NEP29, update to Numpy 1.22[.3]

[IAlibay]

We won't be making a release in the next month, and June 23rd is when we go to Numpy 1.22+

Here I'm actually pinning builds to 1.22.3 because it makes things a bit cleaner (there's issues on a couple of archs and windows which makes life just eaiser to pin there). If we think we should pin as much as we can to 1.22.0, I can do that too if necessary.

PR Checklist

• Tests?
• Docs?
• CHANGELOG updated?
• Issue raised/referenced?

---

📚 Documentation preview 📚: https://mdanalysis--4160.org.readthedocs.build/en/4160/

[github-actions]

Linter Bot Results:

Hi @IAlibay! Thanks for making this PR. We linted your code and found the following:

There are currently no issues detected! 🎉

[codecov]

Codecov Report

Patch coverage: 100.00% and no project coverage change.

Comparison is base (07c9d1f) 93.61% compared to head (fb9c7cf) 93.61%.

Additional details and impacted files

@@           Coverage Diff            @@
##           develop    #4160   +/-   ##
========================================
  Coverage    93.61%   93.61%           
========================================
  Files          193      193           
  Lines        25168    25168           
  Branches      4058     4058           
========================================
  Hits         23560    23560           
  Misses        1092     1092           
  Partials       516      516           

Impacted Files	Coverage Δ
package/MDAnalysis/lib/mdamath.py	100.00% <100.00%> (ø)

☔ View full report in Codecov by Sentry.
📢 Do you have feedback about the report comment? Let us know in this issue.

[IAlibay]

This mypy error makes absolutely no sense to me - the numpy docs specifically say that input is expected to be array_like..

[RMeli]

Odd... From a quick glance at the source code it seems that it requires arrays though? Explicitly converting (if needed) seems to solve the issue:

diff --git a/package/MDAnalysis/lib/mdamath.py b/package/MDAnalysis/lib/mdamath.py
index 260bf0b87..d89805dca 100644
--- a/package/MDAnalysis/lib/mdamath.py
+++ b/package/MDAnalysis/lib/mdamath.py
@@ -104,7 +104,7 @@ def normal(vec1: npt.ArrayLike, vec2: npt.ArrayLike) -> npt.NDArray:
     .. versionchanged:: 0.11.0
        Moved into lib.mdamath
     """
-    normal: npt.NDArray = np.cross(vec1, vec2)
+    normal: npt.NDArray = np.cross(np.asarray(vec1), np.asarray(vec2))
     n = norm(normal)
     if n == 0.0:
         return normal  # returns [0,0,0] instead of [nan,nan,nan]
@@ -171,7 +171,7 @@ def stp(vec1: npt.ArrayLike, vec2: npt.ArrayLike, vec3: npt.ArrayLike) -> float:
     .. versionchanged:: 0.11.0
        Moved into lib.mdamath
     """
-    return np.dot(vec3, np.cross(vec1, vec2))
+    return np.dot(vec3, np.cross(np.asarray(vec1), np.asarray(vec2)))


 def dihedral(ab: npt.ArrayLike, bc: npt.ArrayLike, cd: npt.ArrayLike) -> float:

Proof: RMeli#3

[IAlibay]

@tylerjereddy as the numpy expert here, do you have any ideas why the array_like type check would be falling here?

[tylerjereddy]

I'm on vacation with just my phone until Sunday, but is there any code path where the arguments might be object arrays? I think static analysis prohibits that while runtime allows it, and in 1.22 release notes typing improvements are noted. @BvB93 implemented many of the NumPy static analysis improvements so they may know more.

[tylerjereddy]

Seems to expect bool type, I wonder why, can't check till Monday though

[BvB93]

Looking at the errors this seems to be another case of python/mypy#11347, an upstream bug wherein mypy fails to properly deal with overload ambiguity (due to a statically unknown dtype). There was a PR from ~1.5 years ago to fix the issue, but it seems to have gotten stuck in limbo.

[IAlibay]

Thanks so much for looking at this @BvB93 !

I'm going to treat this for what it is, and just type ignore. Hopefully that upstream issue will get addressed at some point!