Fix of ImageMobject opacity bug

[henrikmidtiby]

Overview: What does this pull request change?

Closes #4072
The issue was introduced in #3960 as part of an optimization.

Further Information and Comments

I would suggest to add a unittest that ensures that the bug is not reintroduced.

The problem was related to handling of unsigned integers.
The code below demonstrates the problem.

import numpy as np

def main():
    start = np.array(127, dtype=np.uint8)
    end = np.array(25, dtype=np.uint8)
    alpha = 0.09739

    val1 = start + alpha * (end - start)
    val2 = (1 - alpha) * start + alpha * end
    print(val1)
    print(val2)
    print(141)
    print(117)

main()

Which outputs

141.99806
117.06622
141
117

Reviewer Checklist

• The PR title is descriptive enough for the changelog, and the PR is labeled correctly
• If applicable: newly added non-private functions and classes have a docstring including a short summary and a PARAMETERS section
• If applicable: newly added functions and classes are tested

[chopan050]

Thanks for noticing this! As mentioned in Discord, this was caused by an underflow when the end value is less than the start value and both of them are unsigned integers. In your code, 25 - 127 = -102, but, because 0 is the lowest representable integer, it goes all the way around from 255 and down to 154 (which is -102 + 256) instead.

I'll think about a unit test. Any suggestions are well appreciated.

[henrikmidtiby]

I have added two unittests for the interpolate function. See commit 580e027.

[behackl]

Thanks!