Skip to content

Test Driven Development in R

3 messages · Jose Quesada, Tobias Verbeke, Spencer Graves

#
Hi,
I wonder what kind of interest there is on Test Driven Development (TDD) in R.

Test Driven Development consists of writing the test before the function, and 
iteratively build the function until it passes the test.

Python and Ruby (specially Ruby) have very strong test-oriented cultures. In 
fact, in Ruby at least the custom is to do TDD and lately Behavior-driven 
development (BDD). In BDD, one writes a story of what one would want the code to 
do. This story is almost native English, and then the test suite converts it 
into something that the language understand as tests.

There are some posts on the list about this, but they are about testing in 
general (Runit), not TDD. Example:
http://thread.gmane.org/gmane.comp.lang.r.general/85047

Recently, I found there is an alpha, but working implementation of TDD for 
Komodo edit:
www.sciviews.org/SciViews-K/index.html

The editor has a green bar that becomes red as soon as one edits a function, and 
that edit breaks the tests. This is tremendously useful.

Using Gmane search, the only mention I could find on svUnit was:
http://thread.gmane.org/gmane.comp.lang.r.general/136632/focus=136662

I think this could make a great UseR 2009 talk. Ideally, by someone with more R 
experience than me, and even more ideally by Philippe Grosjean :), but it push 
comes to shove, I could prepare such a talk.

Would this be interesting at all? Are there any resources that I have missed? 
How many people are using this kind of testing or Komodo -sciViews at all?

Best,
-Jose
#
Jose Quesada wrote:

            
There is the RUnit package which is a mature xUnit implementation for R.

I don't know of a tight integration into an editor (apart from
that it is _planned_ for StatET, the Eclipse R plug-in), but
as such it is very useful already.

http://cran.r-project.org/web/packages/RUnit/index.html

HTH,
Tobias
#
Hi, Jose: 

      How does the following

      I have not studied "Test Driven Development", but I routinely 
write documentation "*.Rd" files before I write code, and the 
documentation files include test cases in the "\examples" section.  With 
"\dontshow", I hide tests that I don't think would benefit a user. 

      The source for the 'fda' package includes numerous constructions 
like the following: 


\dontshow{stopifnot(}
all.equal(fRegressout, knee.hip.f)
\dontshow{)}


      Here, "fRegressout" and "knee.hip.f" are two different objects 
that should theoretically be equal.  I show the user that they are in 
fact equal, but I don't bother the user with the "stopifnot" call that 
will stop "R CMD check" to let me know that the code fails this test.  I 
have not used "Runit", but it's on my "to do" list.  "Sciviews-K" sounds 
interesting, but I have not used it. 

      Also, Chambers (2008) Software for Data Analysis (Springer) says 
that "trustworthy software" is "The Prime Directive".  TDD sounds like 
it would support that goal. 

      I will not be able to attend useR 2009 but would be interested in 
a further discussion of this issue. 

      Best Wishes,
      Spencer Graves
Tobias Verbeke wrote: