流式排序的聚集结果集 StreamingOrderByReducerResultSet性能优化

[iamhuzl]

流式排序的聚集结果集 StreamingOrderByReducerResultSet处理流程多次创建OrderByResultSetRow造成额外性能消耗
流式排序在排序比对时每次都从上次排序后会获取当前第一个(最大或最小)ResultSet的下一条数据，其他结果集都不会变化，不用每次创建OrderByResultSetRow

private void setDelegateResultSet() throws SQLException {
        OrderByResultSetRow chosenOrderByValue = null;
        for (ResultSet each : resultSets) {
           //此段代码有重复创建对象的额外消耗
            OrderByResultSetRow eachOrderByValue = new OrderByResultSetRow(each, orderByColumns);
            if (null == chosenOrderByValue || chosenOrderByValue.compareTo(eachOrderByValue) > 0) {
                chosenOrderByValue = eachOrderByValue;
                setDelegate(each);
            }
        }
        log.trace("Chosen order by value: {}, current result set hashcode: {}", chosenOrderByValue, getDelegate().hashCode());
    }

建议的新功能行为

• 构造StreamingOrderByReducerResultSet对象时包装ResultSet，调用next方法时获取排序值
• 在setDelegateResultSet方法中直接比对，并设置委托ResultSet即可