Can't render bokeh figure (jupyter_bokeh.js import fails)

[yreynaud]

Hello,

I met an issue trying to plot a bokeh figure in a shiny app. Indeed, the jupyter_bokeh.js GET request failed (404 error).
I already downloaded shinywidgets, bokeh, jupyter_bokeh.

I tried the the ipywidget example (https://github.com/rstudio/py-shinywidgets/blob/main/examples/ipywidgets/app.py) and it worked; so maybe the problem is only for bokeh (I also arrived to plot plotly figures).

The problem doesn't come from the widget itself, because the plot works in a jupyter notebook.

I don't know if it is an environment issue (like explained here : https://shiny.posit.co/py/docs/jupyter-widgets.html#troubleshooting )
Here are the concerned libraries versions :

• shinywidgets : 0.3.3
• shiny : 1.0.0
• bokeh : 3.4.3
• jupyter_bokeh : 4.0.5

Here is a Minimal Reproducible Example (MRE):

from shiny import *
from shiny import (
    ui,
    Inputs,
    Outputs,
    Session,
    App,
)
from shinywidgets import output_widget, render_bokeh
from starlette.applications import Starlette
from starlette.routing import Mount


panel_sidebar = []
panel_main = [
    output_widget("figure_bokeh", fill=True, fillable=True)
]

app_ui = ui.page_fluid(
    ui.layout_sidebar(
                ui.sidebar(
                    *panel_sidebar,
                    open='open',
                    position='left'
                ),
                *panel_main
            ),
)

def server(input: Inputs, output: Outputs, session: Session):
    @output
    @render_bokeh
    def figure_bokeh():
        from bokeh.plotting import figure

        x = [1, 2, 3, 4, 5]
        y = [6, 7, 2, 4, 5]
        p = figure(title="Simple line example", x_axis_label="x", y_axis_label="y")
        p.line(x, y, legend_label="Temp.", line_width=2)
        return p


app = App(app_ui, server, debug=True)

routes = [
    Mount('/anonymous', app=app),
]

app = Starlette(routes=routes)

Here is the error notification that appears on the page :

In the logs, A message says that the .js file was not found :
INFO: 172.30.0.5:43362 - "GET /anonymous/%40bokeh/jupyter_bokeh.js HTTP/1.1" 404 Not Found

[yreynaud]

It finally worked using render_widget instead of render_bokeh, and adding bokeh_dependency() call

[kampai-shp]

It finally worked using render_widget instead of render_bokeh, and adding bokeh_dependency() call

Thanks for this! I have been trying to get a bokeh figure to embed in my shiny python app for a couple of days. Would be great if py-shiny would update the docs to reflect.

[gadenbuie]

I'm re-opening this issue. I would expect @render_bokeh to work without work-arounds.

Here's a smaller reproducible example where @render_bokeh works once bokeh_dependency() is included in the UI, which should not be necessary. (shinylive link)

from shinywidgets import bokeh_dependency, output_widget, render_bokeh

from shiny import App, ui

app_ui = ui.page_fluid(
    output_widget("figure_bokeh", fill=True, fillable=True),
    bokeh_dependency()  # << should not be required
)


def server(inputs):
    @render_bokeh
    def figure_bokeh():
        from bokeh.plotting import figure

        x = [1, 2, 3, 4, 5]
        y = [6, 7, 2, 4, 5]
        p = figure(title="Simple line example", x_axis_label="x", y_axis_label="y")
        p.line(x, y, legend_label="Temp.", line_width=2)
        return p


app = App(app_ui, server, debug=True)