C++ Cache implementation
This project implements a simple thread-safe cache with several page replacement policies:
- Least Recently Used
- First-In/First-Out
- Least Frequently Used
More about cache algorithms and policy you could read on Wikipedia
Usage
Using this library is simple. It is necessary to include header with the cache implementation (cache.hpp
file)
and appropriate header with the cache policy if it is needed. If not then the non-special algorithm will be used (it
removes the last element which key is the last in the internal container).
Currently there is only three of them:
fifo_cache_policy.hpp
lfu_cache_policy.hpp
lru_cache_policy.hpp
Example for the LRU policy:
#include <string>
#include "cache.hpp"
#include "lru_cache_policy.hpp"
// alias for easy class typing
template <typename Key, typename Value>
using lru_cache_t = typename caches::fixed_sized_cache<Key, Value, caches::LRUCachePolicy>;
void foo(...) {
constexpr std::size_t CACHE_SIZE = 256;
lru_cache_t<std::string, int> cache(CACHE_SIZE);
cache.Put("Hello", 1);
cache.Put("world", 2);
std::cout << cache.Get("Hello") << cache.Get("world") << '\n';
// "12"
}
Custom hashmap usage
You can use a custom hashmap implementation for the caches::fixed_sized_cache
class which has the same interface
std::unordered_map
has.
For example, you can declare LRU cache type like that:
template <typename Key, typename Value>
using lru_cache_t = typename caches::fixed_sized_cache<Key, Value, caches::LRUCachePolicy,
phmap::node_hash_map<Key, Value>>;
// ...
lru_cache_t<std::string, std::size_t> cache{16};
cache.Put("Hello", 1);
std::cout << cache.Get("Hello") << '\n';
See test
implementation which uses parallel-hashmap
.
Requirements
The only requirement is a compatible C++11 compiler.
This project was tested in the environments listed below:
- MinGW64 (MSYS2 project)
- Clang 3.8.0
- GCC 5.3.0
- MSVC (VS 2015)
- FreeBSD
- Clang 3.4.1
If you have any issues with the library building, let me know please.
Contributing
Please fork this repository and contribute back using pull requests. Features can be requested using issues. All code, comments, and critiques are greatly appreciated.