matplotlib

Use plt.subplots(nrows, ncols) to create a grid of axes, or GridSpec and subplot_mosaic for complex layouts where panels span multiple rows or columns. The object-oriented interface with explicit axes references is recommended over the pyplot state machine.

The pyplot interface maintains implicit global state and suits quick interactive plots, while the object-oriented API explicitly creates Figure and Axes objects. The OO interface is recommended for production code, multi-panel figures, and maintainable scripts.

Call plt.savefig with dpi=300 and bbox_inches='tight' to remove excess whitespace. Use PDF or SVG for vector output that scales without quality loss, and PNG with facecolor='white' for raster output.

Overlapping labels happen when the default layout lacks spacing. Fix it by creating figures with constrained_layout=True, calling tight_layout(), or saving with bbox_inches='tight' so all text fits within the exported bounds.

Use perceptually uniform colormaps like viridis, plasma, or cividis for sequential data, and diverging maps like coolwarm for data centered on zero. Avoid jet, which distorts perception and is not colorblind-friendly.

Use seaborn for quick statistical plots with sensible defaults and plotly for interactive web-based visualizations. Matplotlib is the right choice when you need fine-grained control over every element or custom plot types.