Bokeh (software)

Bokeh (software)
Name	Bokeh
Developer	Anaconda, Inc.
Released	2013
Programming language	Python
Operating system	Cross-platform
Genre	Visualization
License	BSD-3-Clause

Bokeh (software) is an interactive visualization library for creating web-ready plots and dashboards from Python (programming language), designed to produce scalable, high-performance graphics for modern browsers. It targets analysts and developers who work with NumPy, Pandas, SciPy, and Jupyter Notebook, enabling publication-quality visualizations in the context of projects using Anaconda (software distribution), Dask, and Flask (web framework).

Overview

Bokeh provides declarative and programmatic approaches to generate interactive plots, allowing connections to Jupyter Notebook, JupyterLab, Google Colab, and standalone HTML for sharing with teams using GitHub, Bitbucket, and Docker. It emphasizes linking visual representations to data structures from NumPy, Pandas, xarray, and Dask while integrating with plotting paradigms employed by Matplotlib, Seaborn, and Altair. The project addresses visualization needs in workflows involving SciPy, scikit-learn, TensorFlow, PyTorch, and visualization pipelines deployed on Amazon Web Services, Microsoft Azure, and Google Cloud Platform.

History and development

Origins trace to efforts at Continuum Analytics (now Anaconda, Inc.) in the early 2010s, part of a broader move toward interactive, browser-based tools alongside Plotly (company), Vega (visualization grammar), and D3.js. Early development intersected with contributors from UC Berkeley, Los Alamos National Laboratory, and practitioners familiar with Matplotlib and Seaborn. Major milestones include integration with Jupyter Notebook after the IPython split, support for server-driven apps influenced by Shiny (web framework) and adoption by users at NASA, CERN, University of Washington, and NYC Department of Education. Maintenance and governance evolved through community contributors from NumFOCUS, PSF, and corporate backers like Anaconda, Inc..

Features and architecture

Bokeh's architecture separates a Python client API from a JavaScript rendering layer based on HTML5 Canvas and SVG patterns employed by D3.js. Core components include the Python model layer compatible with NumPy and Pandas, a JSON protocol used by WebSocket servers like Tornado (web framework), and the client-side library that works with React (JavaScript library) and TypeScript. Interactive tools include selections, hover tools, linked brushing, zooming, and custom callbacks using JavaScript or Python via a Bokeh Server process similar in spirit to Node.js-driven dashboards and influenced by reactive models seen in Vue.js and React. Output targets include static PNGs, interactive HTML, and embedded widgets for JupyterLab and Voila.

Usage and API

The public API exposes objects such as Figures, Glyphs, ColumnDataSource, and Widgets, designed for users familiar with Matplotlib and Seaborn idioms. Typical workflows use dataframes from Pandas and computations from NumPy or Dask with rendering in Jupyter Notebook or served through Flask or Django (web framework). Users create layouts with Rows, Columns, Tabs, and integrate callbacks using JavaScript or Python callbacks executed on a Bokeh Server hosted with Gunicorn or behind reverse proxies like NGINX. Examples in academic and enterprise settings demonstrate interoperability with ML stacks such as scikit-learn, XGBoost, LightGBM, and deep learning frameworks like TensorFlow and PyTorch.

Integration and ecosystem

Bokeh integrates with scientific packages and visualization tools including Pandas, NumPy, SciPy, scikit-learn, Dask, HoloViews, Datashader, and Panel (software), enabling large-scale rendering and declarative plotting within application frameworks like Flask, Django, and interactive notebooks like Jupyter Notebook and JupyterLab. It complements plotting libraries such as Matplotlib, Seaborn, Plotly, Altair, and systems like Vega-Lite by offering server-driven interactivity and embedding in platforms including Streamlit and Voila. Enterprises use Bokeh within deployment pipelines orchestrated by Kubernetes, Docker Swarm, and CI systems like Travis CI and GitHub Actions.

Performance and scalability

For large datasets, Bokeh delegates aggregation and streaming to tools like Datashader, and parallel processing backends such as Dask and Ray (software), while serving interactive sessions with Tornado or WSGI servers like Gunicorn. Rendering optimizations exploit HTML5 Canvas batching and downsampling strategies comparable to techniques used in Datashader and high-density visualization systems at institutions like Lawrence Berkeley National Laboratory and Argonne National Laboratory. Benchmarks in scientific computing environments compare Bokeh's interactivity and throughput against Plotly and Matplotlib in contexts of geospatial rendering used by OpenStreetMap consumers and telemetry systems at NASA and NOAA.

Licensing and community

Bokeh is distributed under the BSD license family via the Python Package Index and maintained by contributors affiliated with Anaconda, Inc., NumFOCUS, and independent developers from academia and industry. Community activity occurs on platforms including GitHub, Stack Overflow, Discourse (software), and conferences like SciPy, PyData, and Strata Data Conference, with sponsorship and adoption by organizations such as U.S. Department of Energy labs, universities like UC Berkeley, MIT, and corporations including Anaconda, Inc. and cloud providers like Amazon Web Services.

Category:Data visualization software