Plots.jl: using Plots unconditionally inserts 16MB of javascript into notebooks

Created on 5 Sep 2018 · 10Comments · Source: JuliaPlots/Plots.jl

When Plots is loaded for the first time __init__() is run which unconditionally includes the backends/plotly.jl file. If we are in a notebook, this injects 16MB of Javascript from the following lines:

https://github.com/JuliaPlots/Plots.jl/blob/80151ac1a30c498592ac94ee03d5a7af6d2df8de/src/backends/plotly.jl#L82-L96

Seems unnecessary, especially if one does not use Plotly at all.

Source

KristofferC

👍9

Most helpful comment

It is likely still true because those include don't fire on the second using either (because the package is already loaded).

KristofferC on 30 Sep 2018

👍2

All 10 comments

Given that plotly isn't the default, could we just display this HTML when calling plotly()?

mbauman on 25 Sep 2018

Yes that seems like the best solution.

mkborregaard on 26 Sep 2018

For those that hit this, a workaround is to evaluate the using Plots cell twice. The second time, the 16MB javascript will be overwritten since this injection only happen on __init__ time (which doesnt run on the second evaluation of the cell).

KristofferC on 30 Sep 2018

❤1

this injection only happen on __init__ time

That's no longer true with https://github.com/JuliaPlots/Plots.jl/pull/1772, I guess. But this issue will be my next project, so I hope a fix will be available soon.

daschw on 30 Sep 2018

It is likely still true because those include don't fire on the second using either (because the package is already loaded).

KristofferC on 30 Sep 2018

👍2

See also this issue https://github.com/JuliaPlots/Plots.jl/issues/1780 where @jlperla has some relevant considerations.

mkborregaard on 5 Oct 2018

Yes, sorry I missed this when posting. Also, @sglyon has some trick with PlotlyJS that uses WebIO that he said could work.

That said, I am using GR, so don't really care

Does the plotly backend even work with IJulia right now? It never has for me. If not, can this whole thing be removed temporarily so it doesn't break working backends, and a quick tag put in for everyone teaching with jupyter this term?

jlperla on 5 Oct 2018

I think master does not have this problem.

KristofferC on 5 Oct 2018

master seems to have that issue, but it is platform dependent or something - I at least don't get plotly to work for my linux + windows PC on Plots.jl master - but it works for e.g. @daschw.
The good news is, that I fixed it in https://github.com/JuliaPlots/Plots.jl/pull/1777.
If anyone can try & confirm that this branch fixes all issues, that'd be great :)