function hp_row_sort!(ds::Dataset, cols = names(ds, Union{Missing, Number}); kwargs...)

colsidx = index(ds)[cols]

T = mapreduce(eltype, promote_type, view(_columns(ds),colsidx))

m = Matrix{T}(ds[!, colsidx])

Threads.@threads for i in 1:size(m, 1)

@views sort!(m[i, :]; kwargs...)

end

# TODO no parallel is needed here to minimise memory

for i in 1:length(colsidx)

_columns(ds)[colsidx[i]] = m[:, i]

end

removeformat!(ds, cols)

any(index(ds).sortedcols .∈ Ref(colsidx)) && _reset_grouping_info!(ds)

_modified(_attributes(ds))

ds

end

"""

row_sort!(ds::AbstractDataset[, cols]; kwargs...)

sort `cols` in each row.

"""

function hp_row_sort(ds::AbstractDataset, cols = names(ds, Union{Missing, Number}); kwargs...)

dscopy = copy(ds)

hp_row_sort!(dscopy, cols; kwargs...)

dscopy

end

function hp_row_generic(ds::AbstractDataset, f::Function, cols::MultiColumnIndex)

colsidx = multiple_getindex(index(ds), cols)

if length(colsidx) == 2

try

allowmissing(f.(_columns(ds)[colsidx[1]], _columns(ds)[colsidx[2]]))

catch e

if e isa MethodError

_hp_row_generic(ds, f, colsidx)

else

rethrow(e)

end

end

else

_hp_row_generic(ds, f, colsidx)

end

end

function _hp_row_generic(ds::AbstractDataset, f::Function, colsidx)

T = mapreduce(eltype, promote_type, view(_columns(ds),colsidx))

inmat = Matrix{T}(undef, length(colsidx), min(1000, nrow(ds)))

all_data = view(_columns(ds), colsidx)

_fill_matrix!(inmat, all_data, 1:min(1000, nrow(ds)), colsidx)

res_temp = allowmissing(f.(eachcol(inmat)))

if !(typeof(res_temp) <: AbstractVector)

throw(ArgumentError("output of `f` must be a vector"))

end

# if length(res_temp[1]) > 1

# throw(ArgumentError("The matrix output is not supported"))

# res = similar(res_temp, nrow(ds), size(res_temp,2))

# elseif length(res_temp[1]) == 1

res = similar(res_temp, nrow(ds))

# else

# throw(ArgumentError("the result cannot be with zero dimension"))

if nrow(ds)>1000

if size(res, 2) == 1

view(res, 1:1000) .= res_temp

_hp_row_generic_vec!(res, ds, f, colsidx, Val(T))

else

view(res, 1:1000, :) .= res_temp

# not implemented yet

_hp_row_generic_mat!(res, ds, f, colsidx)

end

else

return res_temp

end

return res

end

function _hp_row_generic_vec!(res, ds, f, colsidx, ::Val{T}) where T

nt = Threads.nthreads()

loopsize = div(length(res) - 1000, 1000)

all_data = view(_columns(ds), colsidx)

if loopsize == 0

st = 1001

en = length(res)

inmat = Matrix{T}(undef, length(colsidx), en - st + 1)

_fill_matrix!(inmat, all_data, st:en, colsidx)

view(res, st:en) .= f.(eachcol(inmat))

return

end

max_cz = length(res) - 1000 - (loopsize - 1)*1000

inmat_all = [Matrix{T}(undef, length(colsidx), max_cz) for i in 1:nt]

# make sure that the variable inside the loop are not the same as the out of scope one

Threads.@threads :static for i in 1:loopsize

t_st = i*1000 + 1

i == loopsize ? t_en = length(res) : t_en = (i+1)*1000

_fill_matrix!(inmat_all[Threads.threadid()], all_data, t_st:t_en, colsidx)

for k in t_st:t_en

res[k] = f(view(inmat_all[Threads.threadid()], :, k - t_st + 1))

end

end

end