make StatisticalMeasures a [weakdep]; add DefaultMeasuresExt.jl

Author: ablaom

Project.toml

@@ -26,27 +26,35 @@ Reexport = "189a3867-3050-52da-a836-e630ba90ab69"
 ScientificTypes = "321657f4-b219-11e9-178b-2701a2544e81"
 Serialization = "9e88b42a-f829-5b0c-bbe9-9e923198166b"
 StatisticalMeasures = "a19d573c-0a75-4610-95b3-7071388c7541"
+StatisticalMeasuresBase = "c062fc1d-0d66-479b-b6ac-8b44719de4cc"
 StatisticalTraits = "64bff920-2084-43da-a3e6-9bb72801c0c9"
 Statistics = "10745b16-79ce-11e8-11f9-7d13ad32a3b2"
 StatsBase = "2913bbd2-ae8a-5f71-8c99-4fb6c76f3a91"
 Tables = "bd369af6-aec1-5ad0-b16a-f7cc5008161c"
 
+[weakdeps]
+StatisticalMeasures = "a19d573c-0a75-4610-95b3-7071388c7541"
+
+[extensions]
+DefaultMeasuresExt = "StatisticalMeasures"
+
 [compat]
 CategoricalArrays = "0.9, 0.10"
 CategoricalDistributions = "0.1"
 ComputationalResources = "0.3"
 Distributions = "0.25.3"
 InvertedIndices = "1"
+LearnAPI = "0.1"
 MLJModelInterface = "1.7"
 Missings = "0.4, 1"
-LearnAPI = "0.1"
 OrderedCollections = "1.1"
 Parameters = "0.12"
 PrettyTables = "1, 2"
 ProgressMeter = "1.7.1"
 Reexport = "1.2"
 ScientificTypes = "3"
-StatisticalMeasures = "0.1"
+StatisticalMeasures = "0.1.1"
+StatisticalMeasuresBase = "0.1.1"
 StatisticalTraits = "3.2"
 StatsBase = "0.32, 0.33, 0.34"
 Tables = "0.2, 1.0"
@@ -60,8 +68,9 @@ Logging = "56ddb016-857b-54e1-b83d-db4d58db5568"
 MultivariateStats = "6f286f6a-111f-5878-ab1e-185364afe411"
 NearestNeighbors = "b8a86587-4115-5ab1-83bc-aa920d37bbce"
 StableRNGs = "860ef19b-820b-49d6-a774-d7a799459cd3"
+StatisticalMeasures = "a19d573c-0a75-4610-95b3-7071388c7541"
 Test = "8dfed614-e22c-5e08-85e1-65c5234f0b40"
 TypedTables = "9d95f2ec-7b3d-5a63-8d20-e2491e220bb9"
 
 [targets]
-test = ["DataFrames", "DecisionTree", "Distances", "Logging", "MultivariateStats", "NearestNeighbors", "StableRNGs", "Test", "TypedTables"]
+test = ["DataFrames", "DecisionTree", "Distances", "Logging", "MultivariateStats", "NearestNeighbors", "StableRNGs", "StatisticalMeasures", "Test", "TypedTables"]

ext/DefaultMeasuresExt.jl

@@ -0,0 +1,15 @@
+module DefaultMeasuresExt
+
+using MLJBase
+import MLJBase:default_measure, ProbabilisticDetector, DeterministicDetector
+using StatisticalMeasures
+using StatisticalMeasures.ScientificTypesBase
+
+default_measure(::Deterministic, ::Type{<:Union{Continuous,Count}}) = l2
+default_measure(::Deterministic, ::Type{<:Finite}) = misclassification_rate
+default_measure(::Probabilistic, ::Type{<:Union{Finite,Count}}) = log_loss
+default_measure(::Probabilistic, ::Type{<:Continuous}) = log_loss
+default_measure(::ProbabilisticDetector, ::Type{<:OrderedFactor{2}}) = area_under_curve
+default_measure(::DeterministicDetector, ::Type{<:OrderedFactor{2}}) = balanced_accuracy
+
+end # module

src/MLJBase.jl

@@ -89,8 +89,7 @@ import Distributions: pdf, logpdf, sampler
 const Dist = Distributions
 
 # Measures
-@reexport using StatisticalMeasures
-import StatisticalMeasures.StatisticalMeasuresBase
+import StatisticalMeasuresBase
 
 # from Standard Library:
 using Statistics, LinearAlgebra, Random, InteractiveUtils
@@ -312,4 +311,10 @@ export default_measure
 export pdf, sampler, mode, median, mean, shuffle!, categorical, shuffle,
    levels, levels!, std, Not, support, logpdf, LittleDict
 
+# for julia < 1.9
+if !isdefined(Base, :get_extension)
+    include(joinpath("..","ext", "DefaultMeasuresExt.jl"))
+    @reexport using .DefaultMeasuresExt.StatisticalMeasures
+end
+
 end # module

src/composition/models/stacking.jl

@@ -425,7 +425,7 @@ function internal_stack_report(
                 model_results.operation,
             ))
                 ypred = operation(mach, Xtest)
-                measurements = StatisticalMeasures.measurements(measure, ypred, ytest)
+                measurements = StatisticalMeasuresBase.measurements(measure, ypred, ytest)
 
                 # Update per observation:
                 model_results.per_observation[i][foldid] = measurements

src/default_measures.jl

@@ -9,14 +9,15 @@ reliably inferred.
 For Julia 1.9 and higher, `nothing` is returned, unless StatisticalMeasures.jl is
 loaded.
 
+# New implementations
+
+This method dispatches `default_measure(model, observation_scitype)`, which has
+`nothing` as the fallback return value. Extend `default_measure` by overloading this
+version of the method. See for example the MLJBase.jl package extension,
+DefaultMeausuresExt.jl.
+
 """
 default_measure(m) = nothing
 default_measure(m::Union{Supervised,Annotator}) =
     default_measure(m, nonmissingtype(guess_model_target_observation_scitype(m)))
 default_measure(m, S) = nothing
-default_measure(::Deterministic, ::Type{<:Union{Continuous,Count}}) = l2
-default_measure(::Deterministic, ::Type{<:Finite}) = misclassification_rate
-default_measure(::Probabilistic, ::Type{<:Union{Finite,Count}}) = log_loss
-default_measure(::Probabilistic, ::Type{<:Continuous}) = log_loss
-default_measure(::ProbabilisticDetector, ::Type{<:OrderedFactor{2}}) = area_under_curve
-default_measure(::DeterministicDetector, ::Type{<:OrderedFactor{2}}) = balanced_accuracy

test/composition/learning_networks/deprecated_machines.jl

@@ -9,6 +9,7 @@ using MLJBase
 using Tables
 using StableRNGs
 using Serialization
+using StatisticalMeasures
 rng = StableRNG(616161)
 
 # A dummy clustering model:

test/composition/learning_networks/nodes.jl

@@ -6,6 +6,7 @@ using MLJBase
 using ..Models
 using ..TestUtilities
 using CategoricalArrays
+using StatisticalMeasures
 import Random.seed!
 seed!(1234)
 

test/composition/learning_networks/signatures.jl

@@ -7,6 +7,7 @@ using Tables
 using Test
 using MLJModelInterface
 using OrderedCollections
+using StatisticalMeasures
 
 @testset "signatures - accessor functions" begin
     a = source(:a)

test/composition/models/network_composite.jl

@@ -1,4 +1,4 @@
-module TestNetowrkComposite
+module TestNetoworkComposite
 
 using Test
 using MLJBase
@@ -9,6 +9,7 @@ using Tables
 using MLJModelInterface
 using CategoricalArrays
 using OrderedCollections
+using StatisticalMeasures
 using Serialization
 
 const MMI = MLJModelInterface

test/composition/models/stacking.jl

@@ -2,11 +2,11 @@ module TestStacking
 
 using Test
 using MLJBase
+using StatisticalMeasures
 using MLJModelInterface
 using ..Models
 using Random
 using StableRNGs
-
 import Distributions
 
 rng = StableRNGs.StableRNG(1234)