scaling factor into consideration.

Didier Bur

Hi Todd,
You know, groups are almost like component instances, they are stored in the model like "internal definitions" and a group is like an unique occurence of its definition, with or without a transformation applied. This is where the 'group' object in SU is wrong defined IMHO.
It's a shame we can't apply a transformation to a face, this will solve quickly the problem. I've written Face::transform myself, but for some unknown reason it sometimes raises an error "points are not coplanar" when applying the method to the edges of the face. Go figure...

PS: great presentation at 3DBC, but perhaps you need a little bit more training with the compo spray ?

Didier Bur

@AdamB: just let me call you the ruby wizard Voilà, c'est touT !

AdamB

(FWIW I'd say face.area giving objects space area is the correct behaviour.)

But people, this is all the wrong approach. You don't want to be building geometry, you just want to run through each face in the object and scale the area by the transformation. The trick is to project the scaling into the face's basis frame.

So the code in my book is:

def objectsurfacearea(container, xform)

	total = 0
	for ent in container.entities
		next unless ent.kind_of? Sketchup;;Face	

		# get face basis frame	
		normal = ent.normal
		binormal = Geom;;Vector3d.new(normal.y,normal.z,normal.x)
		tangent = (normal * binormal).normalize
		
		areascale = (xform * binormal).length * (xform * tangent).length

		total += ent.area * areascale
	end

	return total
end

So you'd do:
objectsurfacearea(mygroup, mygroup.transformation)

or

objectsurfacearea(myinstance.definition, myinstance.transformation)

Voila! C'est tous.

Adam

AdamB

@didier bur said:

Voilà, c'est touT !

Enfin! Apres vingt ans avec une epouse Francais, plus ca change avec mon Franglais..

Didier Bur

@unknownuser said:

une epouse FrancaisE

En effet, en effet...

thomthom

Sweet! Just what I needed.
How does it handle nested scaled groups/components? Do y ou have to add up the total transformation?

kat

Hey thanks a lot AdamB for your super professional approach!
Finally found what I was looking for, now my code works fantasticly.

I did some additional reading about the Frennet-Serret/TNB frame but I don't really understand everything.
If I get it wel, the normal(N) is perpendicular to the face, the tangent(T)'touches' the face and de binormal(B) is perpendicular to both normal and tangent. But then how do you know on which side of the face formed by N and T, is B? I don't get the
@unknownuser said:

tangent = (normal * binormal)
multiplication. Isn't B supposed to be the cross product of N and T?
And why is gives
@unknownuser said:

(xform * binormal).length * (xform * tangent).length

you the scale factor?

Could you give me some extra matematical information about this?
I would like to understand why this code works,

thanks, kat

AdamB

@kat said:

I don't get the
tangent = (normal * binormal)
multiplication. Isn't B supposed to be the cross product of N and T?

Makes no odds. Binormal and tangent both lie in the plane.

@kat said:

And why is gives

@adamb said:

(xform * binormal).length * (xform * tangent).length

you the scale factor?

I'm taking the unit vectors in the plane of the face and transforming them by the Group transform which may scale them. Getting the length in the binormal and tangent direction allows me to figure out how much to scale the area of the face accounting for the transform.

Adam