Skip to content

R CMD check: "T used instead of TRUE"

2 messages · Lutz Prechelt, Brian Ripley

Original
Lutz Prechelt
# 
Hello everybody,
 
I am just trying to put together my first own CRAN-able package.
I have completed the programming and my code works OK.
But now when I perform
  R CMD check agsemisc_1.0-2.tar.gz
and R gets to the step
  * checking examples ...
I receive the following error message:
  > print(bwplot(Species~Sepal.Length, data=iris, panel=panel.bwstrip))
  Error in rep(T, length(x)) : T used instead of TRUE
which apparently complains about a use of T that occurs not 
in the example, but rather in the implementation of the function 
called by the example.

I have searched the R-devel archives for a reference to that
error message and in
https://stat.ethz.ch/pipermail/r-devel/2006-February/036520.html
found a mention of function cleanEx() that is
apparently generated by R CMD check(?) and that contains
  delayedAssign("T", stop("T used instead of TRUE"),
    assign.env = .CheckExEnv)
  delayedAssign("F", stop("F used instead of FALSE"),
    assign.env = .CheckExEnv)
(BTW: A rather fascinating feature. I would not have thought
 this possible.)

My questions:

- Is my perception correct that this is meant as a check
  for robust programming style? 
  (Because T might change while TRUE cannot?)
- Is it correct that I should change ALL occurrences of 
  T or F in my package code independent of their context?
  Or are there different kinds of contexts?
- Are there any other similar checks that are performed by
  R CMD check?
- I cannot find such a thing mentioned in Secion 1.3.1 of 
  R-exts.pdf that describes the actions of R CMD check.
  The most pertinent item appears to be number 13:
    "The examples provided by the package's documentation are run."
  Would you say this is a documentation defect that I ought
  to report to R-bugs?

Any help is appreciated.
Thanks in advance,

  Lutz Prechelt

Prof. Dr. Lutz Prechelt; prechelt at inf.fu-berlin.de
Institut fuer Informatik; Freie Universitaet Berlin
Takustr. 9; 14195 Berlin; Germany
+49 30 838 75115; http://www.inf.fu-berlin.de/inst/ag-se/
Brian Ripley
#
On Thu, 18 May 2006, Lutz Prechelt wrote:

            

      
      
      
    

        
Yes.

            

      
      
      
    

        
All those where T and F mean TRUE/FALSE and not variable names.

This is in An Introduction to R, section 2.4.

            

      
      
      
    

        
It is the pertinent item (but it is item 14 in current R).  It does not 
say how they are run (and they are for example each run in as clean an 
environment as possible).  I think it would be wrong to assume that the 
description there is exhaustive.  But we'll add a cross-reference.