Skip to content
New issue

Have a question about this project? Sign up for a free GitHub account to open an issue and contact its maintainers and the community.

Sign up for GitHub

By clicking “Sign up for GitHub”, you agree to our terms of service and privacy statement. We’ll occasionally send you account related emails.

Already on GitHub? Sign in to your account

Preserve non-missingness during non-inner joins #1316

Merged
merged 2 commits into from
Dec 18, 2017
Merged

Preserve non-missingness during non-inner joins #1316

Show file tree
Hide file tree
Changes from all commits
Commits
Show all changes
2 commits
Select commit Hold shift + click to select a range
d01bd20
join: try harder to preserve non-missingness
alyst Dec 14, 2017
c6bc53b
add join() test for frames with Any column
alyst Dec 14, 2017
File filter

Filter by extension

Filter by extension
Conversations
Failed to load comments.
Loading
Jump to
Jump to file
Failed to load files.
Loading
Diff view
Diff view
24 changes: 20 additions & 4 deletions src/abstractdataframe/join.jl
View file
Open in desktop
Original file line number Diff line number Diff line change
Expand Up @@ -85,27 +85,43 @@ function compose_joined_table(joiner::DataFrameJoiner, kind::Symbol,
@assert nrow == length(all_orig_right_ixs) + loil
ncleft = ncol(joiner.dfl)
cols = Vector{Any}(ncleft + ncol(dfr_noon))
_similar = kind == :inner ? similar : similar_missing
# inner and left joins preserve non-missingness of the left frame
_similar_left = kind == :inner || kind == :left ? similar : similar_missing
for (i, col) in enumerate(columns(joiner.dfl))
cols[i] = _similar(col, nrow)
cols[i] = _similar_left(col, nrow)
copy!(cols[i], view(col, all_orig_left_ixs))
end
# inner and right joins preserve non-missingness of the right frame
_similar_right = kind == :inner || kind == :right ? similar : similar_missing
for (i, col) in enumerate(columns(dfr_noon))
cols[i+ncleft] = _similar(col, nrow)
cols[i+ncleft] = _similar_right(col, nrow)
copy!(cols[i+ncleft], view(col, all_orig_right_ixs))
permute!(cols[i+ncleft], right_perm)
end
res = DataFrame(cols, vcat(names(joiner.dfl), names(dfr_noon)))

if length(rightonly_ixs.join) > 0
# some left rows are missings, so the values of the "on" columns
# some left rows are missing, so the values of the "on" columns
# need to be taken from the right
for (on_col_ix, on_col) in enumerate(joiner.left_on)
# fix the result of the rightjoin by taking the nonmissing values from the right table
offset = nrow - length(rightonly_ixs.orig) + 1
copy!(res[on_col], offset, view(joiner.dfr_on[on_col_ix], rightonly_ixs.orig))
end
end
if kind ∈ (:right, :outer) && !isempty(rightonly_ixs.join)
# At this point on-columns of the result allow missing values, because
# right-only rows were filled with missing values when processing joiner.dfl
# However, when the right on-column (plus the left one for the outer join)
# does not allow missing values, the result should also disallow them.
for (on_col_ix, on_col) in enumerate(joiner.left_on)
LT = eltype(joiner.dfl_on[on_col_ix])
RT = eltype(joiner.dfr_on[on_col_ix])
if !(RT >: Missing) && (kind == :right || !(LT >: Missing))
res[on_col] = disallowmissing(res[on_col])
end
end
end
return res
end

Expand Down
73 changes: 52 additions & 21 deletions test/join.jl
View file
Open in desktop
Original file line number Diff line number Diff line change
Expand Up @@ -186,46 +186,43 @@ module TestJoin
fid = [1, 3, 5],
fid_1 = [1, 3, missing])
@test typeof.(l(on).columns) ==
[Vector{Union{T, Missing}} for T in (Int, Float64, Float64)]
[Vector{Int}, Vector{Float64}, Vector{Union{Float64, Missing}}]
@test r(on) ≅ DataFrame(id = [1, 3, 0, 2, 4],
fid = [1, 3, missing, missing, missing],
fid_1 = [1, 3, 0, 2, 4])
@test typeof.(r(on).columns) ==
[Vector{Union{T, Missing}} for T in (Int, Float64, Float64)]
[Vector{Int}, Vector{Union{Float64, Missing}}, Vector{Float64}]
@test o(on) ≅ DataFrame(id = [1, 3, 5, 0, 2, 4],
fid = [1, 3, 5, missing, missing, missing],
fid_1 = [1, 3, missing, 0, 2, 4])
@test typeof.(o(on).columns) ==
[Vector{Union{T, Missing}} for T in (Int, Float64, Float64)]
[Vector{Int}, Vector{Union{Float64, Missing}}, Vector{Union{Float64, Missing}}]

on = :fid
@test i(on) == DataFrame(Any[[1, 3], [1.0, 3.0], [1, 3]], [:id, :fid, :id_1])
@test typeof.(i(on).columns) == [Vector{Int}, Vector{Float64}, Vector{Int}]
@test l(on) ≅ DataFrame(id = [1, 3, 5],
fid = [1, 3, 5],
id_1 = [1, 3, missing])
@test typeof.(l(on).columns) == [Vector{Union{T, Missing}} for T in (Int,Float64,Int)]
@test typeof.(l(on).columns) == [Vector{Int}, Vector{Float64}, Vector{Union{Int, Missing}}]
@test r(on) ≅ DataFrame(id = [1, 3, missing, missing, missing],
fid = [1, 3, 0, 2, 4],
id_1 = [1, 3, 0, 2, 4])
@test typeof.(r(on).columns) == [Vector{Union{T, Missing}} for T in (Int,Float64,Int)]
@test typeof.(r(on).columns) == [Vector{Union{Int, Missing}}, Vector{Float64}, Vector{Int}]
@test o(on) ≅ DataFrame(id = [1, 3, 5, missing, missing, missing],
fid = [1, 3, 5, 0, 2, 4],
id_1 = [1, 3, missing, 0, 2, 4])
@test typeof.(o(on).columns) == [Vector{Union{T, Missing}} for T in (Int,Float64,Int)]
@test typeof.(o(on).columns) == [Vector{Union{Int, Missing}}, Vector{Float64}, Vector{Union{Int, Missing}}]

on = [:id, :fid]
@test i(on) == DataFrame(Any[[1, 3], [1, 3]], [:id, :fid])
@test typeof.(i(on).columns) == [Vector{Int}, Vector{Float64}]
@test l(on) == DataFrame(id = [1, 3, 5], fid = [1, 3, 5])
@test typeof.(l(on).columns) == [Vector{Union{Int, Missing}},
Vector{Union{Float64, Missing}}]
@test typeof.(l(on).columns) == [Vector{Int}, Vector{Float64}]
@test r(on) == DataFrame(id = [1, 3, 0, 2, 4], fid = [1, 3, 0, 2, 4])
@test typeof.(r(on).columns) == [Vector{Union{Int, Missing}},
Vector{Union{Float64, Missing}}]
@test typeof.(r(on).columns) == [Vector{Int}, Vector{Float64}]
@test o(on) == DataFrame(id = [1, 3, 5, 0, 2, 4], fid = [1, 3, 5, 0, 2, 4])
@test typeof.(o(on).columns) == [Vector{Union{Int, Missing}},
Vector{Union{Float64, Missing}}]
@test typeof.(o(on).columns) == [Vector{Int}, Vector{Float64}]
end

@testset "all joins with CategoricalArrays" begin
Expand Down Expand Up @@ -276,17 +273,17 @@ module TestJoin
fid = [1, 3, 5],
fid_1 = [1, 3, missing])
@test all(isa.(l(on).columns,
[CategoricalVector{Union{T, Missing}} for T in (Int,Float64,Float64)]))
[CategoricalVector{T} for T in (Int,Float64,Union{Float64, Missing})]))
@test r(on) ≅ DataFrame(id = [1, 3, 0, 2, 4],
fid = [1, 3, missing, missing, missing],
fid_1 = [1, 3, 0, 2, 4])
@test all(isa.(r(on).columns,
[CategoricalVector{Union{T, Missing}} for T in (Int,Float64,Float64)]))
[CategoricalVector{T} for T in (Int,Union{Float64, Missing},Float64)]))
@test o(on) ≅ DataFrame(id = [1, 3, 5, 0, 2, 4],
fid = [1, 3, 5, missing, missing, missing],
fid_1 = [1, 3, missing, 0, 2, 4])
@test all(isa.(o(on).columns,
[CategoricalVector{Union{T, Missing}} for T in (Int,Float64,Float64)]))
[CategoricalVector{T} for T in (Int,Union{Float64,Missing},Union{Float64, Missing})]))

on = :fid
@test i(on) == DataFrame(Any[[1, 3], [1.0, 3.0], [1, 3]], [:id, :fid, :id_1])
Expand All @@ -296,17 +293,17 @@ module TestJoin
fid = [1, 3, 5],
id_1 = [1, 3, missing])
@test all(isa.(l(on).columns,
[CategoricalVector{Union{T, Missing}} for T in (Int, Float64, Int)]))
[CategoricalVector{T} for T in (Int, Float64, Union{Int, Missing})]))
@test r(on) ≅ DataFrame(id = [1, 3, missing, missing, missing],
fid = [1, 3, 0, 2, 4],
id_1 = [1, 3, 0, 2, 4])
@test all(isa.(r(on).columns,
[CategoricalVector{Union{T, Missing}} for T in (Int, Float64, Int)]))
[CategoricalVector{T} for T in (Union{Int, Missing}, Float64, Int)]))
@test o(on) ≅ DataFrame(id = [1, 3, 5, missing, missing, missing],
fid = [1, 3, 5, 0, 2, 4],
id_1 = [1, 3, missing, 0, 2, 4])
@test all(isa.(o(on).columns,
[CategoricalVector{Union{T, Missing}} for T in (Int, Float64, Int)]))
[CategoricalVector{T} for T in (Union{Int, Missing}, Float64, Union{Int, Missing})]))

on = [:id, :fid]
@test i(on) == DataFrame(Any[[1, 3], [1, 3]], [:id, :fid])
Expand All @@ -315,15 +312,15 @@ module TestJoin
@test l(on) == DataFrame(id = [1, 3, 5],
fid = [1, 3, 5])
@test all(isa.(l(on).columns,
[CategoricalVector{Union{T, Missing}} for T in (Int, Float64)]))
[CategoricalVector{T} for T in (Int, Float64)]))
@test r(on) == DataFrame(id = [1, 3, 0, 2, 4],
fid = [1, 3, 0, 2, 4])
@test all(isa.(r(on).columns,
[CategoricalVector{Union{T, Missing}} for T in (Int, Float64)]))
[CategoricalVector{T} for T in (Int, Float64)]))
@test o(on) == DataFrame(id = [1, 3, 5, 0, 2, 4],
fid = [1, 3, 5, 0, 2, 4])
@test all(isa.(o(on).columns,
[CategoricalVector{Union{T, Missing}} for T in (Int, Float64)]))
[CategoricalVector{T} for T in (Int, Float64)]))
end

@testset "maintain CategoricalArray levels ordering on join - non-`on` cols" begin
Expand Down Expand Up @@ -465,4 +462,38 @@ module TestJoin
@test join(left, right, on = [:id => :ID, :sid => :SID], kind=:anti) ==
DataFrame(id = 1:2, sid = string.(1:2))
end

@testset "join with a column of type Any" begin
l = DataFrame(a=Any[1:7;], b=[1:7;])
r = DataFrame(a=Any[3:10;], b=[3:10;])

# join by :a and :b (Any is the on-column)
@test join(l, r, on=[:a, :b], kind=:inner) ≅ DataFrame(a=Any[3:7;], b=3:7)
@test eltypes(join(l, r, on=[:a, :b], kind=:inner)) == [Any, Int]

@test join(l, r, on=[:a, :b], kind=:left) ≅ DataFrame(a=Any[1:7;], b=1:7)
@test eltypes(join(l, r, on=[:a, :b], kind=:left)) == [Any, Int]

@test join(l, r, on=[:a, :b], kind=:right) ≅ DataFrame(a=Any[3:10;], b=3:10)
@test eltypes(join(l, r, on=[:a, :b], kind=:right)) == [Any, Int]

@test join(l, r, on=[:a, :b], kind=:outer) ≅ DataFrame(a=Any[1:10;], b=1:10)
@test eltypes(join(l, r, on=[:a, :b], kind=:outer)) == [Any, Int]

# join by :b (Any is not on-column)
@test join(l, r, on=:b, kind=:inner) ≅ DataFrame(a=Any[3:7;], b=3:7, a_1=Any[3:7;])
@test eltypes(join(l, r, on=:b, kind=:inner)) == [Any, Int, Any]

@test join(l, r, on=:b, kind=:left) ≅
DataFrame(a=Any[1:7;], b=1:7, a_1=[fill(missing, 2); 3:7;])
@test eltypes(join(l, r, on=:b, kind=:left)) == [Any, Int, Any]

@test join(l, r, on=:b, kind=:right) ≅
DataFrame(a=[3:7; fill(missing, 3)], b=3:10, a_1=Any[3:10;])
@test eltypes(join(l, r, on=:b, kind=:right)) == [Any, Int, Any]

@test join(l, r, on=:b, kind=:outer) ≅
DataFrame(a=[1:7; fill(missing, 3)], b=1:10, a_1=[fill(missing, 2); 3:10;])
@test eltypes(join(l, r, on=:b, kind=:outer)) == [Any, Int, Any]
end
end