PairsMacros.jl

Make it easier to construct pairs in DataFrames.jl
Author matthieugomez
Github
Popularity
7 Stars
Updated Last
3 Years Ago
Started In
August 2020

Build status

PairsMacros.jl

This package exports two macros:

  • @cols to make it easier to construct pairs of the form source => function => target.
  • @rows to make it easier to construct pairs of the form source => ByRow(function) => target.

These macros can be used within DataFrames.jl transform, select, subset, etc. It is a minimal alternative to DataFramesMeta.jl.

Syntax

using  PairsMacros
@cols z = sum(x)
#> [:x] => sum => :z
@rows z = x + y
#> [:x, :y] => ByRow(+) => :z

Use $ to substitute the name of certain columns by symbols

u = :y
@cols z = sum($u)
#> [:y] => sum => :z
@cols $u = sum(x)
#> [:x] => sum => :y
u = "my variable name"
@cols z = sum($u)
"my variable name" => sum => :z

Use esc to denote variables that do not refer to columns

u = [0.25, 0.75]
@cols z = quantile(y, esc(u))
#> [:y] => (x -> quantile(x, u)) => :z
@cols z = map(esc(cos), y)
#> [:y] => (x -> map(cos, x)) => :z
@rows z = tryparse(esc(Float64), y)
#> [:y] => ByRow(x -> tryparse(Float64, x)) => :z

Example

Use PairsMacros in conjunction with DataFrames

using DataFrames, Chain, Statistics, PairsMacros
df = DataFrame(a = repeat(1:5, outer = 20),
               b = repeat(["a", "b", "c", "d"], inner = 25),
               x = repeat(1:20, inner = 5))
transform(@rows y = 10 * x)

Use Chain for a sequence of transformations:

x_thread = @chain df begin
    transform(@rows y = 10 * x)
    subset(@rows a > 2)
    groupby(:b)
    combine(@cols meanX = mean(x) meanY = mean(y))
    sort(:meanX)
    select(@cols meanX meanY var = b)
end

Details

All symbols are assumed to refer to columns, with the exception of:

  • symbol missing
  • first args of a :call or :. expression (e.g. function calls)
  • arguments inside of a splicing/interpolation expression $()
  • arguments inside esc()