报表分组通用 sql 解决方案已出

2020-11-02 10:53:34 +08:00
 madworks

接上一主题 https://www.v2ex.com/t/719136

通过 dba 的苦思冥想,sql 已经出来了,大家有这方面的可以参考下 sql 效果图见下,懂的人已懂

https://vkceyugu.cdn.bspapp.com/VKCEYUGU-imgbed/9a586ec3-41ea-4190-9b03-37ceb074c486.png

行分组字段 row1,row2,列分组字段 column1,column2,column1 有 012 三种状态,column2 有 01 两种状态,sql 可以呈现 column1,column2 所有组合的值并小计总计,而且可以随意扩展,sql 也是很简洁,稍微变换下可以解决行的分组,可以完美呈现类似下面的报表

https://vkceyugu.cdn.bspapp.com/VKCEYUGU-imgbed/0dfbb436-d5b1-493e-b0d9-2918aecd145e.png 不用自己再逻辑处理数据,动态分组处理逻辑代码简直写的崩溃,不信的人可以去试下

2448 次点击
所在节点    Java
17 条回复
madworks
2020-11-02 10:56:38 +08:00
建表 sql
SET FOREIGN_KEY_CHECKS=0;
DROP TABLE IF EXISTS `xx`;
CREATE TABLE `xx` (
`id` int(11) NOT NULL,
`fee` int(255) DEFAULT NULL,
`column1` varchar(255) DEFAULT NULL,
`column2` varchar(255) DEFAULT NULL,
`row1` varchar(255) DEFAULT NULL,
`row2` varchar(255) DEFAULT '',
PRIMARY KEY (`id`)
) ENGINE=InnoDB DEFAULT CHARSET=utf8;

INSERT INTO `xx` VALUES ('1', '454', '0', '1', '李', '湖北');
INSERT INTO `xx` VALUES ('2', '35', '1', '0', '王', '上海');
INSERT INTO `xx` VALUES ('3', '23', '1', '1', '李', '上海');
INSERT INTO `xx` VALUES ('4', '567', '1', '0', '李', '上海');
INSERT INTO `xx` VALUES ('5', '43', '0', '2', '王', '湖北');
INSERT INTO `xx` VALUES ('6', '456', '0', '1', '李', '河南');
zoharSoul
2020-11-02 10:57:06 +08:00
听起来就是行转列?
madworks
2020-11-02 10:57:12 +08:00
实现 sql
SELECT
row1,
row2,
IFNULL(column1,'总计'),
IFNULL(column2,'小计'),
sum(fee) fee
FROM
(
SELECT
row1,
row2,
column1,
column2,
fee
FROM
(
SELECT
*
FROM
(
SELECT
x.row1,
x.row2,
y.column1,
y.column2,
0 AS fee
FROM
(
SELECT
row1,
row2
FROM
xx
GROUP BY
row1,
row2,
column1,
column2
) x,
(
SELECT
*
FROM
(
SELECT
0 AS column1
UNION
SELECT
1 AS column1
) AS t1
LEFT JOIN (
SELECT
0 AS column2
UNION
SELECT
1 AS column2
UNION
SELECT
2 AS column2
) AS t2 ON 1 = 1
) y
WHERE
(3 * y.column1 + y.column2) NOT IN (
SELECT
3 * m.column1 + m.column2
FROM
xx m
WHERE
m.row1 = x.row1
AND m.row2 = x.row2
)
GROUP BY
row1,
row2,
column1,
column2
) n
UNION
SELECT
row1,
row2,
column1,
column2,
sum(fee) fee
FROM
xx
GROUP BY
row1,
row2,
column1,
column2
) temp
GROUP BY
row1,
row2,
column1,
column2
) AS temp
GROUP BY
row1,
row2,
column1,
column2 WITH ROLLUP
madworks
2020-11-02 11:00:20 +08:00
@zoharSoul,不是行转列,有 2 个行,2 个列,每行每列都有固定的几个状态,一些行列组合肯定没有记录,现在写个 sql 可以展现所有组合的值,没有记录的赋值 0,你怎么写?听起来很简单,但是实现起来还是比较难的
weizhen199
2020-11-02 11:06:53 +08:00
pivot 应该是可以的

还有 dba 帮忙写业务 sql 的,太羡慕。。。/哭
madworks
2020-11-02 11:12:16 +08:00
@weizhen199 是啊,pbi,tableau 这种写这个不要太简单,可惜不开源,不能搞在项目里,只能单独的搭建一套报表系统
madworks
2020-11-02 11:17:47 +08:00
@weizhen199 @zoharSoul 你们说的好像是一个,我了解下,我以为你说的 bi pivot
l00t
2020-11-02 12:05:11 +08:00
这样不还是存在你在原帖中自己评论的问题?值多了怎么办?打算写多长?

不如搞个字典表,直接关联上字典表,比你这么挨个 select 要简洁很多。而且字典表也方便管理这种状态值,甚至可以把中文说明都给直接带出来。
madworks
2020-11-02 13:53:14 +08:00
@l00t 哈哈,是这样的,多谢
madworks
2020-11-02 14:25:54 +08:00
SELECT
row1,
row2,
IFNULL(column1, '总计'),
IFNULL(column2, '小计'),
sum(fee) fee
FROM
(
SELECT
row1,
row2,
column1,
column2,
fee
FROM
(
SELECT
*
FROM
(
SELECT
x.row1,
x.row2,
y.column1,
y.column2,
0 AS fee
FROM
(
SELECT
row1,
row2
FROM
xx_index
GROUP BY
row1,
row2
) x,
(
SELECT
column1,
column2
FROM
xx_index
GROUP BY
column1,
column2
) y
WHERE
(3 * y.column1 + y.column2) NOT IN (
SELECT
3 * m.column1 + m.column2
FROM
xx m
WHERE
m.row1 = x.row1
AND m.row2 = x.row2
)
GROUP BY
row1,
row2,
column1,
column2
) n
UNION
SELECT
row1,
row2,
column1,
column2,
sum(fee) fee
FROM
xx
GROUP BY
row1,
row2,
column1,
column2
) temp
GROUP BY
row1,
row2,
column1,
column2
) AS temp
GROUP BY
row1,
row2,
column1,
column2 WITH ROLLUP
wysnylc
2020-11-02 14:29:54 +08:00
代码可能不会超过 10 行(lambda)
madworks
2020-11-02 14:43:32 +08:00
@wysnylc 太自信了
zifangsky
2020-11-02 15:29:14 +08:00
你这个 SQL 写得也太复杂了吧,先不说在数据量大的情况下执行速度怎么样,光是后期维护也非常困难吧。而且,一般来说复杂逻辑都是放在业务代码里手动实现,SQL 只做简单高效的查询就可以了
madworks
2020-11-02 16:04:59 +08:00
@zifangsky 这 sql 只是看起来长,不复杂啊,修改的话就套模板,很快的。执行速度的话只要行列分组不超过 5 个都在 1 分钟之类,这种报表用逻辑代码写才是噩梦,我写出来要几百行代码,各种递归,小计总计还要判断位置,没写过想象不到的复杂
madworks
2020-11-02 16:10:32 +08:00
@zifangsky 随便给你看下插入小计的位置的代码
public void getInsertPosition(int levelFlag, List<List<Integer>> groupOrderInfoNumCollection,
List<OrderInfo> orderInfoList, int len, List<List<Integer>> nodeIndexList, List<List<Integer>> nestOrderInfoRepeatCollection)
{
int insertPos = 0;
int position = 0;
int[] listLen = new int[levelFlag];

if (levelFlag >= 1)
{
int[] index = new int[levelFlag];
for (int i = 0; i < levelFlag - 1; i++)
{
index[i] = groupOrderInfoNumCollection.get(i + 1).get(0);
}
for (Integer integer : groupOrderInfoNumCollection.get(0))
{
position += len * integer;
orderInfoList.remove(position);
insertPos += integer;
nodeIndexList.get(0).add(insertPos);
if(integer == 1){
nestOrderInfoRepeatCollection.get(0).add(insertPos);
}
for (int i = 0; i < levelFlag - 1; i++)
{
if (insertPos == index[i])
{
nodeIndexList.get(i + 1).add(insertPos);
orderInfoList.remove(insertPos * len);
listLen[i]++;
if (listLen[i] < groupOrderInfoNumCollection.get(i + 1).size())
{
index[i] += groupOrderInfoNumCollection.get(i + 1).get(listLen[i]);
}
}
}
}
if(orderInfoList.size() > 1){
orderInfoList.remove(orderInfoList.size() - 1);
}
}
}
madworks
2020-11-02 16:14:26 +08:00
把结果集整理成需要的形式的代码
public void resultSetHandler(int rowLen, int levelFlag, Map<String, Object> map,
Map<String, Object> rowGroupMap, String[] rowParamArr,
List<List<OrderInfo>> nestOrderInfoCollection, int len, int i)
{
if (rowLen >= 1)
{
List<List<Integer>> temp = new ArrayList<>();
levelFlag--;
map.remove(rowParamArr[rowLen]);
rowGroupMap.remove(rowParamArr[rowLen]);
nestOrderInfoCollection.get(i).addAll(performanceService.getOrderInfo(map));
deleteNodeTotal(nestOrderInfoCollection.get(i), len, rowGroupMap, rowParamArr, levelFlag,temp,temp);
rowLen--;
i++;
resultSetHandler(rowLen, levelFlag, map, rowGroupMap, rowParamArr, nestOrderInfoCollection, len, i);
}
}
madworks
2020-11-02 16:15:09 +08:00
删除节点统计的代码
public void deleteNodeTotal(List<OrderInfo> orderInfoList, int len,
Map<String, Object> rowGroupMap, String[] rowParamArr, int levelFlag, List<List<Integer>> nestOrderInfoIndexCollection, List<List<Integer>> nestOrderInfoRepeatCollection)
{

List<List<Integer>> groupOrderInfoNumCollection = new ArrayList<List<Integer>>();
for (int i = 0; i < levelFlag; i++)
{
nestOrderInfoIndexCollection.add(new ArrayList<Integer>());
nestOrderInfoRepeatCollection.add(new ArrayList<Integer>());
groupOrderInfoNumCollection.add(new ArrayList<Integer>());
}

if (rowParamArr.length > 1)
{

getGroupOrderInfoNum(levelFlag, rowGroupMap, rowParamArr, groupOrderInfoNumCollection);

if (levelFlag >= 1)
{
getInsertPosition(levelFlag, groupOrderInfoNumCollection, orderInfoList, len, nestOrderInfoIndexCollection, nestOrderInfoRepeatCollection);
}
}

}

这是一个专为移动设备优化的页面(即为了让你能够在 Google 搜索结果里秒开这个页面),如果你希望参与 V2EX 社区的讨论,你可以继续到 V2EX 上打开本讨论主题的完整版本。

https://www.v2ex.com/t/720911

V2EX 是创意工作者们的社区,是一个分享自己正在做的有趣事物、交流想法,可以遇见新朋友甚至新机会的地方。

V2EX is a community of developers, designers and creative people.

© 2021 V2EX