Skip to content

% symbol in .Rd files?

7 messages · Martin Maechler, Dominick Samperi

Original
Dominick Samperi
# 
Hello,

I am encountering problems when there is a '%' character in
package .Rd files. More precisely, CRAN is having problems
with these characters when running check, whereas I am not
seeing any problems under Linux and Windows.  I did see
these problems, and found that escaping the %'s fixes the
problem in my testing, but apparently this does not fix the
problem when check is run at CRAN.

I suspect that there are some compatibility problems between shells,
TeX, or perl versions, or perhaps the check on CRAN goes
through more levels of preprocessing?

I can probably fix this problem by removing all '%' chars
from my .Rd files, but this doesn't seem like a very
natural solution.

Any ideas?

Thanks,
ds
2 days later
Martin Maechler
# 
Dominick> Hello, I am encountering problems when there is a
    Dominick> '%' character in package .Rd files. More
    Dominick> precisely, CRAN is having problems with these
    Dominick> characters when running check, whereas I am not
    Dominick> seeing any problems under Linux and Windows.  I
    Dominick> did see these problems, and found that escaping
    Dominick> the %'s fixes the problem in my testing, but
    Dominick> apparently this does not fix the problem when
    Dominick> check is run at CRAN.

    Dominick> I suspect that there are some compatibility
    Dominick> problems between shells, TeX, or perl versions, or
    Dominick> perhaps the check on CRAN goes through more levels
    Dominick> of preprocessing?

    Dominick> I can probably fix this problem by removing all
    Dominick> '%' chars from my .Rd files, but this doesn't seem
    Dominick> like a very natural solution.

definitely not.  Other packages have them as well,
without giving any CRAN hickups

No, you should find and fix the problem.

    Dominick> Any ideas?

We need examples which -- at least for you -- produce the
problem.  You could post a small tar file (mime-types 'application/x-tar'
or 'application/x-compressed-tar' are accepted for R-devel) if
you want.

Martin Maechler, ETH Zurich
Dominick Samperi
#
Martin Maechler wrote:
An example of the '% problem' is provided by the RcppExample.Rd file that
is part of the RcppTemplate package (V4.2). There you will find:

# Note that we need to escape the percent sign here...
datestr <- c('2006-6-10', '2006-7-12', '2006-8-10')
datevec <- as.Date(datestr, "\%Y-\%m-\%d") # date vector

Without the escaped '%' signs I get an obscure error message when
I run:
 > example(RcppExample)

Here is the error message:
Error in parse(file, n = -1, NULL, "?") : syntax error at
36:
 
37:
 >

There was another problem with this package (now fixed) that
was showing up on CRAN, but not in my testing. This was due to
the fact that the latest version of r-devel includes some new
type checking, so an error was caught at CRAN, but not in
my testing (I use the standard version of R). This problem was
unrelated to the '% problem'.


ds
Martin Maechler
# 

        
DoSa> Martin Maechler wrote:

        
>>>>>>> "DoSa" == Dominick Samperi <dsamperi at decisionsynergy.com>
    >>>>>>> on Mon, 24 Jul 2006 13:07:29 -0400 writes:
    >>>>>>> 
    >> 
    DoSa> Hello, I am encountering problems when there is a
    DoSa> '%' character in package .Rd files. More
    DoSa> precisely, CRAN is having problems with these
    DoSa> characters when running check, whereas I am not
    DoSa> seeing any problems under Linux and Windows.  I
    DoSa> did see these problems, and found that escaping
    DoSa> the %'s fixes the problem in my testing, but
    DoSa> apparently this does not fix the problem when
    DoSa> check is run at CRAN.
    >> 
    DoSa> I suspect that there are some compatibility
    DoSa> problems between shells, TeX, or perl versions, or
    DoSa> perhaps the check on CRAN goes through more levels
    DoSa> of preprocessing?
    >> 
    DoSa> I can probably fix this problem by removing all
    DoSa> '%' chars from my .Rd files, but this doesn't seem
    DoSa> like a very natural solution.
    >> 
    >> definitely not.  Other packages have them as well,
    >> without giving any CRAN hickups
    >> 
    >> No, you should find and fix the problem.
    >> 
    DoSa> Any ideas?
    >> 
    >> We need examples which -- at least for you -- produce the
    >> problem.  You could post a small tar file (mime-types 'application/x-tar'
    >> or 'application/x-compressed-tar' are accepted for R-devel) if
    >> you want.
    >> 
    >> Martin Maechler, ETH Zurich
    >> 
    >> 

    DoSa> An example of the '% problem' is provided by the
    DoSa> RcppExample.Rd file that is part of the RcppTemplate
    DoSa> package (V4.2). There you will find:

    DoSa> # Note that we need to escape the percent sign here...
    DoSa> datestr <- c('2006-6-10', '2006-7-12', '2006-8-10')
    DoSa> datevec <- as.Date(datestr, "\%Y-\%m-\%d") # date vector

which is perfect...
"Writing R Extensions" has always said you must escape them...

    DoSa> Without the escaped '%' signs

you really get invalid examples, not just for example()

    DoSa> I get an obscure error message when
    DoSa> I run:
    >> example(RcppExample)

    DoSa> Here is the error message:
    DoSa> Error in parse(file, n = -1, NULL, "?") : syntax error at
    DoSa> 36:
 
    DoSa> 37:

Now if I run  'R CMD check RcppTemplate' I see no problem.
In my eyes, everything is ok there.

So where exactly does CRAN (i.e. the CRAN machine that does the
daily checks) have a problem?
Maybe it's really a problem on that machine, than rather with
your *.Rd files ...  :-)

Hmm, actually I now start wondering: Why did we ever start
talking about this at all?  Isn't this something you have to settle
with Kurt, Fritz, or whomever.. ?

Regards,
Martin Maechler

  
  


  


      

    

    

      
      

        


  

        

      Dominick Samperi
    

    

      
      #
    

  


  

      
Martin Maechler wrote:

            

      
      
      
    

        
Sorry if I was unclear, but yes, it was settled with the help of Kurt. As I
said, the latest r-devel includes some new type checking (to be sure that
STRING_ELT is applied where it makes sense, for example), and this
is why CRAN was picking up problems that I was not seeing.

I guess I missed that fact about the '%' character in the docs.

ds

  
  


  


      

    

    

      
      

        


  

        

      Martin Maechler
    

    

      
      #
    

  


  

          

      
      
      
    

        

        
Dominick> Martin Maechler wrote:

        
>> Hmm, actually I now start wondering: Why did we ever start
    >> talking about this at all?  Isn't this something you have to settle
    >> with Kurt, Fritz, or whomever.. ?
    >> Regards,
    >> Martin Maechler
    Dominick> Sorry if I was unclear, but yes, it was settled with the help of Kurt. As I
    Dominick> said, the latest r-devel includes some new type checking (to be sure that
    Dominick> STRING_ELT is applied where it makes sense, for example), and this
    Dominick> is why CRAN was picking up problems that I was not seeing.

but not related to "%" ???

Also, with R-devel "as of last night"
I (i.e., "R-devel CMD check")  don't see any problems in
RcppTemplate or its help files.

    Dominick> I guess I missed that fact about the '%' character in the docs.

  
  


  


      

    

    

      
      

        


  

        

      Dominick Samperi
    

    

      
      #
    

  


  

      
Martin Maechler wrote:

            

      
      
      
    

        
This is a somewhat old post, and I didn't mean to suggest that there are 
any remaining
problems (the package wouldn't have been released otherwise). The 
trivial topic about
'%' signs never would have come up if it wasn't mixed up with the other 
unrelated
and non-trivial problem that is now resolved.

Thanks for the feedback.

ds