Outstanding review issues from #3096 (gridcell rotations)

[pp-mo]

Premature merge.
Outstanding comments :

• lots of stickler complaints
• Gridcell rotations #3096 (review)
• maybe reconsider Gridcell rotations #3096 (comment)
• ? Gridcell rotations #3096 (comment)
• ? consider : combine with other code for 3d latlon/xyz transform ? Gridcell rotations #3096 (comment)
• inherited + other comments to clarify
  • Gridcell rotations #3096 (comment)
  • Gridcell rotations #3096 (review)
  • Gridcell rotations #3096 (comment)
• should remove make-discontiguous routine : see Pin Dask to avoid 0.18.2 bug with masked arrays. #3127 (comment)

---

Missing test aspects : originally from #3103

CODE CHANGES

• angles routine should handle non-2d coords with meshgrid
  • Too tricky : ABANDONED ? but this reduces the generality of "rotate_grid_angles"

ANGLES TEST ASPECTS : testing of 'grid_angles'

• THINK THIS IS ALL NOW COVERED BY Vector plots 2 #3120 (check when that is accepted)
• (cube)
• unrecognised inputs (no shapes)
• missing y + x not cube
• coords
  • not 2d : (support?)
  • different shapes : error (except 1d?)
  • different cube dims (2d : ok for 1d)
  • different cs : error
  • no cs, but not angles, i.e. plain X+Y coords : ??
  • X/Y = coord/not, and not/coord
  • cell_angle_boundpoints unrecognised

ROTATE TEST ASPECTS : testing of 'rotate_grid_vectors'

• u+v
• u+v with angles supplied
• u+v with angles derived from u
• angles method key
• check that magnitudes remain unchanged
• check actual angles-rotation calculation (known vector + angles)
• check that non-2d cubes work (not supported now ?)
• check masking of invalid points
• check behaviour with 0-vectors ?
• Exceptions:
  • angles not angular (bad unit)
  • u+v not same shape (implied by arithmetic)
  • angles not same dims as u+v (implied by arithmetic)
• ?possibly~ : NaN angles handling ?

[pp-mo]

Checklist of things above

Outstanding comments :

• lots of stickler complaints (presumably, all fixed in Pin Dask to avoid 0.18.2 bug with masked arrays. #3127 : DONE)
• Gridcell rotations #3096 (review)
  • "Can aa i.e. the grid cell angles also be masked, or just nans? ... should we [not] return masked arrays in every case?"
  • "Can this be more explicit" (DONE)
  • "Any chance it would ever need to be upper-left, upper right?"
  • "Kinda confusing names 03_to_12 and 0_to_1"
  • " general purpose functions ... could we not put them somewhere else" / combine with other code for 3d latlon/xyz transform ?
• inherited + other comments to clarify
  • "Why are v and T used here? ... docs strings are not quite finalised yet?" / clarify inherited docstrings
  • "a bit confusing" / comments in _grid_angles. _latlon_from_xyz
  • "so many commas" / comments in _grid_angles. _latlon_from_xyz
• "remove this now that we're not doing contiguity checks" / remove make-discontiguous routine

---

[lbdreyer]

All of these comments have been dealt with so I am going to close this.