mathutils.geometry.intersect_line_line does not behave consistently for 2D vectors on Linux #101591

New Issue

Dion Moult · 2022-10-04T13:46:08+02:00

Dion Moult commented

2022-10-04 13:46:08 +02:00

Apologies if I missed something in the documentation, but from what I read I assumed intersect_line_line to take two lines defined by 4 points and find their intersection. Therefore I would expect any 2D lines that aren't parallel to always have a point of intersection.

However the behaviour seems to depend on a few factors - is this documented anywhere, and is this intended?

import bpy
from bpy import data as D
from bpy import context as C
from mathutils import *
from math import *

a = [Vector((0,0)), Vector((0,1))]
b = [Vector((0.8,1)), Vector((0,1))]

# Returns nan vector
print(geometry.intersect_line_line(a[0], a[1], b[0], b[1]))

# Returns correct first vector
print(geometry.intersect_line_line(a[1], a[0], b[0], b[1]))

# Returns correct both vectors
print(geometry.intersect_line_line(a[1], a[0], b[1], b[0]))

# Returns 0,1 as expected
print(geometry.intersect_line_line(a[0].to_3d(), a[1].to_3d(), b[0].to_3d(), b[1].to_3d()))

I am able to recreate it but only on Linux. I've tested the latest 3.3 stable downloaded from Blender.org. It seems sporadic - if I restart Blender, run the same code (in the text editor) sometimes it will work, sometimes I will get nans or different results in the 2D vectors vs the 3D vectors.

Here are a few examples of what might be printed:

(Vector((0.0, 1.0)), Vector((0.0, 1.0)))
(Vector((0.0, 1.0)), Vector((0.0, 1.0)))
(Vector((0.0, 1.0)), Vector((0.800000011920929, 1.0)))
(Vector((0.0, 1.0, 0.0)), Vector((0.0, 1.0, 0.0)))
# ... restart blender and try again ...
(Vector((0.0, 1.0)), Vector((0.0, 1.0)))
(Vector((0.0, 1.0)), Vector((0.0, 1.0)))
(Vector((0.0, 1.0)), Vector((1.7022413225232148e-40, 1.0)))
(Vector((0.0, 1.0, 0.0)), Vector((0.0, 1.0, 0.0)))
# ... restart blender and try again ...
(Vector((nan, nan)), Vector((nan, nan)))
(Vector((0.0, 1.0)), Vector((nan, nan)))
(Vector((0.0, 1.0)), Vector((0.0, 1.0)))
(Vector((0.0, 1.0, 0.0)), Vector((0.0, 1.0, 0.0)))

Here's another snippet that has odd results:

import bpy
from bpy import data as D
from bpy import context as C
from mathutils import *
from math import *
 
a = [Vector((-0.08, -0.88)), Vector((3.47, -1.6))]
b = [Vector((0, -1.6)), Vector((0, -0.6))]
 
- Shouldn't these two results be the same?
- print((a[0], a[1], b[0], b[1]))
print(geometry.intersect_line_line(a[0], a[1], b[0], b[1]))
- When using 3D vectors, I get the expected result (i.e. returns 2 same coords, where X ordinate = 0)
- print((a[0].to_3d(), a[1].to_3d(), b[0].to_3d(), b[1].to_3d()))
print(geometry.intersect_line_line(a[0].to_3d(), a[1].to_3d(), b[0].to_3d(), b[1].to_3d()))

I've tested it on my Linux (Gentoo 64bit) machine and am able to recreate weird results now and then. Another user helped test on another Linux machine (Ubuntu 22.04 64bit - also downloaded binary from Blender.org website) and was able to get nan vectors about 10% of the time. A Windows user was not able to reproduce. I was also not able to reproduce on another Windows machine.

I've noticed that when using 3D vectors, the results tend to always be correct, whereas the weird result only occurs with 2D vectors.

Apologies if I missed something in the documentation, but from what I read I assumed intersect_line_line to take two lines defined by 4 points and find their intersection. Therefore I would expect any 2D lines that aren't parallel to always have a point of intersection. However the behaviour seems to depend on a few factors - is this documented anywhere, and is this intended? ``` import bpy from bpy import data as D from bpy import context as C from mathutils import * from math import * a = [Vector((0,0)), Vector((0,1))] b = [Vector((0.8,1)), Vector((0,1))] # Returns nan vector print(geometry.intersect_line_line(a[0], a[1], b[0], b[1])) # Returns correct first vector print(geometry.intersect_line_line(a[1], a[0], b[0], b[1])) # Returns correct both vectors print(geometry.intersect_line_line(a[1], a[0], b[1], b[0])) # Returns 0,1 as expected print(geometry.intersect_line_line(a[0].to_3d(), a[1].to_3d(), b[0].to_3d(), b[1].to_3d())) ``` I am able to recreate it but _only_ on Linux. I've tested the latest 3.3 stable downloaded from Blender.org. It seems sporadic - if I restart Blender, run the same code (in the text editor) sometimes it will work, sometimes I will get nans or different results in the 2D vectors vs the 3D vectors. Here are a few examples of what might be printed: ``` (Vector((0.0, 1.0)), Vector((0.0, 1.0))) (Vector((0.0, 1.0)), Vector((0.0, 1.0))) (Vector((0.0, 1.0)), Vector((0.800000011920929, 1.0))) (Vector((0.0, 1.0, 0.0)), Vector((0.0, 1.0, 0.0))) # ... restart blender and try again ... (Vector((0.0, 1.0)), Vector((0.0, 1.0))) (Vector((0.0, 1.0)), Vector((0.0, 1.0))) (Vector((0.0, 1.0)), Vector((1.7022413225232148e-40, 1.0))) (Vector((0.0, 1.0, 0.0)), Vector((0.0, 1.0, 0.0))) # ... restart blender and try again ... (Vector((nan, nan)), Vector((nan, nan))) (Vector((0.0, 1.0)), Vector((nan, nan))) (Vector((0.0, 1.0)), Vector((0.0, 1.0))) (Vector((0.0, 1.0, 0.0)), Vector((0.0, 1.0, 0.0))) ``` Here's another snippet that has odd results: ``` import bpy from bpy import data as D from bpy import context as C from mathutils import * from math import * a = [Vector((-0.08, -0.88)), Vector((3.47, -1.6))] b = [Vector((0, -1.6)), Vector((0, -0.6))] - Shouldn't these two results be the same? - print((a[0], a[1], b[0], b[1])) print(geometry.intersect_line_line(a[0], a[1], b[0], b[1])) - When using 3D vectors, I get the expected result (i.e. returns 2 same coords, where X ordinate = 0) - print((a[0].to_3d(), a[1].to_3d(), b[0].to_3d(), b[1].to_3d())) print(geometry.intersect_line_line(a[0].to_3d(), a[1].to_3d(), b[0].to_3d(), b[1].to_3d())) ``` I've tested it on my Linux (Gentoo 64bit) machine and am able to recreate weird results now and then. Another user helped test on another Linux machine (Ubuntu 22.04 64bit - also downloaded binary from Blender.org website) and was able to get nan vectors about 10% of the time. A Windows user was not able to reproduce. I was also not able to reproduce on another Windows machine. I've noticed that when using 3D vectors, the results tend to always be correct, whereas the weird result only occurs with 2D vectors. ![image.png](https://archive.blender.org/developer/F13613644/image.png)

Dion Moult commented

2022-10-04 13:46:08 +02:00

Added subscriber: @Moult