Automatic Adaptive Plotting of Functions Needs Work

[oscarrutt]

The adaptive function plotter in Julia is not very robust and can't be trusted even with simple functions. Consider the following simple cases made with Julia 1.6.3 and the gr backend.

Consider the following function:

function p(x::Float64)
    if x < 0.0 || x > 1.0
        return 0.0;
    else
        return 1.0 + x;
    end
end;

Then the following bit works OK:

k = 2.0;
s = 2.0;
plot(x -> p(k * (x - s)))

with this result:

but this bit

k = 2.0;
s = 3.0;
plot(x -> p(k * (x - s)))

fails as such

For another failure on a simple plot, consider

plot(sinc, xlims = (0.0, 30.0))

which looks nice

but simply changing the plot limits

plot(sinc, xlims = (0.0, 40.0))

causes this failure:

[t-bltg]

Related: #3894.

[oscarrutt]

@t-bltg Thanks. Let's make sure we don't get hung up on sinc. It's not hard to find failing cases.

plot(x -> cos(20.0 * x) * exp(-x), xlims = (0.0, 10.0))

And it's not just oscillating functions; my first example didn't display a plot at all, and that's a super simple function.

I saw a thread from a few years ago where the devs were trying to mimic Mathematica's adaptve plotter. This isn't finished work.

[BeastyBlacksmith]

That function lives here and improvements are welcome!

[oscarrutt]

@BeastyBlacksmith How is this not a bug? Please reconsider.

[t-bltg]

@BeastyBlacksmith How is this not a bug? Please reconsider.

This is not a bug per se, but a limitation in the mentioned algorithm: we only need a finer / better distributed 1D grid for evaluating the functions.