Dfyjkt

I have a vbar_stack chart that I am updating with CheckboxGroup selections. When I change the accounts being selected, the data does change, but the chart doesn't re-sort. How can I make the accounts re-sort in descending order?

Right now, I am using this code for the CheckboxGroup onchange command. The custom_make_dataset function does re-sort the filtered dataset.

def custom_make_dataset(df, accounts):

 remove_accounts = [i for i in df["acc"].unique() if i not in accounts]

 new_df = df 
 .groupby(["acc","product"])["cost"] 
 .sum() 
 .reset_index() 
 .pivot(columns="acc", index="product", values="cost") 

 for account in remove_accounts:
 new_df[account] = 0

 new_df["Total"] = new_df.sum(axis=1)
 new_df.sort_values(by="Total", ascending=False, inplace=True)
 del new_df["Total"]

 return ColumnDataSource(new_df)

def update_checkbox(attr, old, new):
 accounts_to_plot = [account_selection.labels[i] for i in account_selection.active]
 new_src = custom_make_dataset(df,accounts_to_plot)
 source.data.update(new_src.data)

account_selection = CheckboxGroup(name="Customer Accounts", labels=initial_accounts, active = [i for i,j in enumerate(initial_accounts)])
account_selection.on_change('active', update_checkbox)

Full Selection
After 1 Account Selection

I figured out the problem. While I was updating the data, I wasn't updating the x-axis itself. You can put something like this in the onchange command to update the plot x-axis.

p.x_range.factors = (new_src.data['product'])

Maybe bokeh is using the dataframe index to determine on which x position the values are placed.

Try resetting your new_df index before returning the new ColumnDataSource.
new_df.reset_index(drop=True, inplace=True)