Question

This is a question for a SQL specialist!

I’m using SQL SERVER 2008 R2 Express.

I have a table named [myTable] consist of 2 types of records.

The 1st type of records is a master record and the second type is a [Relative] record.

Each master record might have several related record.

I would like to SLELECT TOP 10 * FROM [myTable] master records, and the union sub records like SELECT TOP 4 of [Relative] records for each.

Each record has a [PKID] NO NULL IDENTITY(1,1) PRIMARY KEY CLUSTERED column.

I think I need something like that:

SELECT TOP 10 * FROM [myTable]  WHERE [Relative]=0 
UNION
For each (SELECT TOP 10 [PKID] as Master, * FROM [myTable] WHERE [Relative]=0 )
{SELECT TOP 4 * FROM [myTable] WHERE [Relative] = Master}

How should I correct the query in order to achieve my goal?

I have a secondary question which I doubt there is a simple solution for it:

Relative records comes in two flavors as described by the column [IsImportant].

Is there a way to make sure only 1 important relative record will be selected for each master record?

Is there a way to skip a master record if there are less then 4 relative records for it while only 1 of them is important?

Answer 1

EDIT

Is this a correct paraphrase of your requirements?

[myTable] has 3 relevant fields…

• PKID : Unique identifier for each row
• relative : PKID of the row’s parent (0 if the row has no parent [AKA – a master])

• IsImportant : 0/1 flag

  1. Return no more than 10 master records
  2. For each master record, return no more than 4 relative records
  3. Of the 4 relative records, return no more than 1 IsImportant record
  4. Skip any master that has less than 4 relatives (treating all IsImportant as just 1)

Questions:
– If a master has 3 relative records, and none of them are isImportant, still skip?
– If a master has 4 relative records, and more than 1 is isImportant, still skip?

Best guess answer…

WITH
  master_metadata
AS
(
  SELECT
    relative         AS MasterID,
    COUNT(*)         AS Relatives,
    SUM(isImportant) AS IsImportantRelatives
  FROM
    [myTable]
  WHERE
    relative <> 0
  GROUP BY
    relative
  HAVING
    COUNT(*) - SUM(isImportant) + MAX(isImportant) >= 4
)
,
  master
AS
(
  SELECT TOP 10
    NULL AS sequence_id,
    [myTable].*,
    [master_metadata].Relatives,
    [master_metadata].IsImportantRelatives
  FROM
    [myTable]
  INNER JOIN
    [master_metadata]
      ON [master_metadata].MasterID = [myTable].PKID
  ORDER BY
    [myTable].Selector
)
,
  relative
AS
(
  SELECT
    ROW_NUMBER() OVER (PARTITION BY relative, IsImportant ORDER BY Selector)  AS sequence_id,
    *
  FROM
    [myTable]
)
,
  data
AS
(
  SELECT
    PKID AS MasterID,
    *
  FROM
    [master]

  UNION ALL

  SELECT
    [master].PKID AS MasterID,
    [relative].*, Relatives, IsImportantRelatives
  FROM
    [master]
  INNER JOIN
    [relative]
      ON  ([relative].relative = [master].PKID)
      AND (  ([relative].isImportant = 1 AND [relative].sequence_id  = 1)
          OR ([relative].isImportant = 0 AND [relative].sequence_id <= 3)
          OR ([relative].isImportant = 0 AND [relative].sequence_id  = 4 AND [master].IsImportantRelatives = 0)
          )
)

SELECT
  *
FROM
  [data]
ORDER BY
  MasterID,
  CASE WHEN MasterID = PKID THEN 0 ELSE 1 END,
  IsImportant DESC,
  relative