Question

Let’s say I have a table with the following data.

• Table name [Data].
• PrimaryID: The primary id of the table.
• ParentID: The table references itself; this is a FK constraint to PrimaryID.

• DateTime: The time that this was last updated.

  PrimaryID        ParentID          Date
      1            null      1/1/2013
      2               1      1/2/2013
      3               1      1/3/2013
      4            null      1/4/2013
      5               4      1/5/2013
      6            null      1/6/2013
  

I want to select results that look like this:

PrimaryID        ParentID
        3               1
        5               4
        6               6

For each “group” (defined as all the entries with the same ParentID and that parent), I want to select the row that is the most recent, and also replace a null ParentID (which usually indicates that this row is the parent) with the row’s own PrimaryID.

I’m really lost on where to even begin generating a query like this.

I have an inner select that looks something like this:

(SELECT PrimaryID, ISNULL(ParentID, PrimaryID) as ParentID, Date FROM [Data])

This looks like it’s the right direction to start, but I don’t know where to go from here.

Answer 1

With the help of @ypercube, something like this should work:

SELECT t.PrimaryId, coalesce(t.ParentId,t.PrimaryId) as parent
FROM YourTable t 
JOIN (
   SELECT coalesce(ParentId, PrimaryId) as parent, Max(DateField) as dtMax
   FROM YourTable
   GROUP BY coalesce(ParentId, PrimaryId)
) t2 ON coalesce(t.ParentId,t.PrimaryId) = parent AND t.DateField = t2.dtMax

And here is the updated Fiddle.