Bokeh 0.12.5 Released

/ Bryan Van de Ven

We’re pleased to announce the release of Bokeh 0.12.5!

This update has the following highlights:

  • New general capability for Python/JS events
  • Bokeh apps easily viewable inline in Jupyter Notebooks
  • Confusing --host Bokeh Server CLI parameter no longer necessary
  • Interactive legends can now control glyph visibility
  • Many fixes and improvements to GMapPlot
  • CustomJSTransform now available for CDS columns
  • Sophisticated “pivot” example app contributed by mmowers
  • Themes now work with components and in Jupyter notebooks
  • New glyphs can be added to running Bokeh server applications
  • A Community Code of Conduct has been adopted

As well as many other small bugfixes and docs additions. For full details, see the CHANGELOG and Release Notes.

As a reminder, we now package and upload examples for each Bokeh release to our CDN. Click here to download.

This release can most easily be installed from the Bokeh channel on by executing the command conda install -c bokeh bokeh if you are using Anaconda or pip install bokeh otherwise.

There are several great fixes, improvements, and features to show off below. But first I should get one important announcement out of the way. Sometimes there are things that we might like to do, but must ultimately admit that we do not have the ability to do. This is one of those instances:

Support for Matplotlib compatibility is deprecated.

The Bokeh team unfortunately does not have the resources to continue supporting this functionality in its current form. It was always extremely limited in the first place, and often produced substandard results. In order to support the long term health of the Bokeh project, all MPL compatibility support will be removed on the occasion of a 1.0 release. Any code that currently uses to_bokeh will continue to work with a deprecation warning until that time.

Improvements to Google Maps plots

For quite some time there have been a number of lingering issues with using GMapPlot. Not all of them were “big” on their own, but they could add up to some severe usability problems. I am happy to report that a major pass was made over this corner of Bokeh:

#2822—GMapPlot Resets to (0, 0)

#2940—GMapPlot Coordinate Axes

#3737—GMapPlot doesn’t display anything in Jupyter v.4

#4835—Multiple GMap plots whiting out in notebook.

#5592—GMapOptions are not dynamically applied

#5801—GMap underflowing Bokeh figured

#5826—Add bokeh.plotting.gmap to create GMapPlot correctly and easily

#5845—add support for GMap scale control

In particular, it’s now easier than ever to create Google Maps plots. Previously, map plots required writing a fair chunk of code using the low level bokeh.models API. But now all it takes is one call to the new bokeh.plotting.gmap function:

map_options = GMapOptions(lat=30.2861, lng=-97.7394, map_type="roadmap", zoom=13)

plot = gmap("GOOGLE_API_KEY", map_options)


When you create map plots this way, they will automatically be configured with the new axes that display human-friendly latitude and longitude values. Additionally, now map options like map styles can be updated dynamically.

The example plot below expands on this snippet to demonstrate more of the fixes and features in the list:

Interactive Legends for Hiding Glyphs

One long-requested feature has been to have interactive legends that users could click on to hide, or possibly mute the corresponding glyphs. That has been added in this release. To have the legend completely hide glyphs, set the click_policy to hide:

plot.legend.click_policy = "hide"

You can also set the policy to mute the associate glyph in some way. Typically in this case you’ll want to configure the parameters such as muted_line_alpha to control exactly how the muted glyph appear:

plot.line(x='x', y='y', source=source, legend="Stuff", muted_line_alpha=0.1)
plot.legend.click_policy = "mute"

You can see these interactive legend in action in the plot example for the next section.

Custom Client Side Data Transformations

v_func = """\
    var first = xs[0]
    var norm = new Float64Array(xs.length)
    for (i = 0; i < xs.length; i++) { norm[i] = xs[i] / first }
    return norm

normalize = CustomJSTransform(func="", v_func=v_func)

plot.line(x='date', y=dict(field='close', transform=normalize), source=aapl_source)

Oh, and be sure to click on the legend to mute the visibility of the different series!

JS and Python Callbacks for General UI Events

Another often-requested feature has been a way to respond to events in a more fine-grained way. I am happy to say that JS and Python callbacks are now possible on many different kinds of events: low level mouse and wheel actions, higher level pan, tap and button click actions, as well as responding to the start and stop of “Level of Detail” mode:

Image of events

More events are planned for future releases. You can see how these events are used with CustomJS callbacks in this snippet:

s = ColumnDataSource(data=dict(x=[], y=[]))'x', 'y', color='c', size=20, alpha=0.4, source=s)

p.js_on_event(Tap, CustomJS(args=dict(s=s1), code="""
    event = cb_obj;

The example below uses this code and some additional callbacks for pan and double tap events to create a simple drawing canvas:

These examples show how to use these new events with CustomJS callbacks that run in the browser. But they work just as well in Bokeh server apps as well. Just use the on_event method instead of js_on_event. Of course, Bokeh is extensible, so it’s possible to define your own custom events as well.

Bokeh Server Options Simplified

The –host parameter to the Bokeh Server CLI was a common source of confusion. Recent changes to the Bokeh server have rendered it no longer necessary, so it is now deprecated. In order not to break existing deployments, the parameter is still accepted, but will be ignored. If you are embedding a Bokeh server application in a page that has a different public URL than the Bokeh server, you will want to supply the –allow-websockets-origin parameter to allow connections originating from the public URL.

Bokeh Apps Fully Support in Jupyter Notebooks

The previous Proof-of-Concept demonstrated for embedding Bokeh applications Jupyter notebooks that was demonstrated inthe 0.12.4 release has been fully fleshed out. It’s now easy to create and show Bokeh applications directly inline in Jupyter notebooks, using the same show function that is used for standalone Bokeh plots:

Image of notebook app

You can find the notebook above in the GitHub repository.

Code of Conduct Adopted

With the help of input from several people, BEP 3: Community Code has been implemented. This is an important and required step in order to pursue NumFocus Fiscal Sponsorship.

The details can be found in the link, but as mentioned there, the Bokeh community has been fantastic since day one, and a one-line summary of the CoC is simply a reflection of the community values already on display:

Please be always kind and courteous. There’s never a need to be mean or rude or disrespectful.

What’s Next?

One of the focus areas for the next 0.12.6 release is to clean up and refactor the layout system to make it more robust, maintainable, and performant. Additionally, we plan to clean up the handling of categorical data as well as support nested coordinate systems.

We hope to make 0.12.6 a smaller milestone with a shorter release cycle.

One other important announcement for 0.12.6: the bokeh.charts API will be moved into a separate GitHub project. The new project will be called bkcharts. For now, Bokeh will depend on the new bkcharts package, and import its contents automatically. All existing code will work with a warning message until the 1.0 release, when the dependency will be dropped and bkcharts charts will have to be explicitly installed to be used.

Finally, while I can’t make definite statements about timelines, I can say that resources have been allocated specifically to work on all the following important tasks:

  • Programmatic static image (PNG) generation using Selenium
  • Large graph support in Datashader / Small graph support in Bokeh
  • Core improvements to support the RBokeh project

As we’re getting closer to a Bokeh 1.0 release, I’d like to thank the 216 total contributors who have helped make Bokeh such an amazing project. As always, for questions, technical assistance or if you’re interested in contributing, please ask the Bokeh mailing list or join the chat on Gitter.


Bryan Van de Ven