HyperSpy on pyodide (and jupyterlite)

[ericpre]

Required changes to get HyperSpy to work on pyodide: https://ericpre.github.io/jupyterlite-hyperspy
@CSSFrancis 😃

Progress of the PR

• Make numba and numexpr optional,
• Fallback to synchronous dask scheduler (dask.get) on pyodide, because dask.threaded doesn't exist - pyodide currently doesn't support threading,
• Replace dill by cloudpickle - dill doesn't work on pyodide,
• [n/a] update docstring (if appropriate),
• [n/a] update user guide (if appropriate),
• add an changelog entry in the upcoming_changes folder (see upcoming_changes/README.rst),
• Check formatting changelog entry in the readthedocs doc build of this PR (link in github checks)
• add tests,
• ready for review.

[codecov]

Codecov Report

Attention: 21 lines in your changes are missing coverage. Please review.

Comparison is base (b685684) 80.78% compared to head (ad43ad2) 80.82%.

Additional details and impacted files

@@                  Coverage Diff                   @@
##           RELEASE_next_major    #3255      +/-   ##
======================================================
+ Coverage               80.78%   80.82%   +0.04%     
======================================================
  Files                     140      140              
  Lines                   20360    20395      +35     
  Branches                 4820     4825       +5     
======================================================
+ Hits                    16447    16485      +38     
+ Misses                   2840     2834       -6     
- Partials                 1073     1076       +3     

Files	Coverage Δ
hyperspy/_components/bleasdale.py	100.00% <100.00%> (ø)
hyperspy/_components/exponential.py	82.35% <100.00%> (ø)
hyperspy/_components/expression.py	98.42% <100.00%> (+0.07%)	⬆️
hyperspy/_components/gaussian.py	97.33% <100.00%> (ø)
hyperspy/_components/gaussian2d.py	100.00% <ø> (ø)
hyperspy/_components/gaussianhf.py	96.22% <100.00%> (ø)
hyperspy/_components/logistic.py	100.00% <100.00%> (ø)
hyperspy/_components/lorentzian.py	97.29% <100.00%> (ø)
hyperspy/_components/polynomial.py	77.77% <100.00%> (ø)
hyperspy/_components/power_law.py	94.44% <ø> (ø)
... and 12 more

... and 1 file with indirect coverage changes

☔ View full report in Codecov by Sentry.
📢 Have feedback on the report? Share it here.

[CSSFrancis]

@ericpre Super cool! Let me know if you want me to test anything or review.

[ericpre]

@CSSFrancis, if you can have a look, that would be great! Now that numba is optional, we can also test python 3.12 without to wait for numba supporting it and I will follow up with a PR to add python 3.12!

[francisco-dlp]

LGTM other than the comments above.

[ericpre]

I have tried to use the following code to generate a file with a saved custom component to dumps the class and I can't even saved the file:

import hyperspy.api as hs
from hyperspy.component import Component

class CustomComponent(Component):

    def __init__(self, p1=1, p2=2):
        Component.__init__(self, ('p1', 'p2'))

        self.p1.value = p1
        self.p2.value = p2

        self.p1.grad = self.grad_p1
        self.p2.grad = self.grad_p2

    def function(self, x):
        p1 = self.p1.value
        p2 = self.p2.value
        return p1 + x * p2

    def grad_p1(self, x):
        return 0

    def grad_p2(self, x):
        return x


s = hs.signals.Signal1D(range(10))
m = s.create_model()

c = CustomComponent()
m.append(c)
m.store('a')

import hyperspy
version = hyperspy.__version__
s.save(f"hs{version}_custom_component.hspy")

It seems that saving custom component is broken., but surprisingly, the code is covered according to codecov...

[francisco-dlp]

@ericpre, would you like to fill a bug report for the pickling issue that you report above. I wonder if traits is to blame.