BUG: Fix typo in vector operations and enable plot modifications for spherical geometry

[jisuoqing]

• Fixed typos in vector operations in spherical geometry
• Enable Streamline and LineIntegralConvolution callbacks for spherical geometry
• Callback tests added

Some further thoughts -- when using Quiver, Streamline and LineIntegralConvolution callbacks for curvilinear geometries, we assume users know what vector fields to plot (_cartesian_x, _conic_x, _cylindrical_radius etc.), while this is obscure and hidden in the source code. Maybe improvement could be made in future to generate a warning message when users try to use inappropriate vector fields in those callbacks.

[neutrinoceros]

Thanks ! could you provide a simple demo script so we can verify that Streamlines and LIC callbacks just work ?

[neutrinoceros]

oof, thank you for catching these !

[jisuoqing]

Sure! Let me add a test for this recently.

[neutrinoceros]

Some further thoughts -- when using Quiver, Streamline and LineIntegralConvolution callbacks for curvilinear geometries, we assume users know what vector fields to plot (_cartesian_x, _conic_x, _cylindrical_radius etc.), while this is obscure and hidden in the source code.
Maybe improvement could be made in future to generate a warning message when users try to use inappropriate vector fields in those callbacks.

+1/2 on this. Available fields could certainly be better documented, but I think Quiver, Streamline and LineIntegralConvolution callbacks should in principle accept any pair of scalar fields (i.e., not flag any usage as "inappropriate").

[jisuoqing]

+1/2 on this. Available fields could certainly be better documented, but I think Quiver, Streamline and LineIntegralConvolution callbacks should in principle accept any pair of scalar fields (i.e., not flag any usage as "inappropriate").

I agree that we won't throw an error, while it's unphysical, say, to plot velocity_r and velocity_theta on a plot along the phi-axis in spherical geometry (and the physical vector fields should be velocity_cylindrical_radius and velocity_cylindrical_z). Should a warning message appear to give a hint for the users?

[neutrinoceros]

Should a warning message appear to give a hint for the users?

IMHO it shouldn't. While velocity vectors and magnetic fields are by far the most common use cases for these callbacks, I don't think that we should invalidate creative usage in any way (no errors, no warnings).

[jisuoqing]

Well I just browsed the public dataset list on yt website, and cannot find a 3D spherical one with a non-zero coherent velocity/magnetic field, so I display a plot from my data as follows, and update some testing script as well.

[neutrinoceros]

pre-commit.ci autofix