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

Jump to bottom

Matchmissing == :notequal #2724

Merged
merged 12 commits into from
Jun 3, 2021
Merged

Matchmissing == :notequal #2724

Show file tree
Hide file tree
Changes from 8 commits
Commits
Show all changes
12 commits
Select commit Hold shift + click to select a range
03721c2
Matchmissing == :notequal
pstorozenko Apr 17, 2021
432560f
`matchmissing==:notequals`
pstorozenko May 30, 2021
076589d
Merge branch 'main' into ps/match_notequal
pstorozenko May 30, 2021
9afafae
Apply suggestions from code review
pstorozenko May 31, 2021
8d82161
Update NEWS
pstorozenko May 31, 2021
94da602
More tests
pstorozenko May 31, 2021
7a86c06
Add `copycols=true` to selects
pstorozenko Jun 1, 2021
9ba9d9b
Changes in tests
pstorozenko Jun 1, 2021
1edef1b
Apply suggestions from code review
pstorozenko Jun 2, 2021
aec82da
Create `dfl_on` only once
pstorozenko Jun 2, 2021
e24c7f2
bug fix
pstorozenko Jun 2, 2021
c534633
Apply suggestions from code review
pstorozenko Jun 2, 2021
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
9 changes: 9 additions & 0 deletions NEWS.md
View file
Open in desktop
Original file line number Diff line number Diff line change
@@ -1,5 +1,14 @@
# DataFrames.jl changes on main since last release notes

## New functionalities

* add option `matchmissing=:notequal` in joins;
in `leftjoin`, `semijoin` and `antijoin` missings are dropped in right data frame,
but preserved in left; in `rightjoin` missings are dropped in left data frame,
but preserved in right; in `innerjoin` missings are dropped in both data frames;
in `outerjoin` this value of keyword argument is not supported
([#2724](https://github.com/JuliaData/DataFrames.jl/pull/2724))

## Bug fixes

* fix bug in how `issorted` handles custom orderings and improve performance
Expand Down
47 changes: 36 additions & 11 deletions src/join/composer.jl
View file
Open in desktop
Original file line number Diff line number Diff line change
Expand Up @@ -23,7 +23,8 @@ struct DataFrameJoiner

function DataFrameJoiner(dfl::AbstractDataFrame, dfr::AbstractDataFrame,
on::Union{<:OnType, AbstractVector},
matchmissing::Symbol)
matchmissing::Symbol,
kind::Symbol)
on_cols = isa(on, AbstractVector) ? on : [on]
left_on = Symbol[]
right_on = Symbol[]
Expand All @@ -45,8 +46,8 @@ struct DataFrameJoiner
"Symbol or Pair{Symbol, Symbol}."))
end
end
dfl_on = dfl[!, left_on]
dfr_on = dfr[!, right_on]
dfl_on = select(dfl, left_on, copycols=true)
dfr_on = select(dfr, right_on, copycols=true)
pstorozenko marked this conversation as resolved.
Show resolved Hide resolved

if matchmissing === :error
for df in (dfl_on, dfr_on), col in eachcol(df)
Expand All @@ -55,8 +56,27 @@ struct DataFrameJoiner
"when matchmissing == :error"))
end
end
elseif matchmissing === :notequal
if kind in (:left, :semi, :anti)
dfr = dropmissing(dfr, right_on, view=true)
dfr_on = select(dfr, right_on, copycols=true)
pstorozenko marked this conversation as resolved.
Show resolved Hide resolved
elseif kind === :right
dfl = dropmissing(dfl, left_on, view=true)
dfl_on = select(dfl, left_on, copycols=true)
pstorozenko marked this conversation as resolved.
Show resolved Hide resolved
elseif kind === :inner
# it possible to drop only left or right df
# to gain some performance but needs more testing, see #2724
dfl = dropmissing(dfl, left_on, view=true)
dfl_on = select(dfl, left_on, copycols=true)
pstorozenko marked this conversation as resolved.
Show resolved Hide resolved
dfr = dropmissing(dfr, right_on, view=true)
dfr_on = select(dfr, right_on, copycols=true)
pstorozenko marked this conversation as resolved.
Show resolved Hide resolved
elseif kind === :outer
throw(ArgumentError("matchmissing == :notequal for `outerjoin` is not allowed"))
else
throw(ArgumentError("matchmissing == :notequal not implemented for kind == :$kind"))
end
elseif matchmissing !== :equal
throw(ArgumentError("matchmissing allows only :error or :equal"))
throw(ArgumentError("matchmissing allows only :error, :equal, or :notequal"))
end

for df in (dfl_on, dfr_on), col in eachcol(df)
Expand Down Expand Up @@ -311,7 +331,7 @@ function _join(df1::AbstractDataFrame, df2::AbstractDataFrame;
throw(ArgumentError("Missing join argument 'on'."))
end

joiner = DataFrameJoiner(df1, df2, on, matchmissing)
joiner = DataFrameJoiner(df1, df2, on, matchmissing, kind)

# Check merge key validity
left_invalid = validate[1] ? any(nonunique(joiner.dfl, joiner.left_on)) : false
Expand Down Expand Up @@ -485,7 +505,8 @@ change in future releases.
data frame and left unchanged.
- `matchmissing` : if equal to `:error` throw an error if `missing` is present
in `on` columns; if equal to `:equal` then `missing` is allowed and missings are
matched (`isequal` is used for comparisons of rows for equality)
matched; if equal to `:notequal` then missings are dropped in `df1` and `df2`
`on` columns; `isequal` is used for comparisons of rows for equality

It is not allowed to join on columns that contain `NaN` or `-0.0` in real or
imaginary part of the number. If you need to perform a join on such values use
Expand Down Expand Up @@ -626,7 +647,8 @@ change in future releases.
data frame and left unchanged.
- `matchmissing` : if equal to `:error` throw an error if `missing` is present
in `on` columns; if equal to `:equal` then `missing` is allowed and missings are
matched (`isequal` is used for comparisons of rows for equality)
matched; if equal to `:notequal` then missings are dropped in `df2` `on` columns;
`isequal` is used for comparisons of rows for equality

All columns of the returned data table will support missing values.

Expand Down Expand Up @@ -772,7 +794,8 @@ change in future releases.
data frame and left unchanged.
- `matchmissing` : if equal to `:error` throw an error if `missing` is present
in `on` columns; if equal to `:equal` then `missing` is allowed and missings are
matched (`isequal` is used for comparisons of rows for equality)
matched; if equal to `:notequal` then missings are dropped in `df1` `on` columns;
`isequal` is used for comparisons of rows for equality

All columns of the returned data table will support missing values.

Expand Down Expand Up @@ -923,7 +946,7 @@ This behavior may change in future releases.
data frame and left unchanged.
- `matchmissing` : if equal to `:error` throw an error if `missing` is present
in `on` columns; if equal to `:equal` then `missing` is allowed and missings are
matched (`isequal` is used for comparisons of rows for equality)
matched; `isequal` is used for comparisons of rows for equality

All columns of the returned data table will support missing values.

Expand Down Expand Up @@ -1071,7 +1094,8 @@ The order of rows in the result is undefined and may change in the future releas
By default no check is performed.
- `matchmissing` : if equal to `:error` throw an error if `missing` is present
in `on` columns; if equal to `:equal` then `missing` is allowed and missings are
matched (`isequal` is used for comparisons of rows for equality)
matched; if equal to `:notequal` then missings are dropped in `df2` `on` columns;
`isequal` is used for comparisons of rows for equality

It is not allowed to join on columns that contain `NaN` or `-0.0` in real or
imaginary part of the number. If you need to perform a join on such values use
Expand Down Expand Up @@ -1176,7 +1200,8 @@ The order of rows in the result is undefined and may change in the future releas
By default no check is performed.
- `matchmissing` : if equal to `:error` throw an error if `missing` is present
in `on` columns; if equal to `:equal` then `missing` is allowed and missings are
matched (`isequal` is used for comparisons of rows for equality)
matched; if equal to `:notequal` then missings are dropped in `df2` `on` columns;
`isequal` is used for comparisons of rows for equality

It is not allowed to join on columns that contain `NaN` or `-0.0` in real or
imaginary part of the number. If you need to perform a join on such values use
Expand Down
141 changes: 141 additions & 0 deletions test/join.jl
View file
Open in desktop
Original file line number Diff line number Diff line change
Expand Up @@ -32,6 +32,8 @@ anti = left[Bool[ismissing(x) for x in left.Job], [:ID, :Name]]

@test_throws ArgumentError innerjoin(name, job)
@test_throws ArgumentError innerjoin(name, job, on = :ID, matchmissing=:errors)
@test_throws ArgumentError innerjoin(name, job, on = :ID, matchmissing=:weirdmatch)
@test_throws ArgumentError outerjoin(name, job, on = :ID, matchmissing=:notequal)

@test innerjoin(name, job, on = :ID) == inner
@test outerjoin(name, job, on = :ID) ≅ outer
Expand Down Expand Up @@ -1557,4 +1559,143 @@ end
c="c", d="d")
end

@testset "matchmissing :notequal correctness" begin
Random.seed!(1337)
names = [
DataFrame(ID=[1, 2, missing],
Name=["John Doe", "Jane Doe", "Joe Blogs"]),
DataFrame(ID=Int[],
Name=String[]),
pstorozenko marked this conversation as resolved.
Show resolved Hide resolved
DataFrame(ID=Union{Int, Missing}[missings(3)...],
pstorozenko marked this conversation as resolved.
Show resolved Hide resolved
Name=String["John Doe", "Jane Doe", "Joe Blogs"]),
pstorozenko marked this conversation as resolved.
Show resolved Hide resolved
DataFrame(ID=[1, 2, 3],
Name=[missing, "Jane Doe", missing]),
DataFrame(ID=[1:100; missings(100)],
Name=repeat(["Jane Doe"], 200)),
DataFrame(ID=[missings(100); 1:100],
Name=repeat(["Jane Doe"], 200)),
DataFrame(ID=[1:50; missings(100); 51:100],
Name=repeat(["Jane Doe"], 200)),
DataFrame(ID=[1:64; missings(64); 129:200],
Name=repeat(["Jane Doe"], 200)),
DataFrame(ID=[1:63; missings(65); 129:200],
Name=repeat(["Jane Doe"], 200)),
DataFrame(ID=rand([1:1000; missing], 10000),
Name=rand(["John Doe", "Jane Doe", "Joe Blogs", missing], 10000)),
]
jobs = [
DataFrame(ID=[1, 2, 2, 4],
Job=["Lawyer", "Doctor", "Florist", "Farmer"]),
DataFrame(ID=[missing, 2, 2, 4],
Job=["Lawyer", "Doctor", "Florist", "Farmer"]),
DataFrame(ID=[missing, 2, 2, 4],
Job=["Lawyer", "Doctor", missing, "Farmer"]),
DataFrame(ID=Union{Int, Missing}[],
Job=Union{String, Missing}[]),
pstorozenko marked this conversation as resolved.
Show resolved Hide resolved
DataFrame(ID=[1:100; missings(100)],
Job=repeat(["Lawyer"], 200)),
DataFrame(ID=[missings(100); 1:100],
Job=repeat(["Lawyer"], 200)),
DataFrame(ID=[1:50; missings(100); 51:100],
Job=repeat(["Lawyer"], 200)),
DataFrame(ID=[1:64; missings(64); 129:200],
Job=repeat(["Lawyer"], 200)),
DataFrame(ID=[1:63; missings(65); 129:200],
Job=repeat(["Lawyer"], 200)),
DataFrame(ID=rand([1:1000; missing], 10000),
Job=rand(["Lawyer", "Doctor", "Florist", missing], 10000)),
]
for name in names, job in jobs
@test leftjoin(name, dropmissing(job, :ID), on=:ID, matchmissing=:equal) ≅
leftjoin(name, job, on=:ID, matchmissing=:notequal)
@test semijoin(name, dropmissing(job, :ID), on=:ID, matchmissing=:equal) ≅
semijoin(name, job, on=:ID, matchmissing=:notequal)
@test antijoin(name, dropmissing(job, :ID), on=:ID, matchmissing=:equal) ≅
antijoin(name, job, on=:ID, matchmissing=:notequal)
@test rightjoin(dropmissing(name, :ID), job, on=:ID, matchmissing=:equal) ≅
rightjoin(name, job, on=:ID, matchmissing=:notequal)
@test innerjoin(dropmissing(name, :ID), dropmissing(job, :ID), on=:ID, matchmissing=:equal) ≅
innerjoin(name, job, on=:ID, matchmissing=:notequal)
end

rl(n) = rand(["a", "b", "c"], n)
names2 = [
DataFrame(ID1=[1, 1, 2],
ID2=["a", "b", "a"],
Name=["John Doe", "Jane Doe", "Joe Blogs"]),
DataFrame(ID1=[1, 1, 2, missing],
ID2=["a", "b", "a", missing],
Name=["John Doe", "Jane Doe", "Joe Blogs", missing]),
DataFrame(ID1=[missing, 1, 2, missing],
ID2=["a", "b", missing, missing],
Name=[missing, "Jane Doe", "Joe Blogs", missing]),
DataFrame(ID1=[missing, 1, 2, missing],
ID2=["a", "b", missing, missing],
Name=Union{String, Missing}[missings(4)...]),
pstorozenko marked this conversation as resolved.
Show resolved Hide resolved
DataFrame(ID1=[missing, 1, 2, missing],
ID2=Union{String, Missing}[missings(4)...],
pstorozenko marked this conversation as resolved.
Show resolved Hide resolved
Name=["John Doe", "Jane Doe", "Joe Blogs", missing]),
DataFrame(ID1=[1:100; missings(100)],
ID2=[rl(100); missings(100)],
Name=rand(["Jane Doe", "Jane Doe"], 200)),
DataFrame(ID1=[missings(100); 1:100],
ID2=[missings(100); rl(100)],
Name=rand(["Jane Doe", "Jane Doe"], 200)),
DataFrame(ID1=[1:50; missings(100); 51:100],
ID2=[rl(50); missings(100); rl(50)],
Name=rand(["Jane Doe", "Jane Doe"], 200)),
DataFrame(ID1=[1:64; missings(64); 129:200],
ID2=[rl(64); missings(64); rl(200 - 128)],
Name=rand(["Jane Doe", "Jane Doe"], 200)),
DataFrame(ID1=[1:63; missings(65); 129:200],
ID2=[rl(64); missings(65); rl(200 - 129)],
Name=rand(["Jane Doe", "Jane Doe"], 200)),
DataFrame(ID1=rand([1:100; missing], 10000),
ID2=rand(["a", "b", "c", missing], 10000),
Name=rand(["John Doe", "Jane Doe", "Joe Blogs", missing], 10000)),
]
jobs2 = [
DataFrame(ID1=[1, 2, 2, 4],
ID2=["a", "b", "b", "c"],
Job=["Lawyer", "Doctor", "Florist", "Farmer"]),
DataFrame(ID1=[1, 2, 2, 4, missing],
ID2=["a", "b", "b", "c", missing],
Job=["Lawyer", "Doctor", "Florist", "Farmer", missing]),
DataFrame(ID1=[1, 2, missing, 4, missing],
ID2=["a", "b", missing, "c", missing],
Job=[missing, "Doctor", "Florist", "Farmer", missing]),
DataFrame(ID1=[1:100; missings(100)],
ID2=[rl(100); missings(100)],
Job=rand(["Doctor", "Florist"], 200)),
DataFrame(ID1=[missings(100); 1:100],
ID2=[missings(100); rl(100)],
Job=rand(["Doctor", "Florist"], 200)),
DataFrame(ID1=[1:50; missings(100); 51:100],
ID2=[rl(50); missings(100); rl(50)],
Job=rand(["Doctor", "Florist"], 200)),
DataFrame(ID1=[1:64; missings(64); 129:200],
ID2=[rl(64); missings(64); rl(200 - 128)],
Job=rand(["Doctor", "Florist"], 200)),
DataFrame(ID1=[1:63; missings(65); 129:200],
ID2=[rl(64); missings(65); rl(200 - 129)],
Job=rand(["Doctor", "Florist"], 200)),
DataFrame(ID1=rand([1:100; missing], 10000),
ID2=rand(["a", "b", "c", missing], 10000),
Job=rand(["Doctor", "Florist", "Farmer", missing], 10000)),
]
k = [:ID1, :ID2]
for name in names2, job in jobs2
@test leftjoin(name, dropmissing(job, k), on=k, matchmissing=:equal) ≅
leftjoin(name, job, on=k, matchmissing=:notequal)
@test semijoin(name, dropmissing(job, k), on=k, matchmissing=:equal) ≅
semijoin(name, job, on=k, matchmissing=:notequal)
@test antijoin(name, dropmissing(job, k), on=k, matchmissing=:equal) ≅
antijoin(name, job, on=k, matchmissing=:notequal)
@test rightjoin(dropmissing(name, k), job, on=k, matchmissing=:equal) ≅
rightjoin(name, job, on=k, matchmissing=:notequal)
@test innerjoin(dropmissing(name, k), dropmissing(job, k), on=k, matchmissing=:equal) ≅
innerjoin(name, job, on=k, matchmissing=:notequal)
end
end

end # module