超级麻烦的、基于“乐观锁”的mongoDB原子更新操作(在老值上加上一个新值)

MongDBAtomicUpdate.java

// 1.查找相同主键对象
        DBObject old = col.findOne(qo);
        if (old == null) {
            // 2.若不存在，尝试插入, 若出错"duplicate 主键"，则转到3.更新操作
            try {
                col.insert(newObj, WriteConcern.SAFE);
            } catch (DuplicateKey e) {
                // 2.1 瞬间又被插了一条记录，尝试更新操作
                BasicDBObject upObj = new BasicDBObject();
                WriteResult rst = null;
                int countDown = MAX_TRY_COUNT;
                do {
                    old = col.findOne(qo);
                    upObj.putAll(unit.merge(old.toMap()));
                    rst = col.update(old, upObj);
                    countDown--;
                } while (!(Boolean) rst.getField("updatedExisting") && countDown > 0);
                if (countDown <= 0) log.error("Update failed after " + MAX_TRY_COUNT + " times retry, tableName: "
                                              + tableName + ", _id: " + _id);
            }
        } else {
            // 3.新老两条记录的合并、更新操作，且要对照老对象做update操作，若老对象已被更改，则要重新查出老对象、再循环尝试这一步操作
            BasicDBObject upObj = new BasicDBObject();
            upObj.putAll(unit.merge(old.toMap()));
            WriteResult rst = col.update(old, upObj);
            int countDown = MAX_TRY_COUNT;
            while (!(Boolean) rst.getField("updatedExisting") && countDown > 0) {
                old = col.findOne(qo);
                upObj.putAll(unit.merge(old.toMap()));
                rst = col.update(old, upObj);
                countDown--;
            }
            if (countDown <= 0) log.error("Update failed after " + MAX_TRY_COUNT + " times retry, tableName: "
                                          + tableName + ", _id: " + _id);
        }