Technology Sharing

한어Русский языкEnglishFrançaisIndonesianSanskrit日本語DeutschPortuguêsΕλληνικάespañolItalianoSuomalainenLatina

SPI stands for Service Provider Interface, which can be understood as an interface specifically provided to service providers or developers who want to extend framework functions. SPI separates the service interface from the specific service implementation, decouples the service caller from the service implementer, and can improve the scalability and maintainability of the program. Modifying or replacing the service implementation does not require modifying the caller. Many frameworks use Java's SPI mechanism, such as the Spring framework, database loading driver, log interface, and Dubbo's extended implementation. Refer to Dubbo's SPI mechanism to implement the SPI part of this RPC framework.

For example, when the client communicates with the server, it needs to serialize the message. There are many serialization algorithms that can be used for serialization, including Hessian, Kryo, and ProtoStuff. The system requires that the method in the class corresponding to the relevant serialization algorithm be called according to the serialization algorithm name in the message to perform serialization and deserialization. In addition, in order to facilitate expansion, SPI needs to be used for decoupling.

The usage of SPI is as follows:

Serializer serializer = ExtensionLoader.getExtensionLoader(Serializer.class)
                    .getExtension(codecName);

codecName is the name of the serialization algorithm, and the corresponding class needs to be loaded according to the name.

    private final Class<?> type;
 
    private ExtensionLoader(Class<?> type) {
        this.type = type;
    }
 
    // 每个SPI接口都有自身的ExtensionLoader
    public static <S> ExtensionLoader<S> getExtensionLoader(Class<S> type) {
        if (type == null) {
            throw new IllegalArgumentException("Extension type should not be null.");
        }
        if (!type.isInterface()) {
            throw new IllegalArgumentException("Extension type must be an interface.");
        }
        if (type.getAnnotation(SPI.class) == null) {
            throw new IllegalArgumentException("Extension type must be annotated by @SPI");
        }
        // firstly get from cache, if not hit, create one
        ExtensionLoader<S> extensionLoader = (ExtensionLoader<S>) EXTENSION_LOADERS.get(type);
        if (extensionLoader == null) {
            EXTENSION_LOADERS.putIfAbsent(type, new ExtensionLoader<S>(type));
            extensionLoader = (ExtensionLoader<S>) EXTENSION_LOADERS.get(type);
        }
        return extensionLoader;
    }

Each SPI interface has its own ExtensionLoader. When getExtensionLoader is called, a series of legal checks will be performed first, and then an attempt will be made to obtain the ExtensionLoader of the interface. First, an attempt will be made to obtain it from the local cache CHM. If it cannot be obtained, a Loader object will be created.

Then get the instance through getExtension. The instance is also cached locally. If it is not in the cache, create the instance again.

    private final Map<String, Holder<Object>> cachedInstances = new ConcurrentHashMap<>();
 
    public T getExtension(String name) {
        if (StringUtil.isBlank(name)) {
            throw new IllegalArgumentException("Extension name should not be null or empty.");
        }
        // firstly get from cache, if not hit, create one
        // 缓存holder
        Holder<Object> holder = cachedInstances.get(name);
        if (holder == null) {
            cachedInstances.putIfAbsent(name, new Holder<>());
            holder = cachedInstances.get(name);
        }
        // create a singleton if no instance exists
        // holder为空，双重检查锁创建示例
        Object instance = holder.get();
        if (instance == null) {
            synchronized (holder) {
                instance = holder.get();
                if (instance == null) {
                    instance = createExtension(name);
                    holder.set(instance);
                }
            }
        }
        return (T) instance;
    }

After obtaining the Class object of the class, you can create this object through reflection.

   // 缓存   
   private static final Map<Class<?>, Object> EXTENSION_INSTANCES = new ConcurrentHashMap<>();
 
   private T createExtension(String name) {
        // load all extension classes of type T from file and get specific one by name
        // SPI接口对应的实现类，其标识名与class文件的映射，根据标识名获取class
        Class<?> clazz = getExtensionClasses().get(name);
        if (clazz == null) {
            throw new RuntimeException("No such extension of name " + name);
        }
        T instance = (T) EXTENSION_INSTANCES.get(clazz);
        if (instance == null) {
            try {
                // 缓存中不存在，则创建实例
                EXTENSION_INSTANCES.putIfAbsent(clazz, clazz.newInstance());
                instance = (T) EXTENSION_INSTANCES.get(clazz);
            } catch (Exception e) {
                log.error(e.getMessage());
            }
        }
        return instance;
    }

The key is the process of obtaining the Class object, that is, the getExtensionCalsses method:

    // 该SPI接口所有实现类的标识与其Class对象的缓存
    private final Holder<Map<String, Class<?>>> cachedClasses = new Holder<>(); 
 
    private static final String SERVICE_DIRECTORY = "META-INF/extensions/";  
 
    private Map<String, Class<?>> getExtensionClasses() {
        // get the loaded extension class from the cache
        // 根据Interface实现类的类名获取对应类的缓存
        Map<String, Class<?>> classes = cachedClasses.get();
        // double check
        if (classes == null) {
            synchronized (cachedClasses) {
                classes = cachedClasses.get();
                if (classes == null) {
                    classes = new HashMap<>();
                    // load all extensions from our extensions directory
                    loadDirectory(classes);
                    // 将Map集合存储在Holder中进行缓存
                    cachedClasses.set(classes);
                }
            }
        }
        return classes;
    }
 
    private void loadDirectory(Map<String, Class<?>> extensionClasses) {
        // 固定路径下的文件，SPI接口的类名作为文件名，在此文件中规定需要加载的实现类
        String fileName = ExtensionLoader.SERVICE_DIRECTORY + type.getName();
        try {
            Enumeration<URL> urls;
            // 系统类加载器，它能够加载用户类路径（ClassPath）上的类和资源。对于SPI机制尤为重要，因为SPI的实现类通常是由应用程序提供并放置在应用程序的类路径下的
            ClassLoader classLoader = ExtensionLoader.class.getClassLoader();
            // 获取当前类加载器加载的URL资源，文件名确定一般urls是唯一的
            urls = classLoader.getResources(fileName);
            if (urls != null) {
                while (urls.hasMoreElements()) {
                    URL resourceUrl = urls.nextElement();
                    // 使用classLoader加载资源，资源目标在resourceUrl下，加载后的class存储在extensionClasses Map集合当中
                    loadResource(extensionClasses, classLoader, resourceUrl);
                }
            }
        } catch (IOException e) {
            log.error(e.getMessage());
        }
    }
 
    private void loadResource(Map<String, Class<?>> extensionClasses, ClassLoader classLoader, URL resourceUrl) {
        try (BufferedReader reader = new BufferedReader(new InputStreamReader(resourceUrl.openStream(), UTF_8))) {
            String line;
            // read every line
            // #是注释，截取注释之前的部分
            while ((line = reader.readLine()) != null) {
                // get index of comment
                final int ci = line.indexOf('#');
                if (ci >= 0) {
                    // string after # is comment so we ignore it
                    line = line.substring(0, ci);
                }
                line = line.trim();
                if (line.length() > 0) {
                    try {
                        final int ei = line.indexOf('=');
                        // 标识与类名
                        String name = line.substring(0, ei).trim();
                        String clazzName = line.substring(ei + 1).trim();
                        // our SPI use key-value pair so both of them must not be empty
                        if (name.length() > 0 && clazzName.length() > 0) {
                            // 加载类
                            Class<?> clazz = classLoader.loadClass(clazzName);
                            // 在map中保存
                            extensionClasses.put(name, clazz);
                        }
                    } catch (ClassNotFoundException e) {
                        log.error(e.getMessage());
                    }
                }
 
            }
        } catch (IOException e) {
            log.error(e.getMessage());
        }
    }

kyro=github.javaguide.serialize.kyro.KryoSerializer
protostuff=github.javaguide.serialize.protostuff.ProtostuffSerializer
hessian=github.javaguide.serialize.hessian.HessianSerializer

The layer-by-layer method calls implement the process of loading the corresponding SPI configuration file in the META-INF/extensions/ path to load the Class object and obtain the instance. The important parts can be referred to the comments.

It should be noted that the file name under META-INF/extensions/ must be consistent with the code. The file name specified in the code is the full class name of the SPI interface class. The content of the file also needs to be written according to (implementation class identifier = implementation class full class name) so that it is consistent with the code, the program can correctly parse the file, and use the class loader to load the corresponding Class. Finally, cache according to <implementation class identifier, implementation class Class object>.

Due to the existence of the three caches (class identifier, Class object), (Class object, object instance), and (class identifier, object instance), the identifier can be directly passed in to obtain the instance of the corresponding class, which also optimizes the performance of the RPC framework.