tests/testthat/fixtures/README.md

In TrialSimulator: Clinical Trial Simulator

Test fixtures: design setups for C++/R rollback consistency

Each file in this directory is a self-contained simulation setup used by test-cpp-rollback.R. The tests verify that the C++ lock-time fast path and the pure-R fallback produce identical controller$get_output().

Two file forms

• setup_<name>.R (vignette-derived): sourcing the file leaves a global controller object in the current environment, ready to call $run(). All trial(...) calls use a fixed seed so the runs are reproducible.

• doc_<name>.R (TrialSimulatorDocuments-derived): sourcing the file defines a simulate_<name>(n, seed) function that builds and runs internally and returns the output data frame.

Why no stored expected RDS

The tests run the SAME setup twice on the SAME branch -- once with options(trialsimulator.use_cpp = FALSE) and once with TRUE -- and compare the two outputs with expect_identical. This avoids version-pinning a binary fixture into the repo.

The toggle itself was validated once (offline) against a fresh install of v1.17.1 (pre-C++) on every fixture in this directory, confirming that the R-fallback path is byte-identical to the original v1.17.1 R logic.

setup_stopFollowupAccrual.R is not vignette-derived: it exists to cover the adaptation methods stop_followup() and update_accrual_rate() (v1.25/v1.26) under the C++/R identical-path check.

TrialSimulator documentation built on Sept. 4, 2026, 5:08 p.m.