vizlint

Super-fast linting for Python visualizations so misleading charts never leave your notebook.

Overview

vizlint inspects Matplotlib figures for easy-to-miss quality problems such as truncated bar charts or missing labels. Use it directly from Python, in a notebook, or via a CLI that runs headless by default (Agg backend) so it is safe in CI.

Features

• Single-call API: vizlint.lint(fig) returns a Report you can print or serialize.
• Matplotlib adapter included; install vizlint[mpl] for the plotting extra.
• CLI that executes your plotting script, lints every open figure, and supports per-rule toggles via --disable.
• Notebook helper vizlint.lint_last() to lint the figure you just rendered.
• Default rules catch truncated bar charts, unlabeled axes, stretched ranges, and missing titles.

Install

pip install vizlint
pip install "vizlint[mpl]"  # Matplotlib extra

Quickstart (Python)

import matplotlib.pyplot as plt
import vizlint

fig, ax = plt.subplots()
ax.bar(["A", "B"], [101, 103])
ax.set_ylim(100, 104)

report = vizlint.lint(fig)
print(report.summary())

Example output

vizlint report:
- WARN [bar_zero_baseline] Bar chart y-axis does not include zero, which can exaggerate differences between bars. Hint: Consider starting the y-axis at zero or switch to a different chart type if you need to zoom in.

CLI usage

The CLI runs your plotting script, forces the Agg backend for headless safety, and lints every open figure:

vizlint path/to/script.py

Disable individual rules by name (rule function names) without editing code:

vizlint path/to/script.py --disable axis_labels_missing --disable bar_zero_baseline

Exit code is 0 when all figures are clean, 1 when any warning is emitted, and 2 if the user script or Matplotlib import fails.

Jupyter / notebooks

Lint the most recently drawn Matplotlib figure from a notebook cell:

import vizlint

# after plotting
report = vizlint.lint_last()
display(report.summary())

Current checks

• bar_zero_baseline (warning) – Flags bar charts whose y-axis range excludes zero, making differences look larger than they are.
• axis_labels_missing (warning) – Reports charts missing an x-axis label, y-axis label, or both.
• axis_range_overexpanded (warning) – Detects y-axes that span far beyond the data range, which can minimize apparent variation.
• title_missing (warning) – Reminds you to add a descriptive chart title to aid quick interpretation.

Configuration & custom rules

vizlint.lint(fig, rules=None) uses vizlint.rules.DEFAULT_RULES when rules is None. Pass your own list to customize behavior:

from vizlint import lint
from vizlint.rules import DEFAULT_RULES, axis_labels_missing

custom = [rule for rule in DEFAULT_RULES if rule is not axis_labels_missing]
report = lint(fig, rules=custom)

# Structured output (e.g., for JSON APIs)
payload = report.to_dict()

In the CLI, use --disable rule_name to skip specific default checks. For richer integrations, the Report object exposes .issues, .is_clean(), .summary(), and .to_dict().

Development

git clone https://github.com/jaeday1212/vizlint.git
cd vizlint
pip install -e .
pip install -e .[mpl]
python -m pytest

You can also run the CLI locally via python -m vizlint.cli path/to/script.py.

Roadmap

• Add more visualization rules (e.g., stacked bars, dual-axis charts).
• Support additional backends beyond Matplotlib.
• Provide optional JSON output mode in the CLI.
• Deepen notebook widgets for inline remediation hints.

Looking Ahead

Upcoming explorations include native adapters for Seaborn and Plotly plus hooks that let LLM-based assistants recommend fixes based on the Report output.

Contributing

Bug reports and pull requests are welcome. Please run python -m pytest before submitting and keep changes focused with clear descriptions.

License

vizlint is available under the MIT License.