[Rd] interaction with C++ code (PR#10487)

[ripley at stats.ox.ac.uk]

On Mon, 3 Dec 2007, Simon Urbanek wrote:

> This is a bug, but not in R, it's in your program:
>
> On Dec 3, 2007, at 9:45 AM, I.J.Wilson at ncl.ac.uk wrote:
>
>> Full_Name: Ian Wilson
>> Version: 2.6.1
>> OS: linux
>> Submission from: (NULL) (128.240.229.7)
>>
>>
>> The problem is new to R2.6.?.  The code works as expected in R-2.5.0.
>
> it still does in 2.6.x - but it depends on what you expect. You fail
> to terminate the strings, so you get trailing garbage - it just may
> have happened that by chance the trailing memory was zero in 2.5.

And there is another bug.  What is passed in is 80 zero-length strings, so 
in principle you cannot extend those strings (as only one byte is 
allocated for each).  It happens that R_alloc currently allocates 8 bytes 
(because it rounds up to a multiple of 8) and so the example worked, but 
you cannot rely on such implementation details.

The only valid change to an element of a character vector via the .C 
interface is to replace it by a 'char *' of the same length or shorter.
.Call is needed for almost all real applications that alter character 
vectors.



>
>
>> I get the
>> problem with two different operating systems - an older redhat and
>> new ubuntu
>> and with both g++4.1 and g++3.4.
>>
>> I have a problem with character data that is passed back from C++
>> code.  A small
>> example is the following C++ code and R functions.
>>
>> #include <sstream>
>> extern "C" {
>>  void extracttxt( char **txt, int *nchars) {
>>      std::ostringstream oss;
>>      oss << "abcdefghij";
>>      const char *ltxt=oss.str().c_str();
>>      for (int j=0;j<*nchars;j++) *txt[j]=static_cast<char>(ltxt[j]);
>
> As said above, you fail to terminate the string, you want to add
> something like:
>        txt[j][1]=0;
> here.
>
> Cheers,
> Simon
>
>>
>>    }
>> }
>>
>> # begin R code
>> dyn.load("test.so")
>>
>> "testtxt" <-  function() {
>>  nchars <- 80
>>  txt <- .C("extracttxt"
>>            ,character(nchars)
>>            ,as.integer(nchars))[[1]]
>>
>> txt
>> }
>> a=testtxt()
>> a[1:10]
>>
>> # typically the results are something like
>> [1] "a"                   "b\033\xba\bfaul\030" "cz#\b"
>> [4] "d"                   "e"                   "f"
>> [7] "g"                   "h"                   "i"
>> [10] "j"
>>
>> or
>> [1] "az#\b"            "b"                "c"                "d"
>>
>> [5] "e"                "f"                "g"                "h"
>>
>> [9] "i"                "j1\xba\brese\030"
>>
>> ______________________________________________
>> R-devel at r-project.org mailing list
>> https://stat.ethz.ch/mailman/listinfo/r-devel
>>
>>
>
> ______________________________________________
> R-devel at r-project.org mailing list
> https://stat.ethz.ch/mailman/listinfo/r-devel
>

-- 
Brian D. Ripley,                  ripley at stats.ox.ac.uk
Professor of Applied Statistics,  http://www.stats.ox.ac.uk/~ripley/
University of Oxford,             Tel:  +44 1865 272861 (self)
1 South Parks Road,                     +44 1865 272866 (PA)
Oxford OX1 3TG, UK                Fax:  +44 1865 272595