black's failure gets fixed in #952.

This pull request has been mentioned on PlasmaPy Community Discourse Forum. There might be relevant details there:

https://plasmapy.discourse.group/t/e-cross-b-drift/57/4

Codecov Report

Merging #953 (8d967e7) into master (95aa647) will decrease coverage by 0.39%.
The diff coverage is 44.18%.

@@            Coverage Diff             @@
##           master     #953      +/-   ##
==========================================
- Coverage   96.29%   95.89%   -0.40%     
==========================================
  Files          62       63       +1     
  Lines        5715     5748      +33     
==========================================
+ Hits         5503     5512       +9     
- Misses        212      236      +24     

Continue to review full report at Codecov.

Legend - Click here to learn more
Δ = absolute <relative> (impact), ø = not affected, ? = missing data
Powered by Codecov. Last update 8beba18...d25ce97. Read the comment docs.

This PR would also address a couple of the points raised in issue #936

Merging this now would break Python 3.9 support, because that version broke Numba (numba/numba#6345) and we won't get that fixed until RC 2021.

For now, it is probably the best call to make Numba optional, so people can still install PlasmaPy on 3.9, and conditionally xfail the test on 3.9, so we can get this PR moving. Does that sound reasonable?

Could you possibly have a numba and non-numba version by doing something like....

try:
    import numba
    _range = numba.prange
except ImportError:
    warn(
        "If running Python < 3.9 consider installing numba.  This will allow parallel "
        "computing and speed up ParticleTracker"
    )
    _range = np.arange

def foo(*args):
    # use _range for the loops so the function is automatically a numba or 
    # non-numba version
    ...

if "numba" in globals:
    foo = numba.njit()(foo)

Then we might be able to use environment markers so numba is only installed on python < 3.9. (see also https://stackoverflow.com/a/32643122) I've never done this, so I'm not sure how well it will work.

It's still a for loop. If we do "optional Numba" here, the performance will suffer greatly over the current implementation.

It's still a for loop. If we do "optional Numba" here, the performance will suffer greatly over the current implementation.

I get that. The idea is that, with the environmental markers we can make numba a dependency for installs on python < 3.9 and thus utilized the numba functionality you're integrating here. If someone is using python >= 3.9, then numba will not be installed, the particle tracker functionality will revert to a slower form, and the user is warned about what is happening. However, the UI is made consistent across the python versions so when the numba with python39 issue is resolved the user does not have to change anything when make the patch on our end.

I went without numba for now. I figure we can wait until Q1 2021 and introduce less complexity then.

In the meantime, I went with refactoring the numpy version out of the class, and lo and behold, I fixed the ExB drift direction bug. I've got the numba version backed up, but for now, this should be good.

This pull request has been mentioned on PlasmaPy Community Discourse Forum. There might be relevant details there:

https://plasmapy.discourse.group/t/e-cross-b-drift/57/5

Should be good to go!

@namurphy here's the promised reminder :)