plot() and WebIO failure in Jupyter notebook

[jjstickel]

Based on documentation (README.md of this package and docs of WebIO), PlotlyJS uses a Jupyter extension (webio_jupyter_extension) to render plots in Jupyter. However, that extension (v0.1.0) is 4 years old (based on release date in pypi) and does not work with Jupyterlab v4 as reported in JuliaGizmos/WebIO.jl#514. Consequently, the figure object created with plot() will not display and instead generates the error WebIO not detected.

However, as I report in that same issue, it is possible to display the lower-level Plot() object. Specifically,

fig = plot()
display(fig.plot)

will show the plot without the webio-jupyter extension installed. So, I'm asking the developers, is that a reasonable workaround that would be expected to continue to work in the future? If so, I can spend a few minutes to create a PR to update the readme and docs to promote this usage in Jupyter notebooks.

[MHellmund]

I use Jupyter remotely, via Jupyterhub. For me, this trick only works after I explicitly load the plotlyjs javascript library via

display("text/html", "<script src='https://cdn.plot.ly/plotly-3.0.1.min.js' charset='utf-8'></script>")

Without this, my firefox debug console shows

Uncaught ReferenceError: Plotly is not defined

[jjstickel]

FWIW, display(fig) and display(fig.plot) are both working for me now. I'm not sure what changed on my system. Likely some package or subpackge was updated, or Julia itself.

The warning,
The WebIO Jupyter extension was not detected. See the WebIO Jupyter integration documentation for more information.
still appears after using PlotlyJS, even though it is not needed. And so I will keep this issue open for now.

[jd-foster]

Thanks for raising the issue.
I guess you're seeing the difference between the Plot function / types from PlotlyBase and the plot function using the SyncPlot type in PlotlyJS. They have different methods to show and display, with the SyncPlot version trying to wrap some extra JS and HTML code around it first.
It would be useful to have a PR if you're willing to highlight the difference in your experience.
We should also address the WebIO dependency in the medium term to prevent the warning occurring at all.