How to escape % in roxygen literate programming?

The default value of a parameter of my function contains a "%". This seems to be a problem for roxygen, it produces a lot of warnings and R CMD check fails when trying to build latex documentation.

How can I make this function (and its documentation) work? Using %% or \% instead of % does not help.

#' Test escape \% from in-source documentation (roxygen).
#'
#' What happens when parameters contain special latex charac开发者_开发知识库ters? 
#'
#' @param x unsuspicious parameter 
#' @param format sprintf format string (default "\%5.0f")
#'
#' @return formatted string
#' @export
#' @author Karsten Weinert
testroxy <- function(x, format = "%5.0f") {
  sprintf(format,x)
}

Code similar to the one in the question works for me without a problem, and generates the correct documentation. As multiple comments say, the reason must be that I am using the roxygen2 package. So just use roxygen2 and then escape "%" in the documentation using backslash:

#' The percentage sign is written in the documentation like this: \%.

#!/usr/bin/perl
use strict;
use File::Temp qw/tempfile/;
use File::Copy;

my $usage = <<EOD

  $0 file1.Rd [file2.Rd ...]

  When using roxygen to generate documentation for an R pacakge, if a default
  argument has a percent sign in it then roxygen will copy it directly into
  the .Rd file. Since .Rd is basically latex, this will be interpreted as a
  comment and case the file to be parsed incorrectly.

  For percent signs elsewhere in your documentation, for example in the
  description of one of the parameters, you should use "\%" so parse_Rd
  interprets it correctly.

  But you cannot do this in the default arguments because they have to be
  valid R code, too.

  Since the .Rd files are automatically generated they should not have
  any latex comments in them anyway.

  This script escapes every unescaped % within the file.

  The .Rd files are modified in place, since it would be easy to
  generate them again with R CMD roxygen.

EOD
;

my $n_tot = 0;
my $n_file = @ARGV;
my $n_esc_file = 0;
foreach my $fn (@ARGV)  {

  print STDERR ' ' x 100, "\rWorking on $fn\t";
  open my $fh, $fn or die "Couldn't open $fn: $!";
  my ($tmp_fh, $tmp_fn) = tempfile();

  my $n;
  while(<$fh>)  {
    $n += s/(?<!\\)%/\\%/g;  # if % is not preceded with backslash then replace it with '\%'
    print $tmp_fh $_;
  }
  $n_tot += $n;
  $n_esc_file ++ if $n;
  print "Escaped $n '%'\n" if $n;
  close $tmp_fh;
  move($tmp_fn => $fn);
}

print "\n";

print "$n_file files parsed\n";
print "$n_esc_file contained at least one unescaped %\n";
print "$n_tot total % were escaped\n";

This is my inelegant solution. Save the perl script as, for example, escape_percents.pl, then the sequence would be like this:

R CMD roxygen my.package
perl escape_percents.pl my.package.roxygen/man/*.Rd
R CMD install my.package.roxygen

This may introduce more problems, for example if you have example code using %% as the modulus operator, but it has been handy for me.