Feature request: Listing table for multiple columns

[PharmCat]

listingtable is great but it make table with only one variable. will be very useful to make similar table with row grouping and statistics with multiple columns - without 'cols' grouping variable.

Yes, I can make DataFrame stack-unstack but in this case I can't make pretty description for column header (as I can do in table_one).

[jkrumbiegel]

You could change the descriptions by replacing them with a utility function. I'm not sure if I would like to extend the API of this function to handle the multiple-columns case. It seems common enough but I'm not sure what a clean way would be to have this coexist with the current functionality.

using SummaryTables, DataFrames


df = DataFrame(
    group = repeat(["A", "B", "C", "D"], inner = 5),
    id = repeat(1:5, 4),
    value_a = randn(4 * 5),
    value_b = randn(4 * 5),
    value_c = randn(4 * 5),
)

function stack_with_rename(df, cols...)
    _replacer(col::String) = col => col
    _replacer(pair::Pair{String,<:Any}) = pair
    replacers = map(_replacer, cols)
    stacked = stack(df, collect(first.(replacers)))
    stacked.variable .= replace(stacked.variable, replacers...)
    return stacked
end

listingtable(
    stack_with_rename(df, "value_a", "value_b" => "Value B", "value_c" => Multiline("Value", "C")),
    :value,
    rows = [:group, :id],
    cols = :variable,
    sort = false,
    variable_header = false,
)

The sort = false here is necessary because the Multiline value breaks sorting. If you only used strings, that wouldn't be necessary.