Fix caching of TransformForest

[wkentaro]

What is the problem?

Caching in TransformForest seems not so robust.

If we have below program and we run it several times, I got below output, which is unexpected.

program

#!/usr/bin/env python

import numpy as np
import trimesh


scene = trimesh.Scene()
scene.add_geometry(trimesh.creation.axis(0.01))

scene.set_camera()
print('transform (old):\n', scene.camera.transform)
scene.camera.transform = np.eye(4)
print('transform (new):\n', scene.camera.transform)

scene.show()

output

% python spam.py
transform (old):
 [[1.         0.         0.         0.04514312]
 [0.         1.         0.         0.04509278]
 [0.         0.         1.         0.23255776]
 [0.         0.         0.         1.        ]]
transform (new):
 [[1. 0. 0. 0.]
 [0. 1. 0. 0.]
 [0. 0. 1. 0.]
 [0. 0. 0. 1.]]

% python spam.py
transform (old):
 [[1.         0.         0.         0.04514312]
 [0.         1.         0.         0.04509278]
 [0.         0.         1.         0.23255776]
 [0.         0.         0.         1.        ]]
transform (new):
 [[1.         0.         0.         0.04514312]
 [0.         1.         0.         0.04509278]
 [0.         0.         1.         0.23255776]
 [0.         0.         0.         1.        ]]

% python spam.py
transform (old):
 [[1.         0.         0.         0.04514312]
 [0.         1.         0.         0.04509278]
 [0.         0.         1.         0.23255776]
 [0.         0.         0.         1.        ]]
transform (new):
 [[1. 0. 0. 0.]
 [0. 1. 0. 0.]
 [0. 0. 1. 0.]
 [0. 0. 0. 1.]]

% python spam.py
transform (old):
 [[1.         0.         0.         0.04514312]
 [0.         1.         0.         0.04509278]
 [0.         0.         1.         0.23255776]
 [0.         0.         0.         1.        ]]
transform (new):
 [[1.         0.         0.         0.04514312]
 [0.         1.         0.         0.04509278]
 [0.         0.         1.         0.23255776]
 [0.         0.         0.         1.        ]]

The transform (new) must be always identity matrix (I expect) but it's not.
This problem is caused by the caching strategy in trimesh.scene.transforms.TransformForest, and this PR fixes it.

[coveralls]

Coverage remained the same at 84.09% when pulling b21f988 on wkentaro:caching_transform_forest into f42f7a6 on mikedh:master.

[coveralls]

Coverage remained the same at 84.09% when pulling b21f988 on wkentaro:caching_transform_forest into f42f7a6 on mikedh:master.

[coveralls]

Coverage remained the same at 84.09% when pulling b21f988 on wkentaro:caching_transform_forest into f42f7a6 on mikedh:master.

[mikedh]

Good catch! I wonder if _updated should just be a random number (i.e. in a tight loop does it execute faster than the time resolution.) I'll add a test as well- thanks for the PR!

[wkentaro]

Thanks!

On Mon, Apr 1, 2019 at 16:44 Michael Dawson-Haggerty < ***@***.***> wrote: Merged #363 <#363> into master. — You are receiving this because you authored the thread. Reply to this email directly, view it on GitHub <#363 (comment)>, or mute the thread <https://github.com/notifications/unsubscribe-auth/AEHFk-Tp-xws_v1HMR3XvQzLLCLoja7Xks5vcillgaJpZM4cV0FB> .

-- Kentaro Wada / 和田 健太郎 https://wkentaro.com <http://wkentaro.com>