Jxta初始化过程(二)

这篇帖子重点讲NetPeerGroup的创建过程。其中包含两个问题：

1. 缓存管理(Cm)是在哪里初始化的
2. WorldPeerGroup是用哪个类初始化的

接上文，先看下面的代码

java 代码

// create, and Start the default jxta NetPeerGroup
NetPeerGroupFactory factory = new NetPeerGroupFactory(config.getPlatformConfig(), instanceHome);
netPeerGroup = factory.getInterface();

使用工厂模式实现的。在NetPeerGroupFactory的构造函数中，首先创建了WorldPeerGroup(同样使用了工厂模式)，随后使用创建了WorldPeerGroup的子对等组NetPeerGroup

java 代码

/**
 *  Constructs a Net Peer Group and the World Peer Group using the
 *  configuration specified by the provided ConfigParams and using the
 *  specified storeHome location for persistence. The resulting Net Peer
 *  Group instance will use the default ID, Name and Description.
 *
 *  @param config The configuration to use for the newly created World Peer
 *  Group and Net Peer Groups.
 *  @param storeHome The optional location that the World Peer Group, the
 *  Net Peer Group and its' services should use for storing persistent and
 *  transient information. May be {@code null} if the World Peer Group is
 *  not provided a persistent store (though this not currently supported).
 *  @throws PeerGroupException Thrown for problems constructing the Net Peer
 *  Group.
 **/
public NetPeerGroupFactory( ConfigParams config, URI storeHome ) throws PeerGroupException {
    WorldPeerGroupFactory world = new WorldPeerGroupFactory( config, storeHome );
    PeerGroup worldGroup = world.getInterface();

    NetGroupTunables tunables = new NetGroupTunables( ResourceBundle.getBundle("net.jxta.impl.config"), new NetGroupTunables() );

    try {
        net = newNetPeerGroup( worldGroup, tunables.id, tunables.name, tunables.desc, null );
    } finally {
        worldGroup.unref();
    }
}

在WorldPeerGroupFactory的构造函数中，调用了getDefaultWorldPeerGroupClass方法通过配置文件net.jxta.impl.config.properties来获得WorldPeerGroup的类名称，并调用newWorldPeerGroup实例化该类。从配置文件net.jxta.impl.config.properties中，我们看到用于实例化WorldPeerGroup的类名为net.jxta.impl.peergroup.Platform，这样回答了开篇的第二个问题。

java 代码

/**
 *  Constructs the World Peer Group using the specified configuration and
 *  using the specified storeHome location for persistence.
 *
 *  @param config The configuration to use for the World Peer Group.
 *  @param storeHome The optional location that the World Peer Group and its'
 *  services should use for storing persistent and transient information.
 *  May be null if the World Peer Group is not provided a
 *  persistent store (though this not currently supported).
 *  @throws PeerGroupException Thrown for problems constructing the World
 *  Peer Group.
 */
public WorldPeerGroupFactory( ConfigParams config, URI storeHome ) throws PeerGroupException {

    world = newWorldPeerGroup( getDefaultWorldPeerGroupClass(), config, storeHome );
}

第一个问题，缓存管理初始化的具体地点位于Platform的直接父类StdPeerGroup的initFirst中，该方法在newWorldPeerGroup方法中被调用（该调用关系稍微复杂且有些微妙）。

java 代码

// initialize cm before starting services. Do not refer to assignedID, as it could be
// null, in which case the group ID has been generated automatically by super.initFirst()
try {
    cm = new Cm(getHomeThreadGroup(), getStoreHome(), getPeerGroupID().getUniqueValue().toString(), Cm.DEFAULT_GC_MAX_INTERVAL, false);
} catch (Exception e) {
    if (Logging.SHOW_SEVERE && LOG.isLoggable(Level.SEVERE)) {
        LOG.log(Level.SEVERE, "Error during creation of local store", e);
    }
    throw new PeerGroupException("Error during creation of local store", e);
}

通过Cm的构造函数，可以看到如果硬盘上已存在缓存的实例化文件，那么会直接从硬盘上读取，否则重新创建。这也是为什么如果我们只调用DiscoveryService.flushAdvertisement方法（该方法仅作用于内存结构，不改变硬盘上的持久化文件）将某个adv删除后，而当程序重新启动后，该adv依然起作用的原因了（除非将.cache目录全部删除）。

java 代码

rootDir = new File(ROOTDIRBASE, areaName);
rootDir = new File(rootDir.getAbsolutePath());
if (!rootDir.exists()) {
    // We need to create the directory
    if (!rootDir.mkdirs()) {
        throw new RuntimeException("Cm cannot create directory " + rootDir);
    }
}