Struct Predictor

pub struct Predictor<const N_SHARDS: usize> { /* private fields */ }

Cacheability Predictor

Remembers previously uncacheable assets. Allows bypassing cache / cache lock early based on historical precedent.

NOTE: to simply avoid caching requests with certain characteristics, add checks in request_cache_filter to avoid enabling cache in the first place. The predictor’s bypass mechanism handles cases where the request looks cacheable but its previous responses suggest otherwise. The request could be cacheable in the future.

Implementations

impl<const N_SHARDS: usize> Predictor<N_SHARDS>

where [LruShard<()>; N_SHARDS]: Default,

pub fn new( shard_capacity: usize, skip_custom_reasons_fn: Option<CustomReasonPredicate>, ) -> Predictor<N_SHARDS>

Create a new Predictor with N_SHARDS * shard_capacity total capacity for uncacheable cache keys.

• shard_capacity: defines number of keys remembered as uncacheable per LRU shard.
• skip_custom_reasons_fn: an optional predicate used in mark_uncacheable that can customize which Custom NoCacheReasons ought to be remembered as uncacheable. If the predicate returns true, then the predictor will skip remembering the current cache key as uncacheable (and avoid bypassing cache on the next request).

Trait Implementations

impl<const N_SHARDS: usize> CacheablePredictor for Predictor<N_SHARDS>

where [LruShard<()>; N_SHARDS]: Default,

fn cacheable_prediction(&self, key: &CacheKey) -> bool

Return true if likely cacheable, false if likely not.

fn mark_cacheable(&self, key: &CacheKey) -> bool

Mark cacheable to allow next request to cache. Returns false if the key was already marked cacheable.

fn mark_uncacheable( &self, key: &CacheKey, reason: NoCacheReason, ) -> Option<bool>

Mark uncacheable to actively bypass cache on the next request. May skip marking on certain NoCacheReasons. Returns None if we skipped marking uncacheable. Returns Some(false) if the key was already marked uncacheable.