Methodology
R Observatory measures a good deal about every package. Some of it comes from the CRAN source tarball, some from the upstream repository the tarball leaves behind, and some from building the package and analyzing its code. These notes explain how each measurement is taken and what it does and does not tell you.
Detailed method notes
Which development-tooling and community-health files a maintainer keeps in the upstream repository, recorded as a presence snapshot rather than a quality judgment.
Visible, declared signs that an AI coding tool has worked on a repository, and when each one first appeared. A floor on visible tooling, not a ceiling on AI use.
Also measured
These figures appear on each package page. A fuller write-up of how they are computed is on the way.
The share of a package's own R code that its tests exercise, measured by running the package's tests under coverage instrumentation against a current build. Not measured is kept distinct from zero.
Each function a package defines, exported or internal, and which function calls which, read from a static analysis of the source rather than by running it.
Per-version size, documentation coverage, test-to-code ratio, and dependency counts, taken directly from the package source at each release.
Counts aggregated across several mirrors and package managers, each with its own coverage. The headline figure comes from one widely used mirror, so it undercounts total use.