Java的类是自定义的引用类型,是对职责相关的行为与数据的一种封装,用以表现一种业务领域或者技术领域的概念。在不同的场景,类包含的成员可能有所不同,大体可以分为如下五类:
- 数据类:可以视为是持有数据的容器,类的成员只包含了字段,以及与字段有关的get/set方法
- 实体类:既包含了体现状态的字段,又包含了操作这些状态的方法
- 服务类:只有方法(行为)没有字段(状态),可以理解为提供内聚职责的服务
- 函数类:如果定义的公开方法只有唯一一个,可以理解为它封装的其实是一个函数,通常用匿名类或者Lambda表示
- 工具类:只包含一系列静态方法,通常不支持对该类型的实例化
数据类
在Presto框架中定义的ClientSession可以认为是这样一种数据类。除了构造函数外,它只定义了字段与对应的get()方法(实际上,在框架的源代码中,在ClientSession类中还定义了一系列静态工厂方法,但本质上说,ClientSession还是一个数据类),用以持有客户端Session所必须的数据:
public class ClientSession { private final URI server; private final String use; private final String source; private final String clientInfo; private final String catalog; private final String schema; private final TimeZoneKey timeZone; private final Locale locale; private final Map<String, String> properties; private final Map<String, String> preparedStatements; private final String transactionId; private final boolean debug; private final Duration clientRequestTimeout; public ClientSession( URI server, String user, String source, String clientInfo, String catalog, String schema, String timeZoneId, Locale locale, Map<String, String> properties, String transactionId, boolean debug, Duration clientRequestTimeout) { this(server, user, source, clientInfo, catalog, schema, timeZoneId, locale, properties, emptyMap(), transactionId, debug, clientRequestTimeout); } public ClientSession( URI server, String user, String source, String clientInfo, String catalog, String schema, String timeZoneId, Locale locale, Map<String, String> properties, Map<String, String> preparedStatements, String transactionId, boolean debug, Duration clientRequestTimeout) { this.server = requireNonNull(server, "server is null"); this.user = user; this.source = source; this.clientInfo = clientInfo; this.catalog = catalog; this.schema = schema; this.locale = locale; this.timeZone = TimeZoneKey.getTimeZoneKey(timeZoneId); this.transactionId = transactionId; this.debug = debug; this.properties = ImmutableMap.copyOf(requireNonNull(properties, "properties is null")); this.preparedStatements = ImmutableMap.copyOf(requireNonNull(preparedStatements, "preparedStatements is null")); this.clientRequestTimeout = clientRequestTimeout; // verify the properties are valid CharsetEncoder charsetEncoder = US_ASCII.newEncoder(); for (Entry<String, String> entry : properties.entrySet()) { checkArgument(!entry.getKey().isEmpty(), "Session property name is empty"); checkArgument(entry.getKey().indexOf('=') < 0, "Session property name must not contain '=': %s", entry.getKey()); checkArgument(charsetEncoder.canEncode(entry.getKey()), "Session property name is not US_ASCII: %s", entry.getKey()); checkArgument(charsetEncoder.canEncode(entry.getValue()), "Session property value is not US_ASCII: %s", entry.getValue()); } } public URI getServer() { return server; } public String getUser() { return user; } public String getSource() { return source; } public String getClientInfo() { return clientInfo; } public String getCatalog() { return catalog; } public String getSchema() { return schema; } public TimeZoneKey getTimeZone() { return timeZone; } public Locale getLocale() { return locale; } public Map<String, String> getProperties() { return properties; } public Map<String, String> getPreparedStatements() { return preparedStatements; } public String getTransactionId() { return transactionId; } public boolean isDebug() { return debug; } public Duration getClientRequestTimeout() { return clientRequestTimeout; } @Override public String toString() { return toStringHelper(this) .add("server", server) .add("user", user) .add("clientInfo", clientInfo) .add("catalog", catalog) .add("schema", schema) .add("timeZone", timeZone) .add("locale", locale) .add("properties", properties) .add("transactionId", transactionId) .add("debug", debug) .toString(); } }
这样包含数据或状态的对象通常会作为参数在方法调用之间传递,体现了诸如配置、视图模型、服务传输数据、协议数据等概念。除此之外,我们应尽量避免定义这样的对象去体现某种业务概念,因为基于“信息专家”模式,好的面向对象设计应该是将数据与操作这些数据的行为封装在一起。
实体类
这是最为常见的一种类定义,也是符合面向对象设计原则的,前提是定义的类必须是高内聚的,原则上应该满足单一职责原则。例如JDK定义的Vector展现了一种数据结构,因而它持有的字段与方法应该仅仅与队列操作与状态有关:
public class Vector<E> extends AbstractList<E> implements List<E>, RandomAccess, Cloneable, java.io.Serializable { protected Object[] elementData; protected int elementCount; protected int capacityIncrement; public Vector(int initialCapacity, int capacityIncrement) { super(); if (initialCapacity < 0) throw new IllegalArgumentException("Illegal Capacity: "+ initialCapacity); this.elementData = new Object[initialCapacity]; this.capacityIncrement = capacityIncrement; } public Vector(int initialCapacity) { this(initialCapacity, 0); } public synchronized void setSize(int newSize) { modCount++; if (newSize > elementCount) { ensureCapacityHelper(newSize); } else { for (int i = newSize ; i < elementCount ; i++) { elementData[i] = null; } } elementCount = newSize; } public synchronized int size() { return elementCount; } public synchronized boolean isEmpty() { return elementCount == 0; } public boolean contains(Object o) { return indexOf(o, 0) >= 0; } public synchronized E firstElement() { if (elementCount == 0) { throw new NoSuchElementException(); } return elementData(0); } public synchronized void insertElementAt(E obj, int index) { modCount++; if (index > elementCount) { throw new ArrayIndexOutOfBoundsException(index + " > " + elementCount); } ensureCapacityHelper(elementCount + 1); System.arraycopy(elementData, index, elementData, index + 1, elementCount - index); elementData[index] = obj; elementCount++; } public synchronized void addElement(E obj) { modCount++; ensureCapacityHelper(elementCount + 1); elementData[elementCount++] = obj; } public synchronized boolean removeElement(Object obj) { modCount++; int i = indexOf(obj); if (i >= 0) { removeElementAt(i); return true; } return false; } public synchronized void removeAllElements() { modCount++; // Let gc do its work for (int i = 0; i < elementCount; i++) elementData[i] = null; elementCount = 0; } }
如下类的定义则体现了一种业务概念,方法changePriceTo()实际上表现的是一种业务规则,而它要操作的数据就是Product类自身持有的字段sellingPrice:
public class Product extends Entity<Identity> { private final List<Option> options; private Price sellingPrice; private Price retailPrice; public Product(Identity id, Price sellingPrice, Price retailPrice) { super(id); this.sellingPrice = sellingPrice; if (!sellingPriceMatches(retailPrice) { throw new PricesNotInTheSameCurrencyException("Selling and retail price must be in the same currency"); } this.retailPrice = retailPrice; options = new List<Option>(); } public void changePriceTo(Price newPrice) { if (!sellingPriceMatches(newPrice)) { throw new PricesNotInTheSameCurrencyException("You cannot change the price of this product to a different currency"); } sellingPrice = newPrice; } public Price savings() { Price savings = retailPrice.minus(sellingPrice); if (savings.isGreaterThanZero()) return savings; else return new Price(0m, sellingPrice.currency); } private bool sellingPriceMatches(Price retailPrice) { return sellingPrice.sameCurrency(retailPrice); } public void add(Option option) { if (!this.contains(option)) options.Add(option); else throw new ProductOptionAddedNotUniqueException(string.Format("This product already has the option {0}", option.ToString())); } public bool contains(Option option) { return options.Contains(option); } }
服务类
只有方法没有状态的类定义是对行为的封装,行为的实现要么是通过操作内部封装的不可变私有数据,要么是通过操作传入的参数对象实现对状态的修改。由于参数传入的状态与服务类自身没有任何关系,因此这样的类通常也被视为无状态的类。以下代码是针对升级激活包的验证服务:
public class PreActivePackageValidator { public long validatePreActivePackage(ActiveManifest activeManifest) { validateSamePackageType(activeManifest); validateNoTempPackage(activeManifest); validateNoPackageRunning(activeManifest); validateAllPackagesBeenDownloaded(activeManifest); validateNoFatherPackageBakStatus(activeManifest); validatePackageNum(activeManifest); } private void validateSamePackageType(ActiveManifest activeManifest) { int packakeType = activeManifest.getPackageType(); for (UpagrdePackage pkg : activeManifest.getPackages()) { if (packageType != pkg.getPackageType()) { throw new PackagePreActiveException("pre active exist different type package"); } } } }
服务类还可以操作外部资源,例如读取文件、访问数据库、与第三方服务通信等。例如airlift框架定义的ConfigurationLoader类,就提供加载配置文件内容的服务:
public class ConfigurationLoader { public Map<String, String> loadProperties() throws IOException { Map<String, String> result = new TreeMap<>(); String configFile = System.getProperty("config"); if (configFile != null) { result.putAll(loadPropertiesFrom(configFile)); } result.putAll(getSystemProperties()); return ImmutableSortedMap.copyOf(result); } public Map<String, String> loadPropertiesFrom(String path) throws IOException { Properties properties = new Properties(); try (Reader reader = new FileReader(new File(path))) { properties.load(reader); } return fromProperties(properties); } public Map<String, String> getSystemProperties() { return fromProperties(System.getProperties()); } }
函数类
可以将函数类理解为设计一个类,它仅仅实现了一个接口,且该接口只定义一个方法。使用时,我们会基于依赖倒置原则(DIP)从接口的角度使用这个类。为了重用的目的,这个类可以单独被定义,也可能体现为匿名类,或者Java 8中的Lambda表达式。
单独类形式
例如,在Presto中定义了PagesIndexComparator接口,提供了比较方法以用于支持对页面索引的排序。接口的定义为:
public interface PagesIndexComparator { int compareTo(PagesIndex pagesIndex, int leftPosition, int rightPosition); }
Presto定义了该接口的实现类SimplePagesIndexComparator,该类就是一个函数类:
public class SimplePagesIndexComparator implements PagesIndexComparator { private final List<Integer> sortChannels; private final List<SortOrder> sortOrders; private final List<Type> sortTypes; public SimplePagesIndexComparator(List<Type> sortTypes, List<Integer> sortChannels, List<SortOrder> sortOrders) { this.sortTypes = ImmutableList.copyOf(requireNonNull(sortTypes, "sortTypes is null")); this.sortChannels = ImmutableList.copyOf(requireNonNull(sortChannels, "sortChannels is null")); this.sortOrders = ImmutableList.copyOf(requireNonNull(sortOrders, "sortOrders is null")); } @Override public int compareTo(PagesIndex pagesIndex, int leftPosition, int rightPosition) { long leftPageAddress = pagesIndex.getValueAddresses().getLong(leftPosition); int leftBlockIndex = decodeSliceIndex(leftPageAddress); int leftBlockPosition = decodePosition(leftPageAddress); long rightPageAddress = pagesIndex.getValueAddresses().getLong(rightPosition); int rightBlockIndex = decodeSliceIndex(rightPageAddress); int rightBlockPosition = decodePosition(rightPageAddress); for (int i = 0; i < sortChannels.size(); i++) { int sortChannel = sortChannels.get(i); Block leftBlock = pagesIndex.getChannel(sortChannel).get(leftBlockIndex); Block rightBlock = pagesIndex.getChannel(sortChannel).get(rightBlockIndex); SortOrder sortOrder = sortOrders.get(i); int compare = sortOrder.compareBlockValue(sortTypes.get(i), leftBlock, leftBlockPosition, rightBlock, rightBlockPosition); if (compare != 0) { return compare; } } return 0; } }
我们看到SimplePagesIndexComparator类的逻辑相对比较复杂,构造函数也需要传入三个参数:List<Type> sortTypes,List<Integer> sortChannels和List<SortOrder> sortOrders。虽然从接口的角度看,其实代表的是compare的语义,但由于逻辑复杂,而且需要传入三个对象帮助对PagesIndex进行比较,因而不可能实现为匿名类或者Lambda表达式。在Presto中,对它的使用为:
public class PagesIndexOrdering { private final PagesIndexComparator comparator; public PagesIndexOrdering(PagesIndexComparator comparator) { this.comparator = requireNonNull(comparator, "comparator is null"); } public PagesIndexComparator getComparator() { return comparator; } /** * Returns the index of the median of the three positions. */ private int median3(PagesIndex pagesIndex, int a, int b, int c) { int ab = comparator.compareTo(pagesIndex, a, b); int ac = comparator.compareTo(pagesIndex, a, c); int bc = comparator.compareTo(pagesIndex, b, c); return (ab < 0 ? (bc < 0 ? b : ac < 0 ? c : a) : (bc > 0 ? b : ac > 0 ? c : a)); } }
匿名类形式
同样在该框架下定义的IntComparator接口,它的实现就完全不同了。首先是该接口的定义:
public interface IntComparator { /** Compares the given primitive types. * * @see java.util.Comparator * @return A positive integer, zero, or a negative integer if the first * argument is greater than, equal to, or smaller than, respectively, the * second one. */ int compare(int k1, int k2); } 在针对整型数据提供排序功能时,用到了IntComparator接口: public final class IntBigArray { public void sort(int from, int to, IntComparator comparator) { IntBigArrays.quickSort(array, from, to, comparator); } }
但由于提供整型数据的比较逻辑相对简单,在Presto中并没有定义显式的函数类,而是使用了Lambda表达式:
groupIds.sort(0, groupByHash.getGroupCount(), (leftGroupId, rightGroupId) -> Long.compare(groupByHash.getRawHash(leftGroupId), groupByHash.getRawHash(rightGroupId)));
这里的Lambda表达式其实也可以理解为是一个函数类。
函数重用形式
还有一种特殊的函数类,它的定义形式与后面介绍的工具类非常相似,同样是定义了一组静态方法,但它的目的不是提供工具或辅助功能,而是将其视为函数成为被重用的单元。这时,需要用到Java 8提供的方法引用(method reference)语法。例如我们要对List<Apple>集合进行过滤,过滤条件分别为颜色与重量,这时可以在Apple类中定义两个静态方法:
public class Apple { public static boolean isGreenApple(Apple apple) { return "green".equals(apple.getColor()); } public static boolean isHeavyApple(Apple apple) { return apple.getWeight() > 150; } }
这两个方法实际上满足函数接口Predicate<Apple>的定义,因此可以在filter方法中传入这两个方法的引用:
public List<Apple> filter(Predicate<Apple> predicate) { ArrayList<Apple> result = new ArrayList<>(); for (Apple apple : apples) { if (predicate.test(apple)) { result.add(apple); } } return result; } public List<Apple> filterGreenApples() { return filter(Apple::isGreenApple); } public List<Apple> filterHeavyApples() { return filter(Apple::isHeavyApple); }
此时Apple类可以认为是一个函数类,但准确地说法是一系列可以被重用的函数的容器。与工具类不同的是,这些函数并不是被直接调用,本质上讲,其实是作为“高阶函数”被传递给其他方法而被重用。虽然说实例方法也可以采用这种方式而被重用,但静态方法的调用会更加简单。
工具类
在许多项目或开源项目中,随处可见工具类的身影。无需实例化的特性使得我们使用工具类的方式时变得非常的便利,也不需要考虑状态的维护。然而越是方便,我们越是要警惕工具类的陷阱——设计出臃肿庞大无所不能的上帝工具类。工具类仍然要遵循高内聚的原则,只有强相关的职责才能放到同一个工具类中。
在定义工具类时,通常有三类命名范式:
- 名词复数形式:工具类其实就是一系列工具方法的容器,当我们要针对某种类型(或对象)提供工具方法时,可以直接将工具类命名为该类型的复数形式,例如操作Collection的工具类可以命名为Collections,操作Object的工具类可以命名为Objects,而与前置条件有关的工具类则被命名为Preconditions。
- 以Util为后缀:这体现了工具(Utility)的语义,当我们在类名中看到Util后缀时,就可以直观地了解到这是一个工具类。例如ArrayUtil类是针对数组的工具类,DatabaseUtil是针对数据库操作的工具类,UuidUtil是针对Uuid的工具类。
- 以Helper为后缀:这种命名相对较少,但许多框架也采用这种命名方式来体现“辅助类”的含义。例如在Druid框架中,就定义了JobHelper、GroupByQueryHelper等辅助类。
工具类是无需实例化的,因此在定义工具类时,尽可能将其声明为final类,并为其定义私有的构造函数。例如Guava框架提供的Preconditions工具类:
public final class Preconditions { private Preconditions() { } public static void checkArgument(boolean expression) { if(!expression) { throw new IllegalArgumentException(); } } //other util methods }