Rotating a figure

What are you trying to do?

I am trying to rotate an entire bokeh figure including the axis, labels. My output currently looks like this:

I need it to rotate the entire image anti clockwise by 90 degrees. I am trying to figure if there is a way to save this image as HTML image source and then rotate it by modifying html code. Are there other ways to do this in bokeh itself?

am trying to figure if there is a way to save this image as HTML image source and then rotate it by modifying html code.

Bokeh can output PNGs from Python code, and there are probably CSS transformation that can rotate those images, if you include them in some HTML yourself. But “live” Bokeh plots embedded in HTML are not images, they data plus actual JavaScript code that gets executed to render the data on an HTML canvas. [1]

Are there other ways to do this in bokeh itself?

Yes, you can swap what you plot for “x” vs “y”.

  1. Maybe there are CSS mechanisms for rotating HTML canvas elements, I don’t actually know. But you’d have to muck around with what are normally internal parts of Bokeh plots to try to do that, so YMMV and just being clear that would be totally unsupported usage, even if it works in some fashion. ↩︎

I can swap the x and y for the line plot, no problem. It is the band that is a problem though. In this dataset, if I swap the ‘x’ and ‘y’, the ‘y’ is no more a function of ‘x’, it’s just a relation. That seems to preclude making a band.

There’s various ways to make bands, and it’s certainly possible to rotate those as well. But I can’t offer any concrete advice without actual code, i.e. a complete Minimal Reproducible Example.

That’s great to hear! Rotating the band is the clean solution I am looking for. Here is a code snippet:

import numpy as np
import pandas as pd
from bokeh.plotting import figure
from bokeh.models import ColumnDataSource, Band
from import show 

# Make some simple data
height = np.array([2000.0, 1750.0, 1650.0, 1400.0, 1000.0, 0.0, -300.0, -3000.0, -4000.0])
mean = np.array([130.0, 50., 85.0, 80.0, 70.0, 60.0, 85.0,  65.0, 45.0])
std = np.array([5.0, 10.0, 5.0, 2.0, 2.0, 2.0,  2.0, 2.0, 5.0])
data = pd.DataFrame(np.transpose([height, mean, std]), columns=['Height', 'Mean', 'Std'])
data['Upper'] = data['Mean'] + 3*data['Std']
data['Lower'] = data['Mean'] - 3*data['Std']

# Make a basic bokeh plot
s = figure(plot_width=400, plot_height=600,x_range=[0,130], y_range=[-4200, 2100])
data_source = ColumnDataSource(data)

# Create a plot with 'Mean' in X and 'Height' in Y axis. Note that the independent variable is actually 'Height'
plot = s.line(x='Mean', y='Height', line_color='blue', line_alpha=1.0, line_width=4, source=data_source)

# Add a band; the intention here is to create a confidence interval at each 'Height' value. 'Height' is in X axis. 
# The band below will not create a plot that I am looking for
band = Band(base="Mean", lower="Lower", upper="Upper", source=data_source, fill_alpha=0.3, fill_color="gray", line_color="black")

You need to change the dimension property of the Band

That’s good to know! I set it to ‘width’ and it generates something like this:

I feel like I need to play with the upper and lower fields. Any thoughts on this? Appreciate the help.

I’m not sure, I’ll have to take a closer look. That is the only change I would expect be necessary. I checked, and we do have an image diff test that exercises vertical bands:

So I don’t immediately think it’s a case of there being a bug.

Oh, you need:


The “base” is the coordinates of the dimension along which the band “travels”.

1 Like

Also just noting that there is a decent chance in the near future the the Band annotation is deprecated and instead simpler separate plot.vband and plot.hband glyphs are added.

Thank you very much! This was very useful.

I see that’s good to know

This topic was automatically closed 90 days after the last reply. New replies are no longer allowed.