curdoc().theme = Theme(json=yaml.load()) is not applied to charts when used in Jupyter notebook

[user32000]

Running the color slider example in a Jupyter notebook it seems that the defaults defined in the theme are not applied

[cedricporter]

same problem

[rubik]

Same. I've tried everything, from curdoc().theme = Theme(...) to bokeh.themes.default = Theme(...), but nothing seems to be working.

Any ideas?

[djpugh]

As with #4952 - It looks like it may be the use of with _ModelInEmptyDocument(model): in notebook_div - I assume this is where the embedding occurs

[ghost]

First of all, awesome work. Bokeh sets the current standard thanks to all your efforts.

Second, I am still running into the issue of not having a theme affect the style of the notebook display with version 0.12.9. I would expect the following code to produce a plot with a pink background:

from bokeh.plotting import figure 
from bokeh.io import output_notebook, show, curdoc
from bokeh.themes import Theme, default
import yaml

output_notebook()
theme = Theme(json={'attrs': {'Plot': {'background_fill_color': 'pink'}}})
curdoc().theme = theme

x1 = [0, 1, 2, 3, 4, 5, 6, 7, 8, 9, 10]
y1 = [0, 8, 2, 4, 6, 9, 5, 6, 25, 28, 4]

p = figure(title='PINK PLOT')
p.scatter(x1, y1)
show(p)

Looking at the _ModelInEmptyDocument context manager in embed.py, it appears that the new empty_doc forgets about the applied theme. Changing

    for ref in model.references():
        ref._document = None
    empty_doc = Document()
    empty_doc.add_root(model)

to

    for ref in model.references():
        ref._document = None
    empty_doc = Document(theme=doc.theme)
    empty_doc.add_root(model)

produces the desired results.

First time posting here. I apologize if I am breaking any protocols. Thank you.