fix location_at, position_at and tangent_at with PositionMode.LENGTH for Wire

[danieledapo]

Apparently Wire use a different parametrization than Edge because Edge has its LastParameter being the length of the Edge while Wirehas LastParameter being the numer of Edges with the fractional part being the parameter inside the edge.

The way the patch works is to always divide the distance provided by the user by the real length of the Wire/Edge so that we get a parameter in [0..1] which we can then pass straight to param_at to get the real parameter.

Fixes #572

[danieledapo]

Note: this makes the Edge version slower because we're doing calculation that it would not need, but I do not know if it's worth it to have something like the following in each method

if isinstance(self, Wire):
    parameter = self.param_at(distance / self.length)
else: # edge
   parameter = distance

but let me know what you think.

EDIT: this doesn't work because Edge.make_circle(2) has the same parametrization as of Edge.make_circle(1) or any other circle of any radius (it likely uses the start and end angle as the first and last parameter) which means we still have to do the division.

[gumyr]

Without any changes to the code, consider this test:

circle_wire = Wire(
    [
        Edge.make_circle(1, start_angle=0, end_angle=180),
        Edge.make_circle(1, start_angle=180, end_angle=360),
    ]
)
p1 = circle_wire.position_at(pi, position_mode=PositionMode.LENGTH)
p2 = circle_wire.position_at(pi / circle_wire.length)
print(p1, p2)

circle_edge = Edge.make_circle(1)
p3 = circle_edge.position_at(pi, position_mode=PositionMode.LENGTH)
p4 = circle_edge.position_at(pi / circle_edge.length)
print(p3, p4)

Vector(-0.9026853619331, -0.4303012170001, 0) Vector(-1, 0, 0)
Vector(-1, 0, 0) Vector(-1, 0, 0)

From this I conclude that the Edge case is currently correct and the Wire case is incorrect for PositionMode.LENGTH. Do we agree on this?

If so, position_at, tangent_at and location_at need different implementations for the Edge and Wire classes; therefore, these methods shouldn't be in the Mixin1D class.

[danieledapo]

From this I conclude that the Edge case is currently correct and the Wire case is incorrect for PositionMode.LENGTH. Do we agree on this?

Unfortunately, this test passes on my machine so I'm not quite sure how to debug it. I've committed and pushed it, let's see what the CI says.

[gumyr]

I erroneously recalled that the param_at was really expensive but when I measured it that isn't so - looks like a nice fix. Thanks!