Unexpected FlatExtrapolationBoundaryCondition behaviour in validation script cylinder.jl

[samlewin]

I've been playing around with open outflow boundary conditions and noticed that the validation code for uniform flow past a cylinder no longer functions as intended (v0.94.3) https://github.com/CliMA/Oceananigans.jl/blob/main/validation/open_boundaries/cylinder.jl.

The u boundary conditions are currently set as follows: https://github.com/CliMA/Oceananigans.jl/blame/fe4123f92155f5086bce72e366843045606b4bf5/validation/open_boundaries/cylinder.jl#L47-L48

I guess the implementation has changed since the example was written, since the initialized u velocity field is now forced to zero at the east boundary when it should be a uniform flow throughout:

Something similar to the oscillating cylinder flow example, like
u_boundaries = FieldBoundaryConditions( east = FlatExtrapolationOpenBoundaryCondition(u, parameters = U, relaxation_timescale = 1), west = FlatExtrapolationOpenBoundaryCondition(u, parameters = U, relaxation_timescale = 1))
seems to produce a reasonable looking flow. Is there still a way to implement the east boundary outflow without specifying a particular outflow profile, as was originally the case?

[jagoosw]

Hi @samlewin,

Sorry about this! FlatExtrapolation is currently broken and being fixed in this PR #3937

I'll finish it up today so we can get it merged. I'm also working on a proper open boundary condition here if you're interested.

[samlewin]

Ah, sorry I had missed that PR, thanks! Will keep an eye on the WIP you mentioned :)