Plots.jl: histograms fail in the presence of NaNs

[mkborregaard]

julia> a = randn(10);

julia> a[3] = NaN

julia> histogram(a)  #errors on the quantile call in in recipes.jl line 504
ERROR: ArgumentError: quantiles are undefined in presence of NaNs
Stacktrace:
 [1] #quantile!#546(::Bool, ::Function, ::Array{Float64,1}, ::Float64) at ./statistics.jl:688
 [2] (::Base.#kw##quantile!)(::Array{Any,1}, ::Base.#quantile!, ::Array{Float64,1}, ::Float64) at ./<missing>:0
 [3] #quantile#547(::Bool, ::Function, ::Array{Float64,1}, ::Float64) at ./statistics.jl:739
 [4] (::Plots.#_iqr#226)(::Array{Float64,1}) at /Users/michael/.julia/v0.6/Plots/src/recipes.jl:504
 [5] #_auto_binning_nbins#224(::Symbol, ::Function, ::Tuple{Array{Float64,1}}, ::Int64) at /Users/michael/.julia/v0.6/Plots/src/recipes.jl:528
 [6] (::Plots.#kw##_auto_binning_nbins)(::Array{Any,1}, ::Plots.#_auto_binning_nbins, ::Tuple{Array{Float64,1}}, ::Int64) at ./<missing>:0
 [7] map(::Plots.##230#231{Tuple{Array{Float64,1}},Symbol}, ::Tuple{Int64}) at ./tuple.jl:158
 [8] #_make_hist#232(::Bool, ::Void, ::Function, ::Tuple{Array{Float64,1}}, ::Symbol) at /Users/michael/.julia/v0.6/Plots/src/recipes.jl:550
 [9] (::Plots.#kw##_make_hist)(::Array{Any,1}, ::Plots.#_make_hist, ::Tuple{Array{Float64,1}}, ::Symbol) at ./<missing>:0
 [10] macro expansion at /Users/michael/.julia/v0.6/Plots/src/recipes.jl:566 [inlined]
 [11] apply_recipe(::Dict{Symbol,Any}, ::Type{Val{:barhist}}, ::UnitRange{Int64}, ::Array{Float64,1}, ::Void) at /Users/michael/.julia/v0.6/RecipesBase/src/RecipesBase.jl:287
 [12] _process_seriesrecipe(::Plots.Plot{Plots.GRBackend}, ::Dict{Symbol,Any}) at /Users/michael/.julia/v0.6/Plots/src/pipeline.jl:406
 [13] _process_seriesrecipe(::Plots.Plot{Plots.GRBackend}, ::Dict{Symbol,Any}) at /Users/michael/.julia/v0.6/Plots/src/pipeline.jl:415
 [14] _plot!(::Plots.Plot{Plots.GRBackend}, ::Dict{Symbol,Any}, ::Tuple{Array{Float64,1}}) at /Users/michael/.julia/v0.6/Plots/src/plot.jl:231
 [15] (::RecipesBase.#kw##plot)(::Array{Any,1}, ::RecipesBase.#plot, ::Array{Float64,1}) at ./<missing>:0
 [16] #histogram#595(::Array{Any,1}, ::Function, ::Array{Float64,1}, ::Vararg{Array{Float64,1},N} where N) at /Users/michael/.julia/v0.6/RecipesBase/src/RecipesBase.jl:381
 [17] histogram(::Array{Float64,1}, ::Vararg{Array{Float64,1},N} where N) at /Users/michael/.julia/v0.6/RecipesBase/src/RecipesBase.jl:381

julia> histogram(a, bins = :scott) # avoids the quantile call but fails in StatsBase
ERROR: InexactError()
Stacktrace:
 [1] trunc(::Type{Int64}, ::Float64) at ./float.jl:672
 [2] nbitslen(::Float64, ::Int64) at ./twiceprecision.jl:61
 [3] nbitslen at ./twiceprecision.jl:58 [inlined]
 [4] floatrange(::Float64, ::Float64, ::Float64, ::Float64) at ./twiceprecision.jl:119
 [5] histrange(::Float64, ::Float64, ::Int64, ::Symbol) at /Users/michael/.julia/v0.6/StatsBase/src/hist.jl:115
 [6] histrange(::Array{Float64,1}, ::Int64, ::Symbol) at /Users/michael/.julia/v0.6/StatsBase/src/hist.jl:54
 [7] _hist_edge(::Tuple{Array{Float64,1}}, ::Int64, ::Int64) at /Users/michael/.julia/v0.6/Plots/src/recipes.jl:536
 [8] map(::Plots.##230#231{Tuple{Array{Float64,1}},Symbol}, ::Tuple{Int64}) at ./tuple.jl:158
 [9] #_make_hist#232(::Bool, ::Void, ::Function, ::Tuple{Array{Float64,1}}, ::Symbol) at /Users/michael/.julia/v0.6/Plots/src/recipes.jl:550
 [10] (::Plots.#kw##_make_hist)(::Array{Any,1}, ::Plots.#_make_hist, ::Tuple{Array{Float64,1}}, ::Symbol) at ./<missing>:0
 [11] macro expansion at /Users/michael/.julia/v0.6/Plots/src/recipes.jl:566 [inlined]
 [12] apply_recipe(::Dict{Symbol,Any}, ::Type{Val{:barhist}}, ::UnitRange{Int64}, ::Array{Float64,1}, ::Void) at /Users/michael/.julia/v0.6/RecipesBase/src/RecipesBase.jl:287
 [13] _process_seriesrecipe(::Plots.Plot{Plots.GRBackend}, ::Dict{Symbol,Any}) at /Users/michael/.julia/v0.6/Plots/src/pipeline.jl:406
 [14] _process_seriesrecipe(::Plots.Plot{Plots.GRBackend}, ::Dict{Symbol,Any}) at /Users/michael/.julia/v0.6/Plots/src/pipeline.jl:415
 [15] _plot!(::Plots.Plot{Plots.GRBackend}, ::Dict{Symbol,Any}, ::Tuple{Array{Float64,1}}) at /Users/michael/.julia/v0.6/Plots/src/plot.jl:231
 [16] (::RecipesBase.#kw##plot)(::Array{Any,1}, ::RecipesBase.#plot, ::Array{Float64,1}) at ./<missing>:0
 [17] #histogram#595(::Array{Any,1}, ::Function, ::Array{Float64,1}, ::Vararg{Array{Float64,1},N} where N) at /Users/michael/.julia/v0.6/RecipesBase/src/RecipesBase.jl:381
 [18] (::Plots.#kw##histogram)(::Array{Any,1}, ::Plots.#histogram, ::Array{Float64,1}, ::Vararg{Array{Float64,1},N} where N) at ./<missing>:0

[mkborregaard]

This also means that @df calls of tables in StatPlots containing NA's / NULL's will fail, as they are converted to NaNs. All of Plots is designed to work in the presence of NaNs.

[oschulz]

I'll get on it.

[mkborregaard]

I'm guessing the Plots quantile call is easy to fix, but possibly the StatsBase issue more complex?

[oschulz]

Sorry, I didn't find any time to look into it yet - but I will, promise!

[joshualeond]

I just ran into this problem, is there a work around for this? I see there's been a lot of changes with missing values in Julia since this issue was opened.

[mkborregaard]

@oschulz do you plan to fix this?

[oschulz]

Sorry for the delay - yes!

[pgbarletta]

Hi, I've just ran into it too. No big deal isnan() solves it. Just wanted to let you know.

[mkborregaard]

OK, I fixed it - @oschulz will you review? https://github.com/JuliaPlots/Plots.jl/pull/1889

[mkborregaard]

fixed