Skip to content

[Rcpp-devel] Using testthat instead of adding more RUnit tests?

2 messages · Douglas Bates, Dirk Eddelbuettel

#
In other packages I have started using the testthat package instead of
RUnit for unit tests (it appears that I am in danger of becoming a
"Hadley fanboy" as Dirk characterizes such people).  One great
advantage of testthat is its repeating the errors after all the tests
have been run.  RUnit error reports can be lost in the output from R
CMD check on win-builder.  Also, it is more flexible than RUnit.

I am adding tests to RcppEigen in advance of a major release
incorporating Eigen-3.1.0-alpha2 and I think I will switch the tests
to testthat unless Dirk or others complain vociferously.
#
On 7 March 2012 at 10:26, Douglas Bates wrote:
| In other packages I have started using the testthat package instead of
| RUnit for unit tests (it appears that I am in danger of becoming a
| "Hadley fanboy" as Dirk characterizes such people).  One great

Hah! What is the value of ess-toggle-underscore?  ;-)

| advantage of testthat is its repeating the errors after all the tests
| have been run.  RUnit error reports can be lost in the output from R
| CMD check on win-builder.  Also, it is more flexible than RUnit.
| 
| I am adding tests to RcppEigen in advance of a major release
| incorporating Eigen-3.1.0-alpha2 and I think I will switch the tests
| to testthat unless Dirk or others complain vociferously.

I'm pragmatic. If you do the legwork of testing and adapting it to the
package structure I have no particular qualms about switching one day. The
error reporting is a little iffy with RUnit, but it was essentially all we
had at the time...

Another (related) good thing is that Uwe changed how win-builder reports
things, which also seems to help.

Dirk