Slow start about graphmakie.jl HOT 2 CLOSED

Well, you're hitting the infamous time-to-first-plot issue with Julia. Since Julia is a just-in-time compiled language, a lot of code needs to be compiled for the first run of a function. Subsequent runs will be much faster. In short, that's why most people prefer an interactive REPL base workflow (similar to a Jupyter-Notebook), where you keep your Julia session alive while changing code, updating scripts and recreate plots rather than executing full scripts with a new julia processes all the time. Here is a good general introduction to modern julia workflows.

Generally, Julia 1.9 improved that a lot with improved caching of precompiled code. So first, I would recommend using that (if you're not already doing so). With that, I get a TTFP of around 13 seconds, which is not great but not terrible...

julia> import Pkg; Pkg.activate(temp=true)
#  Activating new project at `/var/folders/cx/04cwlxl9287_ndx30k6q3n1h0000gn/T/jl_4Ofxj5`

julia> Pkg.add(["Graphs", "GLMakie", "GraphMakie"])
# this might take ages

julia> @time begin
           using Graphs, GLMakie, GraphMakie
           g = erdos_renyi(10, 0.1)
           graphplot(g)
       end
# 12.330614 seconds (36.68 M allocations: 2.129 GiB, 4.68% gc time, 49.49% compilation time: 8% of which was recompilation)

Subsequent runs of the same code will take less than 7 ms.

PS: The temporary environment is just to test it on a newly created env. In practice one should obviously reuse the environment for specific project. Then you'll only pay the precompilation costs when updating packages, so the perceived startup time will be around ~13 seconds on my machine.

from graphmakie.jl.

Ah, understood, it is also not an specific issue of GraphMakie.

Thanks a lot for the nice explanation.

from graphmakie.jl.