Frequently Asked Questions ¶

Author:

Rohit Goswami

For the marker-table workflow, yes. rsx implements the same command set (process, distrib, signif, depth, freq, map, subset, merge) and preserves the tab-separated output format, so existing R and Python scripts keep working. It adds a Bayesian evidence layer, bindings, and bounded-memory execution on top.

Will my numbers match the original C++ RADSex?¶

When groups are specified explicitly, the frequentist calls match RADSex v1.2.0. rsx does not change the statistical test, the null distribution, or the reported p-value threshold; the Bayesian outputs are additional, not a replacement.

Why is the binary called `rsx` but the crate `radsex-core`?¶

The command-line tool is rsx (crate rsx-cli); the library crate is radsex-core on crates.io. The Python package is pyrsx on PyPI. The repository is rsx-rs.

Installation ¶

How do I install the CLI?¶

cargo install rsx-cli      # installs the `rsx` binary

Or grab a prebuilt binary from the releases page (Linux and macOS), or build from source with cargo build --release.

How do I install the Python bindings?¶

pip install pyrsx

Why no Windows binary?¶

The default map feature links minimap2, which does not build on MSVC. The other commands build on Windows from source with --no-default-features; only map is unavailable there.

Usage ¶

How much memory does rsx need?¶

It is bounded. distrib, freq, signif, and subset scale with the number of individuals; depth, merge, and large inputs use a fixed buffer with external sort; pca scales with the square of the number of individuals. Typical peak is under 500 MB even for tens of millions of markers.

Frequently Asked Questions ¶