Single Session Different Clients

Panel
1

I am trying to deploy Plot with Columns on this URL: Plot With Columns — Panel v0.13.0

When I run the application o a server and visit the address on my local machine I can use the app. The app is also reachable by other machines with an unexpected feature. Whenever I make a change on the sidebar it is reflected to the other user immediately by the WebSocket. But I need each client to have their own session. Also, the graph in the center does not have this property. I can zoom or move the visual, and it does not reflect other users.

My code is here:

import panel as pn
import hvplot.pandas

from bokeh.sampledata.autompg import autompg_clean

pn.extension(sizing_mode="stretch_width")

quant = [None, 'mpg', 'cyl', 'displ', 'hp', 'weight', 'yr']
cat = [None, 'origin', 'mfr', 'yr']
combined = quant+cat[1:]

x = pn.widgets.Select(name='x', value='mpg', options=combined)
y = pn.widgets.Select(name='y', value='cyl', options=combined)
color = pn.widgets.Select(name='color', options=combined)
facet = pn.widgets.Select(name='facet', options=cat)

@pn.depends(x, y, color, facet)
def plot(x, y, color, facet):
    cmap = 'Category10' if color in cat else 'viridis'
    return autompg_clean.hvplot.scatter(
        x, y, color=color or 'green', by=facet, subplots=True, padding=0.1,
        cmap=cmap, responsive=True, min_height=500, size=100
    )

settings = pn.Row(pn.WidgetBox(x, y, color, facet))
pn.Column(
    '### Auto MPG Explorer', 
    settings,
    plot,
    width_policy='max'
)


plot = pn.template.FastListTemplate(
    site="Panel", title="Auto MPG Explorer", sidebar=[settings], 
    main=["This example demonstrates **how to combine multiple columns of components into an overall layout**.", plot]
)

pn.serve(plot, port=3000,  show=True)
1 Like
2

The issue is that you instantiate one plot to serve/ share between all users. It can in some situations be handy (“live-share”). But not here.

Instead you need to provide a function get_plot to pn.serve that returns a new plot for each session.

I would actually advice on 1) using panel serve name_of_script.py 2) remove pn.serve... and 3) add plot.servable(). That should also solve your problem

panel serve is very widely used and in my experience works the best.