A thread-safe object pool with automatic return and attach/detach semantics
The goal of an object pool is to reuse expensive to allocate objects or frequently allocated objects
toml
[dependencies]
object-pool = "0.5"
rust
extern crate object_pool;
The general pool creation looks like this
rust
let pool: Pool<T> = Pool::new(capacity, || T::new());
Example pool with 32 Vec<u8> with capacity of 4096
rust
let pool: Pool<Vec<u8>> = Pool::new(32, || Vec::with_capacity(4096));
Basic usage for pulling from the pool
rust
let pool: Pool<Vec<u8>> = Pool::new(32, || Vec::with_capacity(4096));
let mut reusable_buff = pool.pull().unwrap(); // returns None when the pool is saturated
reusable_buff.clear(); // clear the buff before using
some_file.read_to_end(reusable_buff);
// reusable_buff is automatically returned to the pool when it goes out of scope
Pull from pool and detach()
rust
let pool: Pool<Vec<u8>> = Pool::new(32, || Vec::with_capacity(4096));
let mut reusable_buff = pool.pull().unwrap(); // returns None when the pool is saturated
reusable_buff.clear(); // clear the buff before using
let (pool, reusable_buff) = reusable_buff.detach();
let mut s = String::from(reusable_buff);
s.push_str("hello, world!");
pool.attach(s.into_bytes()); // reattach the buffer before reusable goes out of scope
// reusable_buff is automatically returned to the pool when it goes out of scope
You simply wrap the pool in a [std::sync::Arc]
rust
let pool: Arc<Pool<T>> = Arc::new(Pool::new(cap, || T::new()));
Objects in the pool are not automatically reset, they are returned but NOT reset
You may want to call object.reset() or object.clear()
or any other equivalent for the object that you are using, after pulling from the pool
Check out the [docs] for more examples
The benchmarks compare alloc() vs pool.pull() vs pool.detach() vs lifeguard vs WIP SyncPool.
Check out the [results]
For those who don't like graphs, here's the [raw output]