Question

我有一个 csv文件，如下所示

accont| prod|Item |Tag |LDESC|may13|jun13|jul13|jun16
a |b |c | d | s | 20 | 20.3|30.2 |34
x |y |z | c | s | 21 | 23.3|30.4 |35

我们每个月都会收到这样的文件，但不知道该文件中会有多少个月。

这里的要求是每个月我都要将数据加载到数据库中并将月份列动态地转换为行。

非常感谢任何帮助。

Answer 1

这是一个解决方案（代码注释中的解释）：

USE tempdb
--here we declare variables
DECLARE @string nvarchar(max),  --this will store csv data as a single string
        @xml xml,               --here is a variable that will hold @string converted in xml
        @table_create nvarchar(max), --holds xsxript to drop and create temporary table
        @columns nvarchar(max), --will hold the columns like [may13],[jun13],[jul13],[jun16]
        @sql nvarchar(max)      --here will be generated query with unpivot

--Load csv into a variable
SELECT @string = BulkColumn
FROM OPENROWSET(BULK N'D:\test.csv', SINGLE_CLOB) AS Document

--convert it to xml (we need only first row to get column headers
SELECT @xml =CAST('<row>'+'<col>'+REPLACE(SUBSTRING(@string,1,CHARINDEX(CHAR(10),@string)),'|','</col><col>')+'</col>'+'</row>' as xml)

--generate a table creating script
SELECT @table_create = N'
    USE tempdb; 
    IF (OBJECT_ID(N''##tempCSVload''))IS NOT NULL DROP TABLE ##tempCSVload; 
    CREATE TABLE ##tempCSVload ('+STUFF(@table_create,LEN(@table_create),1,')') +';'

EXEC sp_executesql @table_create

--Now we shall BULK INSERT into temp table
BULK INSERT ##tempCSVload
FROM  N'D:\test.csv'
WITH
(
    FIRSTROW = 2,
    FIELDTERMINATOR = '|',  --CSV field delimiter
    ROWTERMINATOR = '\n',   --Use to shift the control to next row
    TABLOCK
)

--The number and names of monthYY columns are unknown so we take them from sys.columns table
SELECT @columns = ISNULL(@columns,'') + ','+QUOTENAME(name) 
FROM sys.columns
WHERE [OBJECT_ID] = OBJECT_ID(N'##tempCSVload')
AND column_id > 5 --First 5 are skipped as there are not part of pivoting

--Generate the UNPIVOT script
SELECT @sql = N'
SELECT *
FROM ##tempCSVload
UNPIVOT (
    [Values] FOR [Dates] IN ('+STUFF(@columns,1,1,'')+')
) as unpvt'

--And EXECUTE
EXEC sp_executesql @sql

输出：

accont  prod    Item    Tag     LDESC   Values  Dates
a       b       c       d       s        20     may13
a       b       c       d       s        20.3   jun13
a       b       c       d       s       30.2    jul13
a       b       c       d       s       34      jun16
x       y       z       c       s        21     may13
x       y       z       c       s        23.3   jun13
x       y       z       c       s       30.4    jul13
x       y       z       c       s       35      jun16

您可以将INSERT INTO YourTable添加到SELECT @sql = N'，例如：

SELECT @sql = N'
INSERT INTO YourTable
SELECT *
FROM ##tempCSVload
UNPIVOT (
    [Values] FOR [Dates] IN ('+STUFF(@columns,1,1,'')+')
) as unpvt'

在具有固定列和数据类型的表中插入数据。请注意，[Values]存储为nvarchar(max)，夜晚有空格，因此最好使用CAST([Values] as float)，如：

SELECT @sql = N'
SELECT  accont,
        prod,
        Item,
        Tag,
        LDESC,
        CAST([Values] as float) as [Values],
        [Dates]
FROM ##tempCSVload
UNPIVOT (
    [Values] FOR [Dates] IN ('+STUFF(@columns,1,1,'')+')
) as unpvt'

基于Excel列

1 个答案: