With Makie it is easy to create animated plots. Animations work by making changes to data or plot attribute Observables and recording the changing figure frame by frame. You can find out more about the Observables workflow on the Observables & Interaction page.
To create an animation you need to use the
First you create a
Figure. Next, you pass a function that modifies this figure frame-by-frame to
record. Any changes you make to the figure or its plots will appear in the final animation. You also need to pass an iterable which has as many elements as you want frames in your animation. The function that you pass as the first argument is called with each element from this iterator over the course of the animation.
As a start, here is how you can change the color of a line plot:
using GLMakie using Makie.Colors fig, ax, lineplot = lines(0..10, sin; linewidth=10) # animation settings nframes = 30 framerate = 30 hue_iterator = range(0, 360, length=nframes) record(fig, "color_animation.mp4", hue_iterator; framerate = framerate) do hue lineplot.color = HSV(hue, 1, 0.75) end
Passing a function as the first argument is usually done with Julia's
do-notation, which you might not be familiar with. Instead of the above, we could also have written:
function change_function(hue) lineplot.color = HSV(hue, 1, 0.75) end record(change_function, fig, "color_animation.mp4", hue_iterator; framerate = framerate)
Video files are created with
FFMPEG.jl. You can choose from the following file formats:
.mkv(the default, doesn't need to convert)
.mp4(good for web, widely supported)
.webm(smallest file size)
.gif(lowest quality with largest file size)
Often, you want to animate a complex plot over time, and all the data that is displayed should be determined by the current time stamp. Such a dependency is really easy to express with
We can save a lot of work if we create our data depending on a single time
Observable, so we don't have to change every plot's data manually as the animation progresses.
Here is an example that plots two different functions. The y-values of each depend on time and therefore we only have to change the time for both plots to change. We use the convenient
@lift macro which denotes that the
lifted expression depends on each Observable marked with a
time = Observable(0.0) xs = range(0, 7, length=40) ys_1 = @lift(sin.(xs .- $time)) ys_2 = @lift(cos.(xs .- $time) .+ 3) fig = lines(xs, ys_1, color = :blue, linewidth = 4, axis = (title = @lift("t = $(round($time, digits = 1))"),)) scatter!(xs, ys_2, color = :red, markersize = 15) framerate = 30 timestamps = range(0, 2, step=1/framerate) record(fig, "time_animation.mp4", timestamps; framerate = framerate) do t time = t end
You can set most plot attributes equal to
Observables, so that you need only update a single variable (like time) during your animation loop.
For example, to make a line with color dependent on time, you could write:
time = Observable(0.0) color_observable = @lift(RGBf($time, 0, 0)) fig = lines(0..10, sin, color = color_observable) record(fig, "color_animation_2.mp4", timestamps; framerate = framerate) do t time = t end
You can also append data to a plot during an animation. Instead of passing
z) values separately, it is better to make a
Observable with a vector of
Points, so that the number of
y values can not go out of sync.
points = Observable(Point2f[(0, 0)]) fig, ax = scatter(points) limits!(ax, 0, 30, 0, 30) frames = 1:30 record(fig, "append_animation.mp4", frames; framerate = 30) do frame new_point = Point2f(frame, frame) points = push!(points, new_point) end
You can animate a live plot easily using a loop. Update all
Observables that you need and then add a short sleep interval so that the display can refresh:
points = Observable(Point2f[randn(2)]) fig, ax = scatter(points) limits!(ax, -4, 4, -4, 4) fps = 60 nframes = 120 for i = 1:nframes new_point = Point2f(randn(2)) points = push!(points, new_point) sleep(1/fps) # refreshes the display! end
Another example that updates the contents of a heatmap:
using GLMakie function mandelbrot(x, y) z = c = x + y*im for i in 1:30.0; abs(z) > 2 && return i; z = z^2 + c; end; 0 end x = LinRange(-2, 1, 200) y = LinRange(-1.1, 1.1, 200) matrix = mandelbrot.(x, y') fig, ax, hm = heatmap(x, y, matrix) N = 50 xmin = LinRange(-2.0, -0.72, N) xmax = LinRange(1, -0.6, N) ymin = LinRange(-1.1, -0.51, N) ymax = LinRange(1, -0.42, N) # we use `record` to show the resulting video in the docs. # If one doesn't need to record a video, a normal loop works as well. # Just don't forget to call `display(fig)` before the loop # and without record, one needs to insert a yield to yield to the render task record(fig, "heatmap_mandelbrot.mp4", 1:7:N) do i _x = LinRange(xmin[i], xmax[i], 200) _y = LinRange(ymin[i], ymax[i], 200) hm = _x # update x coordinates hm = _y # update y coordinates hm = mandelbrot.(_x, _y') # update data autolimits!(ax) # update limits # yield() -> not required with record end