Color for each bar in a bar plot

General Usage
,
1

Hi! I want to plot a bar graph of a data frame with 1 column corresponding to the height of the bar and another to its color. For example:

julia> df= DataFrame("data"=>rand(10), "color"=>round.(Int,rand(10)*9 .+ 1))
10×2 DataFrame
 Row │ data      color 
     │ Float64   Int64 
─────┼─────────────────
   1 │ 0.94554       5
   2 │ 0.486335      4
   3 │ 0.946361      4
   4 │ 0.585166      7
   5 │ 0.673883      8
   6 │ 0.190704      8
   7 │ 0.740319      9
   8 │ 0.441671      5
   9 │ 0.952198      5
  10 │ 0.199103      8

julia> bar(df.data, color=:BrBG_10)

This gives a bar plot with only 1 color taken from my palette.

fig

How can I use the vector df.color to index into the palette and get a different color for each bar?

Equivalently, I can pick a different color for each point in a scatter plot:

julia> scatter(1:10,df.data, marker_z= df.color, color=:BrBG_10)

fig

but this doesn’t seem to translate to bar plots.

2

Welcome, fellow HEP user :wink:

First you can generate the df with just:

julia> df= DataFrame(data=rand(10), cs=rand(1:10, 10))
10×2 DataFrame
 Row │ data       cs    
     │ Float64    Int64 
─────┼──────────────────
   1 │ 0.312918       7
   2 │ 0.349858       8
   3 │ 0.80407        3
   4 │ 0.702752       3
   5 │ 0.384168       1
   6 │ 0.302115       2
   7 │ 0.727981       8
   8 │ 0.99511       10
   9 │ 0.0483379      6
  10 │ 0.871488       3

The logic of Plots.jl’s bar is that different groups(color) are on different “column” of the input data, compare:

julia> bar([1 10],
           [1 2],
       )

julia> bar([1,10],
           [1,2],
       )

For what you want in this case:

julia> palette = cgrad(:BrBG_10);

julia> bar((1:10)', df.data', color=[palette[i] for i in df.cs]')

image

2 Likes
3

this is slightly different because this is assuming you’re scattering some 3D data in a 2D plane and using color to represent depth. (or maybe something like data point are in 2D but each point has a temperature represented by color). Bar plot usually doesn’t have this structure

4

This could be written without a comprehension as:
bar((1:n)', df.data', color=palette[df.cs]')

1 Like
5

Thank you @jling ! Works perfectly!

Are you at CERN?

1 Like
6

I was at CERN during the summer haha, but now I’m back to the school. Btw, join us at: https://groups.google.com/g/julia-hep

also if you ever work with .root files again, checkout UnROOT.jl (sorry for the shameless plug)

1 Like