What is the current state of Unit testing support in the R language

R is a statistics programming language. Part of R is the use of Packages, which themselves are written in the R language. Programming best practice includes the use of unit-testing to test the functions within these packages while they are being written and when they are used.

I am aware of a few packages for unit testing within R, these being

• RUnit
• Svunit
• Testthat

I'm interested to know;

Are there any other packages out there ? Given peoples experience, do these packages excel at different things ? What's the current state of the art in unit testing for R ?

Solution 1:

Unit testing seems to be more or less a solved problem, so all three packages will likely be adequate for your needs. There are subtle differences between them though:

RUnit is based on xunit, and as such is easy to understand if you've used any versions of it from other languages.

svUnit uses the same tests as RUnit, but includes a GUI to help with interactive use.

testthat isn't compatible with either, but includes much the same features and can check to only execute tests on files that haven't changed, which is useful for testing large projects.

Several years later...

RUnit and svUnit still don't have checks for messages and warnings, nor test caching, and don't seem to be under much development, so testthat should be your first choice for new projects. You can convert RUnit tests to testthat tests using my runittotestthat package.

Solution 2:

here is a few things I've found from Google.

This topic was discussed on the mailing list here which mentioned the packages above. It seems RUnit and svunit will execute the same test code, but Runit also has tracking and code coverage functions.

A comparison of the two is also on the R wiki at http://rwiki.sciviews.org/doku.php?id=developers:runit