Gtk4Makie.jl

Interactive Makie plots in Gtk4 windows.

This package combines GTK's GtkGLArea and the GLMakie backend. Mouse and keyboard interactivity works just like in GLMakie's GLFW-based backend. There are two ways to draw GLMakie plots using Gtk4Makie:

1. As single plots in windows (GTKScreen).
2. In widgets (GtkMakieWidget), which can be placed at will inside other Gtk4 layout widgets.

For the window-based plots, Control-W (or Command-W on a Mac) closes the window and F11 (or Command-Shift-F on a Mac) fullscreens the window. Control-S (or Command-S on a Mac) opens a dialog for saving the figure to a PNG file.

The widget (#2 above) is unfortunately currently not stable. There is an issue with adding new plots to an existing widget (see #14), and not setting up the widget in the right way results in a crash. Users should use option #1 if at all possible.

To install in Julia's REPL, type ']' and then add Gtk4Makie. The following demonstrates how to produce a single GLMakie plot in a Gtk4 window:

using Gtk4Makie, GLMakie
screen = Gtk4Makie.GTKScreen(size=(800, 800))
display(screen, scatter(1:4))

Note that unlike previous versions, with version 0.2 Gtk4Makie can behave like a Makie backend. This is still experimental (and buggy) and is disabled by default. To enable it call Gtk4Makie.enable_backend(true).

Gtk4Makie has been successfully run on Windows, MacOS, and Linux. However, a problem has been reported by one Linux user on NVidia hardware (#7). On Wayland, getting GTK4's OpenGL backend to work may require a bit of configuration (see here). On some modern Linux distributions, GTK4's OpenGL backend (at least the version of the GTK4 library used by Gtk4.jl) does not work on Wayland (JuliaGtk/Gtk4.jl#72).

Users should be aware that this package unavoidably relies on Makie internals and is likely to break from time to time when upgrading Makie.

Finally, since it is based on Gtk4.jl, going beyond simple use of this package requires some knowledge of the GTK API. Those seeking a smoother experience should consider MousetrapMakie.jl, Mousetrap.jl's package for Makie integration.

This associates a Makie screen (which is basically a canvas) to a GtkWindow, much like the GLMakie backend draws its plots one at a time inside GLFW windows. In Gtk4Makie, the Makie plot is shown in a GtkGLArea that is placed inside a GtkGrid. To add other widgets around the Makie plot in the GTKScreen, you can get the GtkGrid using g = grid(screen). Widgets can then be added using, for example, g[1,2] = GtkButton("Do something") (adds a button below the plot) or insert!(g, glarea(screen), :top); g[1,1] = GtkButton("Do something else") (adds a button above the plot).

The constructor for GTKScreen accepts the following keyword arguments:

• size: sets the initial default width and height of the window in pixels
• title: a string to use as the window title
• fullscreen: if true, the window is set to fullscreen mode immediately

By default (except on Mac OS), Gtk4Makie screen windows include a header bar with a menu button. To omit the header bar, create a screen using, for example, GTKScreen(false; resolution=(800, 800)).

For a Gtk4Makie Screen, you can access the GtkGLArea where it draws Makie plots using glarea(screen) and the GTK window it's in using window(screen).

The GtkMakieWidget is a widget (based on GTK's GtkGLArea) that shows a Makie plot:

using Gtk4, Gtk4Makie
win = GtkWindow(;visible=false,title="2 Makie widgets in one window")
p=GtkPaned(:v;position=200)
p[1]=GtkMakieWidget()
p[2]=GtkMakieWidget()
win[]=p

push!(p[1],lines(rand(10)))
push!(p[2],scatter(rand(10)))
show(win)

The push! function adds a Makie Figure to the widget. This plot widget is what one would expect to work in a package like this, but unfortunately it currently has a lot of problems, as mentioned earlier, and it is not recommended at this time.

A window showing the axes and plots in a figure and their attributes can be opened using attributes_window(f=current_figure()). This can be used to experiment with various attributes, or add axis labels and titles before saving a plot. This functionality is experimental, buggy, and likely to grow and evolve over time.