Combining two sql columns with delimited data by collating into a single column with delimited data

问题

I have two sql columns each with delimited data that I want to collate and combine into a single delimited column. The number of items in the column is variable for each row. However there will always be a matching number of items between the two columns of each row. For Example...

*******************************
ORIGINAL SQL TABLE
*******************************
value          *    unit
*******************************
4 ; 5          *   mg ; kg
50             *   mg
7.5 ; 325      *   kg ; mg
100 ; 1.5 ; 50 *   mg ; g ; mg
********************************

*********************************
DESIRED SQL RESULT
*********************************
value-unit
*********************************
4 mg; 5 kg
50 mg
7.5 kg; 325 mg
100 mg; 1.5 g; 50 mg
*********************************

How do I do this with T-SQL? I'm using SQL Server 2012

回答1:

Using only common table expressions, we can get to the required results too, as below:-

First lets set up the data

declare @original table(
[value] varchar(250), 
[unit] varchar(250)
)
insert into @original values 
('4 ; 5','mg ; kg    '),
('50','mg        '                ),
('7.5 ; 325','kg ; mg    '        ),
('100 ; 1.5 ; 50 ','mg ; g ; mg'  )

Now, lets build the common table expressions:-

;with cte as (
    select o.[value]+';' [value],o.[unit]+';' [unit],row_number() over (ORDER BY (Select 0)) [row]   from @original o
),cte2 as (
    select *
    ,1 [ValueStart],CHARINDEX(';',[value]) [ValueEnd]
    ,1 [UnitStart],CHARINDEX(';',[unit]) [UnitEnd]
     from cte
),cte3 as (
    select * from cte2
    union all 
    select [value],[unit],[row]
        ,[ValueEnd]+1 [ValueStart],CHARINDEX(';',[value],[ValueEnd]+1) [ValueEnd] 
        ,[UnitEnd]+1 [UnitStart],CHARINDEX(';',[unit],[UnitEnd]+1) [UnitEnd] 
    from cte3 where [UnitEnd]>0
),cte4 as (
    select *,row_number() over (partition by [row] order by [row]) [subRow]
    , rtrim(ltrim(substring([unit],[UnitStart],[UnitEnd]-[UnitStart])))  [subUnit] 
    , rtrim(ltrim(substring([value],[ValueStart],[ValueEnd]-[ValueStart])))  [subValue] 
    from cte3
    where [UnitEnd]>0
),cte5 as (
    select subRow,[row],[subValue],[subUnit],cast([subValue]+' '+[subUnit] as varchar(max)) [ValueUnit] from cte4 where subRow=1
    union all
    select cte4.subRow,cte4.[row],cte4.[subUnit],cte4.[subValue]
        ,cte5.[ValueUnit]+';'+ cte4.[subValue]+' '+cte4.[subUnit] [ValueUnit] 
            from cte4
        inner join cte5 on (cte5.subRow+1)=cte4.subRow and cte5.[row]=cte4.[row]
),cte6 as (
    select *,row_number() over (partition by [row] order by subRow desc) [selected] from cte5
)
select ValueUnit from cte6
where [selected]=1
order by [row]

Results will be as below:-

ValueUnit
============
4 mg;5 kg 
50 mg  
7.5 kg;325 mg 
100 mg;1.5 g;50 mg

回答2:

STRING_SPLIT is the most current solution. If you're not on 2016 or later and cannot set your db compatibility to that version or later, here is a more old fashioned approach using xml:

I added an Identity to your original table to have something to order by -

SELECT splitNumbers.splitNumber,
       splitValues.splitValue,
       splitNumbers.splitNumber + ' ' + splitValues.splitValue AS combined
FROM
(
    SELECT --numbers,
        LTRIM(RTRIM(m.n.value('.[1]', 'varchar(8000)'))) AS splitNumber,
        ROW_NUMBER() OVER (ORDER BY id) AS rn
    FROM
    (
        SELECT id,
               CAST('<XMLRoot><RowData>' + REPLACE(numbers, ' ; ', '</RowData><RowData>') + '</RowData></XMLRoot>' AS XML) AS xmlNumbers
        FROM #x
    ) fee
        CROSS APPLY xmlNumbers.nodes('/XMLRoot/RowData') m(n)
) splitNumbers
    INNER JOIN
    (
        SELECT LTRIM(RTRIM(m.v.value('.[1]', 'varchar(8000)'))) AS splitValue,
               ROW_NUMBER() OVER (ORDER BY id) AS rn
        FROM
        (
            SELECT id,
                   CAST('<XMLRoot><RowData>' + REPLACE(units, ' ; ', '</RowData><RowData>') + '</RowData></XMLRoot>' AS XML) AS xmlUnits
            FROM #x
        ) fee
            CROSS APPLY xmlUnits.nodes('/XMLRoot/RowData') m(v)
    ) splitValues
        ON splitNumbers.rn = splitValues.rn;

This gives the following results:
================================
splitNumber splitValue  combined
4           mg           4 mg
5           kg           5 kg
50          mg           50 mg
7.5         kg           7.5 kg
325         mg           325 mg
100         mg           100 mg
1.5         g            1.5 g
50          mg           50 mg

来源：https://stackoverflow.com/questions/55655510/combining-two-sql-columns-with-delimited-data-by-collating-into-a-single-column

标签

tsql

sql-server-2012