Implement tidier::fill() function based on Imputation.jl

[christophscheuch]

I really like the tidier::fill() function to fill in missing values with previous or next values, which frequently happens when I build balanced data sets. I found a workaround to implement the logic for single variables using Imputation.jl:

using DataFrames, Impute, Tidier
df = DataFrame(dt1=[0.2, missing, missing, 1, missing, 5, 6], dt2=[0.3, missing, missing, 3, missing, 5, 6])

@chain df begin
    @mutate(dt1 = ~Impute.locf(dt1))
end

@chain df begin
    @mutate(dt1 = ~Impute.nocb(dt1))
end

I suppose it might be a rather low hangig fruit to build a @fill function for Tidier.jl.

[drizk1]

below is the code for fill function. two helper functions were added so that Impute.jl does not need to be libraried in.

function locf(column::AbstractVector)
    last_observation = column[1]
    for i in 1:length(column)
        if ismissing(column[i])
            column[i] = last_observation
        else
            last_observation = column[i]
        end
    end
    return column
end

function nocb(column::AbstractVector)
    next_observation = column[end]
    for i in length(column):-1:1
        if ismissing(column[i])
            column[i] = next_observation
        else
            next_observation = column[i]
        end
    end
    return column
end

function fill(column::AbstractVector, method::String)
    if method == "locf"
        return locf(column)
    elseif method == "nocb"
        return nocb(column)
    else
        error("Unsupported fill method. Choose either 'locf' or 'nocb'.")
    end
end

df = DataFrame(dt1=[missing, 0.2, missing, missing, 1, missing, 5, 6], dt2=[0.3, missing, missing, 3, missing, 5, 6,missing])

# apply the fill function in the DataFrame chain
@chain df begin
    @mutate(dt1 = ~fill(dt1, "locf"))
    @mutate(dt2 = ~fill(dt2, "nocb"))
end

[kdpsingh]

@fill_missing has now been implemented as of #30. Closing this issue.