plotnine

plotnine

This skill targets plotnine 0.15+ with pandas 2.x.

Behavioral Rules

1. Runnability — all generated code is executable as-is. Include imports, data loading, and the full ggplot expression. No pseudocode, no ... elisions, no undefined variables.

2. Idiomatic plotnine — use + layering, aes(), theme_*, labs(), scale_*. Never fall back to matplotlib. If plotnine cannot achieve the result, propose the closest plotnine-idiomatic alternative and explain the limitation.

3. Data completeness — generated code must include all data needed to run: inline pd.DataFrame() construction, plotnine.data datasets, or clearly referenced from the user's existing context. Never reference undefined DataFrames.

4. Minimal transformations — prefer simple pandas ops (groupby, agg, melt) or polars equivalents. Prefer plotnine stat_*/position_* over manual aggregation when possible.

5. Accessibility — always provide descriptive axis labels via labs() and clear legend titles. When the user requests accessible colors, recommend colorblind-safe palettes (Set2, viridis, Okabe-Ito). Provide alt-text guidance when asked.

6. Reproducibility — use random_state=42 for any plotnine method that accepts it (e.g., DataFrame.sample()). Seed synthetic data with numpy.random.default_rng(42).

When to Use What

Task: Create a basic plot (scatter, bar, line, histogram, boxplot) Use: "Plotnine Essentials" below, then references/geoms.md

Task: Map variables to visual properties or customize scales Use: references/aesthetics-and-scales.md

Task: Customize appearance (fonts, backgrounds, gridlines, themes) Use: references/themes-and-styling.md

Task: Choose accessible colors or palettes Use: references/color-and-accessibility.md

Task: Reshape or prepare data for a specific chart Use: references/data-preparation.md