#string-interner #symbols


A string, path, and buffer internment crate with no dependencies

4 releases

0.2.1 Jun 25, 2023
0.2.0 Jun 2, 2023
0.1.1 Feb 8, 2023
0.1.0 Feb 7, 2023

#246 in Encoding

Download history 159/week @ 2023-11-06 175/week @ 2023-11-13 72/week @ 2023-11-20 109/week @ 2023-11-27 54/week @ 2023-12-04 76/week @ 2023-12-11 236/week @ 2023-12-18 124/week @ 2023-12-25 247/week @ 2024-01-01 195/week @ 2024-01-08 109/week @ 2024-01-15 91/week @ 2024-01-22 175/week @ 2024-01-29 188/week @ 2024-02-05 284/week @ 2024-02-12 377/week @ 2024-02-19

1,026 downloads per month
Used in 2 crates




interner forbids unsafe code crate version Live Build Status HTML Coverage Report for main branch Documentation

An interning crate for Rust with no dependencies and no unsafe code (#![forbid(unsafe_code)]). Most existing interning crates only offer interning strings. This crate allows interning paths and byte buffers as well.

How this crate works

This crate uses a HashSet and a Vec to store its entries. When a value is looked up, if it cannot be found in the HashSet, it is assigned a slot in the Vec. Future lookups will return a clone to the Arc-wrapped data.

When the final reference to a Pooled<T> value is dropped, the pool's HashMap has the value removed and the Vec's slot is made available for re-use.

This crate does not do sub-value interning. Each value is stored independently with its own allocation.

Hash for Pooled Values

The Pooled<T> type implements Hash by hashing its internal unique id rather than using T::hash(). This allows pooled objects to be used as efficent keys in hash maps and sets.

Because T::hash() is different than Pooled<T>::hash(), Pooled<T> does not implement Borrow<T>. This prevents using an &str to look up a value in a HashMap.

Another important caveat of using Pooled<T> values as keys in a hash-based collection is that all Pooled<T> values must be from the same pool. Using values from separate pools will cause lookups to be unable to find the contained keys in most situations. When using this unsupported flow, incorrect matches will never occur because Pooled<T>::eq() is implemented to verify the values are from the same pool, otherwise the underlying values are compared.

Globally Interned Strings

use interner::global::{GlobalString, GlobalPool};

static STRINGS: GlobalPool<String> = GlobalPool::new();

let my_string = STRINGS.get("hello");
let other_copy = STRINGS.get(String::from("hello"));

// Both `my_string` and `other_copy` are pointing to the same underlying string.
assert!(GlobalString::ptr_eq(&my_string, &other_copy));

Interned Strings from a StringPool

use interner::shared::{StringPool, SharedString};

let pool = StringPool::default();
let my_string = pool.get("hello");
let other_copy = pool.get(String::from("hello"));

// Both `my_string` and `other_copy` are pointing to the same underlying string.
assert!(SharedString::ptr_eq(&my_string, &other_copy));

Globally Interned Paths

use std::path::{Path, PathBuf};
use interner::global::{GlobalPath, GlobalPool};

static PATHS: GlobalPool<PathBuf> = GlobalPool::new();

let my_path = PATHS.get(Path::new("hello"));
let other_copy = PATHS.get(PathBuf::from("hello"));

// Both `my_path` and `other_copy` are pointing to the same underlying path.
assert!(GlobalPath::ptr_eq(&my_path, &other_copy));

Interned Paths from a PathPool

use std::path::{Path, PathBuf};
use interner::shared::{PathPool, SharedPath};

let pool = PathPool::default();
let my_string = pool.get(Path::new("hello"));
let other_copy = pool.get(PathBuf::from("hello"));

// Both `my_path` and `other_copy` are pointing to the same underlying path.
assert!(SharedPath::ptr_eq(&my_string, &other_copy));

Globally Interned Byte Buffers

use interner::global::{GlobalBuffer, GlobalPool};

static BUFFERS: GlobalPool<Vec<u8>> = GlobalPool::new();

let my_buffer = BUFFERS.get(&b"hello"[..]);
let other_copy = BUFFERS.get(b"hello".to_vec());

// Both `my_buffer` and `other_copy` are pointing to the same underlying path.
assert!(GlobalBuffer::ptr_eq(&my_buffer, &other_copy));

Interned Byte Buffers from a BufferPool

use interner::shared::{BufferPool, SharedBuffer};

let pool = BufferPool::default();
let my_buffer = pool.get(&b"hello"[..]);
let other_copy = pool.get(b"hello".to_vec());

// Both `my_path` and `other_copy` are pointing to the same underlying path.
assert!(SharedBuffer::ptr_eq(&my_buffer, &other_copy));

Open-source Licenses

This project, like all projects from Khonsu Labs, are open-source. This repository is available under the MIT License or the Apache License 2.0.

To learn more about contributing, please see CONTRIBUTING.md.

No runtime deps