I'm trying to achieve the following o/p from bokeh which is deprecated now. In the new structure, I need to layer everything like working in HTML. Is there any simplified example like this bokeh pie chart(deprecated) ?
1 Answer
You can combine wegde and annular_wedge
The example code below creates this figure:
Example code
Pandas
Lets say we have this data table for the medals of a sport event.
import numpy as np import pandas as pd df = pd.DataFrame({ 'Country':['Fra', 'Deu', 'Can', 'USA'], 'Gold': [0,1,2,3], 'Silver': [1,2,2,2], 'Bronze':[3,2,1,3] }) >>> df Country Gold Silver Bronze 0 FRA 0 1 3 1 GER 1 2 2 2 CAN 2 2 1 3 USA 3 2 3 Now we have to calculate the angles to use the bokeh wedges. We use a copy to not overwrite the original data.
# get a copy and calculate the angles _df = df.copy() _df['Country_total'] = _df[['Gold', 'Silver', 'Bronze']].sum(axis=1) total_sum = _df['Country_total'].sum() _df["End"] = _df['Country_total'].div(total_sum).mul(2 * np.pi).cumsum().round(6) _df["Start"] = _df["End"].shift(1).fillna(0) _df['Color'] = ['blue', 'red', 'green', 'magenta'] _df Bokeh
Here we loop over all rows of the DataFrame and draw the annular wedge first and then the inner wedge to get the correct color in the legend.
from bokeh.models import Legend from bokeh.plotting import show, figure, output_notebook output_notebook() p = figure(width=400, height=300, match_aspect=True, x_range=(-2.2,2.2), y_range=(-2.2,2.2)) p.add_layout( Legend( click_policy="hide", margin=1, ), "right", ) p.rect(x=0, y=0, width=2, height=2, alpha=0) wedges = [] for i, item in _df.iterrows(): start = item["Start"] end = item["End"] start_angle = start end_angle = start for value, color in zip(['Gold', 'Silver', 'Bronze'], ['gold', 'silver', '#bf8970']): angle = np.round(item[value] / total_sum * 2 * np.pi, 6) end_angle += angle p.annular_wedge( x=0, y=0, inner_radius=1, outer_radius=2, start_angle=start_angle, end_angle=end_angle, color=color, line_color='white', legend_label=item['Country'], ) start_angle += angle p.wedge( x=0, y=0, radius=1, start_angle=start, end_angle=end, color=item["Color"], line_color='white', legend_label=item['Country'], ) p.xgrid.visible = False p.ygrid.visible = False show(p) Comment
I hope this is close tou your wanted solution.
1 Comment
ArulRulzz
thanks for your reply. I'm expecting something similar to the previous one. As I said earlier, Here we're doing everything like a designer. Instead, if we have a codebase like the previous one, it'll be easy to use.
bokeh.chartswhich was barely functional and did not have a maintainer, so it had to be jettisoned from the main library. If you want to simplify things, it's probably possible to useannular_wedgewithinner_radius=0for the inner wedges, so that all levels can be created consistently. There might be higher level tools (e.g. holoviews or pandas-bokeh) that offer a multi-level pie chart, but I'm not sure.