lru

Scalable LRU caches
Manual

Semantics cleanup.

Breaking:

  • find drops ?promote and never changes the ordering.

  • add drops ?trim and never drops bindings.

  • size -> weight

  • items -> size

  • unadd -> pop

  • F.S.fold and F.S.iter iterate in LRU order.

  • F.M.fold and F.M.iter drop ?dir and always iterate in LRU order.

Other:

  • add F.S.fold_k and F.S.iter_k

To fix client code:

  • replace find ~promote:false with find;

  • replace find and find ~promote:true with find and promote;

  • replace add ~trim:false with add;

  • replace add and add ~trim:true with add and trim;

  • s/size/weight/g, s/items/size/g, s/unadd/pop/g;

  • audit uses of fold and iter for order-sensitivity.

v0.2.0 2017-03-31

Breaking changes:

  • resize no longer drops bindings if the new size pushes the queue over capacity.

  • of_list has simpler semantics; dropped the cap parameter.

Other changes:

  • Replace Lru.M.cache with more general Lru.memo.

  • Queues with 0 initial capacity are legal.

  • Add trim to shrink a queue to its capacity, as queues are no longer guaranteed to
    have size smaller than capacity.

  • find gets the promote parameter, allowing queries that do not change the order.

  • add gets the trim parameter, allowing insertions that do not drop old entries.

v0.1.1 2016-11-28

  • Fix missing dep on psq in META.

v0.1.0 2016-11-22

First release.