ThreeD bug fixes

[BrendanMartin]

This is a fix for #569.

Added import for new location of extract_mobject_family_members and removed self. from invocation in method add_fixed_in_frame_mobjects.

Acknowledgement

• I have read the Contributing Guidelines

[kolibril13]

Nice, thanks a lot for fixing this! :)
I just tested it with this script:

from manim import *

class Text3D3(ThreeDScene):
    def construct(self):
        axes = ThreeDAxes()
        self.set_camera_orientation(phi=75 * DEGREES,theta=-45*DEGREES)
        text3d=TextMobject("This is a 3D text")
        self.add_fixed_in_frame_mobjects(text3d)
        text3d.to_corner(UL)
        self.add(axes)

and it behaves like expected.
Great work!
Would you like to maybe add a test as well for this function?
That would be very beneficial for the further manimce dev!
Read the Manim CE wiki if you are not familiar with these tests.
In case you don't want to add the test, I would also give it an approve if you say that you don't have time for that and we add it to out todo list for a future pr.

[BrendanMartin]

Some other bugs I found while rendering my ThreeDScene's:

1. Error using move_camera with frame_center parameter:

class Example(ThreeDScene):
    def construct(self):
        cube = Cube()
        self.play(Animation(cube))
        self.move_camera(phi=PI / 4, theta=PI / 4, frame_center=[0, 0, -2])

Traceback (most recent call last):
  File "c:\users\brendan\appdata\local\programs\python\python38\lib\site-packages\manim\scene\three_d_scene.py", line 168, in move_camera
    ApplyMethod(self.renderer.camera.frame_center.move_to, frame_center)
AttributeError: 'numpy.ndarray' object has no attribute 'move_to'

Fixed by changing frame_center to _frame_center.

2. After working on a fix for the above, I found that camera.py was missing import for reduce function
   Fixed by adding import for reduce.

3. Running test_threed.CameraMoveTest showed that camera.py was referencing config["default_pixel_height"], config["default_pixel_width"] but there was no default_ for these values. Removed default_

Finally, made addition to CameraMoveTest by adding frame_center parameter.

The last commit fixes these new bugs.

[kolibril13]

There seems to be a small differnce in the created test data and the reference data: First unmatched index is at [ 30 168]: [153 153 153 255] != [154 154 154 255].
Is the test passing locally at your machine?

[BrendanMartin]

@kolibril13 yeah, the test passes locally.

[kolibril13]

ok, I will test this also locally on my machine right now.

[kolibril13]

Result, when running pytest it gives an error here:

E           AssertionError: The frames don't match. FixedInFrameMObject has been modified.
E           Please ignore if it was intended.
E           First unmatched index is at [29 54]: [ 15  15  15 255] != [ 20  20  20 255]

Same one as the one we got on github.
What operating system are you using?
I'm on Ubuntu 20.04.
As an idea to solve this, I could recreate the test data on my machine and commit it to your branch

[BrendanMartin]

I'm on Windows 10 64. If you have time to make the test data and commit that would be great. I'm not sure why my side passes and yours doesn't

[behackl]

TextMobject has been deprecated; please use Text.

Without specifying the font to use, there is no chance this test will pass on all of our supported operating systems: different systems use different default fonts, this is my guess why the tests are failing.

[kolibril13]

Good catch!
Thanks for this!
Therefore, I think it would be best if you could just update your test, @BrendanMartin.

[behackl]

Actually, this is wrong as has been pointed out on Discord: TextMobject was deprecated, but it is the old equivalent to what is Tex nowadays. I have no idea why the test outputs are different in that case, unfortunately.

[kolibril13]

And one more idea:
It would be super nice to have this example in the example gallery:
https://docs.manim.community/en/latest/examples/3d.html

[BrendanMartin]

@kolibril13 the circle is definitely a better idea. In the test you removed self.wait(), but the only way for the test to pass for me is to include that. Otherwise the result is empty pixels.

[kolibril13]

@kolibril13 the circle is definitely a better idea. In the test you removed self.wait(), but the only way for the test to pass for me is to include that. Otherwise the result is empty pixels.

Thanks.
Oh, can you maybe push these changes with this new test data?
Recently I made a fix for this missing self.wait() but never tested it with 3d scenes: #653

[kolibril13]

Wuhu, tests are not failing anymore!

[BrendanMartin]

Thanks for the help!