Do intrinsic_vector and extrinsic_vector belong in Operators?

[glwagner]

Originally, "operators" were regarded as difference, derivative, and reconstruction operators (eg the atomic operators required for finite volume arithmetic and calculus). Here we also have vector rotations:

Oceananigans.jl/src/Operators/vector_rotation_operators.jl

Line 23 in 8eaba36

@inline intrinsic_vector(i, j, k, grid::AbstractGrid, uₑ, vₑ, wₑ) =

I had expected to find these in Grids. So just wanted to have this discussion.

[navidcy]

I think you are right.

Perhaps the thinking was that these apply an "operation" to the vector quantities (rotating etc where appropriate). But even so, indeed I think it's a different sort of operation and they are grid-specific so probably fit in the Grids module better.

[glwagner]

What do you think @simone-silvestri

[simone-silvestri]

Hmmm, as a first impression, to me, they belong in the operators module because they are operators that enact a pointwise operation (vector rotation) on fields (like the other operators). However, they are probably more entangled with the concept of domain and grid than the other operators, so they might be ok to be in the grid module also.

[glwagner]

I think the question is not whether they are "an operator" (as in the dictionary definition of "operator") but rather whether they are similar to the other functions defined in that module

[simone-silvestri]

Having these functions in the operator module is what makes more sense to me. My logic for this opinion is that these functions are much more similar in structure and objective to the other operators rather than any function of the grid module which typically only require the grid itself as an input.
However, I am ok also to switch them in the grids module because I recognize that a rotation happens relative to a particular grid.

[glwagner]

We do have to mention the notion of the extrinsic and intrinsic coordinate system in the grids module though yes?

My concern is that we don't want a lot of distance between the definition of the grid, and the definition of its extrinsic and intrinsic coordinate system. If the Operators module were deleted, the grids mostly still stand alone (you just can't use them as easily). I feel like if that happened I would still want to know what the coordinate system of the grid was, even if I couldn't do calculus on it.

[simone-silvestri]

Agreed, we definitely have to mention it in the grids module.
Maybe we can come up with some abstraction like ExtrinsicCoordinateSystem(grid) that is used in the operators module to rotate the vectors.

[glwagner]

In the case that these operators serve to define the coordinate systems (which they do now), I think they have to belong in the Grids module because they are part of the definition of each grid.