SQL中如何将单行数据拆分成多行？

Question

I have a table in the following format

ProjectID           LocationID
1                   [1,2,3,4]
2                   [2,3]

Can I split the data in the column LocationID into multiple rows like below?

ProjectID           LocationID
1                   1
1                   2
1                   3
1                   4
2                   2
2                   3

I need to get the data loaded to Power-Bi using the SQL only. is it possible?

Answer 1

If data type of locationID is varchar then:

 create table projects (ProjectID int, LocationID varchar(50));
 insert into projects values(1,                  '[1,2,3,4]');
 insert into projects values(2,                  '[2,3]');

Query:

select projectid, value 
 from projects 
 CROSS APPLY STRING_SPLIT(replace(replace(locationid,'[',''),']',''),',')

Output:

projectid	value
1	1
1	2
1	3
1	4
2	2
2	3

db<fiddle here

Solution for SQL Server 2014

 create table projects (ProjectID int, LocationID nvarchar(max));
 insert into projects values(1,                  '[1,2,3,4]');
 insert into projects values(2,                  '[2,3]');

Query:

     WITH tmp AS
(
    SELECT
        ProjectID,
        LEFT(replace(replace(locationid,'[',''),']',''), CHARINDEX(',', replace(replace(locationid,'[',''),']','') + ',') - 1) LocationID,
        STUFF(replace(replace(locationid,'[',''),']',''), 1, CHARINDEX(',', replace(replace(locationid,'[',''),']','') + ','), '') b
    FROM projects
    
    UNION all

    SELECT
        ProjectID,
        LEFT(b, CHARINDEX(',', b + ',') - 1),
        STUFF(b, 1, CHARINDEX(',', b + ','), '')
    FROM tmp
    WHERE
        b > ''
)

SELECT
    ProjectID, LocationID
FROM tmp
ORDER BY projectid
 

Output:

ProjectID	LocationID
1	1
1	2
1	3
1	4
2	2
2	3

db<fiddle here

Answer 2

In SQL Server 2014, you can use a recursive CTE -- which Kazi also proposes. I think this is a slightly simpler version:

with cte as (
      select projectId, convert(varchar(max), null) as locationid,
             convert(varchar(max), substring(LocationId, 2, len(locationId) - 2)) + ',' as rest
      from t
      union all
      select projectId,
             left(rest, charindex(',', rest) - 1),
             stuff(rest, 1, charindex(',', rest), '')
      from cte
      where rest <> ''
     )
select projectid, locationid
from cte
where locationid is not null;

Here is a db<>fiddle.

In particular, the anchor part just sets up the data -- it does not extract any elements from the string. So, all the logic is in the recursive part, which I find easier to maintain.

Answer 3

I used STRING_SPLIT() which is a table valued function supports SQL server 2016 and higher versions. You need to provide the formatted string into this function and use cross apply to join and generate the desired output.

SELECT
    projectID
   , REPLACE(REPLACE(locationId,'[',''),']','') as [locationid]  
INTO #temp_table
FROM split_string -- Add your table name here


SELECT 
    projectID
    ,VALUE [locationid]  
FROM     #temp_table
CROSS APPLY STRING_SPLIT( [locationid] , ',');

Answer 4

Thanks everyone for helping me out. This solution works for my scenario.

Select projectID,locationid_list from project CROSS APPLY OPENJSON(locationid, '$') WITH (locationid_list int '$')