Skip to Content

Foreign {base}

Foreign Function Interface
Package: 
base
Version: 
R 3.0.2

Description

Functions to make calls to compiled code that has been loaded into R.

Usage

.C(.NAME, ..., NAOK = FALSE, DUP = TRUE, PACKAGE, ENCODING)
 .Fortran(.NAME, ..., NAOK = FALSE, DUP = TRUE, PACKAGE, ENCODING)

Arguments

.NAME
a character string giving the name of a C function or Fortran subroutine, or an object of class "NativeSymbolInfo", "RegisteredNativeSymbol" or "NativeSymbol" referring to such a name.
...
arguments to be passed to the foreign function. Up to 65.
NAOK
if TRUE then any NA or NaN or Inf values in the arguments are passed on to the foreign function. If FALSE, the presence of NA or NaN or Inf values is regarded as an error.
DUP
if TRUE then arguments are duplicated before their address is passed to C or Fortran.
PACKAGE
if supplied, confine the search for a character string .NAME to the DLL given by this argument (plus the conventional extension, ‘.so’, ‘.dll’, ...).

This is intended to add safety for packages, which can ensure by using this argument that no other package can override their external symbols, and also speeds up the search (see ‘Note’).

ENCODING
For back-compatibility, accepted but ignored.

Details

These functions can be used to make calls to compiled C and Fortran 77 code. Later interfaces are .Call and .External which are more flexible and have better performance.

These functions are both primitive, and .NAME is always matched to the first argument supplied (which should not be named). The other named arguments follow ... and so cannot be abbreviated. For clarity, should avoid using names in the arguments passed to ... that match or partially match .NAME.

Values

A list similar to the ... list of arguments passed in (including any names given to the arguments), but reflecting any changes made by the C or Fortran code.

Argument types

The mapping of the types of R arguments to C or Fortran arguments is

R C Fortran
integer int * integer
numeric double * double precision
-- or -- float * real
complex Rcomplex * double complex
logical int * integer
character char ** [see below]
raw unsigned char * not allowed
list SEXP * not allowed
other SEXP not allowed

Numeric vectors in R will be passed as type double * to C (and as double precision to Fortran) unless the argument has attribute Csingle set to TRUE (use as.single or single). This mechanism is only intended to be used to facilitate the interfacing of existing C and Fortran code.

The C type Rcomplex is defined in ‘Complex.h’ as a typedef struct {double r; double i;}. It may or may not be equivalent to the C99 double complex type, depending on the compiler used.

Logical values are sent as   (FALSE), 1 (TRUE) or INT_MIN = -2147483648 (NA, but only if NAOK = TRUE), and the compiled code should return one of these three values: however non-zero values other than INT_MIN are mapped to TRUE.

Note: The C types corresponding to integer and logical are int, not long as in S. This difference matters on most 64-bit platforms, where int is 32-bit and long is 64-bit (but not on 64-bit Windows).

Note: The Fortran type corresponding to logical is integer, not logical: the difference matters on some Fortran compilers.

Missing (NA) string values are passed to .C as the string "NA". As the C char type can represent all possible bit patterns there appears to be no way to distinguish missing strings from the string "NA". If this distinction is important use .Call.

.Fortran passes the first (only) character string of a character vector is passed as a C character array to Fortran: that may be usable as character*255 if its true length is passed separately. Only up to 255 characters of the string are passed back. (How well this works, and even if it works at all, depends on the C and Fortran compilers and the platform.)

Lists, functions are other R objects can (for historical reasons) be passed to .C, but the .Call interface is much preferred. All inputs apart from atomic vectors should be regarded as read-only, and all apart from vectors (including lists), functions and environments are now deprecated.

Warning

DUP = FALSE is dangerous.

People concerned about memory usage are strongly recommended to use the .Call interface instead of these interfaces.

If you pass a local variable to .C/.Fortran with DUP = FALSE, your compiled code can alter the local variable and not just the copy in the return list. Worse, if you pass a local variable that is a formal parameter of the calling function, you may be able to change not only the local variable but the variable one level up. This will be very hard to trace.

With DUP = FALSE, character vectors cannot be used, and single precision values will not be returned.

It is safe and useful to set DUP = FALSE if you do not change any of the variables that might be affected, e.g.,

.C("Cfunction", input = x, output = numeric(10)).

In this case the output variable did not exist before the call so it cannot cause trouble (but as from R 2.15.1 it is not copied even with DUP = TRUE). If the input variable is not changed in the C code of Cfunction you are safe.

Fortran symbol names

All Fortran compilers known to be usable to compile R map symbol names to lower case, and so does .Fortran.

Symbol names containing underscores are not valid Fortran 77 (although they are valid in Fortran 9x). Many Fortran 77 compilers will allow them but may translate them in a different way to names not containing underscores. Such names will often work with .Fortran (since how they are translated is detected when R is built and the information used by .Fortran), but portable code should not use Fortran names containing underscores.

Use .Fortran with care for compiled Fortran 9x code: it may not work if the Fortran 9x compiler used differs from the Fortran 77 compiler used when configuring R, especially if the subroutine name is not lower-case or includes an underscore. It is also possible to use .C and do any necessary symbol-name translation yourself.

Copying of arguments

If DUP = TRUE there are up to two copies made of each argument in ....

Prior to R 2.15.1 there were always two for vectors (one before calling the compiled code and one to collect the results), and this is still the case for character vectors. For other atomic vectors, the argument is not copied before calling the compiled code if it is not otherwise used in the calling code (such as output in the example above). Non-atomic-vector objects are read-only to the C code and are never copied.

This behaviour can be changed by setting options(CBoundsCheck = TRUE). In that case raw, logical, integer, double and complex vector arguments are copied both before and after calling the compiled code. The first copy made is extended at each end by guard bytes, and on return it is checked that these are unaltered. For .C, each element of a character vector uses guard bytes.

References

Becker, R. A., Chambers, J. M. and Wilks, A. R. (1988) The New S Language. Wadsworth & Brooks/Cole.

Note

If one of these functions is to be used frequently, do specify PACKAGE (to confine the search to a single DLL) or pass .NAME as one of the native symbol objects. Searching for symbols can take a long time, especially when many namespaces are loaded.

You may see PACKAGE = "base" for symbols linked into R. Do not use this in your own code: such symbols are not part of the API and may be changed without warning.

PACKAGE = "" used to be accepted (but was undocumented): it is now an error.

The way pairlists were passed by .C prior to R 2.15.0 was not as documented. This has been corrected, but the .Call and .External interfaces are much preferred for passing anything other than atomic vectors.

See Also

dyn.load, .Call.

The ‘Writing R Extensions’ manual.

Documentation reproduced from R 3.0.2. License: GPL-2.