I am trying to put together a Plots.jl user recipe, that would create several scatter plots and histograms arranged in a grid. The scatter plots should occupy the lower triangle of the grid. Histograms go on the diagonals. Each scatter plot should have several series, shown in different colours. The colours should be consistent across subplots.
My problem is when I allow the user to specify custom colours, the series colours in the scatter subplots get messed up.
Here is the (abridged) recipe code. data is is a vector of size r. Each element of data is a matrix of size n_r x d. n_r may vary, d stays the same. There should be d x (d-1) / 2 scatter plots, with r series on each plot, each series having n_r points.
@recipe function my_func(data::my_type; custom_colors=nothing)
# get d, r, ...
for i in i:d
for j in 1:d
@series begin
subplot := (i - 1) * d + j
if i == j
seriestype := :histogram
plot_data = # ... prepare data for histograms
elseif j < i
# scatter subplot recipe
seriestype := :scatter
if custom_colors !== nothing
color := reshape(custom_colors, (1, r))
end
x = Vector()
y = Vector()
for r in runs
ser = data[r]
append!(x, [ser[:, j]])
append!(y, [ser[:, i]])
end # for r
plot_data = (x, y)
else
# leave empty
plot_data = [0]
end # if/else
plot_data # return from the macro function
end # @series
end # for j
end # for i
end # @recipe
My problem is that whenever I supply custom colours, I get inconsistency across subplots:
plot(my_data, custom_colors=["blue", "green", "black"])
Note how subplot (2, 1) has black dots in the middle, whereas all other scatter subplots have black on the outside
If I do the plot without custom colors:
plot(my_data)
I get consistent colours in all scatter subplots:
Any clues why the colours are mixed up on the first plot?