- Published on
A Rust-Backed Super-Fast Python Charting Library That Keeps 100 Million Point Charts Interactive
- Authors
- Name
- DeepHorus
Reflex AI has open-sourced XY, a new Apache-2.0 Python charting library designed to keep interactive 2D visualizations responsive even with 100 million data points. The library leverages a native Rust core and WebGL2 rendering to sidestep the performance collapse that plagues most Python charting stacks. Early benchmarks show render times holding at roughly 0.08 seconds from 10,000 to 100 million points, a dramatic improvement over tools like Matplotlib and Plotly. XY is available now via pip install xy and requires Python 3.11 or newer.
Most Python charting libraries create one drawable object per row, so interactions like hover, zoom, and render degrade sharply beyond a few hundred thousand points. XY addresses this by moving the heavy lifting into a Rust core and sending typed binary buffers to the browser instead of JSON. It also uses screen-bounded representationsâlike M4 decimation for ordered lines and density grids for scatter plotsâso the rendering workload stays constant regardless of source row count. Exact f64 values remain stored in Python, enabling precise hover and selection that resolves original rows even in decimated views.
Version 0.0.1 is early alpha, so the deployment envelope is clear: startups and mid-size data teams can adopt it today for internal analytics, notebooks, and shareable artifacts, while regulated enterprises should pilot it before any customer-facing critical path. The strongest use cases are domains where row counts are the real bottleneck: quantitative finance tick data, genomics Manhattan plots, observability telemetry, astronomy, and geospatial analytics. Benchmark results on an Apple M5 Pro show a 34Ă speedup at 10 million points and 177Ă at 50 million versus established alternatives.
The performance numbers are striking. A 10-million-point interactive export shrinks to 258 KiB of HTML, compared with 259 MiB for a Plotly equivalent, and the payload stays near that size from 1 million to 100 million rows. Peak Python-side memory at 10 million points is 0.32 GiB for XY versus 0.84 GiB for Matplotlib and 1.86 GiB for Plotly. Reflex also reports rendering the entire OpenStreetMap datasetâ10 billion pointsâin the same interactive framework. The library exports charts as shareable HTML files with no JavaScript required for embedding.
XYâs API is declarative, with marks, axes, legends, and tooltips composed into 14 chart families including scatter, line, area, histogram, box, violin, ECDF, heatmap, hexbin, and contour. Styling accepts CSS and Tailwind classes via stable DOM slots, and an import shim xy.pyplot as plt runs common Matplotlib code, though not everything is supported. A separate reflex-xy adapter turns any chart into a Reflex component, avoiding iframes entirely.
Industry reaction has been cautious but optimistic. Data visualization practitioners note that the 258 KiB HTML payload is a game-changer for sharing large datasets, while some analysts point out that alpha software requires careful validation before production use. The open-source release has already drawn attention from quantitative finance and bioinformatics communities, where sampling before plotting has long been the unwanted default. Reflex is clear that current decimation and density thresholds are pre-1.0 policies, not API guarantees.
Looking ahead, the roadmap suggests hardening the API, expanding Matplotlib compatibility, and refining the density and decimation heuristics. The success of XY could push established libraries like Plotly and Bokeh to adopt similar Rust-based or GPU-accelerated rendering pipelines. For now, teams with massive row counts have a practical, open-source option that keeps every point accessible and every interaction smooth.
XY fits a broader trend toward native-code performance in Python tooling, from Polars to Ruff to PyO3-based libraries. By keeping exact data in Python and shipping GPU-drawn binary buffers to the browser, Reflex has created a hybrid architecture that may redefine what âinteractiveâ means for large-scale visualization. The projectâs GitHub repo and technical documentation are now public, and the library is ready for real-world stress tests.
