It's a small (<200 LOC) simple cache implementation. What makes it curious is how it approaches item expiration. Usually in-memory cache libraries either scan entire key space from time to time or maintain priority queue to kick out oldest items. This library handles item expiration differently: on each new item addition it performs fixed small number of eviction attempts against keys selected randomly. This way it is able to maintain stable high ratio of valid elements in cache.
What's more important is that validity of item is decided by user-provided function. That allows to bring in your own notion of item validity: item age, frequency of the item use, internal state of shared object and so on. E.g. I use it, among other things, to evict per-user instances of "golang.com/x/time/rate".Limiter as soon as token bucket recovered to initial state and limiter has no shared lock being held on it.
0 comments
—
Show HN: Secache – Sampling Eviction Cache · HackerTrans
It's a small (<200 LOC) simple cache implementation. What makes it curious is how it approaches item expiration. Usually in-memory cache libraries either scan entire key space from time to time or maintain priority queue to kick out oldest items. This library handles item expiration differently: on each new item addition it performs fixed small number of eviction attempts against keys selected randomly. This way it is able to maintain stable high ratio of valid elements in cache.
What's more important is that validity of item is decided by user-provided function. That allows to bring in your own notion of item validity: item age, frequency of the item use, internal state of shared object and so on. E.g. I use it, among other things, to evict per-user instances of "golang.com/x/time/rate".Limiter as soon as token bucket recovered to initial state and limiter has no shared lock being held on it.