fix: apply subset of fixes for downstream dask usage

[agoose77]

TL;DR

• Adds cumsum to TypeTracer
• Removes to_rectilinear from nplikes, replaces with layout.to_backend_array

[jpivarski]

This looks good, although there's at least one use of to_rectilinear in the tests, which is where this is failing right now.

awkward/tests/test_1247_numpy_to_rectilinear_ndarray.py

Line 13 in 6cde4b3

assert nplike.to_rectilinear(array).tolist() == ak_array.to_list()

[jpivarski]

Good catch!

[agoose77]

This looks good, although there's at least one use of to_rectilinear in the tests, which is where this is failing right now.

awkward/tests/test_1247_numpy_to_rectilinear_ndarray.py

Line 13 in 6cde4b3

assert nplike.to_rectilinear(array).tolist() == ak_array.to_list()

I could have sworn I ran the test suite locally. Oh, well, thank goodness for CI!

[jpivarski]

It came up in the Linux-ROOT test; it might be a test that requires ROOT in the specific environment to run. Although I see that another one failed, too, so maybe not.

[jpivarski]

In c569a71, I simply removed the failing test. It was testing an aspect of to_rectilinear, but since the function no longer exists, I'd say that the property we intended to test is now obsolete.

[codecov]

Codecov Report

Merging #2283 (8716fc8) into main (495f592) will decrease coverage by 0.02%.
The diff coverage is 63.63%.

Additional details and impacted files

Impacted Files	Coverage Δ
src/awkward/_nplikes/cupy.py	40.84% <ø> (-0.26%)	⬇️
src/awkward/_nplikes/jax.py	83.78% <ø> (+1.73%)	⬆️
src/awkward/_nplikes/numpy.py	62.50% <ø> (-1.50%)	⬇️
src/awkward/_nplikes/numpylike.py	73.80% <ø> (+0.06%)	⬆️
src/awkward/_nplikes/typetracer.py	73.82% <27.27%> (-0.59%)	⬇️
src/awkward/_connect/numpy.py	92.99% <100.00%> (ø)
src/awkward/contents/unionarray.py	84.06% <100.00%> (+0.02%)	⬆️
src/awkward/operations/ak_concatenate.py	96.22% <100.00%> (ø)
src/awkward/operations/ak_unflatten.py	95.52% <100.00%> (ø)