rft: improve parameter constructors by haakon-e · Pull Request #700 · CliMA/CloudMicrophysics.jl

haakon-e · 2026-03-11T22:07:26Z

Summary

This pull request refactors the type hierarchy and usage of parameterized types throughout the codebase to simplify type signatures and improve code clarity. The main change is the removal of the floating-point type parameter (FT) from many struct definitions and function signatures, in favor of storing the floating-point type only where necessary (typically as fields rather than in the type itself). This results in shorter, more readable type signatures and reduces the need for repeated type parameters in function definitions. Additionally, some keyword argument usage is improved for clarity.

The PR is organized with two commits:

The first introduces the struct improvements
The second contains the show.jl methods

Content

Type System Simplification and Refactoring:

The abstract ParametersType no longer has the FT field in type space. For backward compatibility, the eltype function remains, and recursively looks up the first argument of the concrete struct. Within microphysics parameterizations, FT should instead be fetched from the state.
Relatedly, structs that do not explicitly store elements of type FT, no longer carry this information in type space.

API and Usability Improvements:

Construction of SB2006 objects in plotting scripts now uses keyword arguments for clarity (is_limited = false), replacing positional boolean arguments. [1] [2] [3]

Struct printing

Added generic show methods (see show.jl) that structs can use to enable pretty-printing of their contents.
By default, all parameter structs (i.e. subtypes of ParametersType) with more than 7 elements will be verbosely printed, and otherwise tersely printed. Printing is recursive, as the example below illustrates.
Optionally, you can attach units to elements of the structure for more informative printing.

julia> import CloudMicrophysics.Parameters as CMP

julia> CMP.SB2006(Float32)
SB2006
├─ pdf_c: CloudParticlePDF_SB2006(νc = 1.0f0, μc = 1.0f0, xc_min = 4.2f-15, xc_max = 2.6f-10, ρw = 1000.0f0)
├─ pdf_r: RainParticlePDF_SB2006_limited
│  ├─ νr = -0.6666667
│  ├─ μr = 0.33333334
│  ├─ xr_min = 2.6e-10
│  ├─ xr_max = 5.0e-6
│  ├─ N0_min = 250000.0
│  ├─ N0_max = 2.0e7
│  ├─ λ_min = 1000.0
│  ├─ λ_max = 10000.0
│  ├─ ρw = 1000.0
│  └─ ρ0 = 1.225
├─ acnv: AcnvSB2006(kcc = 4.44f9, x_star = 2.6f-10, ρ0 = 1.225f0, A = 400.0f0, a = 0.7f0, b = 3.0f0)
├─ accr: AccrSB2006(kcr = 5.25f0, τ0 = 5.0f-5, ρ0 = 1.225f0, c = 4.0f0)
├─ self: SelfColSB2006(krr = 7.12f0, κrr = 60.7f0, d = -5.0f0)
├─ brek: BreakupSB2006(Deq = 0.0009f0, Dr_th = 0.00035f0, kbr = 1000.0f0, κbr = 2300.0f0)
├─ evap: EvaporationSB2006(av = 0.78f0, bv = 0.308f0, α = 159.0f0, β = 0.266f0, ρ0 = 1.225f0)
└─ numadj: NumberAdjustmentHorn2012(τ = 100.0f0)

julia> CMP.CloudLiquid(Float32)  # the fields for this struct has units defined
CloudLiquid(τ_relax = 10.0f0 [s], ρw = 1000.0f0 [kg/m³], r_eff = 1.4f-5 [m], N_0 = 5.0f8 [1/m³])

Other Minor Improvements:

Remove Base. prefix from @kwdef macro as it is directly available.

haakon-e · 2026-03-11T22:07:28Z

This change is part of the following stack:

Update the CI to Julia v1.12 #704
- ci: update tests, JuliaFormatter v2 #706
  - rft: improve parameter constructors #700 ◀
    - fix: 2M tendencies, params #694
      - feat(P3): implement additional rates + some fixes #681

_{Change managed by git-spice.}

codecov · 2026-03-11T22:23:28Z

Codecov Report

❌ Patch coverage is 71.25000% with 92 lines in your changes missing coverage. Please review.
✅ Project coverage is 92.28%. Comparing base (610618b) to head (c909c1a).
⚠️ Report is 7 commits behind head on main.

Additional details and impacted files

@@            Coverage Diff             @@
##             main     #700      +/-   ##
==========================================
- Coverage   94.98%   92.28%   -2.70%     
==========================================
  Files          53       54       +1     
  Lines        2371     2269     -102     
==========================================
- Hits         2252     2094     -158     
- Misses        119      175      +56

Components	Coverage Δ
src	`93.28% <71.25%> (-2.77%)`	⬇️
ext	`69.47% <ø> (ø)`

🚀 New features to boost your workflow:

❄️ Test Analytics: Detect flaky tests, report on failures, and find test suite problems.

haakon-e

comments to assist review

haakon-e · 2026-03-12T17:00:49Z

docs/src/API.md

+# Developer docs
+
+```@autodocs
+Modules = [ShowMethods]
+```


The additional method signatures can be viewed in the doc preview

haakon-e · 2026-03-12T17:41:55Z

docs/src/API.md

-CMP.ParametersP3{Float64}
-CMP.ParametersP3(::Float64)
-CMP.ParametersP3(::Parameters.CP.ParamDict)
+CMP.ParametersP3


got some doc errors with this, so simplifying for now, will return improve this in a later PR

haakon-e · 2026-03-12T17:42:51Z

src/parameters/AbstractTypes.jl

+
+
+# Temporary fallback until we stop checking eltype from the parameters
+Base.eltype(p::ParametersType) = fieldcount(typeof(p)) > 0 ? eltype(getfield(p, 1)) : Any


will remove this method in a future breaking PR

haakon-e · 2026-03-12T17:45:27Z

src/parameters/Aerosol_H2SO4_Solution.jl

-H2SO4SolutionParameters(::Type{FT}) where {FT <: AbstractFloat} =
-    H2SO4SolutionParameters(CP.create_toml_dict(FT))
-


Definitions such as this is moved to src/parameters/Parameters.jl, which defines this for all subtypes of ParametersType

haakon-e · 2026-03-12T17:46:17Z

src/parameters/Aerosol_H2SO4_Solution.jl

-    FT = CP.float_type(td)
-    return H2SO4SolutionParameters{FT}(; parameters...)
+    return H2SO4SolutionParameters(; parameters...)


Removed these type flags where I could find them as they're unnecessary since all fields returned from CP.get_parameter_values have uniform FT type already.

haakon-e · 2026-03-12T17:46:51Z

src/parameters/Aerosol_H2SO4_Solution.jl

 # Fields
 $(DocStringExtensions.FIELDS)
 """
-Base.@kwdef struct H2SO4SolutionParameters{FT} <: ParametersType{FT}


@kwdef is always available, no need for Base. prefix. Did a search and replace to remove them.

haakon-e · 2026-03-12T17:54:55Z

src/parameters/Parameters.jl

+### Create methods that enables the creation of parameter types with a given float type
+### for all subtypes of `ParametersType`
+### e.g. `Microphysics1MParams(Float32)`
+import InteractiveUtils


InteractiveUtils is already shipped with the julia standard library, so this is not in reality a new dependency. We do need to add it to the Project.toml file though.

haakon-e · 2026-03-12T18:03:08Z

src/parameters/Parameters.jl

+    return concrete_types
+end
+for T in get_concrete_subtypes(ParametersType)
+    @eval (::Type{$T})(::Type{FT}; kw...) where {FT} = $T(CP.create_toml_dict(FT); kw...)


For future reference: The reason I needed to define these in a loop over concrete subtypes of ParametersType as opposed to simply defining the method

(::Type{T}(::Type{FT}; kw...) where {T <: ParametersType, FT} = T(CP.create_toml_dict(FT); kw...)

is that this results in Aqua/Test complaining about ambiguous method definitions for structs with one field. For example,

@kwdef struct NumberAdjustmentHorn2012{FT} <: ParametersType "Number concentration adjustment timescale [s]" τ::FT end

defines the method

(::Type{NumberAdjustmentHorn2012})(τ::Any)

Here, the struct is more concrete, while the argument is less concrete. With the method above, the struct is less concrete while the argument (::Type{FT}) is more concrete. In practice, it appears that julia is able to resolve this apparent ambiguity, so it may be a bug in Test/Aqua.

The @eval loop instead defines

(::Type{NumberAdjustmentHorn2012})(::Type{FT}) where {FT} = ...

for every subtype of ParametersType, circumventing this potential method ambiguity problem.

.github/workflows/ci.yml

.github/workflows/JuliaFormatter.yml

test/performance_tests.jl

trontrytel · 2026-03-13T22:26:53Z

Amazing! Thank you so much for carrying out for the whole repo.

trontrytel

Thank you!

I was wondering if we can replace some of the parameter printing here: https://clima.github.io/CloudMicrophysics.jl/dev/guides/literated/Parameters/

with the new show. But I can also do it in a separate PR

trontrytel · 2026-03-13T22:31:14Z

Is this a breaking release from ClimaAtmos point of view in any way?

trontrytel · 2026-03-14T05:11:09Z

Closes #573

- add constructor using float type (e.g. Float32) for all param structs - remove explicit type declarations where not needed - clean up some constructors

haakon-e · 2026-03-18T21:30:21Z

Thank you!

I was wondering if we can replace some of the parameter printing here: https://clima.github.io/CloudMicrophysics.jl/dev/guides/literated/Parameters/

with the new show. But I can also do it in a separate PR

Let's do it separately, since the Rain struct does contain quite a bit more information than is currently printed. It's not obvious to me if printing the entire struct is always useful?

Is this a breaking release from ClimaAtmos point of view in any way?

I don't think so, but I won't make a release here until the P3 PR is merged anyways, and I have a corresponding PR in Atmos

haakon-e force-pushed the he/rft-parameter-structs branch from 6b4ed07 to 9215200 Compare March 11, 2026 22:47

This was referenced Mar 11, 2026

feat(P3): implement additional rates + some fixes #681

Open

fix: 2M tendencies, params #694

Merged

haakon-e force-pushed the he/rft-parameter-structs branch 4 times, most recently from 1f4a917 to 7f380dc Compare March 12, 2026 17:40

haakon-e commented Mar 12, 2026

View reviewed changes

haakon-e force-pushed the he/rft-parameter-structs branch 2 times, most recently from e0cb5b6 to 35f69b6 Compare March 12, 2026 20:00

haakon-e requested a review from trontrytel March 12, 2026 20:13

haakon-e force-pushed the he/rft-parameter-structs branch from 35f69b6 to 62241f0 Compare March 12, 2026 21:03

haakon-e commented Mar 12, 2026

View reviewed changes

.github/workflows/ci.yml Outdated Show resolved Hide resolved

haakon-e commented Mar 12, 2026

View reviewed changes

.github/workflows/JuliaFormatter.yml Show resolved Hide resolved

haakon-e force-pushed the he/rft-parameter-structs branch 2 times, most recently from 0459d67 to 3bc6212 Compare March 12, 2026 22:53

trontrytel added the Needs review Please review my pull request label Mar 13, 2026

trontrytel reviewed Mar 13, 2026

View reviewed changes

test/performance_tests.jl Show resolved Hide resolved

trontrytel self-requested a review March 13, 2026 22:28

trontrytel approved these changes Mar 13, 2026

View reviewed changes

trontrytel removed the Needs review Please review my pull request label Mar 13, 2026

trontrytel added the Approved 🍀 label Mar 16, 2026

haakon-e changed the base branch from main to he/ci-update-tests-juliaformatter March 18, 2026 20:35

haakon-e force-pushed the he/rft-parameter-structs branch from be4ca3d to 6fa6924 Compare March 18, 2026 20:35

haakon-e mentioned this pull request Mar 18, 2026

ci: update tests, JuliaFormatter v2 #706

Merged

Base automatically changed from he/ci-update-tests-juliaformatter to aj/julia_v12 March 18, 2026 20:52

Base automatically changed from aj/julia_v12 to main March 18, 2026 20:53

rft: improve parameter constructors

9e24cc0

- add constructor using float type (e.g. Float32) for all param structs - remove explicit type declarations where not needed - clean up some constructors

haakon-e force-pushed the he/rft-parameter-structs branch from 6fa6924 to 1080882 Compare March 18, 2026 20:57

feat: add general show methods for param structs

c909c1a

haakon-e force-pushed the he/rft-parameter-structs branch from 1080882 to c909c1a Compare March 18, 2026 21:19

haakon-e merged commit 3aa2461 into main Mar 19, 2026
16 of 26 checks passed

haakon-e deleted the he/rft-parameter-structs branch March 19, 2026 21:24

haakon-e mentioned this pull request Mar 20, 2026

Make modules that depend on ClimaComms into an extension? CliMA/ClimaUtilities.jl#205

Open

giordano mentioned this pull request Mar 27, 2026

Bump CloudMicrophysics in / NumericalEarth/Breeze.jl#586

Open



		# Temporary fallback until we stop checking eltype from the parameters
		Base.eltype(p::ParametersType) = fieldcount(typeof(p)) > 0 ? eltype(getfield(p, 1)) : Any

		H2SO4SolutionParameters(::Type{FT}) where {FT <: AbstractFloat} =
		H2SO4SolutionParameters(CP.create_toml_dict(FT))

Conversation

haakon-e commented Mar 11, 2026 • edited Loading Uh oh! There was an error while loading. Please reload this page.

Uh oh!

Summary

Content

Uh oh!

haakon-e commented Mar 11, 2026 • edited Loading Uh oh! There was an error while loading. Please reload this page.

Uh oh!

Uh oh!

codecov bot commented Mar 11, 2026 • edited Loading Uh oh! There was an error while loading. Please reload this page.

Uh oh!

Codecov Report

Uh oh!

haakon-e left a comment

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!

Choose a reason for hiding this comment

Uh oh!

Uh oh!

Uh oh!

Uh oh!

trontrytel commented Mar 13, 2026

Uh oh!

trontrytel left a comment

Choose a reason for hiding this comment

Uh oh!

trontrytel commented Mar 13, 2026

Uh oh!

trontrytel commented Mar 14, 2026

Uh oh!

haakon-e commented Mar 18, 2026

Uh oh!

Uh oh!

Reviewers

Assignees

Labels

Projects

Milestone

Development

Uh oh!

2 participants

haakon-e commented Mar 11, 2026 •

edited

Loading

haakon-e commented Mar 11, 2026 •

edited

Loading

codecov bot commented Mar 11, 2026 •

edited

Loading