In order to force the axes to be in sync (so you can add data, change limits, etc.) you can use “parasite axes”. In principle one can create them with a transform, but it is easier to just tinker with tick locations and labels as follows:
using PyPlot
using PyCall
const axgr1 = PyNULL()
copy!(axgr1, pyimport("mpl_toolkits.axes_grid1"))
function host_subplot(args...; kwargs...)
pycall(axgr1["host_subplot"],PyAny, args... ; kwargs...)
end
# make some data
n=100
x=randn(n)
y=exp(x+0.25*rand(n))
# set up figure
f = figure()
ax1 = host_subplot(111)
ax2 = ax1[:twin]() # create parasite axes: defaults to identity transform
# show your data
semilogy(x,y,"o") # just to prove a point
ax1[:set_ylabel]("Custom scale")
ax2[:set_ylabel]("Normal scale")
ax2[:axis]["top"][:toggle](all=false)
# regular plot functions apply to ax1
ytvals = [0.25,0.5,1,2,4,8,16] # usually construct based on extrema(y)
# you can use an arbitrary transformation and format however you like
yticks(ytvals,([string(log2(i)) for i in ytvals]))
ylim(0.05,50)
# default selection isn't very nice
ax2[:set_yticks]([0.1,0.2,0.4,0.8,1,2,4,8,10,20,40])
