[BUG] Colorbar doesn’t reflect ColorGradient scale as specified in cgrad(…, scale = ) -works on gr

[casasgomezuribarri]

Details

I have built the following composite plot:

# Data
x = rand([1, 2, 3, 4, 5], 50)
y = rand(50)
z = rand(50)

# Vector with labels to show on hover
hovector = []
for i in 1:length(x)
    label = "$(x[i]), $(y[i]), $(z[i])"
    push!(hovector, label)
 end

# Frequency histograms and blank plot for top right corner
plotly()
x_hist = Plots.plot(x,
    st = :histogram,
    nb = 50,
    leg = false,
    yguide = "Frequency",
    xticks = false,
    guide_position = :left,
    color = RGB{Float64}(0.031,0.318,0.612))
y_hist = Plots.plot(y,
    st = :histogram,
    nb = 50,
    leg = false,
    color = RGB{Float64}(0.031,0.318,0.612),
    yticks = false,
    xguide = "Frequency",
    orientation = :h)
blank = Plots.plot(grid = false, showaxis = false)

# Main plot
bubblecolor = cgrad(:inferno, scale = :exp)
bubbles = Plots.plot(x, y, st = :scatter,     
    marker_z = z,
    ms = 10,
    hover = hovector,
    color = bubblecolor, 
    label = nothing,
    cb = false)

# Workaround for title (plot_title not currently implemented, see http://docs.juliaplots.org/latest/generated/attributes_plot/)
title = Plots.plot(ones(3), line = nothing, size = (600, 80),
    grid = false, showaxis = false, leg = false,
    annotations=(2, 1.5, "Main Title"))

# Merge plots
l = @layout [x_hist              blank
            bubbles{0.8w,0.8h}   y_hist]
plots = Plots.plot(x_hist, blank, bubbles, y_hist, 
        layout = l, link = :both,
        size = (700, 500))
superplot = Plots.plot(title, plots, layout = grid(2,1,heights=[0.1,0.9]))

# Add colorbar
Plots.plot!([0], [0], zcolor=[NaN], cbtitle="Intensity", leg = false,
        clims = extrema(z),
        color = bubblecolor,
        background_color_subplot=:transparent,
        markerstrokecolor=:transparent, framestyle=:none, 
        cb = :right,
        inset=bbox(0.1, 0, 0.05, 0.9, :center, :right), 
        subplot=6)

which gives me:

My question is related to the last block # Add colorbar (note the color gradient bubblecolor defined just under # Main plot). I am using an exponential scale for my color gradient to make lower Intensity values stand out more clearly -darkly-, and although this is reflected in the color of the bubbles themselves, the colorbar still shows the default linear scale of the :inferno color gradient. This obviously affects the readability of the main scatter.

If I change the line where bubblecolor is defined to

bubblecolor = cgrad(:inferno, rev = true, scale = :exp)

then I get:

So, the color attribute in the last block is sensitive to changes in rev within cgrad, but not scale. Is this an expected limitation, or am I doing something wrong?

PS: Any idea why the cbtitle “Intensity” doesn’t show up? This is less of an issue but would be great to be able to include it.

Backends

This bug occurs on ( insert x below )

Backend	yes	no	untested
gr (default)		x
pyplot			error
plotly	x
plotlyjs	x
pgfplotsx			x
inspectdr			x

Versions

Plots.jl version:
Backend version (]st -m):
Output of versioninfo():

[hervasa2]

I have also found this same bug. The colorbar in pyplot() does not respect the cgrad scale. This however works in gr().
This used to work in v1.10.2. I noticed the bug once I upgraded to v1.16.4.
An example follows:
pyplot()
A = rand(0:10,10,10)
grad = cgrad(:inferno, scale =:log)
heatmap(A, c = grad)

grad2 = cgrad(:inferno, scale =:exp)
heatmap(A, c = grad2)

And now in gr()
gr()
heatmap(A, c = grad)

heatmap(A, c = grad2)

[fhagemann]

This bug was introduced from v1.10.5 to v1.10.6, so I guess that its source is either commit d59f6c9 or commit 100da5c.

I hope that this is useful for someone. I'm lacking the experience to judge which line might be the source of the problem...

[schustermartin]

I encountered this as well!

[isentropic]

I'm aware of the logscale problem, though I need help.
#3432

[isentropic]

let's focus on this in #3174