The C API of the R language and environment is robust, yet a little awkward for C++ programmers. This package wraps a tidy layer of header functions around it.
A recent motivating example is from the dang package which includes one unexported Base R function written in C. The original version is here. While short enough, we found working with a tidier version easier and more enjoyable. We include an abbreviated version (minus the copyright header) here:
#include <tidyCpp>
extern "C" {
// cf src/library/tools/src/text.c
(SEXP text, SEXP ignore_quotes) {
SEXP _check_nonASCII/* Check if all the lines in 'text' are ASCII, after removing
comments and ignoring the contents of quotes (unless ignore_quotes)
(which might span more than one line and might be escaped).
This cannot be entirely correct, as quotes and \ might occur as
part of another character in a MBCS: but this does not happen
in UTF-8.
*/
int i, nbslash = 0; /* number of preceding backslashes */
const char *p;
char quote= '\0';
, inquote = FALSE;
Rboolean ign
if (!R::isString(text)) R::error("invalid input");
= (Rboolean) R::asLogical(ignore_quotes);
ign if (ign == NA_LOGICAL) R::error("'ignore_quotes' must be TRUE or FALSE");
for (i = 0; i < R::length(text); i++) {
= R::charPointer(R::stringElement(text, i)); // ASCII or not not affected by charset
p = FALSE; /* avoid runaway quotes */
inquote for (; *p; p++) {
if (!inquote && *p == '#') break;
if (!inquote || ign) {
if ((unsigned int) *p > 127) {
("%s\n", R::charPointer(R::stringElement(text, i)));
Rprintf("found %x\n", (unsigned int) *p);
Rprintfreturn R::scalarLogical(TRUE);
}
}
if ((nbslash % 2 == 0) && (*p == '"' || *p == '\'')) {
if (inquote && *p == quote) {
= FALSE;
inquote } else if(!inquote) {
= *p;
quote = TRUE;
inquote }
}
if (*p == '\\') nbslash++; else nbslash = 0;
}
}
return R::scalarLogical(FALSE);
}
} // extern "C"
Additions from the tidyCpp
wrapper can be identified
easily via the R::
namespace. While the code is not hiding
its C heritage, we find it overall a little easier to work with. Other
examples (see below) show how using tidyCpp
can help avoid
the very manual and error-prone PROTECT
and
UNPROTECT
and pairs.
For additional examples, please see the Motivation vignette for a longer discussion with more examples, and the files in the snippets/ directory for directly callable code examples.
Dirk Eddelbuettel
GPL (>= 2)