For a 1.4 release

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"

README.md

@@ -1,7 +1,7 @@
 ## MLJBase
 
 Repository for developers that provides core functionality for the
-[MLJ](https:/alan-turing-institute/MLJ.jl) machine
+[MLJ](https:/JuliaAI/MLJ.jl) machine
 learning framework.
 
 | Branch | Julia | Build | Coverage |
@@ -20,7 +20,7 @@ learning framework.
 
 [![Stable](https://img.shields.io/badge/docs-stable-blue.svg)](https://juliaai.github.io/MLJBase.jl/stable/)
 
-[MLJ](https:/alan-turing-institute/MLJ.jl) is a Julia
+[MLJ](https:/JuliaAI/MLJ.jl) is a Julia
 framework for combining and tuning machine learning models. This
 repository provides core functionality for MLJ, including:
 
@@ -37,7 +37,7 @@ repository provides core functionality for MLJ, including:
 - basic utilities for **manipulating datasets** and for **synthesizing datasets** (src/data)
 
 - a [small
- interface](https://alan-turing-institute.github.io/MLJ.jl/dev/evaluating_model_performance/#Custom-resampling-strategies-1)
+ interface](https://JuliaAI.github.io/MLJ.jl/dev/evaluating_model_performance/#Custom-resampling-strategies-1)
  for **resampling strategies** and implementations, including `CV()`, `StratifiedCV` and
  `Holdout` (src/resampling.jl). Actual performance evaluation measures (aka metrics), which previously
  were provided by MLJBase.jl, now live in [StatisticalMeasures.jl](https://juliaai.github.io/StatisticalMeasures.jl/dev/).

docs/src/index.md

@@ -2,7 +2,7 @@
 
 These docs are bare-bones and auto-generated. Complete MLJ
 documentation is
-[here](https://alan-turing-institute.github.io/MLJ.jl/dev/). 
+[here](https://JuliaAI.github.io/MLJ.jl/dev/).
 
 For MLJBase-specific developer information, see also the [README.md
 file](https:/JuliaAI/MLJBase.jl#readme).

src/composition/learning_networks/nodes.jl

@@ -182,7 +182,7 @@ function ScientificTypes.elscitype(
 end
 
 # TODO after
-# https:/alan-turing-institute/ScientificTypesBase.jl/issues/102 :
+# https:/JuliaAI/ScientificTypesBase.jl/issues/102 :
 # Add Probabilistic case to above
 
 ScientificTypes.scitype(N::Node) = CallableReturning{elscitype(N)}

src/composition/models/pipelines.jl

@@ -397,7 +397,7 @@ end
 
 # # LEARNING NETWORK INTERFACE
 
-# https://alan-turing-institute.github.io/MLJ.jl/dev/composing_models/#Learning-network-machines
+# https://JuliaAI.github.io/MLJ.jl/dev/composing_models/#Learning-network-machines
 
 
 # ## Methods to extend a pipeline learning network
@@ -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

@@ -264,19 +264,20 @@ 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:/alan-turing-institute/MLJBase.jl"
+MLJBase.package_url(::Type{<:Stack}) = "https:/JuliaAI/MLJBase.jl"
 MLJBase.package_license(::Type{<:Stack}) = "MIT"
 
 ###########################################################

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:/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/interface/data_utils.jl

@@ -89,7 +89,7 @@ MMI.selectcols(::FI, ::Val{:table}, X, ::Colon) = X
 function MMI.selectrows(::FI, ::Val{:table}, X, r)
  r = r isa Integer ? (r:r) : r
  # next uncommented line is a hack; see
- # https:/alan-turing-institute/MLJBase.jl/issues/151
+ # https:/JuliaAI/MLJBase.jl/issues/151
  isdataframe(X) && return X[r, :]
  cols = Tables.columntable(X)
  new_cols = NamedTuple{keys(cols)}(tuple((c[r] for c in values(cols))...))

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:/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:/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/_models/Constant.jl

@@ -171,7 +171,7 @@ metadata_pkg.((ConstantRegressor, ConstantClassifier,
  DeterministicConstantRegressor, DeterministicConstantClassifier),
  name="MLJModels",
  uuid="d491faf4-2d78-11e9-2867-c94bc002c0b7",
- url="https:/alan-turing-institute/MLJModels.jl",
+ url="https:/JuliaAI/MLJModels.jl",
  julia=true,
  license="MIT",
  is_wrapper=false)

test/_models/simple_composite_model.jl

@@ -54,7 +54,7 @@ for model in COMPOSITE_MODELS
 
  MLJBase.metadata_pkg(
  $(model);
- package_url = "https:/alan-turing-institute/MLJBase.jl",
+ package_url = "https:/JuliaAI/MLJBase.jl",
  is_pure_julia = true,
  is_wrapper = true
  )

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]