Codestin Search App

jeremy-lilly · 2026-06-18T21:26:46Z

This PR adds diagnostic functions for calculating subgrid fluxes for tracers that are determined by the configured closure.

The user exposed routines DiffusiveFluxX, DiffusiveFluxY, and DiffusiveFluxZ use the native Oceananigans kernels diffusive_flux_x, diffusive_flux_y, and diffusive_flux_z respectively.

jeremy-lilly · 2026-06-18T21:29:46Z

Hey @tomchor -- Just a heads up that I will be on vacation and unresponsive for the next ~week, but I wanted to get out for review before I left. Thanks!

tomchor · 2026-06-19T01:11:34Z

Hey @tomchor -- Just a heads up that I will be on vacation and unresponsive for the next ~week, but I wanted to get out for review before I left. Thanks!

That's alright, I'll also be out of office next week so that's pretty good timing lol

I'll try to review it tomorrow. Thanks for the PR!

tomchor · 2026-06-20T02:57:49Z

+const DiffusiveFluxX = CustomKFO{<:typeof(diffusive_flux_x)}
+const DiffusiveFluxY = CustomKFO{<:typeof(diffusive_flux_y)}
+const DiffusiveFluxZ = CustomKFO{<:typeof(diffusive_flux_z)}


Can we name these to XDiffusiveFlux etc? It matches previous notation in the repo and it sounds more like how one would say these in plain english

My rational for this naming convention was to match the diffusive_flux_* kernels from Oceananigans. If you still want it changed, I don't mind.

I see your reasoning, but we made the decision a while ago to standard the user interfaces like this even though the underlying kernels might be different. For example:

Oceanostics.jl/src/TurbulentKineticEnergyEquation.jl

Lines 270 to 276 in c678675

function TurbulentKineticEnergyZShearProductionRate(u′, v′, w′, U, V, W;

grid = u′.grid,

location = (Center, Center, Center))

validate_location(location, "TurbulentKineticEnergyZShearProductionRate")

return KernelFunctionOperation{Center, Center, Center}(shear_production_rate_z_ccc, grid,

u′, v′, w′, U, V, W)

end

Make sense to me! Done in dabec1a.

tomchor · 2026-06-20T02:59:39Z

+    DIFF_FLUX = TracerEquation.DiffusiveFluxX(model, model.grid, model.closure, model.closure_fields, 
+                                    Val(:a), model.tracers.a, model.clock, fields(model), model.buoyancy)


Can we align these? I think the suggestion below does it but it's kinda hard to tell on github

Suggested change

DIFF_FLUX = TracerEquation.DiffusiveFluxX(model, model.grid, model.closure, model.closure_fields,

Val(:a), model.tracers.a, model.clock, fields(model), model.buoyancy)

DIFF_FLUX = TracerEquation.DiffusiveFluxX(model, model.grid, model.closure, model.closure_fields,

Val(:a), model.tracers.a, model.clock, fields(model), model.buoyancy)

Happy to fix this as well, but wanted to check if you wanted all similar lines in this file changed similarly? It looks like the existing function calls are also indented like this.

Yeah, usually we align all these lines to improve readability. I'm open to other ways of alignment as well, but I think before it wasn't aligned with anything

Done in b579203.

tomchor

Nice! Made some very minor comments but I'm happy to merge after they're done. Thanks for the contribution!

...And also after tests pass... :)

jeremy-lilly · 2026-07-02T21:02:44Z

Nice! Made some very minor comments but I'm happy to merge after they're done. Thanks for the contribution!

...And also after tests pass... :)

No problem at all! Happy to contribute!

As for tests passing, just copying this here for myself:

[ Info:         Testing tracer terms
Tracer variance diagnostics tests: Error During Test at /home/runner/work/Oceanostics.jl/Oceanostics.jl/test/test_tracer_diagnostics.jl:158
  Got exception outside of a @test
  MethodError: no method matching diffusive_flux_x(::Int64, ::Int64, ::Int64, ::ImmersedBoundaryGrid{Float64, Periodic, Periodic, Bounded, RectilinearGrid{Float64, Periodic, Periodic, Bounded, Oceananigans.Grids.StaticVerticalDiscretization{OffsetArrays.OffsetVector{Float64, Vector{Float64}},

Probably a typo somewhere -- I will update you when I track it down : )

tomchor · 2026-07-04T00:43:08Z

@jeremy-lilly if it's easier I can also add you to the repo. That way you can run the tests yourself online and don't have to be running them locally (I assumed that's what the last post was about?)

Copilot

Pull request overview

This PR extends Oceanostics.TracerEquation with new diagnostic kernel-function operations for computing closure-determined subgrid diffusive tracer fluxes (diffusive_flux_x/y/z), and adds corresponding unit tests.

Changes:

Add DiffusiveFluxX, DiffusiveFluxY, and DiffusiveFluxZ diagnostics backed by Oceananigans’ diffusive_flux_x/y/z kernels.
Introduce tracer-specific aliases TracerDiffusiveFluxX/Y/Z.
Extend tracer diagnostics tests to cover the new diffusive flux diagnostics and their Field wrapping.

Reviewed changes

Copilot reviewed 2 out of 2 changed files in this pull request and generated 11 comments.

File	Description
`src/TracerEquation.jl`	Adds diffusive-flux diagnostic constructors, type aliases, and docstrings for tracer subgrid fluxes.
`test/test_tracer_diagnostics.jl`	Adds tests asserting the new diffusive-flux diagnostics construct correctly and can be wrapped as `Field`s.

💡 Add Copilot custom instructions for smarter, more guided reviews. Learn how to get started.

tomchor · 2026-07-05T14:21:21Z

+                                    Val(:a), model.tracers.a, model.clock, fields(model), model.buoyancy)
+    DIFF_FLUX_field = Field(DIFF_FLUX)
+    @test DIFF_FLUX isa TracerEquation.DiffusiveFluxX
+    @test DIFF_FLUX isa TracerDiffusiveFluxX
+    @test DIFF_FLUX_field isa Field


The convention in Oceanostics is that diagnostics can be available in the namespace as

Oceanostics.TracerEquation.TracerDiffusiveFluxX Oceanostics.TracerEquation.DiffusiveFluxX

or

Oceanostics.TracerDiffusiveFluxX

for all diagnostics.

But you can't do

Oceanostics.DiffusiveFluxX

because then it's ambiguous what is being fluxed.

So I suggest reviewing the exports to make sure they're in line.

Done in 3d63544.

tomchor · 2026-07-05T14:23:15Z

+function DiffusiveFluxX(model, tracer_name; kwargs...)
+    tracer_index = findfirst(x -> x == tracer_name, keys(model.tracers))
+    tracer = model.tracers[tracer_index]
+    return DiffusiveFluxX(model, model.grid, model.closure, model.closure_fields, Val(tracer_index), tracer, model.clock, fields(model), model.buoyancy; kwargs...)
+end


I agree with this comment! (And same for the other directions)

Done in 67ccb49.

tomchor · 2026-07-05T14:24:05Z

+function DiffusiveFluxX(model, grid, closure, closure_fields, val_tracer_index, tracer, clock, model_fields, buoyancy; location = (Face, Center, Center))
+    validate_location(location, "DiffusiveFluxX", (Face, Center, Center))
+    return KernelFunctionOperation{Face, Center, Center}(diffusive_flux_x, grid, closure, closure_fields, val_tracer_index, tracer, clock, model_fields, buoyancy)
+end


I think I agree with this too. Can you please double check @jeremy-lilly ?

Agreed, this is a good change to make -- see be529b9.

@tomchor, the other routines in TracerEquation.jl have a similar issue where there are redundant fields required -- I'm happy to clean those up as well if you want.

tomchor · 2026-07-04T00:51:23Z

Asked copilot to give a review but I guess it went a little crazy lol
I can go over each comment next week to weed out the non-relevant comments

tomchor · 2026-07-05T14:25:37Z

Resolved the comments by copilot that were repetitive and that I didn't agree with. The remainder comments are the relevant ones

jeremy-lilly · 2026-07-06T16:02:26Z

Resolved the comments by copilot that were repetitive and that I didn't agree with. The remainder comments are the relevant ones

Thank you @tomchor! I will look through these comments and resolve them as I have time -- I will ping you when ready for another look : )

…sure

jeremy-lilly · 2026-07-07T20:59:40Z

Nice! Made some very minor comments but I'm happy to merge after they're done. Thanks for the contribution!
...And also after tests pass... :)

No problem at all! Happy to contribute!

As for tests passing, just copying this here for myself:
[ Info:         Testing tracer terms
Tracer variance diagnostics tests: Error During Test at /home/runner/work/Oceanostics.jl/Oceanostics.jl/test/test_tracer_diagnostics.jl:158
  Got exception outside of a @test
  MethodError: no method matching diffusive_flux_x(::Int64, ::Int64, ::Int64, ::ImmersedBoundaryGrid{Float64, Periodic, Periodic, Bounded, RectilinearGrid{Float64, Periodic, Periodic, Bounded, Oceananigans.Grids.StaticVerticalDiscretization{OffsetArrays.OffsetVector{Float64, Vector{Float64}}, 
Probably a typo somewhere -- I will update you when I track it down : )

This should be fixed by 24fb64c. The diffusive_flux_x,y,z kernels require that model.closure != nothing, so we are now testing these on a separate instance of model, where model.closure == ScalarDiffusivity().

jeremy-lilly added 2 commits June 17, 2026 16:06

Add DiffusiveFluxX,Y,Z for closure determined subgrid fluxes

f6616f1

Add tests for TracerEquation.DiffusiveFluxX,Y,Z

b0051a8

tomchor reviewed Jun 20, 2026

View reviewed changes

Comment thread src/TracerEquation.jl Outdated

tomchor reviewed Jun 20, 2026

View reviewed changes

Fix typo

c186968

tomchor requested a review from Copilot July 4, 2026 00:43

Copilot started reviewing on behalf of tomchor July 4, 2026 00:43 View session

Copilot AI reviewed Jul 4, 2026

View reviewed changes

jeremy-lilly added 2 commits July 7, 2026 14:51

Test subgrid fluxes with a seperate model instance with a defined clo…

24fb64c

…sure

Add proper exports to DiffusiveFluxX,Y,Z and TracerDiffusiveFluxX,Y,Z

3d63544

jeremy-lilly added 4 commits July 7, 2026 15:10

Rename DiffusiveFluxX,Y,Z to X,Y,ZDiffusiveFlux

dabec1a

Clean up indentation

b579203

Better indexing of model.tracers

67ccb49

Simplify public API for X,Y,ZDiffusiveFlux

be529b9

	function TurbulentKineticEnergyZShearProductionRate(u′, v′, w′, U, V, W;
	grid = u′.grid,
	location = (Center, Center, Center))
	validate_location(location, "TurbulentKineticEnergyZShearProductionRate")
	return KernelFunctionOperation{Center, Center, Center}(shear_production_rate_z_ccc, grid,
	u′, v′, w′, U, V, W)
	end

		DIFF_FLUX = TracerEquation.DiffusiveFluxX(model, model.grid, model.closure, model.closure_fields,
		Val(:a), model.tracers.a, model.clock, fields(model), model.buoyancy)

Conversation

jeremy-lilly commented Jun 18, 2026

Uh oh!

jeremy-lilly commented Jun 18, 2026

Uh oh!

tomchor commented Jun 19, 2026

Uh oh!

Uh oh!

Choose a reason for hiding this comment

Uh oh!

Choose a reason for hiding this comment

Uh oh!

Choose a reason for hiding this comment

Uh oh!

Choose a reason for hiding this comment

Uh oh!

Choose a reason for hiding this comment

Uh oh!

Choose a reason for hiding this comment

Uh oh!

Choose a reason for hiding this comment

Uh oh!

Choose a reason for hiding this comment

Uh oh!

tomchor left a comment • edited Loading Uh oh! There was an error while loading. Please reload this page.

Uh oh!

Choose a reason for hiding this comment

Uh oh!

jeremy-lilly commented Jul 2, 2026

Uh oh!

tomchor commented Jul 4, 2026

Uh oh!

Copilot AI left a comment

Choose a reason for hiding this comment

Pull request overview

Reviewed changes

Uh oh!

Uh oh!

Choose a reason for hiding this comment

Uh oh!

Choose a reason for hiding this comment

Uh oh!

Uh oh!

Uh oh!

Uh oh!

Uh oh!

tomchor Jul 5, 2026 • edited Loading Uh oh! There was an error while loading. Please reload this page.

Uh oh!

Choose a reason for hiding this comment

Uh oh!

Choose a reason for hiding this comment

Uh oh!

Uh oh!

Uh oh!

Choose a reason for hiding this comment

Uh oh!

Choose a reason for hiding this comment

Uh oh!

tomchor commented Jul 4, 2026

Uh oh!

tomchor commented Jul 5, 2026

Uh oh!

jeremy-lilly commented Jul 6, 2026

Uh oh!

jeremy-lilly commented Jul 7, 2026

Uh oh!

Reviewers

Assignees

Labels

Projects

Milestone

Development

Uh oh!

3 participants

tomchor left a comment •

edited

Loading

tomchor Jul 5, 2026 •

edited

Loading