Bokeh manual stack bar plot missing level in display - bokeh

Question

I try to manually create a stacked bar plot in bokeh as I can do in matplotlib. The code works but the output plot shows some issues I cannot explain or solve. To me, it seems like the error always appear at the second level, no matter if I have two levels (i.e. number of bars) or three levels stacked. In the three levels plot, the color is miss matched due to the missing second level.
from bokeh.models import ColumnDataSource
from bokeh.plotting import figure, show
from bokeh.io import output_notebook
import numpy as np
output_notebook()
test_data = {
'index' : np.array([1, 2, 3, 4, 5]),
'a' : np.array([10, 20, 30, 40, 50]),
'b' : np.array([10, 10, 10, 10, 10]),
'c' : np.array([10, 10, 10, 10, 10])
}
s = ColumnDataSource(test_data)
alpha_lst = [1, 0.5, 0.1]
keys = list(s.data.keys())[1:-1] # stack 2 levels/bars
# keys = list(s.data.keys())[1:] # stack 3 levels/bars
w_bar = 0.9
level = 0
p = figure(plot_width=600, plot_height=360)
for j in range(len(keys)):
p.vbar(x=s.data['index'], bottom=level,
top=level+s.data[keys[j]], width=w_bar, color='#43a2ca',
line_color='#3A5785', fill_alpha=alpha_lst[j], legend_label=keys[j])
level += s.data[keys[j]]
p.legend.location = 'top_left'
show(p)
Stacked Bar 2 Levels:
Stacked Bar 3 Levels:

Answer 1

Two reasons why level += s.data[keys[j]] is bad:
You change the variable type from int to numpy.array. I maaaaybe OK in this case, but generally it makes it harder to make sense of and debug later on
After it becomes an array (i.e. at the end of the very first loop iteration), the next execution of that line mutates the array in place. Meaning, the VBar that was created by the second call to p.vbar now has bottom with the same value as top.
It's up to you how (and whether) you fix #1. And to fix #2, which will solve your main problem, just replace level += s.data[keys[j]] with level = level + s.data[keys[j]].