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

Improve structure of join benchmarks #2707

Merged
merged 4 commits into from Apr 13, 2021
Merged

Improve structure of join benchmarks #2707

Show file tree
Hide file tree
Changes from all commits
Commits
Show all changes
4 commits
Select commit Hold shift + click to select a range
cca8fe3
improve structure of join benchmarks
almeida-cloves-bcg Apr 10, 2021
eb28492
revert Project.toml changes
almeida-cloves-bcg Apr 10, 2021
5ba575b
restore run.sh script
almeida-cloves-bcg Apr 13, 2021
269d0fa
improve docs
almeida-cloves-bcg Apr 13, 2021
File filter

Filter by extension

Filter by extension
Conversations
Failed to load comments.
Jump to
Jump to file
Failed to load files.
Diff view
Diff view
101 changes: 0 additions & 101 deletions benchmarks/join_performance.jl
View file
Open in desktop

This file was deleted.

180 changes: 96 additions & 84 deletions benchmarks/joins/join_performance.jl
View file
Open in desktop
Expand Up @@ -5,97 +5,109 @@ using Random

fullgc() = (GC.gc(true); GC.gc(true); GC.gc(true); GC.gc(true))

@assert length(ARGS) == 7
@assert ARGS[3] in ["int", "pool", "cat", "str"]
@assert ARGS[4] in ["uniq", "dup", "manydup"]
@assert ARGS[5] in ["sort", "rand"]
@assert ARGS[6] in ["1", "2"]
@assert ARGS[7] in ["inner", "left", "right", "outer", "semi", "anti"]

@info ARGS

llen = parse(Int, ARGS[1])
rlen = parse(Int, ARGS[2])
@assert llen > 1000
@assert rlen > 2000

pad = maximum(length.(string.((llen, rlen))))

if ARGS[3] == "int"
if ARGS[4] == "uniq"
col1 = [1:llen;]
col2 = [1:rlen;]
elseif ARGS[4] == "dup"
col1 = repeat(1:llen ÷ 2, inner=2)
col2 = repeat(1:rlen ÷ 2, inner=2)
else
@assert ARGS[4] == "manydup"
col1 = repeat(1:llen ÷ 20, inner=20)
col2 = repeat(1:rlen ÷ 20, inner=20)
end
elseif ARGS[3] == "pool"
if ARGS[4] == "dup"
col1 = PooledArray(repeat(string.(1:llen ÷ 2, pad=pad), inner=2))
col2 = PooledArray(repeat(string.(1:rlen ÷ 2, pad=pad), inner=2))

function run_bench(llen, rlen, arr_type, dup_mode, sort_mode, num_keys, join_type)


@assert llen > 1000
@assert rlen > 2000

@assert arr_type in ["int", "pool", "cat", "str"]
@assert dup_mode in ["uniq", "dup", "manydup"]
@assert sort_mode in ["sort", "rand"]
@assert num_keys in ["1", "2"]
@assert join_type in ["inner", "left", "right", "outer", "semi", "anti"]

pad = maximum(length.(string.((llen, rlen))))

if arr_type == "int"
if dup_mode == "uniq"
col1 = [1:llen;]
col2 = [1:rlen;]
elseif dup_mode == "dup"
col1 = repeat(1:llen ÷ 2, inner=2)
col2 = repeat(1:rlen ÷ 2, inner=2)
else
@assert dup_mode == "manydup"
col1 = repeat(1:llen ÷ 20, inner=20)
col2 = repeat(1:rlen ÷ 20, inner=20)
end
elseif arr_type == "pool"
if dup_mode == "dup"
col1 = PooledArray(repeat(string.(1:llen ÷ 2, pad=pad), inner=2))
col2 = PooledArray(repeat(string.(1:rlen ÷ 2, pad=pad), inner=2))
else
@assert dup_mode == "manydup"
col1 = PooledArray(repeat(string.(1:llen ÷ 20, pad=pad), inner=20))
col2 = PooledArray(repeat(string.(1:rlen ÷ 20, pad=pad), inner=20))
end
elseif arr_type == "cat"
if dup_mode == "dup"
col1 = categorical(repeat(string.(1:llen ÷ 2, pad=pad), inner=2))
col2 = categorical(repeat(string.(1:rlen ÷ 2, pad=pad), inner=2))
else
@assert dup_mode == "manydup"
col1 = categorical(repeat(string.(1:llen ÷ 20, pad=pad), inner=20))
col2 = categorical(repeat(string.(1:rlen ÷ 20, pad=pad), inner=20))
end
else
@assert ARGS[4] == "manydup"
col1 = PooledArray(repeat(string.(1:llen ÷ 20, pad=pad), inner=20))
col2 = PooledArray(repeat(string.(1:rlen ÷ 20, pad=pad), inner=20))
@assert arr_type == "str"
if dup_mode == "uniq"
col1 = string.(1:llen, pad=pad)
col2 = string.(1:rlen, pad=pad)
elseif dup_mode == "dup"
col1 = repeat(string.(1:llen ÷ 2, pad=pad), inner=2)
col2 = repeat(string.(1:rlen ÷ 2, pad=pad), inner=2)
else
@assert dup_mode == "manydup"
col1 = repeat(string.(1:llen ÷ 20, pad=pad), inner=20)
col2 = repeat(string.(1:rlen ÷ 20, pad=pad), inner=20)
end
end
elseif ARGS[3] == "cat"
if ARGS[4] == "dup"
col1 = categorical(repeat(string.(1:llen ÷ 2, pad=pad), inner=2))
col2 = categorical(repeat(string.(1:rlen ÷ 2, pad=pad), inner=2))

Random.seed!(1234)

if sort_mode == "rand"
shuffle!(col1)
shuffle!(col2)
else
@assert ARGS[4] == "manydup"
col1 = categorical(repeat(string.(1:llen ÷ 20, pad=pad), inner=20))
col2 = categorical(repeat(string.(1:rlen ÷ 20, pad=pad), inner=20))
@assert sort_mode == "sort"
end
else
@assert ARGS[3] == "str"
if ARGS[4] == "uniq"
col1 = string.(1:llen, pad=pad)
col2 = string.(1:rlen, pad=pad)
elseif ARGS[4] == "dup"
col1 = repeat(string.(1:llen ÷ 2, pad=pad), inner=2)
col2 = repeat(string.(1:rlen ÷ 2, pad=pad), inner=2)

joinfun = Dict("inner" => innerjoin, "left" => leftjoin,
"right" => rightjoin, "outer" => outerjoin,
"semi" => semijoin, "anti" => antijoin)[join_type]

if num_keys == "1"
df1 = DataFrame(id1=col1)
df2 = DataFrame(id1=col2)
joinfun(df1[1:1000, :], df2[1:2000, :], on=:id1)
joinfun(df2[1:2000, :], df1[1:1000, :], on=:id1)
fullgc()
t1 = @elapsed joinfun(df1, df2, on=:id1)
fullgc()
t2 = @elapsed joinfun(df2, df1, on=:id1)
else
@assert ARGS[4] == "manydup"
col1 = repeat(string.(1:llen ÷ 20, pad=pad), inner=20)
col2 = repeat(string.(1:rlen ÷ 20, pad=pad), inner=20)
@assert num_keys == "2"
df1 = DataFrame(id1=col1, id2=col1)
df2 = DataFrame(id1=col2, id2=col2)
joinfun(df1[1:1000, :], df2[1:2000, :], on=[:id1, :id2])
joinfun(df2[1:2000, :], df1[1:1000, :], on=[:id1, :id2])
fullgc()
t1 = @elapsed joinfun(df1, df2, on=[:id1, :id2])
fullgc()
t2 = @elapsed joinfun(df2, df1, on=[:id1, :id2])
end

return (t1 + t2) / 2.0
end

Random.seed!(1234)

if ARGS[5] == "rand"
shuffle!(col1)
shuffle!(col2)
else
@assert ARGS[5] == "sort"
end
if abspath(PROGRAM_FILE) == @__FILE__
@info "$ARGS"

const joinfun = Dict("inner" => innerjoin, "left" => leftjoin,
"right" => rightjoin, "outer" => outerjoin,
"semi" => semijoin, "anti" => antijoin)[ARGS[7]]

if ARGS[6] == "1"
df1 = DataFrame(id1 = col1)
df2 = DataFrame(id1 = col2)
joinfun(df1[1:1000, :], df2[1:2000, :], on=:id1)
joinfun(df2[1:2000, :], df1[1:1000, :], on=:id1)
fullgc()
@time joinfun(df1, df2, on=:id1)
fullgc()
@time joinfun(df2, df1, on=:id1)
else
@assert ARGS[6] == "2"
df1 = DataFrame(id1 = col1, id2 = col1)
df2 = DataFrame(id1 = col2, id2 = col2)
joinfun(df1[1:1000, :], df2[1:2000, :], on=[:id1, :id2])
joinfun(df2[1:2000, :], df1[1:1000, :], on=[:id1, :id2])
fullgc()
@time joinfun(df1, df2, on=[:id1, :id2])
fullgc()
@time joinfun(df2, df1, on=[:id1, :id2])
end
llen = parse(Int, ARGS[1])
rlen = parse(Int, ARGS[2])
t = run_bench(llen, rlen, ARGS[3], ARGS[4], ARGS[5], ARGS[6], ARGS[7])
@info("Timing: $t sec")
end