Question

This statement returns 2 rows:

SELECT ts.UnitId,ts.TeststepId, MAX(ts.CreatedAt)as CreatedAt 
FROM Teststep ts
INNER JOIN Unit u ON ts.UnitId = u.UnitId
Where u.TemplateId = 2 
Group by TeststepId, ts.UnitId

How can I get the one row with the highest CreatedAt date ? Should I distinct the unitId?

UPDATE

Aaron and flem aksed for more info:

1 Template has N Units
1 Unit has N Teststeps

The clustered primary key of the Teststep table is TeststepId and CreatedAt.

[TeststepId] [int] NOT NULL,

[Name] [nvarchar](50) NOT NULL,

[PreCondition] [nvarchar](500) NOT NULL,

[TestInstruction] [nvarchar](500) NOT NULL,

[ExpectedResult] [nvarchar](500) NOT NULL,

[CreatedAt] [datetime] NOT NULL,

[CreatedBy] [nvarchar](50) NOT NULL,

[UnitId] [int] NOT NULL,

The Teststep table

sample data with the most important fields in the Teststep table could be:

The teststep shows historized teststeps. From the above image you see that the TeststepId = 4 exist 2 times. A historized version (58 sec.) and a current version (59 sec.).

I want to get only the teststepid 4 with 59 seconds to show in a “current” View.

UPDATE 2

I want to get only the teststepid 4 with 59 seconds AND
all other teststepIds with their current date. If there are only rows with different teststepids take those because they are the current rows.

I am just about to change my integration test and find about the solution.

Answer 1

SELECT TOP (1) ts.UnitId, ts.TeststepId, ts.CreatedAt 
  FROM dbo.Teststep AS ts
  INNER JOIN Unit AS u 
  ON ts.UnitId = u.UnitId
  WHERE u.TemplateId = 2 
  ORDER BY ts.CreatedAt DESC;   

If you want a row per UnitId, and SQL Server 2005+, then:

;WITH x AS 
(
  SELECT ts.UnitId, ts.TestStepId, ts.CreatedAt,
    rn = ROW_NUMBER() OVER (PARTITION BY ts.UnitId ORDER BY ts.CreatedAt DESC)
  FROM dbo.Teststep AS ts
  INNER JOIN Unit AS u 
  ON ts.UnitId = u.UnitId
  WHERE u.TemplateId = 2 
)
SELECT UnitId, TestStepId, CreatedAt
  FROM x
  WHERE rn = 1;