Grouping a dataframe by binning a numeric column

groupbins.jl

function find_bin(x::Float64, limits::Vector{Float64})
    bin = length(limits) + 1
    for i in 1:length(limits)
        if x < limits[i]
            bin = i
            break
        end
    end
    bin
end

# some tests
@assert (find_bin(0.2, [1.0, 2.3, 4.5, 6.7, 9.0]) == 1)
@assert (find_bin(1.0, [1.0, 2.3, 4.5, 6.7, 9.0]) == 2)
@assert (find_bin(1.2, [1.0, 2.3, 4.5, 6.7, 9.0]) == 2)
@assert (find_bin(22.1, [1.0, 2.3, 4.5, 6.7, 9.0]) == 6)
@assert (find_bin(9.0, [1.0, 2.3, 4.5, 6.7, 9.0]) == 6)
@assert (find_bin(6.7, [1.0, 2.3, 4.5, 6.7, 9.0]) == 5)

# return a vector with the bins of all elements in dv (this is a map)
function find_bins(dv::DataVec{Float64}, limits::Vector{Float64})
    vbins = zeros(Int64, length(dv))
    for i in 1:length(dv)
        vbins[i] = find_bin(dv[i], limits)
    end
    vbins
end

# create GroupedDataFrame by binning values in a 
# numeric column according to some binning limits
function group_bins{T}(df::AbstractDataFrame, col::T, limits::Vector{Float64})
    n_bins = length(limits) + 1

    vbins = find_bins(df[col], limits)
    (idx, starts) = groupsort_indexer(vbins, n_bins)

    # Remove zero-length groupings
    starts = _uniqueofsorted(starts) 
    ends = [starts[2:end] - 1]
    GroupedDataFrame(df, [col], idx, starts[1:end-1], ends)
end

# a test
df1 = DataFrame(quote 
    x = [1:8]
    y = [5.4, 10.3, 21.7, 24.5, 16.2, 32.9, 8.2, 29.9]
end)

@assert (length(group_bins(df1, "y", [10.0, 20.0, 30.0])) == 4)

This is fantastic. This should be part of DataFrames or as a supplemental package. I've made a version where limits is also of type ::RangeStepLen{Float64} and it works with things like range(0.0, 0.5, 20) with no code changes. I went looking for this type of binning in DataFrames and came across this gist. Thanks for making this available.