Using iplot with ipython.display duplicates widgets in JupyterLab

[manycoding]

I created the same issue for ipywidgets jupyter-widgets/ipywidgets#2359 because I am not sure where it belongs.

TL;DR If I plot a graph with plotly iplot, any ipywidgets are duplicated.

The reason why I don't use display() and widgets for both is that widgets are not saved if I close a notebook, and I want the graph output to be available after I reopen a notebook.
So I thought I can plot graphs with plotly.offline.iplot() and widgets with display(), but the former prints any available widget again and again.

import ipywidgets
import plotly.graph_objs as go
from plotly.offline import iplot

data = [go.Bar(x=[i for i in range(5)], y=[i for i in range(5)], orientation="h")]
layout = go.Layout(
    bargap=0.1,
    xaxis=go.layout.XAxis(type="log", title="log"),
    template="ggplot2",
    hovermode="y",
    margin=dict(l=200, t=35),
)
iplot(go.Figure(data, layout))

c = ipywidgets.Checkbox(description="%", value=False)
display(c)

plotly 3.7.1
ipython 7.4.0

JupyterLab v0.35.4
Known labextensions:
app dir: /Users/valery/.local/share/virtualenvs/arche-lm24e12g/share/jupyter/lab
@jupyter-widgets/jupyterlab-manager v0.38.1 enabled OK
@jupyterlab/plotly-extension v0.18.2 enabled OK
plotlywidget v0.8.0 enabled OK

iplot_display_dupes.ipynb.zip

[manycoding]

My current workaround - display(HTML(plot(f, include_plotlyjs=False, output_type="div"))).
It keeps the notebook size small, keeps graphs and works fast.
However, the graphs printed in this way are static so it's impossible to update them after they printed.

[jonmmease]

Hi @manycoding,

The reason you're getting multiple displays in the notebook is because iplot and display both show the figure as a side-effect. So this is the expected behavior.

I haven't been following along closely, but it looks like the jupyter-widgets project may have just merged in support for saving/loading the widget state in the JupyterLab Notebook. See jupyter-widgets/ipywidgets#2265. It sounds like this might be the best solution for your usecase.

[manycoding]

@jonmmease So there is no other way to have both ipywidgets and graphs (printed with plotly) in the same cell without duplicating the first ones?
Except for cleaning the widgets.