Seamless R and C++ Integration with Rcpp by Dirk Eddelbuettel

Author:Dirk Eddelbuettel
Language: eng
Format: epub
Publisher: Springer New York, New York, NY

7.4 Case Study: The RcppCNPy Package

Modules are a very powerful tool. They are suited to exposing code from existing class libraries as the RcppBDT package discussed in the previous section illustrates. The Rcpp attributes system uses modules to easily connect the wrappers it generates for the user-supplied code.

Modules can be used to provide simple wrappers to external libraries. A simple example is provided by the RcppCNPy package (Eddelbuettel 2012a). It uses a small stand-alone library provided in a single header and source files in order to access NumPy files used by this popular Python extension.

In the package, two functions npyLoad() and npySave() are defined. They simply transfer data between a given file name and R by relying on the external library provided by the source files cnpy.cpp and cnpy.h.

Listing 7.40 shows simplified versions of these two functions. We have omitted several aspects to keep the exposition shorter: the special case of transposing, the additional layer of transparently dealing with gzip-compressed files, the support for long long types, as well as the support for different underlying data types concentratic on just numeric.

Listing 7.40 NumPy load and save functions defined in RcppCNPy

Rcpp::RObject npyLoad(const std::string & filename,

2 const std::string & type) {

cnpy::NpyArray arr;

4 arr = cnpy::npy_load(filename);

6 std::vector<unsigned int> shape = arr.shape;

SEXP ret = R_NilValue;

8 if (shape.size() == 1) {

if (type == "numeric") {

10 double *p = reinterpret_cast<double*>(arr.data);

ret = Rcpp::NumericVector(p, p + shape[0]);

12 } else {

arr.destruct();

14 Rf_error("Unsupported type in npyLoad");

}

16 } else if (shape.size() == 2) {

if (type == "numeric") {

18 ret = Rcpp::NumericMatrix(shape[0], shape[1],

reinterpret_cast<double*>(arr.data));

20 } else {

arr.destruct();

22 Rf_error("Unsupported type in npyLoad");

}

24 } else {

arr.destruct();

26 Rf_error("Unsupported dimension in npyLoad");

}

28 arr.destruct();

return ret;

30 }

32 void npySave(std::string filename, Rcpp::RObject x,

std::string mode) {

34 if (::Rf_isMatrix(x)) {

if (::Rf_isNumeric(x)) {

36 Rcpp::NumericMatrix mat =

transpose(Rcpp::NumericMatrix(x));

38 std::vector<unsigned int> shape =

Rcpp::as<std::vector<unsigned int> >(

40 Rcpp::IntegerVector::create(mat.ncol(),

mat.nrow()));

42

cnpy::npy_save(filename, mat.begin(),

44 &(shape[0]), 2, mode);

} else {

46 Rf_error("Unsupported matrix type\n");

}

48 } else if (::Rf_isVector(x)) {

if (::Rf_isNumeric(x)) {

50 Rcpp::NumericVector vec(x);

std::vector<unsigned int> shape =

52 Rcpp::as<std::vector<unsigned int> >(

Rcpp::IntegerVector::create(vec.length()));

54 cnpy::npy_save(filename, vec.begin(),

&(shape[0]), 1, mode);

56 } else {

Rf_error("Unsupported vector type\n");

58 }

} else {

60 Rf_error("Unsupported type\n");

}

62 }

Download

Copyright Disclaimer:
This site does not store any files on its server. We only index and link to content provided by other sites. Please contact the content providers to delete copyright contents if any and email us, we'll remove relevant links or contents immediately.