ggsql - A Grammar of Graphics for SQL

ggsql is a grammar of graphics implementation that ties SQL queries and graph construction into a single flow using VISUALIZE, DRAW, PLACE, SCALE, and LABEL clauses.

• Before VISUALIZE is regular SQL, and after it is the visualization query.
• DRAW adds layers like points, bars, histograms, and boxplots.
• PLACE is a clause dedicated to annotations like rule lines and text.
• SCALE converts categorical values into actual colors or scales.
• LABEL controls titles, subtitles, axis labels, and legend labels.

In the examples, a scatter plot with a regression line is created using the penguins dataset, and for astronauts.parquet, a mean line and text annotation are added on top of a histogram of age distribution. SQL query results are passed directly as visualization input, and the necessary layers are aggregated into a single SQL query, which is explained as advantageous for large-scale data.

For the runtime, it aims to be a small executable that works without an R or Python runtime, and can be used in Quarto, Jupyter notebooks, Positron, and VS Code. It is currently in alpha-release, with plans to add a Rust-based writer, themes, interactivity, a language server, a formatter, spatial data support, and more going forward.

As background for its development, it cites support for SQL-centric users, the compatibility between SQL and grammar of graphics, integration with code-based reporting tools and AI analysis assistants, and the goal of applying the design experience of ggplot2 to a new foundation.