如何自动备份删除的对象（3）-优快云博客

到目前为止我们有了一套从Object自动生成Map<String,Object>的工具，下面要作的就是把这个Map中的信息给持久化了，依照不同的持久化方式和后继对删除对象的查询方式，这又有着不同的solution。比如持久化到文件系统和数据库，显然要用不同的方式。

目前这个系统采用的是数据库，OR-Mapping用的是Toplink。对于删除的数据，在前台有能够查询的界面，可供业务人员查询。针对这样的需求，需要解决以下的几个问题。

首先，在何处获得被删除的对象？BaseDao.deleteObject(Object obj)似乎是一个不错的入口，可惜的是，Toplink对于private owner的对象会自动删除，而service developer不会向下面这样去删除一个对象

//如果想在BaseDao.deleteObject(obj)中捕获删除对象，
//那就要求service这样去删除一个具有private owner 属性的Order

for (Item item: order.getItems()){
    getBaseDao().deleteObject(item);
}

getBaseDao().deleteObject(order);

所以，为了透明地得到所有Toplink实际删除了的对象，要在每个UnitOfWork生成时要注册一个Toplink的SessionEventListener

import java.util.Date;
import java.util.Enumeration;
import java.util.Map;

import oracle.toplink.internal.helper.IdentityHashtable;
import oracle.toplink.publicinterface.UnitOfWork;
import oracle.toplink.sessions.SessionEvent;
import oracle.toplink.sessions.SessionEventAdapter;


public class RecordListener extends SessionEventAdapter {
    protected static final FWIIntegrator integrator = FWIntegratorFactory.getIntegrator();

    @Override
    public void postCommitUnitOfWork(SessionEvent event) {
        try {
            UnitOfWork unitOfWork = (UnitOfWork) event.getSession();
            UnitOfWork uow4DeletionRecord = null;
            IdentityHashtable objectsDeletedDuringCommit = unitOfWork.getObjectsDeletedDuringCommit();
            for (Enumeration<?> e = objectsDeletedDuringCommit.keys(); e.hasMoreElements();) {
                Object key = e.nextElement();
                RecordUtils recordUtils = RecordUtils.instance();
                Map<String, Object> content = recordUtils.toContent(key);
                if (content != null) {
                    // since deleted objects are detected in post commit callback
                    // we have to acquire another UOW to write DeletionRecord
                    if (uow4DeletionRecord == null) {
                        uow4DeletionRecord = ParamUtils.getDataSource().getSession().acquireUnitOfWork();
                    }

                    DeletionRecord deleteRecord = recordUtils.toDeleteRecord(content);
                    deleteRecord.setOperatingTime(new Date());
                    uow4DeletionRecord.assignSequenceNumber(deleteRecord);
                    uow4DeletionRecord.registerNewObject(deleteRecord);
                }

            }

            if (uow4DeletionRecord != null) {
                uow4DeletionRecord.commit();
                uow4DeletionRecord.release();
            }
        } catch (Exception e) {
           // You may log it by yourself
        }
    }
}

    private void initSessionAndUnitOfWork(ServerSession serverSession) {
        clientSession = serverSession.acquireClientSession();
        uw = clientSession.acquireUnitOfWork();
        addListenerToUnitOfWork(uw);
    }

    private void addListenerToUnitOfWork(UnitOfWork uw) {
        uw.getEventManager().addListener(new RecordListener());
    }

这个Listener的作用时间是postCommitUnitOfWork，目的是保证UnitOfWork.getObjectsDeletedDuringCommit()有值。当然，由于原有的UnitOfWork已经commit了，所以得新起一个UnitOfWork来写这些DeletionRecord对象

DeletionRecord没有什么特别的东西，它包含一些查询所用到的warehouse,owner,location之类的业务字段（这些字段在数据库中会作索引，同时含有一个content字段，其中存储的是从那个Map转换过来的XML

public class DeletionRecord extends OidEnabledComponent {

    private static interface DeletionRecordSetterCallback {
        void setValue(DeletionRecord obj, Object value);
    }

    public static enum FieldType {

        CLASS_NAME(new DeletionRecordSetterCallback() {
            public void setValue(DeletionRecord obj, Object value) {
                obj.setClassName((String) value);
            }
        }),

        WAREHOUSE_ID(new DeletionRecordSetterCallback() {
            public void setValue(DeletionRecord obj, Object value) {
                obj.setWarehouseId((String) value);
            }

        }), OWNER_ID(new DeletionRecordSetterCallback() {
            public void setValue(DeletionRecord obj, Object value) {
                obj.setOwnerId((String) value);
            }

        }), LOCATION_ID(new DeletionRecordSetterCallback() {
            public void setValue(DeletionRecord obj, Object value) {
                obj.setLocationId((String) value);
            }

        }), ITEM_NO(new DeletionRecordSetterCallback() {
            public void setValue(DeletionRecord obj, Object value) {
                obj.setItemNo((String) value);
            }

        }), ITEM_DESCRIPTION(new DeletionRecordSetterCallback() {
            public void setValue(DeletionRecord obj, Object value) {
                obj.setItemDescription((String) value);
            }

        }), BIZ_TYPE(new DeletionRecordSetterCallback() {
            public void setValue(DeletionRecord obj, Object value) {
                obj.setBizType((Long) value);
            }

        }), ASN_NO(new DeletionRecordSetterCallback() {
            public void setValue(DeletionRecord obj, Object value) {
                obj.setAsnNo((String) value);
            }

        }), RECEIVE_DATE(new DeletionRecordSetterCallback() {
            public void setValue(DeletionRecord obj, Object value) {
                obj.setReceiveDate((Date) value);
            }

        });
        private DeletionRecordSetterCallback setterCallback;

        FieldType(DeletionRecordSetterCallback setterCallback) {
            this.setterCallback = setterCallback;
        };

        public void setValue(DeletionRecord obj, Object value) {
            this.setterCallback.setValue(obj, value);
        }

    }

    private String className;

    private String warehouseId;

    private String ownerId;

    private String locationId;

    private String itemNo;

    private String itemDescription;

    private Long bizType;

    private String asnNo;

    private Date receiveDate;

    private String operatorId;

    private Date operatingTime;

    private String content;

    public String getClassName() {
        return className;
    }

    public void setClassName(String className) {
        this.className = className;
    }

    public String getWarehouseId() {
        return warehouseId;
    }

    public void setWarehouseId(String warehouseId) {
        this.warehouseId = warehouseId;
    }

    public String getOwnerId() {
        return ownerId;
    }

    public void setOwnerId(String ownerId) {
        this.ownerId = ownerId;
    }

    public String getLocationId() {
        return locationId;
    }

    public void setLocationId(String locationId) {
        this.locationId = locationId;
    }

    public String getItemNo() {
        return itemNo;
    }

    public void setItemNo(String id) {
        this.itemNo = id;
    }

    public String getItemDescription() {
        return itemDescription;
    }

    public void setItemDescription(String name) {
        this.itemDescription = name;
    }

    public Long getBizType() {
        return bizType;
    }

    public void setBizType(Long bizType) {
        this.bizType = bizType;
    }

    public String getAsnNo() {
        return asnNo;
    }

    public void setAsnNo(String asnNo) {
        this.asnNo = asnNo;
    }

    public Date getReceiveDate() {
        return receiveDate;
    }

    public void setReceiveDate(Date receiveDate) {
        this.receiveDate = receiveDate;
    }

    public String getOperatorId() {
        return operatorId;
    }

    public void setOperatorId(String operatorId) {
        this.operatorId = operatorId;
    }

    public Date getOperatingTime() {
        return operatingTime;
    }

    public void setOperatingTime(Date operatingTime) {
        this.operatingTime = operatingTime;
    }

    public String getContent() {
        return content;
    }

    public void setContent(String content) {
        this.content = content;
    }

}

需要注意的是，这个转换出来的deletionRecord不是在原有的UnitOfWork中写入的，所以那一段专门去catch了Exception。这是基于以下想法：即使自动备份失败了，也不应该影响正常的删除逻辑。如果认为自动备份失败就不能继续往下作的话，这个catch是要去掉的。

至此，一个自动备份删除对象的简易框架就搭好了。总结一下，核心思想包括以下几点

1）通过Annotation和interface，使得一个对象能够在被删除时指出自己的哪些信息需要被记录

2）通过SessionEventListen得到所有被删除的对象，保持对使用者的透明

3）把结果信息分为两部分，需要查询的用单独的字段记录做索引，其余信息用一个单层的XML保存。

再深入考虑的话，这个框架可以有以下的改进：

一是用XML指定要删除的对象和属性，作到对原始文件的完全无侵入；二是持久化到数据库外其他的媒介如文件系统、网络等，这就可以抽象出一个Appender interface，再实现FileAppender, DBApplender等对象；最后如果性能有需要，可以把持久化被删除对象和删除业务对象的逻辑做成异步的，这样删除N个对象后才持久化一次删除动作，能够减少业务逻辑的延迟。