Question

I’ve been trying to write a Table-Valued function that takes value pairs as a parameter and return a table with two columns.

Below is the function signature I am trying to do.

FUNCTION [dbo].[ValuePairParser]( @DelimitedValuePairs VARCHAR(MAX),
                                  @Delimiter CHAR(1), 
                                  @ValuePairDelimiter CHAR(1) ) 
  RETURNS @ValuePairTable
  TABLE ( Id INT, Code INT ) 

I want to call the method like below

@ValuePairs VARCHAR(MAX) = '1:1, 1:2, 1:4, 2:3, 1000:230, 130:120,'

ValuePairParser (@ValuePairs, ',', ':')

Can you see any nice way to split above ValuePairs sting and create a table with two columns?

Answer 1

Revisiting after a decade there are definitely better ways to do this today.

SQL Server 2022, Azure SQL Database, Managed Instance

(example)

CREATE OR ALTER FUNCTION dbo.SplitWithPairs
(
    @List           nvarchar(max),
    @MajorDelimiter varchar(3) = ',',
    @MinorDelimiter varchar(3) = ':'
)
RETURNS TABLE WITH SCHEMABINDING
AS
  RETURN 
  (
    SELECT LeftItem = [1], RightItem = [2], Position
    FROM
    (
      SELECT Position = o.ordinal, 
             value    = TRIM(i.value),
             i.ordinal
        FROM        STRING_SPLIT(@List,   @MajorDelimiter, 1) AS o
        CROSS APPLY STRING_SPLIT(o.value, @MinorDelimiter, 1) AS i
      WHERE o.value > ''
    ) AS s PIVOT (MAX(value) FOR ordinal IN ([1],[2])) AS p
  );

SQL Server 2016 – 2019

(example)

In SQL Server 2016 through 2019, a few changes:

• STRING_SPLIT doesn’t support ordinal until SQL Server 2022, so we can use OPENJSON instead
• we need to use LTRIM/RTRIM since TRIM was added in SQL Server 2017
• key is 0-based so we bump by 1 to get an equivalent 1-based position

One note about this function is that you can’t use , as the minor delimiter without changing the function, since , is how OPENJSON delimits. There are probably other characters you may want to stay from for both delimiters.

CREATE OR ALTER FUNCTION dbo.SplitWithPairs
(
    @List           nvarchar(max),
    @MajorDelimiter varchar(3) = ',',
    @MinorDelimiter varchar(3) = ':' -- can't be ,
)
RETURNS TABLE WITH SCHEMABINDING
AS
  RETURN 
  (
    SELECT LeftItem = [1], RightItem = [2], Position
    FROM
    (  
      SELECT Position = o.[key] + 1, 
             value    = LTRIM(RTRIM(i.value)),
             ordinal  = i.[key] + 1
      FROM OPENJSON
      (
        CONCAT('["', REPLACE(STRING_ESCAPE(@List, 'JSON'), 
          @MajorDelimiter, '","'), '"]')
      ) AS o
      CROSS APPLY OPENJSON
      (
        REPLACE(CONCAT('[', QUOTENAME(o.value, '"'), ']'), 
          @MinorDelimiter, '","')
      ) AS i
      WHERE o.value > ''
    ) AS s PIVOT (MAX(value) FOR ordinal IN ([1],[2])) AS p
  );

Versions no longer supported

And finally, the original answer, which will work on versions older than SQL Server 2016, I’ll leave as it was, but with a disclaimer that it could be improved (no shortage of reading about there here and, generally, multi-statement table-valued functions – particularly with loops – are a big red flag):

CREATE FUNCTION [dbo].[SplitWithPairs]
(
    @List NVARCHAR(MAX),
    @MajorDelimiter VARCHAR(3) = ',',
    @MinorDelimiter VARCHAR(3) = ':'
)
RETURNS @Items TABLE
(
    Position  INT IDENTITY(1,1) NOT NULL,
    LeftItem  INT NOT NULL,
    RightItem INT NOT NULL
)
AS
BEGIN
    DECLARE
        @Item      NVARCHAR(MAX),
        @LeftItem  NVARCHAR(MAX),
        @RightItem NVARCHAR(MAX),
        @Pos       INT;

    SELECT
        @List = @List + ' ',
        @MajorDelimiter = LTRIM(RTRIM(@MajorDelimiter)),
        @MinorDelimiter = LTRIM(RTRIM(@MinorDelimiter));

    WHILE LEN(@List) > 0
    BEGIN
        SET @Pos = CHARINDEX(@MajorDelimiter, @List);

        IF @Pos = 0 
            SET @Pos = LEN(@List) + LEN(@MajorDelimiter);

        SELECT
            @Item = LTRIM(RTRIM(LEFT(@List, @Pos - 1))),
            @LeftItem = LTRIM(RTRIM(LEFT(@Item,
            CHARINDEX(@MinorDelimiter, @Item) - 1))),
            @RightItem = LTRIM(RTRIM(SUBSTRING(@Item,
            CHARINDEX(@MinorDelimiter, @Item)
            + LEN(@MinorDelimiter), LEN(@Item))));

        INSERT @Items(LeftItem, RightItem)
            SELECT @LeftItem, @RightItem;

        SET @List = SUBSTRING(@List,
            @Pos + LEN(@MajorDelimiter), DATALENGTH(@List));
    END
    RETURN;
END
GO

DECLARE @ValuePairs VARCHAR(MAX) = '1:1, 1:2, 1:4, 2:3,1000:230, 130:120,';

SELECT LeftItem, RightItem
  FROM dbo.SplitWithPairs(@ValuePairs, ',', ':')
  ORDER BY Position;
GO