plotnine started as an effort to improve the scaling functionality in ggpy formally known as "ggplot for python". It was part of a larger goal to evolve the package into one that supported a full plotting grammar. It turned out that to have a grammar of graphics system we had to just about start anew.
The complete guide to what constitutes a "Grammar of Graphics" is Leland Wilkinson's book The Grammar of Graphics. To create ggplot2 Hadley Wickham came up with an interpretation termed A layered grammar of graphics 1. Core to the interpretation is a crucial plot building pipeline 2 in ggplot2 that we adopted 3 for plotnine.
The R programming language has a rich statistical ecosystem that ggplot2 taps into with ease. In plotnine we have done our best to integrate with the rest of the scientific python ecosystem. Though we feel we could do more on that integration, notwithstanding language differences, users familiar with ggplot2 should be comfortable and productive with plotnine.
matplotlib - Plotting backend.
pandas - Data handling.
mizani - Scales framework.
statsmodels - For various statistical computations.
scipy - For various statistical computation procedures.
The Grammar of Graphics has to be interpreted into a form that can be implemented. We were not up to this task.
This is more or less an implementation of what is depicted in Figure 2.2 of The Grammar of Graphics
By adopting a similar pipeline and user API as
other internals referenced within the pipeline look similar.