Discrete colorbar with PlotlyJS

Hi, wonder if there is any good way to make a discrete colour bar when making a heatmap plot with PlotlyJS? It would also be nice to be able to label these.

Here is an example:

The code to produce this is


using PlotlyJS

## Set-up:
# Define the categories and corresponding border levels between them:
categories = ["Undetectable", "Low", "Medium", "High", "Extreme", "Wat"]
levels = [0, 8, 16, 24, 32, 40, 48]
label_levels = levels .+ 4 # note the shift to centre the labels

# Some random data:
data = [
    29  16  15   7  14
    10   4  36  14   2
    25  23   0  29  41
]
# data = rand(minimum(levels):maximum(levels), 3, 5)

# A discrete colour scheme: see https://juliagraphics.github.io/Colors.jl/dev/namedcolors/
colour_pairs = [
    "crimson"      => (220,  20,  60),
    "orangered"    => (255,  69,   0),
    "goldenrod"    => (218, 165,  32),
    "aquamarine3"  => (102, 205, 170),
    "royalblue"    => ( 65, 105, 225),
    "darkviolet"   => (148,   0, 211),
    "black"        => (  0,   0,   0),
]
# Make it into a suitable list:
colour_list = map(x -> "rgb$(x[2])", colour_pairs)

# The tricky part - creating ordered ranges of colours:
norm_levels = (levels .- minimum(levels))/maximum(levels)
colour_marks = [[norm_levels[1], colour_list[1]]]
for i in 2:length(levels)
    push!(colour_marks, [norm_levels[i], colour_list[i-1]])
    push!(colour_marks, [norm_levels[i], colour_list[i]])
end

## Plot:
# Adapting http://juliaplots.org/PlotlyJS.jl/stable/examples/heatmaps/
# See https://plotly.com/julia/reference/heatmap/ for reference to keyword arguments
function heatmap2()
    trace = heatmap(
        x=["Monday", "Tuesday", "Wednesday", "Thursday", "Friday"],
        y=["Morning", "Afternoon", "Evening"],
        z=data,
        autocolorscale=false,
        colorbar=attr(tickmode="array",
                      tickvals=label_levels,
                      ticktext=categories),
        colorscale=colour_marks,
        zmin=minimum(levels),
        zmax=maximum(levels)
    )
    plot(trace)
end
p = heatmap2()

See also this post:

I post a Julia version of the plotly.py code for a heatmap with discrete colorscale
https://chart-studio.plotly.com/~empet/15229/heatmap-with-a-discrete-colorscale/#/

using PlotlyJS
function discrete_colorscale(boundary_vals, mycolors)
    #boundary_vals - vector of values bounding intervals/ranges of interest
    #colors - list of rgb or hex colorcodes for values in [bvals[k], bvals[k+1]],1<=k < length(boundary_vals)
    #returns the plotly  discrete colorscale
    if length(boundary_vals) != length(mycolors)+1
           error("length of boundary values should be equal to  length(mycolors)+1")
    end
    bvals = sort(boundary_vals)     
    nvals = [(v-bvals[1])/(bvals[end]-bvals[1]) for v in bvals]  #normalized values
    
    dcolorscale = [] #discrete colorscale
    for k in 1:length(mycolors)
        append!(dcolorscale, [[nvals[k], mycolors[k]], [nvals[k+1], mycolors[k]]])
    end    
    return dcolorscale 
      
end            

function colorbar_ticks(bvals)
    tickvals = [sum(bvals[k:k+1])/2 for k in 1:length(bvals)-1] #position with respect to bvals, to be placed ticktext 
    ticktext = String[]
    push!(ticktext, "<$(bvals[2])")
    for k in 2:length(bvals)-2
        push!(ticktext, "$(bvals[k])-$(bvals[k+1])")
    end        
    push!(ticktext, ">$(bvals[end-1])")  
    return tickvals, ticktext
end        
       
z = rand(2:90, 20, 20)
bvals = [2, 15, 40, 65, 90]
mycolors = ["#09ffff", "#19d3f3", "#e763fa" , "#ab63fa"]
#mycolors = ["rgb(133,187,77)","rgb(202,216,216)","rgb(238,164,2)", "rgb(221,114,54)"]
dcolorscale = discrete_colorscale(bvals, mycolors)

tvals, ttext= colorbar_ticks(bvals)

pl = Plot(heatmap(z=z, colorscale=dcolorscale,
                  colorbar=attr(tickvals=tvals, ticktext=ttext)), 
          Layout(width=450, height=400))

discrete-colorscale

1 Like

This is really nice; more adaptive than my example by using functions to generate labels, for example. And I really like your colour choice :yum:!

Thanks for you`re advice, by using a combination of these suggestions I ended up with a really nice plot! Thanks!

1 Like