Roadmap#

Last updated: June, 2026

This document outlines the roadmap for ongoing development of the Hiveplotlib Python package. It includes planned features, improvements, and changes that are expected to be implemented in future releases.

The development of Hiveplotlib continues to be a high priority for the developers, and we can promise a quick turnaround on bug fixes. That being said, we have no strict timeline for the development and release of the new features discussed below, as this work is currently constrained to “nights and weekends,” making consistent progress difficult.

Broadly, our vision is to bring hive plots to the mainstream, both as a means of exploratory and explanatory analysis of graphs.

More specifically, we are working on the following features, ordered from highest to lowest priority:

  1. Profiling / improving scaling - Although we wrote our code vectorized with numpy wherever possible, the scaling of Hiveplotlib to larger networks (even when visualizing with the datashader backend) still leaves much to be desired. A formal profiling exercise will dictate the necessary / appropriate next steps, but further possibilities include:

    • Revising pandas usage to polars or dask (which could lead to a narwhals side quest for more tabular data package compatibility).

    • Writing some bottleneck computations in Rust.

  2. Add an interactive example for the docs - panel has supported client-side apps in sphinx docs for a while, making this easy to add once we have a nice example. A simple, interactive example will serve as a nice means of introducing / building intuition around hive plots.

  3. Plotting network subsets - Currently, an instantiated HivePlot instance can only be plotted including all nodes and edges on / between its axes. As a means of drill-down, users should be able to filter based on a subset of nodes / edges and observe the resulting plot.

  4. Highlighting network subsets - Similar to (3), users should be able to highlight a subset of nodes / edges on a plot, but still view them in the context of the entire network dataset. This will allow users to explore specific subsets of interest while still being able to see the full context of the network.

  5. More examples - We are always looking for more examples to add to the documentation. If you have a specific example in mind, please let us know!

  6. Optional igraph backend for compute_graph_metrics - igraph ships several community-detection algorithms that networkx does not (Leiden, Infomap, walktrap, spinglass, fast-greedy modularity), plus faster implementations of some centralities. The graph_features.networkx/ subpackage layout that landed in v0.28.0 leaves room for a sibling igraph backend to slot in alongside it.

Completed#

Recently shipped work, in reverse chronological order. See the Releases for details.

  • NetworkX integration - v0.28.0 - build a HivePlot / HivePlotMatrix directly from a networkx graph (the graph= constructor), compute_graph_metrics / the graph_features package, HivePlot.to_networkx() and JSON export, and the graph_metric_backend dispatch system.

  • HivePlotMatrix class - v0.27.0

  • Edge-construction parallelization + numba JIT on bottleneck computations - v0.27.0

On this page