Frequently asked questions

What makes this matplotlib wrapper different?

There is already a great matplotlib wrapper called seaborn. Also, pandas and xarray both offer convenient matplotlib plotting commands. How does ProPlot compare against these tools?

  • ProPlot, seaborn, pandas, and xarray all offer tools for generating rigid, simple, nice-looking plots from data stored in DataFrames and DataArrays (ProPlot tries to apply labels from these objects, just like pandas and xarray).

  • Unlike seaborn, pandas, and xarray, ProPlot also works for arbitrarily complex subplot grids, and ProPlot provides tools for heavily customizing plots.

  • ProPlot is integrated with cartopy and basemap. You will find plotting geophysical data in ProPlot to be much more concise than working with cartopy and basemap directly.

  • ProPlot expands upon the seaborn tools for working with color and global settings. For example, see Colormap, PerceptuallyUniformColormap, and rc_configurator.

  • ProPlot expands upon matplotlib by fixing various quirks, developing a more sophisticated automatic layout algorithm, simplifying the process of drawing outer colorbars and legends, and much more.

  • ProPlot is built right into the matplotlib API, thanks to special subclasses of the Figure and Axes classes, while seaborn, pandas, and xarray are meant to be used separately from the matplotlib API.

In a nutshell, ProPlot is intended to unify the convenience of seaborn, pandas, and xarray plotting with the power and customizability of the underlying matplotlib API.

Why didn’t you add to matplotlib directly?

Since ProPlot is built right into the matplotlib API, you might be wondering why we didn’t contribute to the matplotlib project directly.

  • Certain features directly conflict with matplotlib. For example, ProPlot’s tight layout algorithm conflicts with matplotlib’s tight layout by permitting fluid figure dimensions, and the new GridSpec class permits variable spacing between rows and columns and uses physical units rather than figure-relative and axes-relative units.

  • Certain features are arguably be too redundant. For example, format is convenient, but the same tasks can be accomplished with existing axes and axis “setter” methods. Also, some of the functionality of subplots can be replicated with axes_grid1. Following TOOWTDI philosophy, these features should probably not be integrated.

Nevertheless, if any core matplotlib think that some of ProPlot’s features should be added to matplotlib, please contact core developer Luke Davis and let him know!

Why do my inline figures look different?

These days, most publications prefer plots saved as vector graphics 1 rather than raster graphics 2. When you save vector graphics, the content sizes should be appropriate for embedding the plot in a document (for example, if an academic journal recommends 8-point font for plots, you should use 8-point font in your plotting code).

Most of the default matplotlib backends make low-quality, artifact-plagued jpegs. To keep them legible, matplotlib uses a fairly large default figure width of 6.5 inches (usually only suitable for multi-panel plots) and a slightly large default font size of 10 points (where most journals recommend 5-9 points). This means your figures have to be downscaled so the sizes used in your plotting code are not the sizes that appear in the document.

ProPlot helps you get your figure sizes correct for embedding them as vector graphics inside publications. It uses a slightly smaller default font size, calculates the default figure size from the number of subplot rows and columns, and adds the journal keyword argument to Figure which can be used to employ figure dimensions from a particular journal standard. To keep the inline figures legible, ProPlot also employs a higher quality default inline backend.


Vector graphics use physical units (e.g. inches, points), are infinitely scalable, and often have much smaller file sizes than bitmap graphics. You should consider using them even when your plots are not destined for publication. PDF, SVG, and EPS are the most common formats.


Raster graphics use pixels and are not infinitely scalable. They tend to be faster to display and easier to view, but they are discouraged by most academic publishers. PNG and JPG are the most common formats.