Easy to use, modular, header only, macro based, generic and type-safe Data Structures in C
- Installation
- Contributing
- Usage
- Features
- Project Structure
- Available Collections
- Other Features
- Design Decisions
No installation is required. The entire library is made of header files and can be directly included into your project.
There is a lot to be done. You can check the TODO
file in the root of the repository or the issues in the github page. Also, tests and documentation are in constant development.
Below is a minimal example that makes use of some core functionalities. And a lot of explanation too!
// The C Macro Collections takes advantage of headers that can be included
// multiple times. You can check the contents of src/cmc/list.h and there are no
// include guards, meaning they can be included multiple times and this is what
// generates the code.
// Before including it though we need to define a few basic macros. You can
// check the documentation to see which ones are required.
#define V int // list value data type
#define PFX intl // list function's prefix
#define SNAME int_list // list struct name (struct int_list)
#define CMC_EXT_STR // Enables an extra functionality (STR) of the list
#include "cmc/list.h"
// This header file includes a lot of functions that can be used as methods to
// basic data types. In this case, cmc_i32_str is used
#include "cmc/utl/futils.h"
int main(void)
{
// Our list type is defined by SNAME and all functions are prefixed by PFX
// (PFX can also be thought as the function's namespace). Also, nodes,
// entries, iterators and other structures are prefixed by SNAME. So PFX is
// for functions and SNAME is for structs, where the main one is simply
// 'struct SNAME'.
// To initialize a list we need to pass in an initial capacity and something
// that is called a function table for the V type.
// This function table is a struct with methods that will extract some
// (sometimes) necessary behaviour from your custom data type. Things like
// hash, comparison and printing. A list doesn't require any of these
// functions. That is, the CORE module doesn't use any of them.
// But since we are using the STR module, we will need to define the 'str'
// function. Luckily, for the 'int' type, the library already provides such
// function (cmc_i32_str), provided by the cmc/utl/futils.h header file.
struct int_list *list = intl_new(32, &(struct int_list_fval){ .str = cmc_i32_str, NULL });
// Check if the list was successfully initialized. It could fail if the
// initial capacity is too big or if 'struct int_list_fval *' is NULL,
// because every data structure must have a valid function table.
if (!list)
return 1;
// Add some items to the list. The CMC data structures are all dynamically
// sized. So there can be as many items as you want as long as you have
// enough memory.
for (int i = 0; i < 100; i++)
{
// Try to add an element to the list. If it fails you can check what
// caused it by getting its flag.
if (!intl_push_back(list, i))
{
enum cmc_flags flag = intl_flag(list);
// Use cmc_flags_to_str (global variable) to map the enum error to a
// string.
fprintf(stderr, "%s : push_back failed\n", cmc_flags_to_str[flag]);
}
}
// Now we will use the STR module, the _print() function. This is where the
// '.str' from the function table comes into play. If we haven't defined it,
// the program would've crashed. We also need to define to which file we
// will be printing the list's content. In this case, the terminal. Also,
// some strings need to be defined. They will be printed: before all
// elements, between each one, and after all elements. This is very useful
// for debugging.
intl_print(list, stdout, "[ ", ", ", " ]\n");
// Free all of its resources
intl_free(list);
}
Now all you have to do is to compile the source code with -I /path/to/library/src
. There is no required installation or pre-compilation.
The C Macro Collections library is organized into many other sub-libraries. The following table is a quick overview.
Library | Name | Description |
---|---|---|
CMC | Macro Collections | The main library with dynamically-sized collections |
COR | Core | Core functionalities used by more than one collection, usually from different sub-libraries |
EXT | Extensions | Extension to collections |
SAC | Statically Allocated Collections | Collections with a fixed sized array that don't use any heap allocation |
TSC | Thread-Safe Collections | Collections that allow multiple operations from multiple threads |
UTL | Utilities | General utilities |
- benchmarks - Where all benchmarks are hosted
- docs - A folder hosting the generated documentation by mdBook
- documentation - The markdowns used by mdBook to generate the website
- examples - Examples using the C Macro Collections Library
- src - All headers part of the C Macro Collections Library
- tests - Where all tests are hosted
The following table is an overview of all the currently available or upcoming data structures:
These are some features within the library that are implemented by all collections.
All collections come with a two-way iterator. You can go back and forwards in constant time and access elements in constant time.
All collections have a cmc_alloc_node
which provides pointers to the four dynamic memory allocation functions in C: malloc
, calloc
, realloc
and free
. These pointers can be customized for each individual collection created or a default can be used, as specified in cmc_alloc_node_default
.
Every function that operates on a collection can be separated in 5 different types. Create, Read, Update, Delete and (an extra one besides CRUD) Resize. You can define one callback function for each operation. Check out the documentation to see when each callback function is called.
Functions table is a struct
of function pointers containing 'methods' for a custom data type. Some methods are optional and others are needed in order for a collection to operate. They are:
A comparator function is used in sorted collections or when an equality is being checked like when trying to find a certain element in a list. It is responsible for taking two arguments of the same data type and comparing them. The return value is an int
with the following definitions:
- Return
1
if the first argument is greater than the second; - Return
0
if the first argument equals the second; - Return
-1
if the first argument is less than the second.
A copy function is used when a collection is being copied. It can be used to make a deep copy of of your custom data type. It must take a single parameter and return a new copy of that same data type. If this function is absent (NULL
) the data type will be copied by assignment (for pointers this is a shallow copy).
A string function is responsible for taking a FILE
pointer and a custom data type and outputting the string representation of that data returning a bool
indication success or failure. It is useful for debugging.
The free function is called when a collection is cleared (all elements removed) or freed (all elements removed and freed from memory) and it is responsible for completely freeing all resources that are usually acquired by your data type.
This function receives a custom data type as parameter and returns a size_t
hash of that data. Used in hashtables.
A priority function works much like the comparator function except that it compares the priority between two elements. It is used in collections whose structure is based on the priority of elements and not in their general comparison.
- Return
1
if the first argument has a greater priority than the second; - Return
0
if the first argument has the same priority as second; - Return
-1
if the first argument has a lower priority than the second.
The following table shows which functions are required, optional or never used for each Collection:
Collection | CMP | CPY | STR | FREE | HASH | PRI |
---|---|---|---|---|---|---|
BitSet | ||||||
Deque | ||||||
HashMap | ||||||
HashBidiMap | ||||||
HashMultiMap | ||||||
HashMultiSet | ||||||
HashSet | ||||||
Heap | ||||||
IntervalHeap | ||||||
List | ||||||
LinkedList | ||||||
Queue | ||||||
SortedList | ||||||
Stack | ||||||
TreeMap | ||||||
TreeSet |
Color | Label |
---|---|
Required for basic functionality. | |
Required for specific functions. | |
Required for non-core specific functions. | |
Optional. | |
Not Used. |
Currently all collections need to be allocated on the heap. Iterators have both options but it is encouraged to allocate them on the stack since they don't require dynamic memory.
Yes, you can use a Deque as a Queue or a List as a Stack without any major cost, but the idea is to have the least amount of code to fulfill the needs of a collection.
Take for example the Stack. It is simple, small and doesn't have many functions. If you generate a List to be used (only) as a Stack (which is one of the bulkiest collections) you'll end up with a lot of code generated and compiled for nothing.
The Deque versus Queue situation is a little less problematic, but again, you need to be careful when generating a lot of code as compilation times might go up to 15 seconds even with modern ultra-fast compilers.
Another example is using a HashMap/TreeMap as a HashSet/TreeSet (with a dummy value that is never used), but I just think that this is a bad thing to do and you would be wasting some memory. Also, the sets generate a lot of code related to set theory, whereas maps don't.
You can use them as Stacks, Queues and Deques, but with modern memory hierarchy models, array-based data structures have a significantly faster runtime due to caching, so I didn't bother to have specific implementations of those aforementioned collections.
Modifying a collection will possibly invalidate all iterators currently initialized by it. Currently, the only collection that allows this is the LinkedList (using the node-based functions, not the iterator).
CMC_NO_ALLOC
- disables"cor/alloc.h"
, useful for statically allocated collections (SAC)CMC_NO_IMPORTS
- disables all imports from the standard library at"cor/core.h"