r-testing

Use test_that() blocks with one behavior per test and descriptive names like "function_name() handles edge case X". For BDD style, use describe() and it() blocks. Always use testthat 3rd edition by declaring Config/testthat/edition: 3 in DESCRIPTION.

Name test files as test- plus the source file name with hyphens, so R/functions/utils_data.R maps to tests/testthat/test-utils-data.R. Place shared setup in helper.R, static test data in fixtures/, and snapshot output in _snaps/.

Use snapshot testing with testthat::expect_snapshot() for printed output and expect_snapshot_file() for saved plot files. Add skip_on_ci() when rendering may differ across platforms, and store snapshots in the auto-generated _snaps/ directory.

This convention sets 80% line coverage as an advisory threshold tracked in CI, plus 100% coverage of exported functions meaning every exported function has at least one test. Measure coverage with covr::package_coverage() and generate reports with covr::report().

Using library() in tests creates undeclared dependencies; instead use pkg::function() or rely on DESCRIPTION. setwd() breaks test isolation, so use withr::local_dir() for temporary directory changes and testthat::test_path() for fixture references.