Do you need tar_target_cpp()? | The two constructors | How a C++ step differs from the Python/Julia steps | Arguments | Installing Rcpp and checking | Object output (iris example) | File output | Extension packages | Parallel computing with RcppParallel | Bringing an R function into C++ | The function is not part of the compiled library | Where the R function can be defined | Dynamic branching | Compile the C++ code only once across branches | They fall back to the plain pattern on Python and Julia | Several C++ steps in one pipeline | Compile once, reuse in other steps | Reusing a tar_target_cpp() companion library elsewhere | Using more than one library in a step | Current limitations | Step output and tar_polyglot_log() | Rtools note (Windows) | Trade-offs vs Python/Julia

The two constructors | Arguments | Scripts, data, and output | Choosing the Julia install / project | The three-script model | Object output | File output | Dynamic branching (iris example) | Parallel computing (Distributed, not Threads) | Choosing the Julia install and project | Use case 1: Julia on PATH (default) | Use case 2: a specific juliaup version | Use case 3: an explicit Julia home | Use case 4: a pinned project (recommended) | Tracking scripts as dependencies | Conversion caveats

The two constructors | Arguments | Scripts | Wiring in data and reading it back | Choosing the interpreter / environment | The three-script model | Object output | File output | Dynamic branching | Parallel computing (multiprocessing) | Choosing the Python environment | Use case 1: global / system default | Use case 2: a pinned uv virtualenv (recommended) | Use case 3: a poetry project | Use case 4: a plain venv or a conda env | Use case 5: pin just the version | Tracking scripts as dependencies | Conversion caveats

Do you need tar_target_rs()? | The two constructors | How a Rust step differs | Arguments | Installing Rust and rextendr | Object output (iris example) | File output | Crate dependencies, features, profile | Dynamic branching | Compile the Rust code only once across branches | They fall back to the plain pattern on Python and Julia | Several Rust steps in one pipeline | Compile once, reuse in other steps | Using more than one library in a step | Current limitations | Windows toolchain note | Trade-offs vs Python/Julia

Motivation | Installation | Limitations to know first | A worked pipeline (Python + iris) | The three-script model | Object output: return a value into the pipeline | File output: return paths to files on disk | Tracking the scripts as dependencies | Parallelism and isolation with crew | Pre- and post-scripts: moving variables in and out | Sending variables in (the pre-script) | Reading variables out (the post-script) | Where to next

The three forms at a glance | 1. A literal path string | 2. A tar_target_path() reference | 3. Inline code with tar_code() | Mixing forms in one call | Which should I use? | Tracking helper files your script imports | Wiring the helper in with inputs | Julia | Why C++ and Rust are different | A note on building code with glue() | See also