How to refresh Bokeh Document

Question

I would like to refresh a bokeh document so I can replace old plots with new ones. However, right now I just get the new plots appended to the document so the old ones don't go away.

#myfile.py
from bokeh.plotting import curdoc, figure
doc = curdoc()
p1 = figure(width=1500, height=230, active_scroll="wheel_zoom")
doc.add_root(p1)
doc.clear()
p2 = figure(width=1500, height=500, active_scroll="wheel_zoom")
doc.add_root(p2)

This results in the second plot being displayed after the first plot, though the expected behavior I am looking for is the second plot replacing the first plot. How can I resolve this? I am running this on a bokeh server via bokeh serve --show myfile.py

Answer 1

The best way to accomplish something like this is to have a top level layout of some kind (e.g. row or column) that has the content you want to replace inside it. Then when you want to replace things, keep the layout container, but change the value of its children property:

from bokeh.plotting import curdoc, figure
from bokeh.layouts import row

doc = curdoc()

p1 = figure(width=1500, height=230, active_scroll="wheel_zoom")

layout = row(p1)
doc.add_root(layout)

p2 = figure(width=1500, height=500, active_scroll="wheel_zoom")

layout.children[0] = p2

You can see a similar technique in the Crossfilter example.

Answer 2

Just in case anyone is struggling on how to set the children for layouts when there are multiple elements (say, widgets, more figures, rows etc), you can do so by wrapping the elements in a layout and assigning the children property directly:

p2 = figure(width=1500, height=500, active_scroll="wheel_zoom")
p3 = figure(width=1500, height=500, active_scroll="wheel_zoom")
new_layout = row(p2, p3)
layout.children = new_layout.children