Question

I have a table of data:

|Key|Field |DateField |
|A  |Value1|2012-06-01|
|A  |Value2|2012-06-02|
|A  |Value3|2012-06-03|
|B  |Value4|2012-06-04|
|B  |Value5|2012-06-05|

I want to aggregate this string data by key – like in a group but I can’t work out how.
What I want to end up with is:

|Key|Field Aggregate       |
|A  |Value1, Value2, Value3|
|B  |Value4, Value5        |

Better still would be to end up with the sequence position inserted into the aggregation where the order of the sequence is determined by the order of the corresponding date field.

|Key|Field Aggregate                |
|A  |1: Value1, 2: Value2, 3: Value3|
|B  |1: Value4, 2: Value5           |

This seems like something I could do with a Common Table Expression, but I don’t seem to be able to work out how.

I wouldn’t ordinarily perform this kind of operation in SQL – I’d normally pull the data into an app and manipulate it as necessary in code.

Unfortunately this does not appear to be an option in this instance.

How can I do this in SQL Server 2005 T-SQL.

Is a CTE the right approach? Is there something more appropriate?

Answer 1

Given this sample data:

CREATE TABLE #t
(
  [Key]     char(1),
  [Field]   varchar(32), 
  DateField datetime
);

INSERT #t([Key],[Field],DateField) 
  SELECT 'A','Value1','20120601'
  UNION ALL SELECT 'A','Value2','20120602'
  UNION ALL SELECT 'A','Value3','20120603'
  UNION ALL SELECT 'B','Value4','20120604'
  UNION ALL SELECT 'B','Value5','20120605';

In versions of SQL Server older than SQL Server 2017, we can do this:

SELECT [Key], [Field Aggregate] = STUFF(
  (SELECT ', ' + CONVERT(varchar(11), 
      ROW_NUMBER() OVER (PARTITION BY [Key] ORDER BY DateField)) 
      + ': ' + [Field]
    FROM #t AS t2 WHERE t2.[Key] = t.[Key]
    ORDER BY DateField
    FOR XML PATH(''), 
    TYPE).value(N'./text()[1]', N'varchar(max)'), 1, 2, '')
  FROM #t AS t
  GROUP BY [Key] ORDER BY [Key];

Results:

Key	Field Aggregate
A	1: Value1, 2: Value2, 3: Value3
B	1: Value4, 2: Value5

• Example db<>fiddle

In SQL Server 2017 or greater, we have a much easier time using STRING_AGG():

;WITH cte AS
(
  SELECT [Key], [Field], rn = ROW_NUMBER() OVER 
    (PARTITION BY [Key] ORDER BY DateField)
    FROM #t
)
SELECT [Key], [Field Aggregate] = STRING_AGG(
    CONCAT(rn, ': ', [Field]), ', ')
    WITHIN GROUP (ORDER BY rn) 
  FROM cte
  GROUP BY [Key] ORDER BY [Key];

• Example db<>fiddle