Quality assurance tests via ExplicitImports + Aqua

[navidcy]

Add quality assurance tests via ExplicitImports

[navidcy]

Question: The type piracy errors for fields and prognostic_fields would be healed if

fields(::Nothing) = NamedTuple()
prognostic_fields(::Nothing) = NamedTuple()

are defined in Oceananigans rather than here?

[navidcy]

@giordano if you can have a look at the Aqua errors and help me with proposed solutions it'd be great!
I'm not sure what to do with the free_drift_u/free_drift_v ambiguities for example.

[giordano]

@giordano if you can have a look at the Aqua errors and help me with proposed solutions it'd be great! I'm not sure what to do with the free_drift_u/free_drift_v ambiguities for example.

In general there isn't a recipe for how to handle ambiguities, that's the point of being ambiguous.

In this case the problem is

ClimaSeaIce.jl/src/SeaIceDynamics/stress_balance_free_drift.jl

Lines 48 to 49 in 9552724

	const TISB = StressBalanceFreeDrift{<:Any, <:SemiImplicitStress}
	const BISB = StressBalanceFreeDrift{<:SemiImplicitStress, <:Any}

ClimaSeaIce.jl/src/SeaIceDynamics/stress_balance_free_drift.jl

Line 53 in 9552724

@inline function free_drift_u(i, j, k, grid, f::TISB, clock, fields)

ClimaSeaIce.jl/src/SeaIceDynamics/stress_balance_free_drift.jl

Line 79 in 9552724

@inline function free_drift_u(i, j, k, grid, f::BISB, clock, fields)

TISB constrains the first parameter of StressBalanceFreeDrift, BISB constrains the last one. What to do where you have the mixed case? 🤷 That's entirely up to who wrote the code.

[navidcy]

Yes. Perhaps @simone-silvestri is better to comment.
Or I can define a method for the mixed case that returns and error and we see if it ever gets triggered by the tests?