Plots.jl: [BUG] GKS: Rectangle definition is invalid in routine SET_VIEWPORT

[darsnack]

Other issues have a similar error messages, but none appear related to the context leading to my error. As far as I can tell from previous issues, this error message is fairly generic and not representative of the core problem.

Details

The following code (previously working for many versions of Plots.jl over the past year or so) produces an error.

# plot raster plot
inspikes = round.(∂t .* spikes; digits = 6)
outspikes = round.(∂t .* output; digits = 6)
scatter(inspikes, fill(1, length(spikes)), label = "Input",
        xlabel = "Time (sec)", ylabel = "", yticks = :none, title = "Raster Plot (\\alpha response)")
raster_plot = scatter!(outspikes, fill(2, length(output)), label = "Output")

The error:

GKS: Rectangle definition is invalid in routine SET_VIEWPORT

The value of the data just to show I am not plotting stuff with crazy precisions.

julia> inspikes
18-element Array{Float64,1}:
  0.7
  1.1
  1.19
  2.34
  4.17
  5.21
  6.3
  7.65
  8.2
  8.63
  9.15
  9.31
 10.36
 10.45
 11.25
 11.77
 11.78
 14.97

julia> outspikes
6-element Array{Float64,1}:
  1.77
  3.66
  6.14
  9.97
 12.64
 14.41

Backends

This bug occurs on ( insert x below )

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

Versions

Plots.jl version: [91a5bcdd] Plots v1.4.3
Backend version (]st -m): [28b8d3ca] GR v0.50.1
Output of versioninfo():

julia> versioninfo()
Julia Version 1.4.0
Commit b8e9a9ecc6 (2020-03-21 16:36 UTC)
Platform Info:
  OS: macOS (x86_64-apple-darwin18.6.0)
  CPU: Intel(R) Core(TM) i7-6700HQ CPU @ 2.60GHz
  WORD_SIZE: 64
  LIBM: libopenlibm
  LLVM: libLLVM-8.0.1 (ORCJIT, skylake)
Environment:
  JULIA_NUM_THREADS = 4

[nfernandez-arias]

Hello,

I am having the same error. I make a bunch of plots and then arrange them into a subplot. This has worked with no issues until today when I tried it (using exactly the same data in the plots). The data does not require a high degree of precision and I have seen them before in exactly the form I want them in previous version of Plots.jl.

Now it gives me a GKS: Rectangle definition is invalid in routine SET_WINDOW error, and then at the end when I call

plot(p1,p2,p3,p4)

To arrange them into a subplot, it plots them all on top of each other.

[daschw]

The issue in the original code is title = "Raster Plot (\\alpha response)" replacing this with e.g. title = "Raster Plot (α response)" works. This seems somehow related to https://github.com/JuliaPlots/Plots.jl/issues/1842 and https://github.com/JuliaPlots/Plots.jl/issues/2715.

[nfernandez-arias]

This seems promising - my plots also involve \\ notation - thanks for the quick update, will lyk

[daschw]

@jheinen the root of this issue seems to be:

julia> GR.jlgr.inqtext(0, 0, "Raster Plot (\\alpha response)")
([NaN, NaN, NaN, NaN], [NaN, NaN, NaN, NaN])

(Plots has a similar logic like GR.jlgr to decide between GR.inqmathtex, GR.inqtextext and GR.inqtext)

[darsnack]

Makes sense. The working “X” for pyplot also switched the \\ notation. I’ll try without that for GR and report back.

[nfernandez-arias]

Removing the \\ notation fixes the problem in my case too

I was using it for purposes like

\\alpha_1

which renders as an alpha with a beta subscript. No longer works with

α_1

which now doesn't render it as a subscript, and

\$ α_1 \$

gives me errors, and doesn't render anything.

Given this bug, is LaTeXStrings.jl the only way to get subscripts / supserscripts?

[jheinen]

I will check this and try to allow both notations ...
BTW: "\$\\alpha_1\$" works, too.

[nfernandez-arias]

There are still some issues that I am having trouble fixing. Previously I had a some code

xAxisLabels = ["r"; "g"; "OI"; "E"; "S"; "RD"; "\$ \\Large \\theta \$"; "\$ \\Large  \\beta \$"; "\$ \\Large  \\psi \$"]
p = bar(xAxisLabels, transpose(invjac) * grad, title = "\$ \\nabla_m \\log (\\textrm{CEV Welfare chg. \\%}) \$")

which seemed to work. Now, it does not recognize the \Large. And more importantly, it does not recognize the \textrm{} anymore.

But the most confusing part is that it gives me no errors when I run it line by line. But when I run a script containing these lines, I get the same GKS: Rectangle definition is invalid in routine SET_VIEWPORT error, and it jumbles up the whole plot.

Apologies if this description of the pathology is not 100% clear. It is difficult to succinctly describe how this is failing. I have spent the last hour trying to figure out what the pattern is the best I can.

[jheinen]

The "internal" mathtex renderer only supports a LaTeX subset. I could add a flag to force the use of the old one (based on "external" dvipng command as part of a LaTeX installation) or (you could) use workarounds like L"\nabla_p \log\ (CEV\ Welfare\ chg. \%)".

[nfernandez-arias]

Just to make sure I understand your suggestion, it only adds spaces, but does not render the text in normal text mode, correct? This is probably fine for now and I really appreciate your help. Will this feature be added at some point? Often it is useful to be able to combine math and non-math expressions in the title or labels of a plot.

[darsnack]

I can confirm that the same type of change \\alpha -> ⍺ works for me.

[jheinen]

The textrm problem is fixed in gr#develop. plot(randn(10), title = "\$ \\nabla_p \\log (\\textrm{CEV Welfare chg. %}) \$") :