Skip to contents

Contributing

Source: CONTRIBUTING.md

Questions and suggestions

Write to . Questions about how to specify a model, about what a number means, or about a model the package does not fit yet are all welcome, and so are suggestions.

If something looks like a defect, an issue is the better place, because the answer stays visible to whoever runs into the same thing next:

https://github.com/phyllype/BreedingR/issues

Reporting a problem

What makes a report easy to act on:

  • a small script that reproduces it. simulate_breeding() ships with the package and produces a pedigree with records, so most reports can be written on simulated data. Please do not send data you are not free to share.
  • the output of sessionInfo() and of br_version().
  • what you expected and what came out. For a fit, summary() of the object together with converged, theta and the iteration count usually says enough.

Convergence deserves a word of its own. A fit reporting converged = FALSE is not necessarily a defect: a component resting on its floor, a correlation at 1, or a covariance group the data does not identify are all legitimate places to stop. Sending the fit along with the design, how many animals, how many groups, how many records per animal, is usually enough to tell which of the two it is.

Sending a change

Fork, branch, open a pull request. Two things are looked at before a merge:

  • R CMD check passes. GitHub Actions runs it on macOS, on Windows and on three versions of R under Ubuntu, for every push and every pull request.
  • new behaviour arrives with a test under tests/testthat/. The bar is that the test fails without the change; one that would pass either way does not gate anything.

The numerics are C++ under src/, compiled by R CMD INSTALL, and the package has no run-time dependency. A change that would add one is a design question rather than a detail, so it is worth raising in an issue before the code gets written.

You keep the copyright on what you write. By opening a pull request you also grant the maintainer the right to license your contribution under the licence this project carries now, GPL-3, and under any other licence the project may adopt later.

The reason is worth stating plainly, because a clause like this is easy to resent. A project whose copyright is spread across many authors cannot change its licence without tracking down every one of them and getting each to agree; projects have been frozen that way, sometimes by a single contributor nobody could reach years later. This keeps that door open. It takes nothing away from you: your name stays on the commit, and whatever version your work was released in stays under the licence it was released under, for everyone who received it. Relicensing is never retroactive.

If you would rather not grant that, say so in the pull request. The contribution can still be discussed, and often the same result is reached by describing the problem well enough that the fix is written here.

Building from a clone

install.packages(".", repos = NULL, type = "source")   # Windows needs Rtools