Merge pull request #37 from ceferisbarov/coerce

Add Coerce

README.md

@@ -151,6 +151,7 @@ Please check the docstrings for additional information.
 | `DropMissing` | Drop missings |
 | `Rename` | Column renaming |
 | `Coalesce` | Replace missings |
+| `Coerce` | Coerce scientific types |
 | `Identity` | Identity transform |
 | `Center` | Mean removal |
 | `Scale` | Interval scaling |

src/TableTransforms.jl

@@ -37,6 +37,7 @@ export
   DropMissing,
   Rename,
   Coalesce,
+  Coerce,
   Identity,
   Center,
   Scale,

src/transforms.jl

@@ -215,6 +215,7 @@ include("transforms/select.jl")
 include("transforms/filter.jl")
 include("transforms/rename.jl")
 include("transforms/coalesce.jl")
+include("transforms/coerce.jl")
 include("transforms/identity.jl")
 include("transforms/center.jl")
 include("transforms/scale.jl")

src/transforms/coerce.jl

@@ -0,0 +1,48 @@
+# ------------------------------------------------------------------
+# Licensed under the MIT License. See LICENSE in the project root.
+# ------------------------------------------------------------------
+
+"""
+    Coerce(pairs, tight=false, verbosity=1)
+
+Return a copy of the table, ensuring that the scientific types of the columns match the new specification.
+
+This transform wraps the ScientificTypes.coerce function. Please see their docstring for more details.
+
+```julia
+Coerce(:col1 => Continuous, :col2 => Count)
+```
+"""
+struct Coerce{P} <: Transform
+  pairs::P
+  tight::Bool
+  verbosity::Int
+end
+
+Coerce(pair::Pair{Symbol,<:Type}...; tight=false, verbosity=1) = 
+  Coerce(pair, tight, verbosity)
+
+isrevertible(::Type{<:Coerce}) = true
+
+function apply(transform::Coerce, table)
+  newtable = coerce(table, transform.pairs...;
+                    tight=transform.tight,
+                    verbosity=transform.verbosity)
+
+  types = Tables.schema(table).types
+
+  newtable, types
+end
+
+function revert(transform::Coerce, newtable, cache)
+  names = Tables.columnnames(newtable)
+  cols = Tables.columns(newtable)
+  oldcols = map(zip(cache, names)) do (T, n)
+    x = Tables.getcolumn(cols, n)
+    collect(T, x)
+  end
+
+  𝒯 = (; zip(names, oldcols)...)
+  𝒯 |> Tables.materializer(newtable)
+end
+

test/Project.toml

@@ -1,11 +1,13 @@
 [deps]
+CategoricalArrays = "324d7699-5711-5eae-9e2f-1d82baa6b597"
 Distributions = "31c24e10-a181-5473-b8eb-7969acd0382f"
 GR = "28b8d3ca-fb5f-59d9-8090-bfdbd6d07a71"
 ImageIO = "82e4d734-157c-48bb-816b-45c225c6df19"
 LinearAlgebra = "37e2e46d-f89d-539d-b4ee-838fcccc9c8e"
 Plots = "91a5bcdd-55d7-5caf-9e0b-520d859cae80"
 Random = "9a3f8284-a2c9-5f02-9a11-845980a1fd5c"
 ReferenceTests = "324d217c-45ce-50fc-942e-d289b448e8cf"
+ScientificTypes = "321657f4-b219-11e9-178b-2701a2544e81"
 Statistics = "10745b16-79ce-11e8-11f9-7d13ad32a3b2"
 Tables = "bd369af6-aec1-5ad0-b16a-f7cc5008161c"
 Test = "8dfed614-e22c-5e08-85e1-65c5234f0b40"

test/runtests.jl

@@ -2,6 +2,8 @@ using TableTransforms
 using Distributions
 using Tables
 using TypedTables
+using CategoricalArrays
+using ScientificTypes: Count, Multiclass
 using LinearAlgebra
 using Statistics
 using Test, Random, Plots

test/transforms.jl

@@ -673,6 +673,31 @@
     @test ttypes == Tables.schema(tₒ).types    
   end
 
+  @testset "Coerce" begin
+    x1 = [1.0, 2.0, 3.0, 4.0, 5.0]
+    x2 = [1.0, 2.0, 3.0, 4.0, 5.0]
+    x3 = [5.0, 5.0, 5.0, 5.0, 5.0]
+    t = Table(;x1, x2, x3)
+
+    T = Coerce(:x1=>Count, :x2=>Count)
+    n, c = apply(T, t)
+    @test eltype(n.x1) == Int
+    @test eltype(n.x2) == Int
+    n, c = apply(T, t)
+    tₒ = revert(T, n, c)
+    @test eltype(tₒ.x1) == eltype(t.x1)
+    @test eltype(tₒ.x2) == eltype(t.x2)
+
+    T = Coerce(:x1=>Multiclass, :x2=>Multiclass)
+    n, c = apply(T, t)
+    @test eltype(n.x1) <: CategoricalValue
+    @test eltype(n.x2) <: CategoricalValue
+    n, c = apply(T, t)
+    tₒ = revert(T, n, c)
+    @test eltype(tₒ.x1) == eltype(t.x1)
+    @test eltype(tₒ.x2) == eltype(t.x2)
+  end
+
   @testset "Identity" begin
     x = rand(4000)
     y = rand(4000)