Add informative error message when FigureWidget not available due to missing dependencies

[vlizanae]

I'm running an alpine python3 container and for some reason FigureWidget is not available. Jupyter is not installed because it's meant to be a Dash app, are there any dependencies I might have overlooked?

[jonmmease]

Hi @vlizanae ,

The FigureWidget class won't be available unless ipywidgets is installed. If you're making a Dash app, then you'll want to use the graph_objs.Figure class instead of FigureWidget. FigureWidget is essentially a subclass of Figure with extra logic and dependencies for integrating with the Jupyter Notebook.

Thanks for reporting this because it reminds me that it would be nice to do something more user friendly when someone tries to access a FigureWidget when the dependencies aren't available.

Alright if I change the title of this issue to "Raise informative error if FigureWidget dependencies are not installed"?

[vlizanae]

sure, thank you!

[astrojuanlu]

I was thinking about this a bit and I don't see a quick & straightforward solution. This is the way FigureWidget is being imported in the graph_objs namespace:

try:
    import ipywidgets
    from distutils.version import LooseVersion
    if LooseVersion(ipywidgets.__version__) >= LooseVersion('7.0.0'):
        from ._figurewidget import FigureWidget
    del LooseVersion
    del ipywidgets
except ImportError:
    pass

The only options I see:

• Keep this code, but instead of silencing the ImportError, display a warning. The problem is that lots of users that are not interested in FigureWidgets will see it.
• Do not import it in graph_objs, and instead make the user do from plotly.graph_objs.widgets import FigureWidget. This is an API breaking change, and also annoying to have this object in a different module.
• Add ipywidgets>7 to the plotly.py requirements and get rid of the conditional import altogether.

Other ideas?

[nicolaskruchten]

Closed by #2445