PharmFlux R source package

The R package uses the shared native Rust engine. Compile a model once with compile_model(), then simulate, fit, scan, or run Morris screening using versioned requests. Results retain units, observation sides, and execution identity. Check fitting status before interpreting parameter estimates.

Start with the R user guide, or download its R Markdown tutorial. Run examples against the same source checkout used to build the package.

Install from the public workspace

R, a compatible C compiler, Rust 1.98 or later, and Cargo are required. From the repository root on macOS or Linux:

mkdir -p "$HOME/R/pharmflux-library"
export R_LIBS_USER="$HOME/R/pharmflux-library"
export R_HOME="$(R RHOME)"
cargo fetch --locked
R CMD INSTALL --library="$R_LIBS_USER" bindings/r
Rscript -e 'library(pharmflux); print(packageVersion("pharmflux"))'

The package builds its native library using the surrounding public Rust workspace. Keep the entire checkout together. Registry dependencies must be cached for its locked offline Cargo build. Restart R after replacing a loaded native library.

A standalone source archive can instead be generated with its own workspace:

python3 scripts/prepare-r-source.py /tmp/pharmflux-r-source
R CMD INSTALL --library="$R_LIBS_USER" /tmp/pharmflux-r-source/pharmflux_0.1.0.tar.gz

The source archive still needs the cached Cargo registry dependencies. These instructions do not assume a CRAN or R-universe release. The documentation examples are checked on macOS ARM64; CI checks the Linux source installation. Windows installation requires separate verification.

Data and estimation boundaries

nonmem_requests() handles an explicit subset of NONMEM-style records with units, dose compartments, and observation sides supplied by the caller. Unsupported records are rejected. fit_data_requests() prepares individual or pooled Gaussian fits. population_fit_data_requests() prepares the native FOCEI or SAEM request and retains source-row mappings. These helpers are not a general NONMEM importer: they do not infer covariates, units, endpoints, or missing-data policy.

The package includes installed R help for each exported function. See the user guide for complete examples, fitting limits, and error handling. The engine and R wrappers use Apache-2.0; dependency licenses remain separate.