up DataAPI.jl to 1.7 and CategoricalArrays.jl to 0.10.0

[bkamins]

Fixes #2788

[bkamins]

@nalimilan - these changes follow CategoricalArrays.jl 0.10.0 release, so a check would be welcome.

[nalimilan]

Interesting. Why aren't levels the same for inner and outer joins? ["d", "c", "b", "a"] would sound like a good choice since levels(A.b) is a superset of levels(B.b).

That said, since the old behavior wasn't ideal either, the new one is OK -- both are kind of arbitrary.

[nalimilan]

Interesting. Why aren't levels the same for inner and outer joins? ["d", "c", "b", "a"] would sound like a good choice since levels(A.b) is a superset of levels(B.b).

That said, since the old behavior wasn't ideal either, the new one is OK -- both are kind of arbitrary.

[bkamins]

Why aren't levels the same for inner and outer joins?

The reason is (under CategoricalArrays.jl 0.10):

julia> x = categorical(["a", "b", "c", "d"])
4-element CategoricalArray{String,1,UInt32}:
 "a"
 "b"
 "c"
 "d"

julia> levels!(x, ["d", "c", "b", "a"])
4-element CategoricalArray{String,1,UInt32}:
 "a"
 "b"
 "c"
 "d"

julia> y = categorical(["a", "b", "c"])
3-element CategoricalArray{String,1,UInt32}:
 "a"
 "b"
 "c"

julia> levels(x)
4-element Vector{String}:
 "d"
 "c"
 "b"
 "a"

julia> levels(y)
3-element Vector{String}:
 "a"
 "b"
 "c"

julia> levels([x; y])
4-element Vector{String}:
 "d"
 "c"
 "b"
 "a"

julia> levels([y; x])
4-element Vector{String}:
 "d"
 "a"
 "b"
 "c"

so as you can see the order of vectors matters (in this example in vcat).

And in the innerjoin the order is different because:

julia> innerjoin(B, A, on=:b)
3×3 DataFrame
 Row │ b     c      a
     │ Cat…  Int64  Int64
─────┼────────────────────
   1 │ a         5      1
   2 │ b         6      2
   3 │ c         7      3

and as you can see the "d" level does not get added (in innerjoin we use the left data frame as a source for columns we join-on, in outerjoin we need to use both).

[nalimilan]

Ah right that's because the orders are not compatible, so we have to make a choice.

[bkamins]

Thank you!