Data model
Metalang99 provides four different types of collections, as well as a means for defining and dealing with compound data types. (Type checking is not performed due to performance reasons.)
Variadics
Representation: x, y, z, ...
.
Documentation: variadics.h
.
Variadics are a comma-separated list of preprocessor tokens. Most of the time, a user will provide variadic arguments to your macro so that you could manipulate them as you want:
#include <metalang99.h>
#include <assert.h>
#define ASSERT_FOR_EACH(...) \
do { \
ML99_EVAL(ML99_variadicsForEach( \
ML99_compose(v(ML99_semicoloned), ML99_reify(v(assert))), \
v(__VA_ARGS__))) \
} while (0)
int main(void) {
ASSERT_FOR_EACH(123 == 123, 2 + 2 == 4, "foo"[1] == 'o');
}
Here, ASSERT_FOR_EACH
accepts variadic arguments, which is communicated by its signature. Internally, it calls ML99_variadicsForEach
to iterate through each argument, resulting in a number of assert(...);
statements instead of one assert(a && b && ...);
(which makes debugging easier because failed assertions will not be collapsed with each other).
Tuple
Representation: (x, y, z, ...)
.
Documentation: tuple.h
.
A tuple is formed by putting variadic arguments into parentheses. Besides a collection-like usage, tuples can simulate C structures; the common pattern is to define a tuple constructor (rect
below) and accessors of the corresponding fields (rectWidth
and rectHeight
):
// Computes the area of a rectangle.
#include <metalang99.h>
#define rect(width, height) ML99_tuple(width, height)
#define rectWidth ML99_tupleGet(0)
#define rectHeight ML99_tupleGet(1)
#define rectArea(rect) ML99_mul(rectWidth(rect), rectHeight(rect))
/*
* 15
* +------------------------------+
* | |
* | |
* | | 7
* | |
* | |
* +------------------------------+
*/
#define RECTANGLE rect(v(15), v(7))
ML99_ASSERT_EQ(rectArea(RECTANGLE), v(15 * 7));
int main(void) {}
In type theory, tuples are known as product types.
Choice
Representation: (tag, ...)
.
Documentation: choice.h
.
A choice type encodes a set of alternatives; each alternative can be constructed by a corresponding value constructor and be inspected by the means of pattern matching:
// Sums all nodes of a binary tree, recursively.
#include <metalang99.h>
#define leaf(x) ML99_choice(v(leaf), x)
#define node(lhs, data, rhs) ML99_choice(v(node), lhs, data, rhs)
#define sumTree(tree) ML99_match(tree, v(sumTree_))
#define sumTree_leaf_IMPL(x) v(x)
#define sumTree_node_IMPL(lhs, data, rhs) ML99_add3(sumTree(v(lhs)), v(data), sumTree(v(rhs)))
/*
* 4
* / \
* / \
* / \
* 2 6
* / \ / \
* 1 3 5 7
*/
#define TREE node(node(leaf(v(1)), v(2), leaf(v(3))), v(4), node(leaf(v(5)), v(6), leaf(v(7))))
ML99_ASSERT_EQ(sumTree(TREE), v(28));
int main(void) {}
In type theory, choice types are usually referred to as sum types.
Cons-list
Representation: choice type.
Documentation: list.h
.
A cons-list is represented as a choice type with two alternatives: an empty list ML99_nil()
and a list constructor ML99_cons(x, xs)
. Cons-list is the most powerful collection:
#include <metalang99.h>
// 3, 3, 3, 3, 3
static int five_threes[] = {
ML99_LIST_EVAL_COMMA_SEP(ML99_listReplicate(v(5), v(3))),
};
// 5, 4, 3, 2, 1
static int from_5_to_1[] = {
ML99_LIST_EVAL_COMMA_SEP(ML99_listReverse(ML99_list(v(1, 2, 3, 4, 5)))),
};
// 9, 2, 5
static int lesser_than_10[] = {
ML99_LIST_EVAL_COMMA_SEP(
ML99_listFilter(ML99_appl(v(ML99_greater), v(10)), ML99_list(v(9, 2, 11, 13, 5)))),
};
Although cons-lists provide many more functions than any other collection, they are also more time and space-consuming. If you can deal with a native representation directly (such as variadics, tuples, or sequences), you should do so. Most of the time, a user will provide variadic arguments to your macro, and the only appropriate operation will be for-each loop (such as ML99_variadicsForEach
).
However, if you still miss a certain function for native collections, you can always convert your representation to a cons-list via ML99_list
, ML99_listFromSeq
, etc.
Sequence
Representation: (x)(y)(z) ...
.
Documentation: seq.h
.
A sequence is yet another fast native collection. A perfect example of using sequences is Interface99, which allows us to define a software interface like this:
#define Shape_IFACE \
vfunc( int, perim, const VSelf) \
vfunc(void, scale, VSelf, int factor)
interface(Shape);
With vfunc
being defined as follows (simplified):
#define vfunc(ret_ty, name, ...) (ret_ty, name, __VA_ARGS__)
Thus, a number of vfunc
invocations forms a complete sequence. Later, Interface99 works with this sequence via ML99_seqForEach
and a few other helpful macros.
Last updated
Was this helpful?