这篇关于SPI的文章是在blogspot上面，被ZF给强奸了，大多数人访问不了，就贴在这里，比较好懂。

From ServiceLoader javadoc: A service is a well-known set of interfaces and classes. A service provider is a specific implementation of a service. The classes in a provider typically implement the interfaces and subclass the classes defined in the service itself.

Since JDK 6, a simple service-provider loading facility is implemented. As is suggested it is simple, you cannot understand that implementation as a complex system for implementing plugins for your application, but can help you in many situations where you want implementations of a service could be discovered by other module automatically.

What I really like about using JDK Services is that your services do not have any dependency to any class. Moreover for registering a new implementation of a service, you just have to put jar file into classpath and nothing more.

Now I will explain the service we are going to implement, and then I will show you how to code it.

We want to implement a system that depending on kind of structured input (comma-separated value, tab-separated value, ...) returns a String[] of each value; so for example you can receive input a,b,c,dor 1<tab>2<tab>3<tab>4 and the system should return an array with [a, b, c, d] or [1, 2, 3, 4].

So our system will have three Java projects.

One defining service contract (an interface) and, because of teaching purpose, a main class where internet media type, for example text/csv, is received with input data. Then using a factory class that I have created, it will ask which registered service can transform input to String[].

And two projects each one implementing a service following defined contract, one for comma-separated values and another one for tab-separated values.

Let's see the code:

Main project (reader) is composed by an interface, a main class and a factory class.

The most important part is Decode interface which defines service contract.

[java]

public interface Decode { boolean isEncodingNameSupported(String encodingName); String[] getContent(String data); }

public interface Decode { boolean isEncodingNameSupported(String encodingName); String[] getContent(String data); }Two operations are defined, one that returns if service supports given input, and another that transforms data to String[].

DecodeFactory class is responsible for finding an implementation service that supports required encoding. In fact, this class encapsulates java.util.ServiceLoader calls. ServiceLoader class is in charge of load registered services.

[java]

public class DecodeFactory {

private static ServiceLoader decodeSetLoader = ServiceLoader.load(Decode.class);

public static Decode getDecoder(String encodingName) throws UnsupportedEncodingException {

for (Decode decode : decodeSetLoader) {

if(decode.isEncodingNameSupported(encodingName)) {

return decode;

}

}

throw new UnsupportedEncodingException();

}

}

public class DecodeFactory { private static ServiceLoader decodeSetLoader = ServiceLoader.load(Decode.class); public static Decode getDecoder(String encodingName) throws UnsupportedEncodingException { for (Decode decode : decodeSetLoader) { if(decode.isEncodingNameSupported(encodingName)) { return decode; } } throw new UnsupportedEncodingException(); } }At line 3 we are loading all services that are registered in classpath. At line 7 we only iterate through all services asking if given encoding name is supported.

And finally main class.

[java]

public class App { public static void main(String[] args) throws UnsupportedEncodingException { String encodeName = args[0]; String data = args[1]; Decode decoder = DecodeFactory.getDecoder(encodeName); System.out.println(Arrays.toString(decoder.getContent(data))); } }

public class App { public static void main(String[] args) throws UnsupportedEncodingException { String encodeName = args[0]; String data = args[1]; Decode decoder = DecodeFactory.getDecoder(encodeName); System.out.println(Arrays.toString(decoder.getContent(data))); } }And now if you run this class with java -jar reader.jar "text/cvs" "a, b, c, d", anUnsupportedEncodingException will be thrown. Now we are going to implement our first service. Note that reader project will not be modified nor recompiled.

First service we are going to implement is one that can support comma-separated values encoding. Only one class and one file are important.

CSV class is an implementation of Decode interface and transforms comma-separated values.

[java]

public class CSVDecoder implements Decode { private static final String DELIMITER = Character.toString(DecimalFormatSymbols.getInstance().getPatternSeparator()); public boolean isEncodingNameSupported(String encodingName) { return "text/csv".equalsIgnoreCase(encodingName.trim()); } public String[] getContent(String data) { List values = new LinkedList(); StringTokenizer parser = new StringTokenizer(data, DELIMITER); while(parser.hasMoreTokens()) { values.add(parser.nextToken()); } return values.toArray(new String[values.size()]); } }

public class CSVDecoder implements Decode { private static final String DELIMITER = Character.toString(DecimalFormatSymbols.getInstance().getPatternSeparator()); public boolean isEncodingNameSupported(String encodingName) { return "text/csv".equalsIgnoreCase(encodingName.trim()); } public String[] getContent(String data) { List values = new LinkedList(); StringTokenizer parser = new StringTokenizer(data, DELIMITER); while(parser.hasMoreTokens()) { values.add(parser.nextToken()); } return values.toArray(new String[values.size()]); } }As you can see a simple StringTokenizer class. Only take care that this class is Locale sensitive, countries where comma (,) is used as decimal delimiter, separation character is semicolon (;).

And next important file is a file that is placed into META-INF. This file contains a pointer to service implementation class.

This file should be in META-INF/services and should be called as interface full qualified name. In this case org.alexsotob.reader.Decode. And its content should be service implementation full qualified name.

[plain]

org.alexsotob.reader.csv.CSVDecoder #Comma-separated value decode.

org.alexsotob.reader.csv.CSVDecoder #Comma-separated value decode.

And now you can package this project, and you can reexecute reader project but with generated jar(csv.jar) into classpath. And now output will be an array with a, b, c and d characters instead of unsupported exception.

See that reader project has not been modified and its behaviour has been changed. Now you can develop new implementations for decoding new inputs, and you should only take care of copying them into classpath.

Only take care that all services should have a default constructor with no arguments.

And for those who use Spring Framework, Services are also supported through three differentFactoryBeans, ServiceFactoryBean, ServiceListFactoryBean, ServiceLoaderFactoryBean.

As I have noted at start of this post, JDK services is a simple (yet powerful) solution if you need to create a simple plugins system. In my case it has been enough with JDK services, and I have never required more complex structure; but in case you are thinking about a complete plugin solution you can use JPF that offers a solution like Eclipse plugins, or even OSGI.

I wish you have found this post useful and now you know (if you didn't know yet), an easy solution to develop modules that are plugged and play.

Download Main Project:

Download Comma-Separated Values Service Project:

Download Tab-Separated Values Service Project: