RapidMiner之Repository源码分析

Repository是存放process配置等元数据的地方，按惯例，先上UML图：

图1：Repository UML图

     下面以LocalRepository为例来说明Repository是如何实现的，其它类型的Repository都是与此类似的：

     首先来看下UML中的接口：

     Entry：表示存储库中的一条记录，既可以是一个目录，也可以是一条数据项（叶子节点）

     这个接口中的方法，看其英文注释就能明白其作用，RepositoryLocation这个类的作用是记录entry在repository中的位置，其格式为//Repository/path/to/object

public interface Entry {

    /** Returns the name, the last part of the location. */
    public String getName();
    
    /** Returns a string describing the type: "folder", "data", "blob", or "process". */
    public String getType();
    
    /** Returns the user name of the owner. */
    public String getOwner();
    
    /** Returns a human readable description. */
    public String getDescription();
    
    /** Returns true if this entry cannot be written to. */
    public boolean isReadOnly();

    /** Changes the name of the entry. The entry stays in the same folder. 
     * @throws RepositoryException */
    public boolean rename(String newName) throws RepositoryException;
    
    /** Needs to be implemented only for folders in the same repository. 
     *  Moving between different repositories is implemented by 
     *  {@link RepositoryManager#move(RepositoryLocation, Folder, com.rapidminer.tools.ProgressListener)}
     *  using a sequence of copy and delete.
     * @throws RepositoryException */
    public boolean move(Folder newParent) throws RepositoryException;
    
    /** Needs to be implemented only for folders in the same repository. 
     *  Moving between different repositories is implemented by 
     *  {@link RepositoryManager#move(RepositoryLocation, Folder, com.rapidminer.tools.ProgressListener)}
     *  using a sequence of copy and delete.
     *  @param newName New name for moved entry. If moved entry shouldn't be renamed: newName=null.
     * @throws RepositoryException */
    public boolean move(Folder newParent, String newName) throws RepositoryException;
    
    /** Returns the folder containing this entry. */
    public Folder getContainingFolder();
    
    /** Subclasses can use this method to signal whether getting information
     *  from this entry will block the current thread, e.g. because information must
     *  be fetched over the network. */    
    public boolean willBlock();

    /** A location, that can be used, e.g. as a parameter in the {@link RepositorySource} or which can be used
     *  to locate the entry using {@link RepositoryManager#resolve(String)}. */
    public RepositoryLocation getLocation();

    /** Deletes the entry and its contents from the repository. 
     * @throws RepositoryException */
    public void delete() throws RepositoryException;

    /** Returns custom actions to be displayed in this entry's popup menu. */
    public Collection<Action> getCustomActions();
}

 

   Folder：包含sub-entries的entry

   Folder扩展了Entry接口，增加了获取该folder下entry、刷新folder、判断是否存在某个entry的方法。

public interface Folder extends Entry {

    public static final String TYPE_NAME = "folder";

    public List<DataEntry> getDataEntries() throws RepositoryException;
    
    public List<Folder> getSubfolders() throws RepositoryException;
    
    public void refresh() throws RepositoryException;

    public boolean containsEntry(String name) throws RepositoryException;
    
    public Folder createFolder(String name) throws RepositoryException;
    
    public IOObjectEntry createIOObjectEntry(String name, IOObject ioobject, Operator callingOperator, ProgressListener progressListener) 

           throws RepositoryException;

    public ProcessEntry createProcessEntry(String name, String processXML) throws RepositoryException;

    public BlobEntry createBlobEntry(String name) throws RepositoryException;

    /** Returns true iff a child with the given name exists and a {@link #refresh()} 
     *  would find this entry (or it is already loaded).
     * @throws RepositoryException 
     */
    public boolean canRefreshChild(String childName) throws RepositoryException;
}

 

     Repository：表示一个完整的存储库，由Folder和Entry组成

     Repository扩展了Forder接口，增加了如下方法：

public interface Repository extends Folder {
    // 添加RepositoryListener，该listener用于监听添加、删除、修改entry事件和刷新folder事件
    public void addRepositoryListener(RepositoryListener l);
    // 删除一个RepositoryListener
    public void removeRepositoryListener(RepositoryListener l);

    /**根据字符串路径获取一个Entry
     * This will return the entry if existing or null if it can't be found.
     */
    public Entry locate(String string) throws RepositoryException;

    /** Returns some user readable information about the state of this repository. */
    public String getState();
    
    /** Returns the icon name for the repository. */
    public String getIconName();

    /** Returns a piece of XML to store the repository in a configuration file. */
    public Element createXML(Document doc);

    public abstract boolean shouldSave();

    /** Called after the repository is added. Currently unused, but may be useful. Was once
     *  used to fetch JDBC connection entries from remote server. */
    public void postInstall();

    public void preRemove();

    /** Returns true if the repository is configurable. In that case,*/
    public boolean isConfigurable();

    /** Creates a configuration panel.*/
    public RepositoryConfigurationPanel makeConfigurationPanel();
}

 

     LocalRepository是用户存放在本地文件系统中的存储库，从上面的UML图可以看出，LocalRepository继承自类SimpleFolder并实现了接口Repository， SimpleFolder继承自类SimpleEntry，所以在研究LocalRepository之前，要先研究下SimpleEntry和SimpleFolder是如何实现的：

     SimpleEntry是Entry接口的抽象实现:

     SimpleEntry定义了如下三个成员变量，并且由子类来调用该类的构造函数进行初始化：

    private String name;
    private LocalRepository repository;
    private SimpleFolder containingFolder;

    SimpleEntry(String name, SimpleFolder containingFolder, LocalRepository repository) {
        this.name = name;
        this.repository = repository;
        this.containingFolder = containingFolder;
    }

      SimpleEntry定义了二个抽象方法，handleRename和handleMove，分别用来处理重命名和移动事件，这二个方法需要在其子类实现，SimpleEntry类对Entry接口方法的实现中，调用到了这二个抽象类：

   protected abstract void handleRename(String newName) throws RepositoryException;
   protected abstract void handleMove(Folder newParent, String newName) throws RepositoryException;

 

我的