连续数分组的SQL查询

假设我有下表：

| ... | orderId | serialNumber | type          | ... |
|-----|---------|--------------|---------------|-----|
| ... | 1       | 01           | someType      | ... |
| ... | 1       | 02           | someType      | ... |
| ... | 1       | 03           | someOtherType | ... |
| ... | 1       | 04           | someOtherType | ... |
| ... | 1       | 05           | someType      | ... |
| ... | 1       | 06           | someType      | ... |
| ... | 2       | 07           | someType      | ... |
| ... | 2       | 08           | someType      | ... |

我希望我的查询生成以下结果：

| orderId | count | min | max | type          |
|---------|-------|-----|-----|---------------|
| 1       | 2     |  01 | 02  | someType      |
| 1       | 2     |  03 | 04  | someOtherType |
| 1       | 2     |  05 | 06  | someType      |
| 2       | 4     |  07 | 08  | someType      |

我需要按类型和医嘱ID分组。当serialNumber未运行时，应创建一个新条目

这是我当前的查询：

SELECT
 orderId,
 count(*) AS count,
 min(serialNumber) AS min,
 max(serialNumber) AS max,
 type
FROM tblMyTable
group by type, orderId
order by orderId

但它造成的结果是错误的：

| orderId | count | min | max | type          |
|---------|-------|-----|-----|---------------|
| 1       | 4     |  01 | 06  | someType      |  <-- this should be 2 entries
| 1       | 2     |  03 | 04  | someOtherType |
| 2       | 2     |  07 | 08  | someType      |

---

这可能需要一些运行，但只是一个开始

with cte as (
      SELECT *, row_number() over (partition by type, orderId order by serialNumber) as rn
      FROM Table1
)
SELECT *
FROM cte c1
LEFT JOIN cte c2
  ON c1.orderId =  c2.orderId
 AND c1.type = c2.type
 AND c1.rn = c2.rn-1
 AND c1.rn % 2  = 1

输出

+----+---------+--------------+---------------+----+---------+--------------+---------------+------+
|    | orderId | serialNumber |     type      | rn | orderId | serialNumber |     type      |  rn  |
+----+---------+--------------+---------------+----+---------+--------------+---------------+------+
|  1 |       1 |            3 | someOtherType |  1 | 1       | 4            | someOtherType | 2    |
|  2 |       1 |            4 | someOtherType |  2 | NULL    | NULL         | NULL          | NULL |
|  3 |       1 |            1 | someType      |  1 | 1       | 2            | someType      | 2    |
|  4 |       1 |            2 | someType      |  2 | NULL    | NULL         | NULL          | NULL |
|  5 |       1 |            5 | someType      |  3 | 1       | 6            | someType      | 4    |
|  6 |       1 |            6 | someType      |  4 | NULL    | NULL         | NULL          | NULL |
|  7 |       2 |            7 | someType      |  1 | 2       | 8            | someType      | 2    |
|  8 |       2 |            8 | someType      |  2 | NULL    | NULL         | NULL          | NULL |
+----+---------+--------------+---------------+----+---------+--------------+---------------+------+

---

这里还有另一个选项，假设序列由

[serialNumber]

决定

这通常称为缺口和孤岛问题

示例

;with cte as (
    Select *
          ,Grp = sum(Flg) over (order by [serialNumber])
        From (
              Select *
                    ,Flg = case when concat(orderid,type)<> lag(concat(orderid,type),1) over (Order by [serialNumber]) then 1 else 0 end
               From  YourTable
             ) A
)
Select [orderid]
      ,[count] = count(*)
      ,[min]   = min ([serialNumber])
      ,[max]   = max ([serialNumber])
      ,[type]  = max([type])
 From  cte
 Group by [orderid],Grp

---

这是一个缺口和孤岛问题，但我会使用行号差异的方法：

select orderid, count(*), min(serialNumber), max(serialNumber), type
from (select t.*,
             row_number() over (partition by orderid order by serialnumber) as seqnum,
             row_number() over (partition by orderid, type order by serialnumber) as seqnum_type
      from t
     ) t
group by orderid, type, (seqnum - seqnum_type)
order by orderid, min(serialNumber);

您的

serialnumber

看起来像一个字符串，但它有数值。如果您可以信任它的顺序，您甚至不需要两个行号值：

select orderid, count(*), min(serialNumber), max(serialNumber), type
from (select t.*,
             row_number() over (partition by orderid, type order by serialnumber) as seqnum_type
      from t
     ) t
group by orderid, type, (serialnumber - seqnum_type)
order by orderid, min(serialNumber);

---

这样做的原因有点难以解释，但如果您运行子查询，这是很明显的。您将看到行号之间的差异如何标识您想要标识的组。

我猜您对SQL分组的解释有点错误。分组是不考虑记录顺序的。它聚合了所有具有相同分组字段值的记录（在您的案例中是orderId和type）。要将一行拆分为两行，您应该找到一些使这行不同的字段。您的rdbms是什么？Sql Server、postgres、oracle？MS Transact-Sqlall类型成对出现？不，也可以出现单个类型。我已经用这些信息更新了我的问题。更干净了。。。我刚把旗子的方法卡在我的小脑袋里+1

select orderid, count(*), min(serialNumber), max(serialNumber), type
from (select t.*,
             row_number() over (partition by orderid order by serialnumber) as seqnum,
             row_number() over (partition by orderid, type order by serialnumber) as seqnum_type
      from t
     ) t
group by orderid, type, (seqnum - seqnum_type)
order by orderid, min(serialNumber);

select orderid, count(*), min(serialNumber), max(serialNumber), type
from (select t.*,
             row_number() over (partition by orderid, type order by serialnumber) as seqnum_type
      from t
     ) t
group by orderid, type, (serialnumber - seqnum_type)
order by orderid, min(serialNumber);