Fix shared array access on init

[banesullivan]

Resolve #2696, depends on #2698, and generally improves how meshes are created without copying source data when possible

@MatthewFlamm, please review as this is a follow up to your #2070

I'd like to add some more/better tests for this scenario

Memory usage comparison

`main`

This branch

Line # Mem usage Increment Occurrences Line Contents ============================================================= 6 138.4 MiB 138.4 MiB 1 @profile 7 def test(): 8 9 2427.2 MiB 2288.9 MiB 1 source = np.random.rand(100_000_000, 3) 10 11 2427.3 MiB 0.1 MiB 1 mesh = pv.PolyData() 12 4716.2 MiB 2288.9 MiB 1 mesh.points = source 13 14 4716.2 MiB 0.0 MiB 1 pts = mesh.points 15 4716.2 MiB 0.0 MiB 1 pts /= 2

Line # Mem usage Increment Occurrences Line Contents ============================================================= 6 138.2 MiB 138.2 MiB 1 @profile 7 def test(): 8 9 2427.0 MiB 2288.9 MiB 1 source = np.random.rand(100_000_000, 3) 10 11 2427.1 MiB 0.1 MiB 1 mesh = pv.PolyData() 12 2427.2 MiB 0.0 MiB 1 mesh.points = source 13 14 2427.2 MiB 0.0 MiB 1 pts = mesh.points 15 2427.2 MiB 0.0 MiB 1 pts /= 2

[codecov]

Codecov Report

Merging #2697 (e3bfa25) into main (975645d) will decrease coverage by 0.03%.
The diff coverage is 100.00%.

@@            Coverage Diff             @@
##             main    #2697      +/-   ##
==========================================
- Coverage   93.82%   93.78%   -0.04%     
==========================================
  Files          75       76       +1     
  Lines       16148    16208      +60     
==========================================
+ Hits        15151    15201      +50     
- Misses        997     1007      +10     

[banesullivan]

I'm working to wrap up https://github.com/pyvista/pyvista-xarray and need this to merge and patch release as soon as we can

[akaszynski]

I'm working to wrap up https://github.com/pyvista/pyvista-xarray and need this to merge and patch release as soon as we can

Per our agreement in our release discussion, we should give everyone 24 hours to review. However, I'm good to merge this sooner if critical.

[adeak]

My docstring changes seem to break docstring testing, but the failure seems to be in jupyter places(?) and I can't reproduce it locally.

Actually, things are falling apart everywhere in CI. What did I break?! 😄

[adeak]

@akaszynski we're almost done here:

1. One discussion about copy=False change, seems resolved to me.
2. One discussion about fixing up StucturedGrid's __init__() and docs, for which I've opened Clean up StructuredGrid documentation and __init__() #2733
3. Fix shared array access on init #2697 (comment) which needs input.

Approving anyway because 3. is still correct as is.

[akaszynski]

1. One discussion about copy=False change, seems resolved to me.

I think this is resolved as-is given @MatthewFlamm's comment:

Although the filters that it was implemented for specifically used copy=False which would further suggest this as the better default.

Furthermore, I'm only seeing one "filter" using this find_containing_cell, which doesn't modify the point. As for datasets that are using it, it makes sense that we don't copy as we're assigning points, as in its use in core/dataset.py

$ git grep _coerce_pointslike_arg
core/dataset.py:from pyvista.utilities.common import _coerce_pointslike_arg
core/dataset.py:        points = _coerce_pointslike_arg(points)
core/dataset.py:        point = _coerce_pointslike_arg(point, copy=False)
core/dataset.py:        point = _coerce_pointslike_arg(point, copy=False)
utilities/common.py:def _coerce_pointslike_arg(
utilities/geometric_objects.py:from pyvista.utilities.common import _coerce_pointslike_arg
utilities/geometric_objects.py:    points = _coerce_pointslike_arg(points)

2. One discussion about fixing up StucturedGrid's __init__() and docs, for which I've opened Clean up StructuredGrid documentation and __init__() #2733

That indeed needs to be cleaned up, but outside this PR.

3. #2697 (comment) which needs input.

Resolved in e552144. Thanks for pointing this out!

[adeak]

Still LGTM.