VLA for C++: dynarray
A header-only library, providing C99 VLA-like class for C++. The VLA extension provide by compiler is not required.
Depencencies
C++ 14 or 17
C++ Standard Library
Purpose of this project
C Language provided VLA since C99. This is my favourite feature in C. A similar library was almost became a part of C++14 standard: N3662 but it's got removed before release year of C++14.
A proposal P0785R0 was proposed after C++14, but it's still not a part of C++ standard.
The most obvious feature of VLA is, a contiguous memory space will be allocated for a dynamic defined array, including multi-dimensional array. The size of this array will not (and cannot) be changed after definition.
vector
can also do the same thing for multi-dimensional array, if use it with custom allocator to allocate contiguous memory space. But if the requirement is ‘I don't want to change the size after definition’, almost nothing we can do for it. The only thing we can do is, write a note or comment to inform everyone ‘Do not use push_back
or emplace_back
’.
Since std:dynarray
was removed before C++14 release, I've modified std::dynarray
and extended it to provide multi-dimensional array (nested-array) support. The behaviour of modified dynarray
is designed between VLA and vector
. It will allocates a contiguous memory space as VLA does, and uses C++ iterator as vector
does.
File Description
Doxyfile
Create documents with doxygen
.
dynarray.hpp
Proterotype version, use the same structure (class) all the time. The size is largest.
Requires C++17.
vla/dynarray.hpp
Template specialised version. Medium size.
Requires C++14.
vla/dynarray_lite.hpp
Lite version, does not guaranteed to provide contiguous memory spaces for multi-dimensional array.
Requires C++17.
vla/dynarray_mini.hpp
Using std::unique_ptr<[]>
inside the dynarray
, does not guaranteed to provide contiguous memory spaces for multi-dimensional array. Custom allocator cannot be used in this version.
Requires C++17.
Version comparison
Version Description | Proterotype version | Partial template specialisation | Lite Version | Mini Version |
---|---|---|---|---|
File1 | dynarray.hpp | vla/dynarray.hpp | vla/dynarray_lite.hpp | vla/dynarray_mini.hpp |
C++ Version | C++17 | C++14 | C++17 | C++17 |
sizeof dynarray (Outermost2)3 | 48 bytes | 48 bytes | 24 bytes | 16 bytes |
sizeof dynarray (middle layer per node2)3 | 48 bytes | 48 bytes | 24 bytes | 16 bytes |
sizeof dynarray (Innermost per node2)3 | 48 bytes | 32 bytes | 24 bytes | 16 bytes |
sizeof dynarray (one-dimensional array)3 | 48 bytes | 32 bytes | 24 bytes | 16 bytes |
contiguous memory spaces for multi-dimensional array | Yes | Yes | No | No |
custom allocator can be used | Yes | Yes | Yes | No |
1 Use one of the .hpp
file only. Please don't use them all at the same time.
2 Multi-dimensional array
3 Aligned
How to use
#include <iostream>
#include "dynarray.hpp"
int main()
{
int x = 100, y = 200;
int value = 5;
vla::dynarray<vla::dynarray<int>> vla_array(x, y, value);
std::cout << vla_array[8][16] << std::endl; // 5
vla_array[8][16] = 20;
std::cout << vla_array[8][16] << std::endl; // 20
}
Create a one-dimensional array
- Create an array with variable
int count = 100;
vla::dynarray<int> vla_array(count);
Equivalent to
int count = 100;
int vla_array[count];
memset(vla_array, 0, sizeof vla_array);
- Create an array with initial value
int count = 100;
vla::dynarray<int> vla_array(count, 256); // initial value 256
Equivalent to
int count = 100;
int vla_array[count];
memset(vla_array, 256, sizeof vla_array);
- Create a zero-size array
vla::dynarray<int> vla_array;
or
vla::dynarray<int> vla_array(0);
- Initialise current dynarray or replace current dynarray with another dynarray
vla::dynarray<int> vla_array(vla::dynarray<int>(100, 256));
vla::dynarray<int> vla_array_a(100);
vla::dynarray<int> vla_array_b(vla_array_a);
vla::dynarray<int> vla_array_a(100);
vla::dynarray<int> vla_array_b;
vla_array_b = vla_array_a;
- Initialization list
vla::dynarray<int> vla_array = {2, 4, 8, 16};
vla::dynarray<int> vla_array;
vla_array = {2, 4, 8, 16};
- Iterator
int raw_array[100] = {};
vla::dynarray<int> vla_array(std::begin(raw_array), std::end(raw_array));
vla::dynarray<int> vla_array_a(100);
vla::dynarray<int> vla_array_b(vla_array_a.begin() + 20, vla_array_a.end());
Create a 2D array
- Create an array with variable
int x = 100, y = 200;
vla::dynarray<vla::dynarray<int>> vla_array(x, y);
Equivalent to
int x = 100, y = 200;
int vla_array[x][y];
memset(vla_array, 0, sizeof vla_array);
- Create an array with initial value
int x = 100, y = 200;
vla::dynarray<vla::dynarray<int>> vla_array(x, y, 256);
Equivalent to
int x = 100, y = 200;
int vla_array[x][y];
memset(vla_array, 256, sizeof vla_array);
- Create a zero-size array
As long as the number of parameters is less than the actual dimension, or one of the size is set as zero, a zero-size array will be created.
vla::dynarray<vla::dynarray<int>> vla_array;
or
vla::dynarray<vla::dynarray<int>> vla_array(0);
or
vla::dynarray<vla::dynarray<int>> vla_array(30, 0);
or
vla::dynarray<vla::dynarray<int>> vla_array(0, 5);
- Initialise current dynarray or replace current dynarray with another dynarray
vla::dynarray<vla::dynarray<int>> vla_array(vla::dynarray<vla::dynarray<int>>(100, 200));
vla::dynarray<vla::dynarray<int>> vla_array_a(100, 300);
vla::dynarray<vla::dynarray<int>> vla_array_b(vla_array_a);
vla::dynarray<vla::dynarray<int>> vla_array_a(100, 200, 10);
vla::dynarray<vla::dynarray<int>> vla_array_b(100, 200);
vla_array_b = vla_array_a; // all elements of vla_array_b have value 10
-
Initialization list
- create 3 × 3 array
vla::dynarray<vla::dynarray<int>> array33 = { {1, 2, 3 }, {3, 2, 1}, {2, 4, 6} };
- create 3 × 3 array
vla::dynarray<vla::dynarray<int>> array33; array33 = { {1, 2, 3 }, {3, 2, 1}, {2, 4, 6} };
- create a variable-length array
vla::dynarray<vla::dynarray<int>> vla_array = { {10, 100, 1000}, {1, 3, 5}, {0, 3} };
In this example:
vla_array.size() == 3
vla_array[0].size() == 3
vla_array[1].size() == 3
vla_array[2].size() == 2
-
Iterator
The useage is similar with one-dimensional array. Examples are omitted here.
Create multi-dimensional array
The useage is similar with one-dimensional array and 2D array. Examples are omitted here.
Allow me to remind you again: as long as the number of parameters is less than the actual dimension, or one of the size is set as zero, a zero-size array will be created.
All of these are zero-size array:
vla::dynarray<vla::dynarray<vla::dynarray<int>>> vla_array;
vla::dynarray<vla::dynarray<vla::dynarray<int>>> vla_array_a(100);
vla::dynarray<vla::dynarray<vla::dynarray<int>>> vla_array_b(vla_array_a);
vla::dynarray<vla::dynarray<vla::dynarray<int>>> vla_array_c(100, 200);
Use a custom allocator
vla::dynarray
uses std::allocator
by default. You will need to use your own allocator if you want to let vla::dynarray
allocate memory on stack.
The usage of allocator in vla::dynarray
is slightly different with container of std.
Assume you have an allocator as below
template<typename T>
class your_allocator { /* ...... */ };
When we're using std container, allocator will be used like this: place your_allocator<T>
in the angle brackets.
your_allocator<int> my_alloc(/* sth */);
std::vector<int, your_allocator<int>> my_vec(100, my_alloc);
But vla::dynarray
is not the same as above. You should place your template name your_allocator
in the angle brackets.
your_allocator<int> my_alloc(/* sth */);
vla::dynarray<int, your_allocator> my_array(100, my_alloc);
It would be much more verbose for multi-dimensional array
your_allocator<int> my_alloc(/* sth */);
your_allocator<vla::dynarray<int, your_allocator>> my_alloc_2(/* sth */);
vla::dynarray<vla::dynarray<int, your_allocator>, your_allocator> my_array(200, my_alloc_2,
100, my_alloc);
vla::dynarray<vla::dynarray<int, your_allocator>, your_allocator> another_array(my_array, my_alloc_2, my_alloc);
You can also do this directly:
template<typename T>
class your_allocator { /* ...... */ };
vla::dynarray<int, your_allocator> my_array_1(200);
vla::dynarray<vla::dynarray<int, your_allocator>, your_allocator> my_array_2(200, 100);
vla::dynarray<vla::dynarray<int, your_allocator>, your_allocator> another_array(my_array_2);
Note: all of the allocators must come from the same source (template, class) or dynarray will not be compiled. Incorrect Example:
std::allocator<int> std_alloc(/* sth */);
your_allocator<vla::dynarray<int, std::allocator>> my_alloc_2(/* sth */);
// cannot compile
vla::dynarray<vla::dynarray<int, std::allocator>, your_allocator> my_array(200, my_alloc_2,
100, std_alloc);
operator=
Behaviour of Using operator=
on vla::dynarray
will only assign values to the left-side array. The size will not be changed.
- If one of the array is a zero-size array on either side of the equal sign,
operator=
will do nothing on any side.
vla::dynarray<int> vla_array;
vla::dynarray<int> vla_array_2(5, 10);
vla_array = vla_array_2; // do nothing
- If the size of right-side array is smaller than left-size's, then only some of the left-side elements will be filled.
- Example 1
vla::dynarray<vla::dynarray<int>> vla_array(6, 6);
vla::dynarray<vla::dynarray<int>> vla_array_2(3, 3, 5);
vla_array | [x][0] | [x][1] | [x][2] | [x][3] | [x][4] | [x][5] |
---|---|---|---|---|---|---|
[0][y] | 0 | 0 | 0 | 0 | 0 | 0 |
[1][y] | 0 | 0 | 0 | 0 | 0 | 0 |
[2][y] | 0 | 0 | 0 | 0 | 0 | 0 |
[3][y] | 0 | 0 | 0 | 0 | 0 | 0 |
[4][y] | 0 | 0 | 0 | 0 | 0 | 0 |
[5][y] | 0 | 0 | 0 | 0 | 0 | 0 |
vla_array_2 | [x][0] | [x][1] | [x][2] |
---|---|---|---|
[0][y] | 5 | 5 | 5 |
[1][y] | 5 | 5 | 5 |
[2][y] | 5 | 5 | 5 |
vla_array = vla_array_2;
vla_array | [x][0] | [x][1] | [x][2] | [x][3] | [x][4] | [x][5] |
---|---|---|---|---|---|---|
[0][y] | 5 | 5 | 5 | 0 | 0 | 0 |
[1][y] | 5 | 5 | 5 | 0 | 0 | 0 |
[2][y] | 5 | 5 | 5 | 0 | 0 | 0 |
[3][y] | 0 | 0 | 0 | 0 | 0 | 0 |
[4][y] | 0 | 0 | 0 | 0 | 0 | 0 |
[5][y] | 0 | 0 | 0 | 0 | 0 | 0 |
- Example 2
vla::dynarray<vla::dynarray<int>> vla_array(6, 6);
vla::dynarray<vla::dynarray<int>> vla_array_2(3, 3, 5);
vla_array | [x][0] | [x][1] | [x][2] | [x][3] | [x][4] | [x][5] |
---|---|---|---|---|---|---|
[0][y] | 0 | 0 | 0 | 0 | 0 | 0 |
[1][y] | 0 | 0 | 0 | 0 | 0 | 0 |
[2][y] | 0 | 0 | 0 | 0 | 0 | 0 |
[3][y] | 0 | 0 | 0 | 0 | 0 | 0 |
[4][y] | 0 | 0 | 0 | 0 | 0 | 0 |
[5][y] | 0 | 0 | 0 | 0 | 0 | 0 |
vla_array_2 | [x][0] | [x][1] | [x][2] |
---|---|---|---|
[0][y] | 5 | 5 | 5 |
[1][y] | 5 | 5 | 5 |
[2][y] | 5 | 5 | 5 |
vla_array[2] = vla_array_2[2];
vla_array | [x][0] | [x][1] | [x][2] | [x][3] | [x][4] | [x][5] |
---|---|---|---|---|---|---|
[0][y] | 0 | 0 | 0 | 0 | 0 | 0 |
[1][y] | 0 | 0 | 0 | 0 | 0 | 0 |
[2][y] | 5 | 5 | 5 | 0 | 0 | 0 |
[3][y] | 0 | 0 | 0 | 0 | 0 | 0 |
[4][y] | 0 | 0 | 0 | 0 | 0 | 0 |
[5][y] | 0 | 0 | 0 | 0 | 0 | 0 |
vla_array_2[0] = vla_array[0];
vla_array_2 | [x][0] | [x][1] | [x][2] |
---|---|---|---|
[0][y] | 0 | 0 | 0 |
[1][y] | 5 | 5 | 5 |
[2][y] | 5 | 5 | 5 |
Other APIs
-
at(n)
vla::dynarray<int> vla_array(5, 10); int number = vla_array.at(2);
-
[]
vla::dynarray<vla::dynarray<int>> vla_array(5, 5, 10); int number = vla_array[2][2];
-
front()
vla::dynarray<int> vla_array(5, 10); int number = vla_array.front();
-
back()
vla::dynarray<int> vla_array(5, 10); int number = vla_array.back();
-
data()
vla::dynarray<int> vla_array(5, 10); int *raw_array = vla_array.data();
data()
will return a first level pointer even if the array is multi-dimensional array. The pointer is pointing to a contiguous memory space.vla::dynarray<vla::dynarray<int>> vla_array(5, 5, 10); int *raw_array = vla_array.data();
-
get()
vla::dynarray<vla::dynarray<int>> vla_array(5, 5, 10); vla::dynarray<int> *raw_array = vla_array.get();
-
empty()
vla::dynarray<int> vla_array(5, 10); bool is_empty = vla_array.empty(); // is_empty == false
-
size()
vla::dynarray<int> vla_array(5, 10); std::size_t array_size = vla_array.size(); // array_size == 5
-
max_size()
Returns the maximum number of elements the container (dynarray) is able to hold.
vla::dynarray<int> vla_array(5, 10); std::size_t max_size = vla_array.max_size(); // std::numeric_limits<std::ptrdiff_t>::max()
-
swap()
Swap internal values only.
dynarray
itself keeps unchanged.vla::dynarray<vla::dynarray<int>> vla_array_a(6, 6, 1); vla::dynarray<vla::dynarray<int>> vla_array_b(3, 3, 5);
vla_array_a [x][0] [x][1] [x][2] [x][3] [x][4] [x][5] [0][y] 1 1 1 1 1 1 [1][y] 1 1 1 1 1 1 [2][y] 1 1 1 1 1 1 [3][y] 1 1 1 1 1 1 [4][y] 1 1 1 1 1 1 [5][y] 1 1 1 1 1 1 vla_array_b [x][0] [x][1] [x][2] [0][y] 5 5 5 [1][y] 5 5 5 [2][y] 5 5 5
vla_array_a.swap(vla_array_b);
vla_array_a [x][0] [x][1] [x][2] [x][3] [x][4] [x][5] [0][y] 5 5 5 1 1 1 [1][y] 5 5 5 1 1 1 [2][y] 5 5 5 1 1 1 [3][y] 1 1 1 1 1 1 [4][y] 1 1 1 1 1 1 [5][y] 1 1 1 1 1 1 vla_array_b [x][0] [x][1] [x][2] [0][y] 1 1 1 [1][y] 1 1 1 [2][y] 1 1 1
-
fill()
vla::dynarray<int> vla_array(100); vla_array.fill(256); // all elements have value 256
vla::dynarray<vla::dynarray<int>> vla_array(100, 100); vla_array.fill(256); // all elements in all dimension have value 256
Iterators
- begin()
- cbegin()
- end()
- cend()
- rbegin()
- crbegin()
- rend()
- crend()
Internal Design
Proterotype version
For a multilayer dynarray, allocates a contiguous memory space on the outermost layer first. The size is provided by user. And then allocate each level's management nodes inside. These dynarray
s have head/tail pointers pointing to the right place, according to the sizes and sequences.
Single-layer dynarray
is a simplified version of multiple dynarray.
vla/dynarray.hpp
vla/dynarray_lite.hpp
The most important line of the code
friend class dynarray<dynarray<T, _Allocator>, _Allocator>;
Extremely simple. No lying. No mystery.
Thanks for friend
, the outer layer can access any data of internal layer, including private members.
It's still safe. The users outside this library are still not able to use private members.
License
The code in this repository is licensed under the BSD-3-Clause License