A hash table supporting full concurrency of retrievals, adjustable expected concurrency for updates, and a maximum capacity to bound the map by. This implementation differs from ConcurrentHashMap in that it maintains a page replacement algorithm that is used to evict an entry when the map has exceeded its capacity. Unlike the Java Collections Framework, this map does not have a publicly visible constructor and instances are created through a Builder.

An entry is evicted from the map when the entry size exceeds its maximum capacity threshold.

An EvictionListener may be supplied for notification when an entry is evicted from the map. This listener is invoked on a caller's thread and will not block other threads from operating on the map. An implementation should be aware that the caller's thread will not expect long execution times or failures as a side effect of the listener being notified. Execution safety and a fast turn around time can be achieved by performing the operation asynchronously, such as by submitting a task to an java.util.concurrent.ExecutorService.

The concurrency level determines the number of threads that can concurrently modify the table. Using a significantly higher or lower value than needed can waste space or lead to thread contention, but an estimate within an order of magnitude of the ideal value does not usually have a noticeable impact. Because placement in hash tables is essentially random, the actual concurrency will vary.

This class and its views and iterators implement all of the optional methods of the Map and Iterator interfaces.

Like java.util.Hashtable but unlike HashMap, this class does not allow null to be used as a key or value. Unlike java.util.LinkedHashMap, this class does not provide predictable iteration order. A snapshot of the keys and entries may be obtained in ascending and descending order of retention.