Higher level programming in C
C C++
Switch branches/tags
Nothing to show
Pull request Compare This branch is 231 commits behind orangeduck:master.
Fetching latest commit…
Cannot retrieve the latest commit at this time.
Failed to load latest commit information.




** Example C+ Program

#include "C+.h"

int main(int argc, char** argv) {

  /* Stack objects are created using "$" */
  var int_item = $(Int, 5);
  var float_item = $(Real, 2.4);
  var string_item = $(String, "Hello");

  /* Heap objects are created using "new" */
  var items = new(List, 3, int_item, float_item, string_item);
  /* Collections can be looped over */
  foreach(items, item) {
    /* Types are also objects */
    var type = type_of(item);
    printf("Type: '%s'\n", as_str(type));
  /* Heap objects destroyed with "delete" */
  /* Tables require "Eq" and "Hash" on key type */
  var prices = new(Table, String, Int);
  put(prices, $(String, "Apple"),  $(Int, 12)); 
  put(prices, $(String, "Banana"), $(Int,  6)); 
  put(prices, $(String, "Pear"),   $(Int, 55)); 
  var pear_price = get(prices, $(String, "Pear"));
  printf("Price of a 'Pear' is '%li'\n", as_long(pear_price));

  /* Tables also supports iteration */
  foreach(prices, key) {
    var price = get(prices, key);
    printf("Price of '%s' is '%li'\n", as_str(key), as_long(price));
  /* File-like objects can "open" and "close" */
  var file = open($(File, NULL), "prices.bin", "wb");
  /* First class function object */
  lambda(write_pair, args) {
    /* Run time type-checking with "cast" */
    var key = cast(at(args, 0), String);
    var val = cast(get(prices, key), Int);
    /* File implements "put/get" like Hashtable */
    put(file, String, key);
    put(file, Int, val);
    return None;
  /* Higher order functions */
  map(prices, write_pair);


C+ is a C GNU99 library which brings higher level programming tools to C. It takes inspiration from C++, Obj-C, Haskell and Python. Most closely C+ resembles C with interfaces, dynamic typing, and some syntactic sugar. There are a selection of new keywords, and many generically named functions in the namespace are taken, but other than that it should be fully compatible with normal C code.

Although I've made the syntax pleasant, this isn't a library for beginners. It is for C power users, as manual memory management doesn't play nicely with many higher-order concepts.

What I don't enjoy in Haskell is writing small detailed algorithms in a functional style. What I do love about Haskell (and C) is the way programs are structured at a higher level. With this project I never really aimed for Object Orientation in C, but I hope that with C+ I've turned it into something of a dynamic and powerful functional language which it might have once been. This also explains why some terminology is closer to Haskell than C++/Java.

More Examples

Defining a new Class.

In the header...

/** Vector - vector operations */

class {
  float (*dot)(var, var);
  float (*length)(var);
} Vector;

float dot(var self, var obj);
float length(var self);

And in the object file...

float dot(var self, var obj) {
  Vector* ivector = type_class(type_of(self), Vector);
  return ivector->dot(self, obj);

float length(var self) {
  Vector* ivector = type_class(type_of(self), Vector);
  return ivector->length(self);

Defining a new Type.

In the header...

module Vec2;

data {
  var type;
  float x, y;
} Vec2Data;

/** Vec2_New(var self, float x, float y); */
var Vec2_New(var self, va_list* args);
var Vec2_Delete(var self);
void Vec2_Assign(var self, var obj);
var Vec2_Copy(var self);

var Vec2_Eq(var self, var obj);

float Vec2_Dot(var self, var obj);
float Vec2_Length(var self);

instance(Vec2, New) = { sizeof(Vec2Data), Vec2_New, Vec2_Delete };
instance(Vec2, Assign) = { Vec2_Assign };
instance(Vec2, Copy) = { Vec2_Copy };
instance(Vec2, Eq) = { Vec2_Eq };
instance(Vec2, Vector) = { Vec2_Dot, Vec2_Length };

And in the object file...

var Vec2 = methods {
  method(Vec2, New),
  method(Vec2, Assign),
  method(Vec2, Copy),
  method(Vec2, Eq),
  method(Vec2, Vector),

var Vec2_New(var self, va_list* args) {
  Vec2Data* v = cast(self, Vec2);
  v->x = va_arg(*args, double);
  v->y = va_arg(*args, double);
  return self;

var Vec2_Delete(var self) {
  return self;

void Vec2_Assign(var self, var obj) {
  Vec2Data* v1 = cast(self, Vec2);
  Vec2Data* v2 = cast(obj, Vec2);
  v1->x = v2->x;
  v1->y = v2->y;

var Vec2_Copy(var self) {
  Vec2Data* v = cast(self, Vec2);
  return new(Vec2, v->x, v->y);

var Vec2_Eq(var self, var obj) {
  Vec2Data* v1 = cast(self, Vec2);
  Vec2Data* v2 = cast(obj, Vec2);
  return (var)(v1->x is v2->x and v1->y is v2->y);

float Vec2_Dot(var self, var obj) {
  Vec2Data* v1 = cast(self, Vec2);
  Vec2Data* v2 = cast(obj, Vec2);
  return (v1->x * v2->x + v2->y * v2->y);

float Vec2_Length(var self) {
  Vec2Data* v = cast(self, Vec2);
  return sqrt(v->x * v->x + v->y * v->y);

More More Examples

Some functional tools using "lambda" statements.

#include "C+.h"

int main(int argc, char** argv) {
  ** Basic method of lambda construction .
  ** Must be at the statement level.
  ** Cannot be used in/as expression.
  lambda(hello_name, args) {
    /* Input is a list of arguments */
    var name = cast(at(args, 0), String);
    printf("Hello %s!\n", as_str(name));
    /* Always must return */
    return None;
  /* Functions called with "call" */
  call(hello_name, $(String, "Bob"));
  /* Higher order functions supported */
  var names = new(List, 3, $(String, "Dan"), $(String, "Robert"), $(String, "Chris"));
  map(names, hello_name);
  /* Here is an example with two arguments */
  lambda(add_print, args) {
    int fst = as_long(cast(at(args, 0), Int));
    int snd = as_long(cast(at(args, 1), Int));
    printf("%i + %i = %i\n", fst, snd, fst+snd);
    return None;
  ** Notice arguments to "call" in curried form.
  ** Use "call_with" for uncurried calling.
  call(add_print, $(Int, 10), $(Int, 21));
  /* Partially applied Function, neat! */
  lambda_partial(add_partial, add_print, $(Int, 5));
  call(add_partial, $(Int, 1));
  ** We can use normal c-functions too.
  ** If they have all argument types as "var".
  ** Then they can be uncurried.
  var Welcome_Pair(var fst, var snd) {
    printf("Hello %s and %s!\n", as_str(fst), as_str(snd));
    return None;
  lambda_uncurry(welcome_uncurried, Welcome_Pair, 2);
  call(welcome_uncurried, $(String, "John"), $(String, "David"));
  return 0;

There are more examples in the demos folder as well as a large amount of reference material under tests and via the source code. Native class definitions (such as for New, Eq, Ord) can be found in Prelude+.h


The first thing that probably comes into your head viewing the above code is var. This is a typedef of void* and is used via convention in C+ code to allow for overloaded functions. As you can see in the example code type checking/hinting can be done at runtime via the cast function.

This allows for a form of poor-man's duck-typing. If an object looks (or sounds) like it has a length, then you are more than free to use len upon it. One can test if a type implements a certain class with the function type_implements(type, Class). Calling a function on an object which does not implement the appropriate classes will throw a runtime error.

Another thing that may have jumped to your mind in the examples is new, delete and the $ symbol. These are ways to allocate memory for objects. new and delete are used for heap objects and call constructors/destructors. $ is used to allocate objects on the stack. It doesn't call a constructor or destructor, but will initialize the corresponding types data structure with the arguments provided.

Other than these things there is not much surprising in the code which cannot be explained via syntactic sugar.

  • is, not, and, or -> ==, !, &&, ||
  • elif -> else if
  • module -> extern var
  • class -> typedef struct
  • data -> typedef struct

finally the instance and methods macros are helpers for defining Type objects statically. Because Type objects are somewhat complicated the syntax is very awkward otherwise.


To build the just run make

To build the tests run make test

To build the examples run make examples

Returning Values

Having every data object as a reference makes returning data objects with manual memory management annoying. To return a full data object one has to allocate it on the heap and expect the user to delete it once they are done with it. There are a couple alternatives to this when such an approach is unacceptable:

  • Embrace destructive operations - Although not as nice semantically as duplicative ones, consistency is the key issue. How often is it we mean not to act destructively anyway? Assume destructive functions by default, return void.

  • Use output parameters - Use arguments intended for output, passed into a function. Use the assign function or otherwise to assign them data.

Questions & Contributions

Questions, Contributions and Feedback are more than welcome.

You can e-mail me at:


Or get in contact via the IRC channel at #c+ on Freenode