oxistore_cache/lib.rs
1#![forbid(unsafe_code)]
2#![warn(missing_docs)]
3
4//! `oxistore-cache` — Pure-Rust LRU, ARC, LFU, and W-TinyLFU eviction primitives.
5//!
6//! This crate provides four cache implementations behind a unified [`Cache`]
7//! trait:
8//!
9//! - [`LruCache`] — classic Least-Recently-Used cache backed by a
10//! `hashlink::LinkedHashMap`. O(1) amortised operations.
11//!
12//! - [`ArcCache`] — Adaptive Replacement Cache (Megiddo & Modha, FAST'03),
13//! which balances recency and frequency by maintaining four lists (T1, T2,
14//! B1, B2) and an adaptive target `p`. ARC is scan-resistant.
15//!
16//! - [`LfuCache`] — Least-Frequently-Used cache with O(1) operations using
17//! the Shah, Mitra & Matani (2010) algorithm.
18//!
19//! - [`WTinyLfuCache`] — Window TinyLFU with Count-Min Sketch frequency
20//! estimation and a doorkeeper bloom filter, providing near-optimal hit rates
21//! on skewed (Zipfian) workloads.
22//!
23//! All cache implementations support per-entry TTL (time-to-live) via the
24//! [`Cache::put_with_ttl`] method. Expiry is checked lazily on access.
25//!
26//! # Example
27//!
28//! ```rust
29//! use oxistore_cache::{LruCache, ArcCache, LfuCache, WTinyLfuCache, Cache};
30//!
31//! let mut lru = LruCache::new(3);
32//! lru.put(1u32, "a");
33//! lru.put(2u32, "b");
34//! lru.put(3u32, "c");
35//! lru.put(4u32, "d"); // evicts 1 (LRU)
36//! assert!(lru.get(&1u32).is_none());
37//! assert_eq!(lru.get(&2u32), Some(&"b"));
38//!
39//! let mut arc = ArcCache::new(3);
40//! arc.put(1u32, "a");
41//! arc.put(2u32, "b");
42//! arc.put(3u32, "c");
43//!
44//! let mut lfu = LfuCache::new(3);
45//! lfu.put(1u32, "a");
46//! lfu.put(2u32, "b");
47//! lfu.put(3u32, "c");
48//!
49//! let mut wtlfu = WTinyLfuCache::new(3);
50//! wtlfu.put(1u32, "a");
51//! wtlfu.put(2u32, "b");
52//! wtlfu.put(3u32, "c");
53//! ```
54
55/// Adaptive Replacement Cache (ARC) — balances recency and frequency.
56pub mod arc;
57/// Bounded-memory cache wrapper — enforces a hard byte-budget cap.
58pub mod bounded;
59/// Cache builder — ergonomic constructor for all cache policies.
60pub mod builder;
61/// Least-Frequently-Used (LFU) cache — O(1) frequency-based eviction.
62pub mod lfu;
63/// Least-Recently-Used (LRU) cache — evicts the least recently accessed entry.
64pub mod lru;
65/// Sharded concurrent cache — N LRU shards behind a Mutex for low contention.
66pub mod sharded;
67/// Count-Min Sketch and Doorkeeper bloom filter for W-TinyLFU.
68pub mod sketch;
69/// Cache hit/miss statistics tracking.
70pub mod stats;
71/// Thread-safe cache wrapper — wraps any `Cache` impl behind a `Mutex`.
72pub mod sync;
73/// Window TinyLFU — state-of-the-art admission policy with CMS frequency estimator.
74pub mod tinylfu;
75/// Write-through and write-back cache adapters.
76pub mod write_adapter;
77
78/// Caching wrapper for `BlobStore` backends (see `oxistore_blob::BlobStore`).
79#[cfg(feature = "blob")]
80pub mod blob_cache;
81
82/// Caching adapter for hot Parquet row groups from `oxistore-columnar`.
83#[cfg(feature = "columnar")]
84pub mod columnar_cache;
85
86/// SQL query-result and prepared-plan caches backed by `oxisql-core` types.
87#[cfg(feature = "sql")]
88pub mod sql_cache;
89
90pub use arc::ArcCache;
91pub use bounded::BoundedCache;
92pub use builder::{CacheBuilder, CachePolicy};
93pub use lfu::LfuCache;
94pub use lru::LruCache;
95pub use sharded::ShardedCache;
96pub use stats::{CacheStats, StatsCache};
97pub use sync::SyncCache;
98pub use tinylfu::WTinyLfuCache;
99pub use write_adapter::{CacheableKvStore, WriteBackCache, WriteThroughCache};
100
101#[cfg(feature = "blob")]
102pub use blob_cache::BlobCache;
103
104#[cfg(feature = "columnar")]
105pub use columnar_cache::ColumnarRowGroupCache;
106
107#[cfg(feature = "sql")]
108pub use sql_cache::{CachedQueryRunner, SqlPlanCache, SqlQueryCache};
109
110/// A cache entry that optionally expires at a given instant.
111///
112/// Used internally by all cache implementations to support per-entry TTL.
113/// Callers interact with the value type `V` via the [`Cache`] trait methods;
114/// `CacheEntry` is exposed for advanced use cases (e.g. custom wrappers).
115pub struct CacheEntry<V> {
116 /// The stored value.
117 pub value: V,
118 /// Optional expiry time. `None` means the entry never expires.
119 pub expires_at: Option<std::time::Instant>,
120}
121
122impl<V> CacheEntry<V> {
123 /// Create a non-expiring entry.
124 #[must_use]
125 pub fn new(value: V) -> Self {
126 CacheEntry {
127 value,
128 expires_at: None,
129 }
130 }
131
132 /// Create an entry that expires after `ttl` from now.
133 #[must_use]
134 pub fn with_ttl(value: V, ttl: std::time::Duration) -> Self {
135 CacheEntry {
136 value,
137 expires_at: Some(std::time::Instant::now() + ttl),
138 }
139 }
140
141 /// Return `true` if this entry has expired (i.e. its deadline has passed).
142 #[must_use]
143 pub fn is_expired(&self) -> bool {
144 match self.expires_at {
145 Some(deadline) => std::time::Instant::now() >= deadline,
146 None => false,
147 }
148 }
149}
150
151/// Unified interface for key-value caches with bounded capacity.
152///
153/// All four implementations ([`LruCache`], [`ArcCache`], [`LfuCache`],
154/// [`WTinyLfuCache`]) implement this trait, allowing callers to write generic
155/// code against the cache interface.
156pub trait Cache<K, V> {
157 /// Look up `key`, returning a reference to the value if present.
158 ///
159 /// Implementations update internal bookkeeping (e.g. promoting the entry
160 /// to MRU or incrementing frequency counts) as a side effect.
161 ///
162 /// If the entry exists but has expired (TTL), it is removed and `None` is
163 /// returned without updating recency or frequency.
164 fn get(&mut self, key: &K) -> Option<&V>;
165
166 /// Insert or update `key` -> `value` without a TTL.
167 ///
168 /// If inserting a new key would exceed the cache's capacity, the
169 /// implementation evicts one entry (per its policy) and returns the
170 /// evicted value. On a key update, the old value is not returned.
171 fn put(&mut self, key: K, value: V) -> Option<V>;
172
173 /// Insert or update `key` -> `value` with a time-to-live.
174 ///
175 /// After `ttl` has elapsed, any access to `key` will treat it as a miss
176 /// and remove the entry lazily.
177 ///
178 /// Concrete implementations in this crate override this to actually store
179 /// the expiry. External impls that don't need TTL may rely on the default
180 /// which falls back to a plain `put` (no expiry).
181 fn put_with_ttl(&mut self, key: K, value: V, ttl: std::time::Duration) -> Option<V> {
182 let _ = ttl;
183 self.put(key, value)
184 }
185
186 /// Return the number of live entries currently in the cache.
187 fn len(&self) -> usize;
188
189 /// Return the maximum number of entries the cache can hold.
190 fn cap(&self) -> usize;
191
192 /// Return `true` if the cache holds no entries.
193 fn is_empty(&self) -> bool {
194 self.len() == 0
195 }
196
197 /// Remove the entry associated with `key`, returning its value if present.
198 ///
199 /// Unlike eviction, this is an explicit removal requested by the caller.
200 fn remove(&mut self, key: &K) -> Option<V>;
201
202 /// Remove all entries from the cache.
203 fn clear(&mut self);
204
205 /// Look up `key` without updating access metadata (no promotion).
206 ///
207 /// If the entry has expired (TTL), it is removed and `None` is returned.
208 /// Returns `None` if the key is not present or has expired.
209 fn peek(&self, key: &K) -> Option<&V>;
210
211 /// Return `true` if `key` is present in the cache (without promotion).
212 ///
213 /// Expired entries are treated as absent.
214 fn contains_key(&self, key: &K) -> bool;
215
216 /// Dynamically resize the cache capacity.
217 ///
218 /// If `new_cap` is smaller than the current length, excess entries are
219 /// evicted according to the cache's eviction policy.
220 fn resize(&mut self, new_cap: usize);
221
222 /// Return `&V` for `key`, inserting `default()` if the key is absent.
223 ///
224 /// The closure `default` is called at most once. If the key is already
225 /// present the closure is never invoked.
226 ///
227 /// Implementations may override this for efficiency (e.g. to avoid a
228 /// second hash lookup). The default implementation uses [`Cache::peek`]
229 /// after ensuring the key exists.
230 fn get_or_insert(&mut self, key: K, default: impl FnOnce() -> V) -> &V
231 where
232 K: Clone,
233 {
234 if !self.contains_key(&key) {
235 let v = default();
236 self.put(key.clone(), v);
237 }
238 // peek is &self and won't panic since we just ensured the key exists.
239 self.peek(&key).expect("key was just inserted")
240 }
241
242 /// Return all live (non-expired) values currently stored in the cache.
243 ///
244 /// The default implementation returns an empty `Vec`. Concrete
245 /// implementations override this to return actual values.
246 fn values(&self) -> Vec<&V> {
247 Vec::new()
248 }
249
250 /// Pre-populate the cache with `(key, value)` pairs from an iterator.
251 ///
252 /// Each pair is inserted via [`Cache::put`], respecting the cache's
253 /// eviction policy. Pairs that exceed capacity cause earlier entries to
254 /// be evicted according to the policy.
255 fn warm(&mut self, iter: impl IntoIterator<Item = (K, V)>) {
256 for (k, v) in iter {
257 self.put(k, v);
258 }
259 }
260}
261
262// ── Async helper ─────────────────────────────────────────────────────────────
263
264/// Look up `key` in `cache`, returning the cached value if present.
265///
266/// If the key is absent, the async closure `loader` is awaited to produce
267/// a value, which is then inserted into the cache and returned.
268///
269/// The closure is invoked **at most once** per call. If the key is already
270/// present it is never called.
271///
272/// # Example
273///
274/// ```rust
275/// use oxistore_cache::{LruCache, Cache, get_or_insert_async};
276/// use std::sync::Mutex;
277///
278/// # async fn example() {
279/// let cache = Mutex::new(LruCache::<u32, String>::new(4));
280/// let val = get_or_insert_async(
281/// &cache,
282/// 42u32,
283/// || async { "computed".to_string() },
284/// ).await;
285/// assert_eq!(val, "computed");
286/// # }
287/// ```
288pub async fn get_or_insert_async<K, V, C, F, Fut>(
289 cache: &std::sync::Mutex<C>,
290 key: K,
291 loader: F,
292) -> V
293where
294 K: Eq + std::hash::Hash + Clone,
295 V: Clone,
296 C: Cache<K, V>,
297 F: FnOnce() -> Fut,
298 Fut: std::future::Future<Output = V>,
299{
300 // Fast path: key already in cache.
301 {
302 let mut guard = cache.lock().expect("cache mutex poisoned");
303 if let Some(v) = guard.get(&key) {
304 return v.clone();
305 }
306 }
307 // Slow path: key absent — await the loader outside the lock.
308 let value = loader().await;
309 {
310 let mut guard = cache.lock().expect("cache mutex poisoned");
311 // Double-check in case another task inserted the key while we waited.
312 if let Some(existing) = guard.peek(&key) {
313 return existing.clone();
314 }
315 guard.put(key, value.clone());
316 }
317 value
318}
319
320/// Look up `key` in a `tokio::sync::Mutex`-wrapped cache asynchronously.
321///
322/// Identical semantics to [`get_or_insert_async`] but uses an async
323/// `tokio::sync::Mutex` instead of `std::sync::Mutex`.
324///
325/// This variant is suitable when the cache is shared across `tokio` tasks
326/// that run on a multi-threaded executor and cannot use a synchronous lock.
327///
328/// Requires the `async-helpers` feature flag.
329#[cfg(feature = "async-helpers")]
330pub async fn get_or_insert_async_tokio<K, V, C, F, Fut>(
331 cache: &tokio::sync::Mutex<C>,
332 key: K,
333 loader: F,
334) -> V
335where
336 K: Eq + std::hash::Hash + Clone,
337 V: Clone,
338 C: Cache<K, V>,
339 F: FnOnce() -> Fut,
340 Fut: std::future::Future<Output = V>,
341{
342 // Fast path: key already in cache (async lock, no blocking).
343 {
344 let mut guard = cache.lock().await;
345 if let Some(v) = guard.get(&key) {
346 return v.clone();
347 }
348 }
349 // Slow path: await the loader without holding the lock.
350 let value = loader().await;
351 {
352 let mut guard = cache.lock().await;
353 if let Some(existing) = guard.peek(&key) {
354 return existing.clone();
355 }
356 guard.put(key, value.clone());
357 }
358 value
359}