拆分列值并移动到行答案

【问题标题】：Split columns values and move to rows拆分列值并移动到行
【发布时间】：2019-03-01 07:22:59
【问题描述】：

我想将该数据拆分并移动到行中。我有很多列需要拆分并移动到行。单列数据是可能的，但我没有得到多列的解决方案。ms sql 或 Excel 中的解决方案都是可以接受的。请找到屏幕截图。

提前致谢

【问题讨论】：

这在任何 SQL 数据库中都是一件痛苦的事情。如果您可以在 SQL Server 之外处理此问题，那么我建议您这样做。
虽然，总的来说，我同意蒂姆的观点，但仍然添加了一些直接的方法来获得预期的解析结果

标签： sql sql-server excel database

【解决方案1】：

很遗憾，SPLIT_STRING() 没有“序数”或“索引”选项。因此，您无法真正按已知顺序检索值。（实际上，它确实会按顺序返回值，但不能保证。）

而且，SQL 表和结果集是无序的。因此，即使代码似乎可以工作，依赖排序也是一种非常糟糕的做法。

我建议使用递归 CTE：

with t as (
      select 'prod1' as name, '3,5,6,9' as col1, '33,44,66,7' as col2, '66,7' as col3
     ),
     cte as (
      select name, 1 as lev,
             convert(varchar(max), left(col1, charindex(',', col1 + ',') - 1)) as col1,
             convert(varchar(max), stuff(col1, 1, charindex(',', col1 + ','), '')) as col1_rest,
             convert(varchar(max), left(col2, charindex(',', col2 + ',') - 1)) as col2,
             convert(varchar(max), stuff(col2, 1, charindex(',', col2 + ','), '')) as col2_rest,
             convert(varchar(max), left(col3, charindex(',', col3 + ',') - 1)) as col3,
             convert(varchar(max), stuff(col3, 1, charindex(',', col3 + ','), '')) as col3_rest
      from t
      union all
      select name, lev + 1,
             left(col1_rest, charindex(',', col1_rest + ',') - 1) as col1,
             stuff(col1_rest, 1, charindex(',', col1_rest + ','), '') as col1_rest,
             left(col2_rest, charindex(',', col2_rest + ',') - 1) as col2,
             stuff(col2_rest, 1, charindex(',', col2_rest + ','), '') as col2_rest,
             left(col3_rest, charindex(',', col3_rest + ',') - 1) as col3,
             stuff(col3_rest, 1, charindex(',', col3_rest + ','), '') as col3_rest
      from cte
      where col1_rest <> '' or col2_rest <> '' or col3_rest <> ''
     )
select name, col1, col2, col3
from cte;

Here 是一个 dbfiddle。

【讨论】：

【解决方案2】：

一些免责声明：

STRING_SPLIT 需要 SQL Server 2016+
MAXDOP 1 特意添加

SQL：

-- seeding data
DECLARE @data TABLE ([Name] VARCHAR(50), Column1  VARCHAR(50), Column2  VARCHAR(50), Column3  VARCHAR(50))
INSERT @data VALUES ('Prod1', '3,5,6,9', '33,44,66,7', '66,7')

-- splitting strings into coluns
;WITH cte_col1 AS (
SELECT Name,
       Col1.value AS ColumnValue,
       ROW_NUMBER() OVER (ORDER BY (SELECT 0)) AS rn,
       1 AS ColumnID       
FROM @data d
    CROSS APPLY STRING_SPLIT(d.Column1, ',') AS Col1
)
, cte_col2 AS (
SELECT Name,
       Col2.value AS ColumnValue,
       ROW_NUMBER() OVER (ORDER BY (SELECT 0)) AS rn,
       2 AS ColumnID
FROM @data d
    CROSS APPLY STRING_SPLIT(d.Column2, ',') AS Col2
)
, cte_col3 AS (
SELECT Name,
       Col3.value AS ColumnValue,
       ROW_NUMBER() OVER (ORDER BY (SELECT 0)) AS rn,
       3 AS ColumnID
FROM @data d
    CROSS APPLY STRING_SPLIT(d.Column3, ',') AS Col3
)

-- gluing all parts together
SELECT 
    d.Name
,   min (CASE WHEN d.ColumnID = 1 THEN d.ColumnValue END ) AS Column1
,   min (CASE WHEN d.ColumnID = 2 THEN d.ColumnValue END ) AS Column2
,   min (CASE WHEN d.ColumnID = 3 THEN d.ColumnValue END ) AS Column3
FROM (
SELECT Name, ColumnValue, rn, cte_col1.ColumnID FROM cte_col1
UNION ALL
SELECT Name, ColumnValue, rn, cte_col2.ColumnID FROM cte_col2
UNION ALL
SELECT Name, ColumnValue, rn, cte_col3.ColumnID FROM cte_col3
) d
GROUP BY d.Name, d.rn
ORDER BY d.Name, d.rn
OPTION (MAXDOP 1)

结果成：

Name    Column1 Column2 Column3
Prod1   3   33  66
Prod1   5   44  7
Prod1   6   66  NULL
Prod1   9   7   NULL

【讨论】：