cumulative_integrate() method

[johnomotani]

Provides the functionality of scipy.integrate.cumulative_trapezoid.

• Tests added
• Passes pre-commit run --all-files
• User visible changes (including notable bug fixes) are documented in whats-new.rst

[dcherian]

There's a cumtrapz in https://github.com/fujiisoup/xr-scipy/blob/master/xrscipy/integrate.py. Does that help?

[johnomotani]

There's a cumtrapz in https://github.com/fujiisoup/xr-scipy/blob/master/xrscipy/integrate.py. Does that help?

I did a quick check and it seems like applying scipy.integrate.cumtrapz to a dask array returns a numpy array, whereas the duck_array_ops.cumulative_trapezoid that I copied from duck_array_ops.trapz does return a dask array (so I assume it must not force a compute?). Maybe @fujiisoup has thoughts though?

BTW the test failures are in test_cftimeindex.py so I assume they're unrelated to these changes.

[fujiisoup]

There's a cumtrapz in https://github.com/fujiisoup/xr-scipy/blob/master/xrscipy/integrate.py. Does that help?

Now most of xr-scipy functionalities are already implemented in xarray and also I couldn't take time to maintain this package.

I think basic functionalities would be better to be integrated into xarray itself and cumulative_trapezoid would be a good candidate, as integrate is already there.

The implementation looks good to me.
I didn't find any edge cases where duck_array_ops.cumulative_trapezoid behaves differently from scipy.integrate.cumtrapz.

[max-sixty]

If we had a .expanding method (we only have .rolling at the moment), then I guess this could be .expanding(dim).integrate(). Is that right?

(but given we don't yet, no need to delay this)

[max-sixty]

Great — shall we merge?

[shoyer]

This is definitely welcome functionality! My only question is whether it should be a flag in the existing integrate method or a new method altogether, e.g., cumulative_integrate?

I would lean towards a new method because the function signature is different (not removing a dimension).

[johnomotani]

The vote seems to be for a separate cumulative_integrate method - I've made that change.

[max-sixty]

Thanks a lot @johnomotani !