C bindings

The C bindings consist of a single header and a shared library which makes application integration straightforward:

The distribution contains:

• include/dnp3.h: the C API, in a single header
• include/dnp3.hpp and src/dnp3.cpp: the C++ wrapper
• lib/<target>/: the shared library for each platform, plus its import library on Windows
• cmake/: a CMake find package script
• LICENSE.txt, third-party-licenses.txt and the software bill of materials dnp3-ffi.cdx.json

Each platform's folder in lib is named after its Rust target, which is also the value the CMake script expects:

Platform	Target
Windows x64 (Windows 10+)	x86_64-pc-windows-msvc
Windows x86 (Windows 10+)	i686-pc-windows-msvc
Linux x64 (kernel 3.2+, glibc 2.17+)	x86_64-unknown-linux-gnu
Linux x64 (musl)	x86_64-unknown-linux-musl
Linux x86 (kernel 3.2+, glibc 2.17+)	i686-unknown-linux-gnu
Linux AArch64 (kernel 4.1+, glibc 2.17+)	aarch64-unknown-linux-gnu
Linux AArch64 (musl)	aarch64-unknown-linux-musl
Linux ARMv7hf (kernel 3.2+, glibc 2.17+)	armv7-unknown-linux-gnueabihf
Linux ARMv6hf (kernel 3.2+, glibc 2.17+)	arm-unknown-linux-gnueabihf
Linux ARMv6hf (musl)	arm-unknown-linux-musleabihf
Linux ARMv6 soft float (kernel 3.2+, glibc 2.17+)	arm-unknown-linux-gnueabi

tip

The CMake script is just a recipe for how you might integrate the library. You don't have to use it. Assuming that you've installed the shared library where the linker can find it, building a program can be as simple as:

> gcc main.c -I./c-bindings/include -ldnp3_ffi

To use the provided CMake find package script, make it discoverable by adding it to the prefix path. Next, call find_package:

set(CMAKE_PREFIX_PATH ${DISTRIBUTION_PATH}/cmake)
set(DNP3_RUST_TARGET x86_64-unknown-linux-gnu)
find_package(dnp3 REQUIRED)

The DNP3_RUST_TARGET variable can also be set via the command line when configuring cmake. It just tells the script which library to link.

> mkdir build; cd build
> cmake .. -DDNP3_RUST_TARGET=x86_64-unknown-linux-gnu

This call will expose the library target:

• dnp3 which is the library.

Declare a dependency on the library for your executable:

add_executable(my_awesome_project main.c)
target_link_libraries(my_awesome_project PRIVATE dnp3)

You must also make the DLL or .so file available to your executable. You can automatically copy it to the executable target directory after the target is declared:

add_custom_command(TARGET my_awesome_project POST_BUILD
    COMMAND ${CMAKE_COMMAND} -E copy_if_different $<TARGET_FILE:dnp3> $<TARGET_FILE_DIR:my_awesome_project>
)

Mapping

Many concepts built into higher-level languages are just design patterns or idioms in C. Use the idioms below to understand how they map to C.

Classes

C doesn't have classes with restricted member visibility. Instead, you can use opaque types to hide implementation details:

typedef struct opaque_type_t opaque_type_t;

You can then define constructors and destructors as functions:

// constructor
opaque_type_t* create_opaque_type();
//destructor
void destroy_opaque_type(opaque_type_t* instance);

Class "member functions" are simply functions that take a pointer to the opaque type as the first argument:

int opaque_type_get_special_value(opaque_type_t* instance);

Interfaces

C accomplishes function polymorphism using function pointers. Interfaces are simply collections of function pointers along with some optional context. For example, let's look at the following logging interface in the library:

typedef struct dnp3_logger_t
{
    void (*on_message)(dnp3_log_level_t level, const char* message, void* ctx);
    void (*on_destroy)(void* ctx);
    void* ctx;
} dnp3_logger_t;

• ctx is an opaque pointer to some state information required by the interface. It is passed into every method as the final argument.
• on_destroy is the destructor that cleans up the ctx.
• on_message is a function pointer used to dispatch a log message.

This example only contains a single method. Other interfaces contain several methods.

warning

Always use C99 struct initialization when defining interfaces as this will guarantee that fields you don't explicitly initialize are set to NULL.

The FFI library will interpret NULL function pointers as "not-defined" and will use an intelligent default for this method of the interface. The ctx value may be NULL as well if your interface is stateless.

Iterators

Collections in the C bindings are always implemented as an opaque iterator type. Think of them as a class with a single "next" method. For example, let's look at an iterator over bool values:

// opaque iterator
typedef struct bool_iterator_t bool_iterator_t;
// next function
bool* bool_iterator_next(bool_iterator_t* iter);

If you are given this iterator in a callback, you can process the values in a loop:

void my_callback(bool_iterator_t* iter, void* ctx)
{
    bool* value = NULL;
    while((value = bool_iterator_next(iter)))
    {
        // process value
    }
}

warning

Never use an iterator outside the callback. Frequently, the iterator points to memory on the stack and will result in undefined behavior if it is used after the callback is complete.

Error handling

C API error handling is performed using error codes. An error code is always an enum value where the first value is equal to zero, indicating success.

Consider an error enum and a function that parses a string as an int that can fail:

typedef enum my_error_t
{
    MY_ERROR_OK = 0,
    MY_ERROR_BAD_INT_STRING = 1,
} my_error_t;

my_error_t parse_string(const char* string, int* out);

When a function that can fail needs to return a value, it always does so using an out parameter as the last argument. Your code should always check for returned error values:

int value = 0;
my_error_t err = parse_string("not a number", &value);
if(err) {
   printf("unable to parse number: %s\n", my_error_to_string(err));
   // handle error
}

note

Every enum generated by the C code generator also has a generated *_to_string helper method for debugging.