saiku 元数据存储分析

一、介绍

使用saiku的人一定对他的元数据存储都特别感兴趣，特别是有分布式管理需求的项目，更是迫切需要了解。其实它是使用Apache的开源项目Jackrabbit管理文件的！

二、代码跟踪

我也是使用了一段时间，希望深入了解它的元数据存储，下面是代码跟踪：

2.1 ajax

首先还是从ajax入手：ajax请求：/saiku/rest/saiku/admin/datasources?_=1466478165922

对应的脚本：org.saiku.web.rest.resources.AdminResource，方法如下：

    @GET

    @Produces( {"application/json"})

    @Path("/datasources")

    @ReturnType("java.lang.List<SaikuDatasource>")

    public Response getAvailableDataSources() 

发现核心代码是：datasourceService.getDatasources().values() 不过看不出来什么，于是从代码注入开始查起！

2.2 代码注入

AdminResource对象注入：

    <bean id="adminBean" class="org.saiku.web.rest.resources.AdminResource">

        <property name="userService" ref="userServiceBean"/>

        <property name="datasourceService" ref="datasourceServiceBean"/>

        <property name="olapDiscoverService" ref="olapDiscoverServiceBean"/>

        <property name="repositoryDatasourceManager" ref="repositoryDsManager"/>

        <property name="logExtractor" ref="logwriterbean"/>

    </bean>

UserService对象注入：

    <bean id="userServiceBean" class="org.saiku.service.user.UserService">

        <property name="jdbcUserDAO" ref="userDAO"/>

        <property name="datasourceService" ref="datasourceServiceBean"/>

        <property name="iDatasourceManager" ref="repositoryDsManager"/>

        <property name="adminRoles">

            <list>

                <value>ROLE_ADMIN</value>

            </list>

        </property>

        <property name="sessionService" ref="sessionService"/>

    </bean>

DatasourceService对象注入：

    <bean id="datasourceServiceBean" class="org.saiku.service.datasource.DatasourceService">

        <property name="connectionManager" ref="connectionManager"/>

    </bean>

发现部分核心代码：

  private void readObject(ObjectInputStream stream)

      throws IOException, ClassNotFoundException {

    stream.defaultReadObject();

    datasources = connectionManager.getDataSourceManager();

  }

继续跟踪：

connectionManager对象注入：

    <bean id="connectionManager" class="org.saiku.web.core.SecurityAwareConnectionManager" init-method="init"

          destroy-method="destroy" depends-on="mondrianVFS">

        <property name="dataSourceManager" ref="repositoryDsManager"/> --这是注入执行setDataSourceManager方法，传入repositoryDsManager对象

        <property name="sessionService" ref="sessionService"/>

    </bean>

repositoryDsManager对象注入：

    <bean id="repositoryDsManager" class="org.saiku.service.datasource.RepositoryDatasourceManager" init-method="load" destroy-method="unload">

        <property name="userService" ref="userServiceBean"/>

        <property name="configurationpath" value="../../repository/configuration.xml"/>

        <property name="datadir" value="../../repository/data"/>

        <property name="foodmartdir" value="../../data"/>

        <property name="foodmartschema" value="../../data/FoodMart4.xml"/>

        <property name="foodmarturl" value="jdbc:h2:../../data/foodmart;MODE=MySQL"/>

        <property name="earthquakeDir" value="../../data"/>

        <property name="earthquakeSchema" value="../../data/Earthquakes.xml"/>

        <property name="earthquakeUrl" value="jdbc:h2:../../data/earthquakes;MODE=MySQL"/>

        <property name="repoPasswordProvider" ref ="repoPasswordProviderBean"/>

        <property name="defaultRole" value="ROLE_USER"/>

        <!-- If you change the repoPassword set this property for at least 1 restart to update the old repo password-->

        <!--<property name="oldRepoPassword" value="sa!kuanalyt!cs"/>-->

    </bean>

获取所有的元数据，就是获取RepositoryDatasourceManager对象的datasources对象，这个对象是由下面的代码生成的：

   public void load() {

        irm = JackRabbitRepositoryManager.getJackRabbitRepositoryManager(configurationpath, datadir, repopasswordprovider.getPassword(),

            oldpassword, defaultRole);

        try {

            irm.start(userService);

            this.saveInternalFile("/etc/.repo_version", "d20f0bea-681a-11e5-9d70-feff819cdc9f", null);

        } catch (RepositoryException e) {

            log.error("Could not start repo", e);

        }

        datasources.clear();

        try {

            List<DataSource> exporteddatasources = null;

            try {

                exporteddatasources = irm.getAllDataSources();

            } catch (RepositoryException e1) {

                log.error("Could not export data sources", e1);

            }

            if (exporteddatasources != null) {

                for (DataSource file : exporteddatasources) {

                    if (file.getName() != null && file.getType() != null) {

                        Properties props = new Properties();

                        SaikuDatasource.Type t = SaikuDatasource.Type.valueOf(file.getType().toUpperCase());

                        SaikuDatasource ds = new SaikuDatasource(file.getName(), t, props);

                        datasources.put(file.getName(), ds);

                    }

                }

            }

        } catch (Exception e) {

            throw new SaikuServiceException(e.getMessage(), e);

        }

    }

三、终极发现

saiku是使用JackRabbit管理树状元数据的，如果想要扩展，只能在JackRabbit基础上进行扩展，同志们继续努力！

JackRabbitIBM文档：http://www.ibm.com/developerworks/cn/java/j-jcr/

JackRabbit入门文档： http://suigara.iteye.com/blog/1454765