Designing Types for R, Empirically (Dataset)

This dataset is intended to accompany the paper "Designing Types for R, Empirically" (@ OOPSLA'20, link to paper). This data was obtained by running the Typetracer (aka propagatr) dynamic analysis tool (link to tool) on the test, example, and vignette code of a corpus of >400 extensively used R packages.

Specifically, this dataset contains:

function type traces for >400 R packages (raw-traces.tar.gz);

trace data processed into a more readable/usable form (processed-traces.tar.gz), which was used in obtaining results in the paper;

inferred type declarations for the >400 R packages using various strategies to merge the processed traces (see type-declarations-* directories), and finally;

contract assertion data from running the reverse dependencies of these packages and checking function usage against the declared types (contract-assertion-reverse-dependencies.tar.gz).

 A preprint of the paper is also included, which summarizes our findings.

Fair warning Re: data size: the raw traces, once uncompressed, take up nearly 600GB. The already processed traces are in the 10s of GB, which should be more manageable for a consumer-grade computer.