Port to CategoricalArrays 0.8, bump version to 0.6.2 (#602)

The change to `levels!` implies that all reference codes will be updated
at the end to match the new order of levels (unless they were already sorted).
Require DataFrames 0.21 which is the only version to support CategoricalArrays 0.8.

Bump version to 0.6.2

Author: nalimilan

Project.toml

@@ -1,7 +1,7 @@
 name = "CSV"
 uuid = "336ed68f-0bac-5ca0-87d4-7b16caf5d00b"
 authors = ["Jacob Quinn <quinn.jacobd@gmail.com>"]
-version = "0.6.0"
+version = "0.6.2"
 
 [deps]
 CategoricalArrays = "324d7699-5711-5eae-9e2f-1d82baa6b597"
@@ -16,8 +16,8 @@ Unicode = "4ec0a83e-493e-50e2-b9ac-8f72acf5a8f5"
 WeakRefStrings = "ea10d353-3f73-51f8-a26c-33c1cb351aa5"
 
 [compat]
-CategoricalArrays = "0.5,0.6,0.7"
-DataFrames = "0.18,0.19,0.20"
+CategoricalArrays = "0.8"
+DataFrames = "0.21"
 FilePathsBase = "0.6,0.7,0.8"
 Parsers = "1"
 PooledArrays = "0.5"

src/header.jl

@@ -85,7 +85,7 @@ end
         throw(ArgumentError("\"$source\" is not a valid file"))
     (types !== nothing && any(x->!isconcretetype(x) && !(x isa Union), types isa AbstractDict ? values(types) : types)) && throw(ArgumentError("Non-concrete types passed in `types` keyword argument, please provide concrete types for columns: $types"))
     if type !== nothing && typecode(type) == EMPTY
-        throw(ArgumentError("$type isn't supported in the `type` keyword argument; must be one of: `Int64`, `Float64`, `Date`, `DateTime`, `Bool`, `Missing`, `PooledString`, `CategoricalString{UInt32}`, or `String`"))
+        throw(ArgumentError("$type isn't supported in the `type` keyword argument; must be one of: `Int64`, `Float64`, `Date`, `DateTime`, `Bool`, `Missing`, `PooledString`, `CategoricalValue{String, UInt32}`, or `String`"))
     elseif types !== nothing && any(x->typecode(x) == EMPTY, types isa AbstractDict ? values(types) : types)
         T = nothing
         for x in (types isa AbstractDict ? values(types) : types)
@@ -94,7 +94,7 @@ end
                 break
             end
         end
-        throw(ArgumentError("unsupported type $T in the `types` keyword argument; must be one of: `Int64`, `Float64`, `Date`, `DateTime`, `Bool`, `Missing`, `PooledString`, `CategoricalString{UInt32}`, or `String`"))
+        throw(ArgumentError("unsupported type $T in the `types` keyword argument; must be one of: `Int64`, `Float64`, `Date`, `DateTime`, `Bool`, `Missing`, `PooledString`, `CategoricalValue{String, UInt32}`, or `String`"))
     end
     checkvaliddelim(delim)
     ignorerepeated && delim === nothing && throw(ArgumentError("auto-delimiter detection not supported when `ignorerepeated=true`; please provide delimiter like `delim=','`"))
@@ -199,10 +199,10 @@ end
     T = type === nothing ? (streaming ? (STRING | MISSING) : EMPTY) : (typecode(type) | USER)
     if types isa Vector
         typecodes = TypeCode[typecode(T) | USER for T in types]
-        categorical = categorical | any(x->x == CategoricalString{UInt32}, types)
+        categorical = categorical | any(x->x == CategoricalValue{String, UInt32}, types)
     elseif types isa AbstractDict
         typecodes = initialtypes(T, types, names)
-        categorical = categorical | any(x->x == CategoricalString{UInt32}, values(types))
+        categorical = categorical | any(x->x == CategoricalValue{String, UInt32}, values(types))
     else
         typecodes = TypeCode[T for _ = 1:ncols]
     end

src/tables.jl

@@ -135,8 +135,8 @@ function Base.copy(col::Union{Column{T, S}, Column2{T, S}}) where {T <: Union{St
     end
     if catg
         pool = CategoricalPool(refs)
-        levels!(pool, sort(levels(pool)))
         A = CategoricalArray{T, 1}(values, pool)
+        levels!(A, sort(levels(A)))
     else
         A = PooledArray(PooledArrays.RefArray(values), refs)
     end

src/utils.jl

@@ -74,7 +74,7 @@ typecode(::Type{Time}) = TIME
 typecode(::Type{Bool}) = BOOL
 typecode(::Type{<:AbstractString}) = STRING
 typecode(::Type{PooledString}) = POOL
-typecode(::Type{CategoricalString{UInt32}}) = POOL
+typecode(::Type{CategoricalValue{String, UInt32}}) = POOL
 typecode(::Type{Union{}}) = EMPTY
 typecode(::Type{Union{T, Missing}}) where {T} = typecode(T) | MISSING
 typecode(::Type{T}) where {T} = EMPTY

src/write.jl

@@ -402,7 +402,7 @@ end
 writecell(buf, pos, len, io, x, opts) =
     writecell(buf, pos, len, io, Base.string(x), opts)
 
-writecell(buf, pos, len, io, x::CategoricalString, opts) =
+writecell(buf, pos, len, io, x::CategoricalValue{String}, opts) =
     writecell(buf, pos, len, io, Base.string(x), opts)
 
 function writecell(buf, pos, len, io, x::AbstractString, opts)

test/basics.jl

@@ -210,11 +210,11 @@ df = CSV.read(IOBuffer("x\na\nb\na\nb\na\nb\na\nc\nd\ne\nf\ng\nh\ni\nj\nk\nl\nn\
 
 # a few corner cases for escape strings
 df = CSV.read(IOBuffer("\"column name with \"\" escape character inside\"\n1\n"))
-@test names(df)[1] == Symbol("column name with \" escape character inside")
+@test names(df)[1] == "column name with \" escape character inside"
 
 df = CSV.read(IOBuffer("\"column name with \"\" escape character inside\",1\n,2"), transpose=true)
-@test names(df)[1] == Symbol("column name with \" escape character inside")
-@test names(df)[2] == :Column2
+@test names(df)[1] == "column name with \" escape character inside"
+@test names(df)[2] == "Column2"
 
 df = CSV.read(IOBuffer("x\na\nb\n\"quoted field with \"\" escape character inside\"\n"), pool=true)
 @test df.x[1] == "a"
@@ -241,19 +241,19 @@ df = CSV.read(IOBuffer("x\nabc\n"), type=Int)
 
 # transpose corner cases
 df = CSV.read(IOBuffer("x,y,1\nx2,y2,2\n"), transpose=true, header=2)
-@test names(df) == [:y, :y2]
+@test names(df) == ["y", "y2"]
 @test size(df) == (1, 2)
 @test df.y[1] == 1
 @test df.y2[1] == 2
 
 df = CSV.read(IOBuffer("x,y,1\nx2,y2,2\n"), transpose=true, header=1, datarow=3)
-@test names(df) == [:x, :x2]
+@test names(df) == ["x", "x2"]
 @test size(df) == (1, 2)
 @test df.x[1] == 1
 @test df.x2[1] == 2
 
 df = CSV.read(IOBuffer("x,y,1\nx2,y2,2\n"), transpose=true, header=false, datarow=3)
-@test names(df) == [:Column1, :Column2]
+@test names(df) == ["Column1", "Column2"]
 @test size(df) == (1, 2)
 @test df.Column1[1] == 1
 @test df.Column2[1] == 2
@@ -266,7 +266,7 @@ df = CSV.read(IOBuffer(""), transpose=true, header=Symbol[])
 
 # providing empty header vector
 df = CSV.read(IOBuffer("x\nabc\n"), header=Symbol[])
-@test names(df) == [:Column1]
+@test names(df) == ["Column1"]
 
 # Union{Bool, Missing}
 df = CSV.read(IOBuffer("x\ntrue\n\n"))
@@ -380,7 +380,7 @@ df = CSV.read(IOBuffer("thistime\n10:00:00.0\n12:00:00.0"))
 
 # 530
 df = CSV.read(IOBuffer(",column2\nNA,2\n2,3"), missingstrings=["NA"])
-@test names(df) == [:Column1, :column2]
+@test names(df) == ["Column1", "column2"]
 
 # reported on slack from Kevin Bonham
 df = CSV.read(IOBuffer("x\n01:02:03\n\n04:05:06\n"), delim=',')

test/runtests.jl

@@ -23,7 +23,7 @@ end # @testset "CSV.File"
 include("write.jl")
 
 @testset "CategoricalArray levels (including ordering)" begin
-    f = CSV.read(IOBuffer("X\nb\nc\na\nc"), types=[CategoricalString{UInt32}], copycols=true)
+    f = CSV.read(IOBuffer("X\nb\nc\na\nc"), types=[CategoricalValue{String, UInt32}], copycols=true)
     v = f.X[1]
     @test v == "b"
     @test levels(v.pool) == ["a", "b", "c"]
@@ -37,7 +37,7 @@ include("write.jl")
     v = f.X[1]
     @test v == "b"
     @test levels(v.pool) == ["b", "c"]
-    @test typeof(f.X) == CategoricalArray{Union{Missing, String},1,UInt32,String,CategoricalString{UInt32},Missing}
+    @test typeof(f.X) == CategoricalArray{Union{Missing, String},1,UInt32,String,CategoricalValue{String, UInt32},Missing}
 
 end