Altair Plot Utils

A collection of ready-made, interactive chart functions built on Altair and Polars.

Philosophy

Most plotting libraries force you to choose between quick-and-dirty (matplotlib one-liners that look bad) and publication-ready (hours of tweaking). This package sits in between:

One function call produces a complete, styled chart with sensible defaults (grid, axis labels, tooltips).
Polars-native — pass a pl.DataFrame directly, no pandas conversion needed.
Interactive by default — every chart supports hover tooltips, highlighting, or linked selection out of the box.
Composable — combine any chart with hchart / vchart to get faceted layouts with independent axes.

Available plots

Module	Functions	Description
`plotutils.hist`	`plot_grouped_histogram`	Side-by-side grouped histograms
`plotutils.uncertainty`	`plot_confidence_scatter`, `plot_deviations`, `plot_predictions_errors`	Error bars, deviation plots, predicted-vs-true scatter
`plotutils.boxplot`	`plot_bivariate_boxes`, `plot_bivariate_strip`	Doubly-grouped boxplots and jittered strip plots
`plotutils.parallel`	`plot_parallel_coordinates`	Parallel coordinates with normalization and highlighting
`plotutils.auc`	`plot_roc_curve`, `AUCReport`	ROC curves with specificity annotations; linked multi-variable AUC explorer
`plotutils.concat`	`hchart`, `vchart`	Facet-like layouts with independent axes per panel

Quick start

import polars as pl
from plotutils.uncertainty import plot_confidence_scatter

df = pl.DataFrame({
    "category": ["A"] * 20 + ["B"] * 20,
    "value": [1.0 + i * 0.1 for i in range(20)]
           + [2.0 + i * 0.1 for i in range(20)],
})

chart = plot_confidence_scatter(df, x_col="category", y_col="value")
chart.save("chart.html")   # interactive HTML
chart.show()               # Jupyter / notebook

Installation

uv pip install -e .

Building the docs

just docs-img     # generate interactive chart HTML files
just docs-build   # build the site (runs docs-img first)
just docs-serve   # live-reload dev server