Sql 为每个类别N行的批次设置行号

我有一张桌子

+-----+----------+---------+
| id  | name     | phase   |
+-----+----------+---------+
| 101 | Bolt     | PHASE 1 |
| 102 | Nut      | PHASE 1 |
| 103 | Screw    | PHASE 1 |
| 104 | Hex BOLT | PHASE 1 |
| 105 | Rubber   | PHASE 1 |
| 106 | Aluminum | PHASE 2 |
| 107 | Slate    | PHASE 2 |
| 108 | Pen      | PHASE 3 |
| 109 | Pencil   | PHASE 3 |
| 110 | Mouse    | PHASE 3 |
| 111 | Keyboard | PHASE 3 |
+-----+----------+---------+

我想创建另一列，在其中我必须输入行号

逻辑： 对于3行，它应该是相同的，并更改为第4行的下一个值。无论何时相位发生变化，它都应该转到下一个数字，即使前一个数字的3集不完整

预期产量

+-----+----------+---------+-----+
| id  | name     | phase   | SET |
+-----+----------+---------+-----+
| 101 | Bolt     | PHASE 1 | 1   |
| 102 | Nut      | PHASE 1 | 1   |
| 103 | Screw    | PHASE 1 | 1   |
| 104 | Hex BOLT | PHASE 1 | 2   |
| 105 | Rubber   | PHASE 1 | 2   |
| 106 | Aluminum | PHASE 2 | 3   |
| 107 | Slate    | PHASE 2 | 3   |
| 108 | Pen      | PHASE 3 | 4   |
| 109 | Pencil   | PHASE 3 | 4   |
| 110 | Mouse    | PHASE 3 | 4   |
| 111 | Keyboard | PHASE 3 | 5   |
+-----+----------+---------+-----+

我尝试了下面的查询，但它没有给我所需的输出

select *, (row_number() over (order by phase)-1) / 3 as sets
from table_main

实际产量：

+-----+----------+---------+------+
| id  | name     | phase   | sets |
+-----+----------+---------+------+
| 101 | Bolt     | PHASE 1 | 0    |
| 102 | Nut      | PHASE 1 | 0    |
| 103 | Screw    | PHASE 1 | 0    |
| 104 | Hex BOLT | PHASE 1 | 1    |
| 105 | Rubber   | PHASE 1 | 1    |
| 106 | Aluminum | PHASE 2 | 1    |
| 107 | Slate    | PHASE 2 | 2    |
| 108 | Pen      | PHASE 3 | 2    |
| 109 | Pencil   | PHASE 3 | 2    |
| 110 | Mouse    | PHASE 3 | 3    |
| 111 | Keyboard | PHASE 3 | 3    |
+-----+----------+---------+------+ 

我也尝试过使用稠密的_秩，但没有得到预期的输出

select *, (row_number() over (order by phase)-1) / 3 as sets
from table_main

试一试：

WITH CTE
as
(
select *, ROW_NUMBER () OVER (PARTITION BY phase ORDER BY id) as rn
from table_main
)
SELECT *,DENSE_RANK() over (ORDER BY phase ,CEILING(rn/3.0)) as set
FROM CTE

试一试：

WITH CTE
as
(
select *, ROW_NUMBER () OVER (PARTITION BY phase ORDER BY id) as rn
from table_main
)
SELECT *,DENSE_RANK() over (ORDER BY phase ,CEILING(rn/3.0)) as set
FROM CTE

用于获取列的上一行值，如果更改，则添加1。使用%3中提到的行数，然后将这些值相加，如下所示

;WITH tm AS (
    SELECT *, 
            IIF((LAG(phase, 1) OVER (ORDER BY phase, id)) = phase, 0, 1) AS phase_change,
            IIF((ROW_NUMBER() OVER (PARTITION BY phase ORDER BY phase)-1)%3 = 0, 1, 0) AS [set]
    FROM table_main
) 
SELECT id, 
        name, 
        phase,
        SUM(IIF([set] > phase_change, [set], phase_change)) OVER (ORDER BY phase, id) AS [set]
FROM tm

用于获取列的上一行值，如果更改，则添加1。使用%3中提到的行数，然后将这些值相加，如下所示

;WITH tm AS (
    SELECT *, 
            IIF((LAG(phase, 1) OVER (ORDER BY phase, id)) = phase, 0, 1) AS phase_change,
            IIF((ROW_NUMBER() OVER (PARTITION BY phase ORDER BY phase)-1)%3 = 0, 1, 0) AS [set]
    FROM table_main
) 
SELECT id, 
        name, 
        phase,
        SUM(IIF([set] > phase_change, [set], phase_change)) OVER (ORDER BY phase, id) AS [set]
FROM tm

---

这个答案肯定比我的答案更清晰易懂。！！这个答案肯定比我的答案更清晰易懂。！！