Stata command-to-package index: a machine-readable mapping from SSC distribution manifests

A machine-readable mapping from Stata command names to the packages that provide them, reconstructed from the Statistical Software Components (SSC) archive's own distribution manifests. 8,726 mappings · 3,967 packages · 7,468 user commands. Packages and commands both counted excluding internal helper files. Why this exists. R has CRAN and Python has PyPI: given an import, you can look up the package. Stata has no equivalent public mapping from a command to its package, which makes Stata code effectively unmeasurable at scale and is a large part of why studies of research software omit it — despite Stata being the language social science replication code uses most. Method. Every package on the SSC mirror ships a .pkg manifest listing the files it distributes. Crawling those manifests yields the command→package mapping, including the many-commands-per-package case a package-name list cannot express: esttab, eststo, estadd and estpost all belong to estout. Three caveats, which change how you should use this. A shipped file is not necessarily a command. An f foo.ado line says a package distributes a file, not that it exposes a user command. The evidence column separates filename inference from program_define confirmation, and is_helper flags internal subroutines. This is a current snapshot, not a history. A command that was user-written in 2010 and later absorbed into official Stata resolves against its status today, so time-series use will see packages appear to vanish exactly when their commands are absorbed. Ambiguity is preserved, not resolved. 211 commands are claimed by more than one package and are listed as such rather than assigned a winner. The accompanying builtins.json lists official Stata commands, each checked against StataCorp's public help server rather than curated from memory. See README.md in the deposit for the full data descriptor, column definitions and a worked example.