✨ Bump version, add src/convert.jl

Project.toml

@@ -2,7 +2,7 @@ name = "RelationalDatasets"
 uuid = "6b8bdb16-f930-11eb-22b5-7ba1dabdf7a8"
 license = "Apache 2.0"
 authors = ["Alexander L. Hayes <[email protected]>"]
-version = "0.1.0"
+version = "0.1.1"
 
 [deps]
 HTTP = "cd3eb016-35fb-5094-929b-558a96fad6f3"

src/RelationalDatasets.jl

@@ -6,4 +6,7 @@ module RelationalDatasets
 include("request.jl")
 export load
 
+include("convert.jl")
+export from_vector
+
 end # module

src/convert.jl

@@ -0,0 +1,80 @@
+# Copyright © 2021 Alexander L. Hayes
+# Apache 2.0 License
+
+"""Convert (array-based) propositional datasets to a relational/ILP
+representation.
+"""
+
+include("types.jl")
+
+"""
+    from_vector()
+
+Convert a classification dataset to an ILP representation.
+"""
+function from_vector(X::Matrix{Int64}, y::Vector{Int64}, names::Union{Vector{String}, Nothing} = nothing)
+
+    # Make sure the X and y are valid.
+    @assert size(y)[1] == size(X)[2]
+
+    if names == nothing
+        names = ["v$(i)" for i in 1:size(X)[1] + 1]
+    end
+
+    # Make sure the names vector makes sense.
+    @assert length(names) - 1 == size(X)[1]
+
+    pos, neg, facts = String[], String[], String[]
+
+    for (i, row) in enumerate(y)
+        if Bool(row)
+            push!(pos, "$(last(names))(id$(i)).")
+        else
+            push!(neg, "$(last(names))(id$(i)).")
+        end
+    end
+
+    for (i, col) in enumerate(eachrow(X))
+        var = names[i]
+        facts = vcat(facts, ["$(var)(id$(j),$(row))." for (j, row) in enumerate(col)])
+    end
+
+    modes = ["$(name)(+id,#var$(name))." for name in names[1:end-1]]
+    push!(modes, "$(last(names))(+id).")
+
+    return RelationalDataset((pos, neg, facts)), modes
+end
+
+"""
+    from_vector()
+
+Convert a regression dataset to an ILP representation.
+"""
+function from_vector(X::Matrix{Int64}, y::Vector{Float64}, names::Union{Vector{String}, Nothing} = nothing)
+
+    # Make sure the X and y are valid.
+    @assert size(y)[1] == size(X)[2]
+
+    if names == nothing
+        names = ["v$(i)" for i in 1:size(X)[1] + 1]
+    end
+
+    # Make sure the names vector makes sense.
+    @assert length(names) - 1 == size(X)[1]
+
+    pos, neg, facts = String[], String[], String[]
+
+    for (i, row) in enumerate(y)
+        push!(pos, "regressionExample($(last(names))(id$(i)),$(row)).")
+    end
+
+    for (i, col) in enumerate(eachrow(X))
+        var = names[i]
+        facts = vcat(facts, ["$(var)(id$(j),$(row))." for (j, row) in enumerate(col)])
+    end
+
+    modes = ["$(name)(+id,#var$(name))." for name in names[1:end-1]]
+    push!(modes, "$(last(names))(+id).")
+
+    return RelationalDataset((pos, neg, facts)), modes
+end

test/runtests.jl

@@ -9,5 +9,6 @@ using RelationalDatasets
 include("test_base.jl")
 include("test_types.jl")
 include("test_request.jl")
+include("test_convert.jl")
 
 end # module

test/test_convert.jl

@@ -0,0 +1,125 @@
+# Copyright © 2021 Alexander L. Hayes
+# Apache 2.0 License
+
+module TestConvert
+
+using Test
+using RelationalDatasets
+
+@testset "Test converting propositional datasets" begin
+
+    @test RelationalDatasets.from_vector(
+        [[0, 1, 1] [1, 0, 2] [2, 2, 0]],
+        [0, 0, 1],
+    ) == (RelationalDatasets.RelationalDataset((
+            pos=["v4(id3)."],
+            neg=["v4(id1).", "v4(id2)."],
+            facts=[
+                "v1(id1,0).",
+                "v1(id2,1).",
+                "v1(id3,2).",
+                "v2(id1,1).",
+                "v2(id2,0).",
+                "v2(id3,2).",
+                "v3(id1,1).",
+                "v3(id2,2).",
+                "v3(id3,0).",
+            ])),
+        [
+            "v1(+id,#varv1).",
+            "v2(+id,#varv2).",
+            "v3(+id,#varv3).",
+            "v4(+id).",
+        ]
+    )
+
+    @test RelationalDatasets.from_vector(
+        [[0, 1, 1] [1, 0, 2] [2, 2, 0]],
+        [0.1, 0.2, 0.3],
+    ) == (RelationalDatasets.RelationalDataset((
+        pos=[
+            "regressionExample(v4(id1),0.1).",
+            "regressionExample(v4(id2),0.2).",
+            "regressionExample(v4(id3),0.3).",
+        ],
+        neg=[],
+        facts=[
+            "v1(id1,0).",
+            "v1(id2,1).",
+            "v1(id3,2).",
+            "v2(id1,1).",
+            "v2(id2,0).",
+            "v2(id3,2).",
+            "v3(id1,1).",
+            "v3(id2,2).",
+            "v3(id3,0).",
+        ])),
+        [
+            "v1(+id,#varv1).",
+            "v2(+id,#varv2).",
+            "v3(+id,#varv3).",
+            "v4(+id).",
+        ]
+    )
+
+    @test RelationalDatasets.from_vector(
+        [[1, 1] [0, 1]],
+        [0, 1],
+        ["a", "b", "c"],
+    ) == (RelationalDatasets.RelationalDataset((
+        pos=["c(id2)."],
+        neg=["c(id1)."],
+        facts=[
+            "a(id1,1).",
+            "a(id2,0).",
+            "b(id1,1).",
+            "b(id2,1).",
+        ])),
+        [
+            "a(+id,#vara).",
+            "b(+id,#varb).",
+            "c(+id).",
+        ]
+    )
+
+    @test RelationalDatasets.from_vector(
+        [[1, 1] [0, 1]],
+        [0.5, 1.0],
+        ["a", "b", "c"],
+    ) == (RelationalDatasets.RelationalDataset((
+        pos=["regressionExample(c(id1),0.5).", "regressionExample(c(id2),1.0)."],
+        neg=[],
+        facts=[
+            "a(id1,1).",
+            "a(id2,0).",
+            "b(id1,1).",
+            "b(id2,1).",
+        ])),
+        [
+            "a(+id,#vara).",
+            "b(+id,#varb).",
+            "c(+id).",
+        ]
+    )
+end
+
+@testset "Misaligned Classification and Regression" begin
+    @test_throws AssertionError RelationalDatasets.from_vector([[1] [2]], [1])
+    @test_throws AssertionError RelationalDatasets.from_vector([[1] [2]], [0.5])
+end
+
+@testset "Misaligned Custom Variable Names" begin
+    @test_throws AssertionError RelationalDatasets.from_vector(
+        [[1] [2]],
+        [1],
+        ["a", "b", "c", "d"],
+    )
+
+    @test_throws AssertionError RelationalDatasets.from_vector(
+        [[1] [2]],
+        [0.5],
+        ["a", "b", "c", "d"]
+    )
+end
+
+end # module