CODEWITHSUNDEEP
pythonwidgetbokeh
Andrea Ciufo
Andrea Ciufo

Reputation: 369

Interactive Selection on Jupyter Notebook through Select Widget raise "Models must be owned by only a single document" error

How can I create an interactive bar plot with bokeh through a selection widget on Jupyter Notebook?

Why the Notebook raise the following error:

"RuntimeError: Models must be owned by only a single document, Title(id='1044', ...) is already in a doc ERROR:tornado.access:500 GET /autoload.js?bokeh-autoload-element=1002&bokeh-absolute-url=http://localhost:54277&resources=none (::1) 117.01ms"

I read carefully this example from github and this similar situation from a Google Bokeh Group, in the latter they run a bokeh server not a the jupyter kernel

output_notebook()

dct={'Date' : ["2018-01-07", "2018-01-12", "2018-01-13", "2018-01-14", "2018-01-20", "2018-01-24"],'Activity' : ['A','B','A','B','A','B'],'Count' : [1, 2, 5, 3, 7, 1]}
df=pd.DataFrame(dct)

activity_list=df['Activity'].unique().tolist().copy()
activity_selected='A'
def modify_doc(doc):
     def make_plot(cdf):
         plot = figure()
         plot.vbar(x=cdf.Date, top=cdf.Count, width=0.9)
         push_notebook()
         show(plot, notebook_handle = True)
         return plot
     def update_plot(attr, old, new):
         activity =  select.value
         sdf = df.copy()
         sdf = sdf[sdf['Activity'] == activity]
         layout.children[0] = make_plot(sdf)


     select = Select(title='Select Activity', value=activity_selected, options=activity_list)
     select.on_change('value', update_plot)
     p=make_plot(df)
     layout=column(select, p)
     doc.add_root(layout)
show(modify_doc)

What I expect is something like this in the snapshoot:

enter image description here

I am using Bokeh 1.0.4

Upvotes: 0

Views: 1284

Answers (2)

Andrea Ciufo
Andrea Ciufo

Reputation: 369

The correct code for selecting the plot through the Select widget is:

activity_list=df['Activity'].unique().tolist().copy()
df['Date']=pd.to_datetime(df['Date'])

activity_selected='A'

def modify_doc(doc):

    df_r=df.copy()
    source = ColumnDataSource(data=df_r)
    plot=figure(title='Daily Hours',x_axis_type="datetime")
    plot.vbar(x="Date", top="Count",source=source, width=4)
    def update_plot(attr, old, new):
        activity =  select.value
        data = df_r[df_r['Activity'] == activity]
        source.data = ColumnDataSource(data=data).data

    select = Select(title='Select Activity', value=activity_selected, options=activity_list)
    select.on_change('value', update_plot)

    layout=column(select, plot)
    doc.add_root(layout)
show(modify_doc)

And this will be the output that you see:

enter image description here enter image description here

Upvotes: 0

bigreddot
bigreddot

Reputation: 34568

It's possible we need to make some documentation improvements, because there are several parts of your code that do not make sense.

  • push_notebook is complementary to showing a real Bokeh server app (i.e. passing modify_doc to show). I can't think of any situation where it would make sense to use them together. The Bokeh server app capability is a strict superset of push_notebook, so since you are already making a Bokeh server app, you should just update everything in the standard way.

  • show should never be called inside the Bokeh server app code (in this case, inside modify_doc) This is in fact the proximate cause of the exception you are getting. You should

  • You should not make a new plot every time! The entire purpose of both push_notebook and the Bokeh server is that updates can happen efficiently without making a whole new plot every time. You should be updating the data and attributes of the existing plot.

There is a complete wokring example of a Bokeh server app in a notebook here:

https://github.com/bokeh/bokeh/blob/master/examples/howto/server_embed/notebook_embed.ipynb

You should study and emulate that as it represents best practice.

Upvotes: 1

Related Questions