One chart with two different y axis ranges in Bokeh?

Question

I would like a Bar chart with Quantity information on the left y-axis, and then overlay a Scatter/Line plot with Yield % on the right. I can create each of these charts separately, but do not know how to combine them into a single plot.

In matplotlib, we would create a second figure using twinx(), and then use yaxis.tick_left() and yaxis.tick_right() on the respective figures.

Is there a method for doing something similar with Bokeh?

Answer 1

# PEA Y PNEA ARE THE EACH OF THE LINES THAT I WANT

p = bpl.figure(plot_width=1600, plot_height=400, 
x_range=peas['periodos'].astype(str),toolbar_location=None)
p.xgrid.grid_line_color = None
p.ygrid.grid_line_color = None
p.axis.major_label_orientation = 3.14/2
p.yaxis.major_label_orientation = 'horizontal'
p.outline_line_color = "White"

# Eje 1
p.line(x=peas['periodos'].astype(str),y=peas['pea'], line_dash='dashed', 
line_width=2, legend_label='PEA', color='red')
p.y_range = Range1d(peas['pea'].min()-1, peas['pea'].max()+1)

# Eje 2
column2_range = 'segundo y' + "_range"
p.extra_y_ranges = {column2_range: Range1d(peas['pnea'].min()-1, 
peas['pnea'].max()+1)


p.line(x=peas['periodos'].astype(str), y=peas['pnea'], legend_label='PNEA', 
y_range_name=column2_range, color="green")

p.add_layout(LinearAxis(y_range_name=column2_range), "right")
p.legend.location = "top_left"
bpl.show(p)

Answer 2

If you have a pandas Dataframe you can use this template to plot two lines with different axis:

from bokeh.plotting import figure, output_file, show
from bokeh.models import LinearAxis, Range1d
import pandas as pd

# pandas dataframe
x_column = "x"
y_column1 = "y1"
y_column2 = "y2"

df = pd.DataFrame()
df[x_column] = range(0, 100)
df[y_column1] = pd.np.linspace(100, 1000, 100)
df[y_column2] = pd.np.linspace(1, 2, 100)

# Bokeh plot
output_file("twin_axis.html")

y_overlimit = 0.05 # show y axis below and above y min and max value
p = figure()

# FIRST AXIS
p.line(df[x_column], df[y_column1], legend=y_column1, line_width=1, color="blue")
p.y_range = Range1d(
    df[y_column1].min() * (1 - y_overlimit), df[y_column1].max() * (1 + y_overlimit)
)

# SECOND AXIS
y_column2_range = y_column2 + "_range"
p.extra_y_ranges = {
    y_column2_range: Range1d(
        start=df[y_column2].min() * (1 - y_overlimit),
        end=df[y_column2].max() * (1 + y_overlimit),
    )
}
p.add_layout(LinearAxis(y_range_name=y_column2_range), "right")

p.line(
    df[x_column],
    df[y_column2],
    legend=y_column2,
    line_width=1,
    y_range_name=y_column2_range,
    color="green",
)

show(p)

Answer 3

Yes, now it is possible to have two y axes in Bokeh plots. The code below shows script parts significant in setting up the second y axis to the usual figure plotting script.

# Modules needed from Bokeh.
from bokeh.io import output_file, show
from bokeh.plotting import figure
from bokeh.models import LinearAxis, Range1d

# Seting the params for the first figure.
s1 = figure(x_axis_type="datetime", tools=TOOLS, plot_width=1000,
           plot_height=600)

# Setting the second y axis range name and range
s1.extra_y_ranges = {"foo": Range1d(start=-100, end=200)}

# Adding the second axis to the plot.  
s1.add_layout(LinearAxis(y_range_name="foo"), 'right')

# Setting the rect glyph params for the first graph. 
# Using the default y range and y axis here.           
s1.rect(df_j.timestamp, mids, w, spans, fill_color="#D5E1DD", line_color="black")

# Setting the rect glyph params for the second graph. 
# Using the aditional y range named "foo" and "right" y axis here. 
s1.rect(df_j.timestamp, ad_bar_coord, w, bar_span,
         fill_color="#D5E1DD", color="green", y_range_name="foo")

# Show the combined graphs with twin y axes.
show(s1)

And the plot we get looks like this:

If you want to add a label to the second axis, this can be accomplished by editing the call to LinearAxis as follows:

s1.add_layout(LinearAxis(y_range_name="foo", axis_label='foo label'), 'right')

Answer 4

This post helped me to accomplish the effect you are looking for.

Here is the content of that post:

from bokeh.plotting import figure, output_file, show
from bokeh.models.ranges import Range1d
import numpy


output_file("line_bar.html")

p = figure(plot_width=400, plot_height=400)

# add a line renderer
p.line([1, 2, 3, 4, 5], [6, 7, 6, 4, 5], line_width=2)

# setting bar values
h = numpy.array([2, 8, 5, 10, 7])

# Correcting the bottom position of the bars to be on the 0 line.
adj_h = h/2

# add bar renderer
p.rect(x=[1, 2, 3, 4, 5], y=adj_h, width=0.4, height=h, color="#CAB2D6")

# Setting the y  axis range   
p.y_range = Range1d(0, 12)

p.title = "Line and Bar"

show(p)

If you want to add the second axis to the plot do so with p.extra_y_ranges as described in the post above. Anything else, you should be able to figure out.

For example, in my project I have code like this:

s1 = figure(plot_width=800, plot_height=400, tools=[TOOLS, HoverTool(tooltips=[('Zip', "@zip"),('((Rides/day)/station)/capita', "@height")])],
           title="((Rides/day)/station)/capita in a Zipcode (Apr 2015-Mar 2016)")

y = new_df['rides_per_day_per_station_per_capita']
adjy = new_df['rides_per_day_per_station_per_capita']/2

s1.rect(list(range(len(new_df['zip']))), adjy, width=.9, height=y, color='#f45666')
s1.y_range = Range1d(0, .05)
s1.extra_y_ranges = {"NumStations": Range1d(start=0, end=35)}
s1.add_layout(LinearAxis(y_range_name="NumStations"), 'right')
s1.circle(list(range(len(new_df['zip']))),new_df['station count'], y_range_name='NumStations', color='blue')
show(s1)

And the result is: