Start of Tutorial > Start of Trail |
Search
Feedback Form |
Implementations are the data objects used to store collections, which implement the interfaces described in the previous lesson. The sections that follow describe three kinds of implementations:The general-purpose implementations are summarized in the table below.
- General-purpose implementations: the most commonly used implementations, designed for everyday use.
- Special-purpose implementations: designed for use in special situations. They display nonstandard performance characteristics, usage restrictions, or behavior.
- Concurrent implementations: designed to support high concurrency, typically at the expense of single-threaded performance. These implementations are part of the package java.util.concurrent.
- Wrapper implementations: used in combination with other types of implementations (often the general-purpose implementations) to provide added or restricted functionality.
- Convenience implementations: mini-implementations typically made available via static factory methods, that provide convenient, efficient alternatives to the general-purpose implementations for special collections (such as singleton sets).
- Abstract implementations: skeletal implementations that facilitate the construction of custom implementations, which is described in the section Custom Implementations. An advanced topic, it's not particularly difficult, but relatively few people will need to do it
Interfaces General-purpose Implementations Hash Table Resizable array Tree Linked List Hash Table + Linked List Set HashSet TreeSet LinkedHashSet List ArrayList LinkedList Queue   Map HashMap TreeMap LinkedHashMap As you can see from the table, the Collections Framework provides several general-purpose implementations of the Set, List , and Map interfaces. In each case, one implementation HashSet, ArrayList, and HashMap is clearly the one to use for most applications, all other things being equal. Note that the SortedSet and the SortedMap interfaces do not have rows in the table. Each of those interfaces has one implementation ( TreeSet and TreeMap) and is listed in the Set and the Map rows. There are two general-purpose Queue implementations: LinkedList (which is also a List implementation) and PriorityQueue (which is omitted from the table). These two implementations provide very different semantics: LinkedList provides first-in-first-out (FIFO) semantics, while PriorityQueue orders its elements according to their values.
The general-purpose implementations each provide all optional operations contained in their interfaces. All permit null elements, keys, and values. None are synchronized (thread-safe). All have fail-fast iterators, which detect illegal concurrent modification during iteration and fail quickly and cleanly rather than risking arbitrary, nondeterministic behavior at an undetermined time in the future. All are Serializable, and all support a public clone method.
The fact that these implementations are unsynchronized represents a break with the past: The legacy collections Vector and Hashtable are synchronized. The present approach was taken because collections are frequently used when the synchronization is of no benefit. Such uses include single-threaded use, read-only use, and use as part of a larger data object that does its own synchronization. In general, it is good API design practice not to make users pay for a feature they don't use. Further, unnecessary synchronization can result in deadlock under certain circumstances.
If you need thread-safe collections, the synchronization wrappers, described in the Wrapper Implementations, allow any collection to be transformed into a synchronized collection. Thus, synchronization is optional for the general-purpose implementations, whereas it is mandatory for the legacy implementations. Moreover, the java.util.concurrent package provides concurrent implementations of the BlockingQueue interface, which extends Queue and the ConcurrentHashMap interface, which extends Map. These implementations offer much higher concurrency than mere synchronized implementations.
As a rule, you should be thinking about the interfaces, not the implementations. That is why there are no programming examples in this section. For the most part, the choice of implementation affects only performance. The preferred style, as mentioned in the Interfaces, is to choose an implementation when a collection is created and to immediately assign the new collection to a variable of the corresponding interface type (or to pass the collection to a method expecting an argument of the interface type). In this way, the program does not become dependent on any added methods in a given implementation, leaving the programmer free to change implementations anytime it is warranted by performance concerns or behavioral details.
The implementations are briefly discussed here. The performance of the implementations is described with such words as constant-time, log, linear, n log(n), and quadratic. These words refer to the asymptotic upper bound on the time complexity of performing the operation. All this is quite a mouthful, and it doesn't matter much if you don't know what it means. If you're interested, refer to any good algorithms textbook. One thing to keep in mind is that this sort of performance metric has its limitations. Sometimes, the nominally slower implementation may be faster. When in doubt, measure the performance!
Start of Tutorial > Start of Trail |
Search
Feedback Form |
Copyright 1995-2005 Sun Microsystems, Inc. All rights reserved.