Merge pull request #980 from JuliaAI/constructor

Have wrappers overload `constructor` trait

Author: ablaom

Project.toml

@@ -1,7 +1,7 @@
 name = "MLJBase"
 uuid = "a7f614a8-145f-11e9-1d2a-a57a1082229d"
 authors = ["Anthony D. Blaom <[email protected]>"]
-version = "1.3"
+version = "1.4.0"
 
 [deps]
 CategoricalArrays = "324d7699-5711-5eae-9e2f-1d82baa6b597"
@@ -47,7 +47,7 @@ DelimitedFiles = "1"
 Distributions = "0.25.3"
 InvertedIndices = "1"
 LearnAPI = "0.1"
-MLJModelInterface = "1.7"
+MLJModelInterface = "1.10"
 Missings = "0.4, 1"
 OrderedCollections = "1.1"
 Parameters = "0.12"
@@ -58,7 +58,7 @@ Reexport = "1.2"
 ScientificTypes = "3"
 StatisticalMeasures = "0.1.1"
 StatisticalMeasuresBase = "0.1.1"
-StatisticalTraits = "3.2"
+StatisticalTraits = "3.3"
 Statistics = "1"
 StatsBase = "0.32, 0.33, 0.34"
 Tables = "0.2, 1.0"

src/composition/models/pipelines.jl

@@ -599,6 +599,9 @@ end
 
 MMI.target_scitype(p::SupervisedPipeline) = target_scitype(supervised_component(p))
 
+MMI.package_name(::Type{<:SomePipeline}) = "MLJBase"
+MMI.load_path(::Type{<:SomePipeline}) = "MLJBase.Pipeline"
+MMI.constructor(::Type{<:SomePipeline}) = Pipeline
 
 # ## Training losses
 

src/composition/models/stacking.jl

@@ -264,16 +264,17 @@ function Base.setproperty!(stack::Stack{modelnames}, _name::Symbol, val) where m
 end
 
 
+# # TRAITS
+
 MMI.target_scitype(::Type{<:Stack{modelnames, input_scitype, target_scitype}}) where
     {modelnames, input_scitype, target_scitype} = target_scitype
 
 
 MMI.input_scitype(::Type{<:Stack{modelnames, input_scitype, target_scitype}}) where
     {modelnames, input_scitype, target_scitype} = input_scitype
 
-
-MLJBase.load_path(::Type{<:ProbabilisticStack}) = "MLJBase.ProbabilisticStack"
-MLJBase.load_path(::Type{<:DeterministicStack}) = "MLJBase.DeterministicStack"
+MMI.constructor(::Type{<:Stack}) = Stack
+MLJBase.load_path(::Type{<:Stack}) = "MLJBase.Stack"
 MLJBase.package_name(::Type{<:Stack}) = "MLJBase"
 MLJBase.package_uuid(::Type{<:Stack}) = "a7f614a8-145f-11e9-1d2a-a57a1082229d"
 MLJBase.package_url(::Type{<:Stack}) = "https://github.com/JuliaAI/MLJBase.jl"

src/composition/models/transformed_target_model.jl

@@ -10,7 +10,8 @@ const TT_SUPPORTED_ATOMS = (
     :Deterministic,
     :DeterministicUnsupervisedDetector,
     :DeterministicSupervisedDetector,
-    :Interval)
+    :Interval,
+)
 
 # Each supported atomic type gets its own wrapper:
 
@@ -265,6 +266,10 @@ MMI.package_uuid(::Type{<:SomeTT}) = "a7f614a8-145f-11e9-1d2a-a57a1082229d"
 MMI.is_wrapper(::Type{<:SomeTT}) = true
 MMI.package_url(::Type{<:SomeTT}) = "https://github.com/JuliaAI/MLJBase.jl"
 
+MMI.load_path(::Type{<:SomeTT}) = "MLJBase.TransformedTargetModel"
+MMI.constructor(::Type{<:SomeTT}) = TransformedTargetModel
+
+
 for New in TT_TYPE_EXS
     quote
         MMI.iteration_parameter(::Type{<:$New{M}}) where M =

src/resampling.jl

@@ -1548,9 +1548,11 @@ end
         compact=false,
     )
 
+*Private method.* Use at own risk.
+
 Resampling model wrapper, used internally by the `fit` method of `TunedModel` instances
-and `IteratedModel` instances. See [`evaluate!](@ref) for options. Not intended for use by
-general user, who will ordinarily use [`evaluate!`](@ref) directly.
+and `IteratedModel` instances. See [`evaluate!`](@ref) for meaning of the options. Not
+intended for use by general user, who will ordinarily use [`evaluate!`](@ref) directly.
 
 Given a machine `mach = machine(resampler, args...)` one obtains a performance evaluation
 of the specified `model`, performed according to the prescribed `resampling` strategy and
@@ -1592,16 +1594,6 @@ mutable struct Resampler{S, L} <: Model
     compact::Bool
 end
 
-# Some traits are markded as `missing` because we cannot determine
-# them from from the type because we have removed `M` (for "model"} as
-# a `Resampler` type parameter. See
-# https://github.com/JuliaAI/MLJTuning.jl/issues/141#issue-951221466
-
-StatisticalTraits.is_wrapper(::Type{<:Resampler}) = true
-StatisticalTraits.supports_weights(::Type{<:Resampler}) = missing
-StatisticalTraits.supports_class_weights(::Type{<:Resampler}) = missing
-StatisticalTraits.is_pure_julia(::Type{<:Resampler}) = true
-
 function MLJModelInterface.clean!(resampler::Resampler)
     warning = ""
     if resampler.measure === nothing && resampler.model !== nothing
@@ -1787,11 +1779,16 @@ function MLJModelInterface.update(
 
 end
 
-# The input and target scitypes cannot be determined from the type
-# because we have removed `M` (for "model") as a `Resampler` type
-# parameter. See
+# Some traits are marked as `missing` because we cannot determine
+# them from from the type because we have removed `M` (for "model"} as
+# a `Resampler` type parameter. See
 # https://github.com/JuliaAI/MLJTuning.jl/issues/141#issue-951221466
 
+StatisticalTraits.is_wrapper(::Type{<:Resampler}) = true
+StatisticalTraits.supports_weights(::Type{<:Resampler}) = missing
+StatisticalTraits.supports_class_weights(::Type{<:Resampler}) = missing
+StatisticalTraits.is_pure_julia(::Type{<:Resampler}) = true
+StatisticalTraits.constructor(::Type{<:Resampler}) = Resampler
 StatisticalTraits.input_scitype(::Type{<:Resampler}) = Unknown
 StatisticalTraits.target_scitype(::Type{<:Resampler}) = Unknown
 StatisticalTraits.package_name(::Type{<:Resampler}) = "MLJBase"

test/composition/models/pipelines.jl

@@ -95,7 +95,9 @@ end
 
 @testset "public constructor" begin
     # un-named components:
-    @test Pipeline(m, t, u) isa UnsupervisedPipeline
+    flute = Pipeline(m, t, u)
+    @test flute isa UnsupervisedPipeline
+    @test MLJBase.constructor(flute) == Pipeline
     @test Pipeline(m, t, u, p) isa ProbabilisticPipeline
     @test Pipeline(m, t, u, p, operation=predict_mean) isa DeterministicPipeline
     @test Pipeline(u, p, u, operation=predict_mean) isa DeterministicPipeline

test/composition/models/stacking.jl

@@ -202,6 +202,9 @@ end
                     measures=rmse,
                     resampling=CV(;nfolds=3),
                     models...)
+
+    @test MLJBase.constructor(mystack) == Stack
+
     @test mystack.ridge_lambda.lambda == 0.1
     @test mystack.metalearner isa FooBarRegressor
     @test mystack.resampling isa CV

test/composition/models/transformed_target_model.jl

@@ -86,6 +86,7 @@ avg_nonlinear = g(mean(f(y))) # = g(mean(z))
 
     # Test wrapping using f and g:
     model = TransformedTargetModel(atom, transformer=f, inverse=g)
+    @test MLJBase.constructor(model) == TransformedTargetModel
     fr1, _, _ = MMI.fit(model, 0, X, y)
     @test first(predict(model, fr1, X)) ≈ fill(avg_nonlinear, 5)
 

test/resampling.jl

@@ -606,6 +606,9 @@ end
     holdout = Holdout(fraction_train=0.75)
     resampler = Resampler(resampling=holdout, model=ridge_model, measure=mae,
                           acceleration=accel)
+    @test constructor(resampler) == Resampler
+    @test package_name(resampler) == "MLJBase"
+    @test load_path(resampler) == "MLJBase.Resampler"
     resampling_machine = machine(resampler, X, y)
     @test_logs((:info, r"^Training"), fit!(resampling_machine))
     e1=evaluate(resampling_machine).measurement[1]