Saving figures via Plots.savefig and its format-specific siblings like Plots.pdf in Plots v1.20.0 is extremely slow.

[SeSodesa]

The culprit code in question goes as follows:

function draw_scatter_and_line(
    xdata,
    ydata,
    gradient,
    constant_term,
    scatter_title,
    line_title,
    xlabel,
    ylabel,
    filename
)
    img_path = joinpath(@__DIR__, filename)
    
    # Generate plot object
    img = Plots.plot(
        xlabel=xlabel,
        ylabel=ylabel,
        legend=:bottomright)

    # Measurements
    Plots.scatter!(
        img,
        xdata,
        ydata,
        label=scatter_title,
        color=:steelblue)

    # Generate fit
    fit_x = [xdata[begin] - abs(xdata[end] / 10), xdata[end] + xdata[end] / 10]
    fit_y = gradient .* fit_x .+ constant_term

    # Draw fitted line:
    Plots.plot!(
        img,
        fit_x,
        fit_y,
        linestyle=:dash,
        label=line_title,
        color=:steelblue)

    # The bottleneck    
    @time Plots.pdf(img, img_path)
end

When called with

xdata = [i for i in 1:10]
nA = maximum(xdata) / 5
noise = rand(-nA:nA/5:nA, length(xdata))
ydata = xdata + noise

the results are abysmal, especially regarding the number of allocations made:

8.972615 seconds (20.30 M allocations: 1.104 GiB, 4.26% gc time, 2.74% compilation time)

This is with just 10 data points! The issue occurs on Ubuntu 20.04 and Julia version 1.6.1. The used back-end is GR.

[BeastyBlacksmith]

This example isn't actually self contained, so I can't run, but is the timing the same if you run the function again?

[mdkeehan]

julia> draw_scatter_and_line( xdata, ydata, 1.0, 2.0, "scaddle", "lime", "xl", "yl", "bob.png")
  6.001535 seconds (16.79 M allocations: 941.686 MiB, 2.67% gc time, 45.99% compilation time)

julia> @time draw_scatter_and_line( xdata, ydata, 1.0, 2.0, "scaddle", "lime", "xl", "yl", "bob.png")
  0.014293 seconds (237.87 k allocations: 4.500 MiB)
  0.015777 seconds (243.70 k allocations: 4.830 MiB)

julia> @time draw_scatter_and_line( xdata, ydata, 1.0, 2.0, "scaddle", "lime", "xl", "yl", "bob.png")
  0.014303 seconds (237.87 k allocations: 4.500 MiB)
  0.015630 seconds (243.70 k allocations: 4.830 MiB)

julia> @time draw_scatter_and_line( xdata, ydata, 1.0, 2.0, "scattere", "line", "xl", "yl", "bob.png")
  0.015115 seconds (237.87 k allocations: 4.500 MiB)
  0.016502 seconds (243.70 k allocations: 4.831 MiB)

The first call is really slow and once it compiles it gets better...

[t-bltg]

This should be better now with SnoopPrecompile.
But overall, there might be a misunderstanding of the julia mechanism (jit compilation).