Add levels (#12)

* Add levels

Currently lives in Missings.jl, which isn't the most logical place. This will allow
StatsModels to drop the dependency on Missings, while CategoricalArrays will continue
adding a specialized method to it.

* Fix sorting

* Bump version to 1.1.0

Author: nalimilan

Project.toml

@@ -1,7 +1,7 @@
 name = "DataAPI"
 uuid = "9a962f9c-6df0-11e9-0e5d-c546b8b5ee8a"
 authors = ["quinnj <[email protected]>"]
-version = "1.0.1"
+version = "1.1.0"
 
 [compat]
 julia = "1"

src/DataAPI.jl

@@ -73,6 +73,26 @@ definition.
 """
 function describe end
 
+"""
+    levels(x)
+
+Return a vector of unique values which occur or could occur in collection `x`,
+omitting `missing` even if present. Values are returned in the preferred order
+for the collection, with the result of [`sort`](@ref) as a default.
+
+Contrary to [`unique`](@ref), this function may return values which do not
+actually occur in the data, and does not preserve their order of appearance in `x`.
+"""
+function levels(x)
+    T = Base.nonmissingtype(eltype(x))
+    levs = convert(AbstractArray{T}, filter!(!ismissing, unique(x)))
+    try
+        sort!(levs)
+    catch
+    end
+    levs
+end
+
 """
     Between(first, last)
 

test/runtests.jl

@@ -27,6 +27,36 @@ end
 
 end
 
+@testset "levels" begin
+
+    @test DataAPI.levels(1:1) ==
+        DataAPI.levels([1]) ==
+        DataAPI.levels([1, missing]) ==
+        DataAPI.levels([missing, 1]) ==
+        [1]
+    @test DataAPI.levels(2:-1:1) ==
+        DataAPI.levels([2, 1]) ==
+        DataAPI.levels(Any[2, 1]) ==
+        DataAPI.levels([2, missing, 1]) ==
+        [1, 2]
+    @test DataAPI.levels([missing, "a", "c", missing, "b"]) == ["a", "b", "c"]
+    @test DataAPI.levels([Complex(0, 1), Complex(1, 0), missing]) ==
+        [Complex(0, 1), Complex(1, 0)]
+    @test typeof(DataAPI.levels([1])) ===
+        typeof(DataAPI.levels([1, missing])) ===
+        Vector{Int}
+    @test typeof(DataAPI.levels(["a"])) ===
+        typeof(DataAPI.levels(["a", missing])) ===
+        Vector{String}
+    @test typeof(DataAPI.levels(Real[1])) ===
+        typeof(DataAPI.levels(Union{Real,Missing}[1, missing])) ===
+        Vector{Real}
+    @test typeof(DataAPI.levels(trues(1))) === Vector{Bool}
+    @test isempty(DataAPI.levels([missing]))
+    @test isempty(DataAPI.levels([]))
+
+end
+
 @testset "Between" begin
 
     for x in (1, :a), y in (1, :a)