Bindings to programming languages

Through binding of the dyncall library into a scripting environment, the scripting language can gain system programming status to a certain degree.
The dyncall library provides bindings to Erlang[1], Java[2], Lua[3], Python[4], R[5], Ruby[6], Go[7] and the shell/command line.
However, please note that some of these bindings are work-in-progress and not automatically tested, meaning it might require some additional work to make them work.

Common Architecture

The binding interfaces of the dyncall library to various scripting languages share a common set of functionality to invoke a function call.

Dynamic loading of code

The helper library dynload which accompanies the dyncall library provides an abstract interface to operating-system speciﬁc mechanisms for loading and accessing executable code out of, but not limited to, shared libraries.

Functions

All bindings are based on a common interface convention providing a common set of the following 4 functions (exact spelling depending on the binding’s scripting environment):

load: - load a module of compiled code
free: - unload a module of compiled code
ﬁnd: - ﬁnd function pointer by symbolic names
call: - invoke a function call

Signatures

A signature is a character string that represents a function’s arguments and return value types. It is used in the scripting language bindings invoke functions to perform automatic type-conversion of the languages’ types to the low-level C/C++ data types. This is an essential part of mapping the more ﬂexible and often abstract data types provided in scripting languages to the strict machine-level data types used by C-libraries. The high-level C interface functions dcCallF(), dcVCallF(), dcArgF() and dcVArgF() of the dyncall library also make use of this signature string format.

The format of a dyncall signature string is as depicted below: dyncall signature string format

<input parameter type signature character>* ’)’ <return type signature character>

The <input parameter type signature character> sequence left to the ’)’ is in left-to-right order of the corresponding C function parameter type list.
The special <return type signature character> ’v’ speciﬁes that the function does not return a value and corresponds to void functions in C.

Signature character	C/C++ data type

’v’	void
’B’	_Bool, bool
’c’	char
’C’	unsigned char
’s’	short
’S’	unsigned short
’i’	int
’I’	unsigned int
’j’	long
’J’	unsigned long
’l’	long long, int64_t
’L’	unsigned long long, uint64_t
’f’	ﬂoat
’d’	double
’p’	void*
’Z’	const char* (pointing to C string)
’A’	aggregate (struct, union) by-value

Table 2: Type signature encoding for function call data types

Please note that using a ’(’ at the beginning of a signature string is possible, although not required. The character doesn’t have any meaning and will simply be ignored. However, using it prevents annoying syntax highlighting problems with some code editors.

Calling convention modes can be switched using the signature string, as well. A ’_’ in the signature string is followed by a character specifying what calling convention to use, as this aﬀects how arguments are passed. This makes only sense if there are multiple co-existing calling conventions on a single platform. Usually, this is done at the beginning of the string, except in special cases, like specifying where the varargs part of a variadic function begins. The following signature characters exist:

Signature character	Calling Convention

’:’	platform’s default calling convention
’*’	platform’s default C++/thiscall calling convention
’e’	vararg function
’.’	vararg function’s variadic/ellipsis part (...), to be speciﬁed before ﬁrst vararg
’c’	only on x86: cdecl
’s’	only on x86: stdcall
’F’	only on x86: fastcall (MS)
’f’	only on x86: fastcall (GNU)
’+’	only on x86: thiscall (MS)
’#’	only on x86: thiscall (GNU)
’A’	only on ARM: ARM mode
’a’	only on ARM: THUMB mode
’$’	syscall

Table 3: Calling convention signature encoding

Examples of C function prototypes

	C function prototype	dyncall signature

void	f1();	”)v”
int	f2(int, int);	”ii)i”
long long	f3(void*);	”p)L”
void	f3(int**);	”p)v”
double	f4(int, bool, char, double, const char*);	”iBcdZ)d”
void	f5(short, long long, ...);	”_esl_.di)v” (for (promoted) varargs: double, int)
struct A	f6(int, union B);	”iA)A”
short	Cls::f(unsigned char, ...);	”_*p_eC_.i)s” (C++: this-ptr as 1st arg, int as vararg)

Table 4: Type signature examples of function prototypes

Erlang language bindings

The OTP library application erldc implements the Erlang language bindings.

Signature character	accepted Erlang data types

’v’	no return type
’B’	atoms ’true’ and ’false’ converted to bool
’c’, ’C’	integer cast to (unsigned) char
’s’, ’S’	integer cast to (unsigned) short
’i’, ’I’	integer cast to (unsigned) int
’j’, ’J’	integer cast to (unsigned) long
’l’, ’L’	integer cast to (unsigned) long long
’f’	decimal cast to ﬂoat
’d’	decimal cast to double
’p’	binary (previously returned from call_ptr or callf) cast to void*
’Z’	string cast to void*

Table 5: Type signature encoding for Erlang bindings

Go language bindings

A Go binding is provided through the godc package. Since Go’s type system is basically a superset of C’s, the type mapping from Go to C is straightforward.

Signature character	accepted Go data types

’v’	no return type
’B’	bool
’c’, ’C’	int8, uint8
’s’, ’S’	int16, uint16
’i’, ’I’	int, uint
’j’, ’J’	int32, uint32
’l’, ’L’	int64, uint64
’f’	ﬂoat32
’d’	ﬂoat64
’p’, ’Z’	uintptr, unsafe.Pointer

Table 6: Type signature encoding for Go bindings

Note that passing a Go-string directly to a C-function expecting a pointer is not directly possible. However, the binding comes with two helper functions, AllocCString(value string) unsafe.Pointer and FreeCString(value unsafe.Pointer) to help with converting a string to an unsafe.Pointer which then can be passed to ArgPointer(value unsafe.Pointer). Once you are done with this temporary string, free it using FreeCString(value unsafe.Pointer).

Python language bindings

The python module pydc implements the Python language bindings, namely load, find, free, call, new_callback, free_callback.

Signature character	accepted Python 2 types	accepted Python 3 types

’v’	no return type	no return type
’B’	bool	bool
’c’, ’C’	int, string (with single char)	int, string (with single char)
’s’, ’S’	int	int
’i’, ’I’	int	int
’j’, ’J’	int	int
’l’, ’L’	int, long	int
’f’	ﬂoat	ﬂoat
’d’	ﬂoat	ﬂoat
’p’	bytearray, int, long, None, (PyCObject, PyCapsule)	bytearray, int, None, (PyCObject, PyCapsule)
’Z’	string, unicode, bytearray	string, bytes, bytearray

Table 7: Type signature encoding for Python bindings

This is a very brief description that omits many details. For more, refer to the README.txt ﬁle of the binding.

R language bindings

The R package rdyncall implements the R langugae bindings providing the function .dyncall() .

Signature character	accepted R data types

’v’	no return type
’B’	coerced to logical vector, ﬁrst item
’c’	coerced to integer vector, ﬁrst item truncated char
’C’	coerced to integer vector, ﬁrst item truncated to unsigned char
’s’	coerced to integer vector, ﬁrst item truncated to short
’S’	coerced to integer vector, ﬁrst item truncated to unsigned short
’i’	coerced to integer vector, ﬁrst item
’I’	coerced to integer vector, ﬁrst item casted to unsigned int
’j’	coerced to integer vector, ﬁrst item
’J’	coerced to integer vector, ﬁrst item casted to unsigned long
’l’	coerced to numeric, ﬁrst item casted to long long
’L’	coerced to numeric, ﬁrst item casted to unsigned long long
’f’	coerced to numeric, ﬁrst item casted to ﬂoat
’d’	coerced to numeric, ﬁrst item
’p’	external pointer or coerced to string vector, ﬁrst item
’Z’	coerced to string vector, ﬁrst item

Table 8: Type signature encoding for R bindings

Some notes on the R Binding:

Unsigned 32-bit integers are represented as signed integers in R.
64-bit integer types do not exist in R, therefore we use double ﬂoats to represent 64-bit integers (using only the 52-bit mantissa part).

Ruby language bindings

The Ruby gem rbdc implements the Ruby language bindings.

Signature character	accepted Ruby data types

’v’	no return type
’B’	TrueClass, FalseClass, NilClass, Fixnum casted to bool
’c’, ’C’	Fixnum cast to (unsigned) char
’s’, ’S’	Fixnum cast to (unsigned) short
’i’, ’I’	Fixnum cast to (unsigned) int
’j’, ’J’	Fixnum cast to (unsigned) long
’l’, ’L’	Fixnum cast to (unsigned) long long
’f’	Float cast to ﬂoat
’d’	Float cast to double
’p’, ’Z’	String cast to void*

Table 9: Type signature encoding for Ruby bindings

index