Question

From the following three tables, I need to:

1. Return one row for each record in People, don’t duplicate the person if they have more than one associated contact.
2. Include associated ContactValue, where ContactType is HOMEPHONE, and the start of the date range is prior to now, and the end of the date range is later than now or null. If a person does not have a home phone, still show the person, but show NULL as the value for ContactValue.

People

PersonId (PK, int)
FullName (nvarchar)

PeopleContacts

PersonId (PK, int)
ContactId (PK, int)
StartValidDate (PK, datetime)
EndValidDate (datetime, null)

Contacts

ContactId (PK, int)
ContactType (nvarchar)
ContactValue (nvarchar)

The People table contains a unique list of people. PeopleContacts may contain several contact associations for each person, for different contact types, and a date range that the contact is good for. Contacts contains a list of contact values of certain types. For example, ContactType of “WORKPHONE” and ContactValue of “[phone]”.

Requirements:

1. Return one row for each record in People, don’t duplicate the person if they have more than one associated contact.
2. Include associated ContactValue, where ContactType is HOMEPHONE, and the start of the date range is prior to now, and the end of the date range is later than now or null. If a person does not have a home phone, still show the person, but show NULL as the value for ContactValue.

If People contained three rows, with:

1     Jones, Bob
2     Smith, Bob
3     Smith, Fred

And PeopleContacts contained four rows, with:

1     4    01/01/2012     NULL
2     1    01/01/2012     02/01/2012
2     2    02/02/2012     NULL
3     3    01/01/2012     NULL
3     4    01/01/2012     NULL

And Contacts contained four rows, with:

1     HOMEPHONE     [phone]
2     HOMEPHONE     [phone]
3     HOMEPHONE     [phone]
4     WORKPHONE     [phone]

Output from the correct query, if run after 02/02/2012, should look like:

FullName     ContactValue
--------     ------------
Jones, Bob   NULL
Smith, Bob   [phone]
Smith, Fred  [phone]

Jones, Bob has an association to a contact, but it’s to a WORKPHONE, so ContactValue should be NULL. Smith, Bob has associations to two HOMEPHONE records, but only one, [phone], still has a valid date range. Smith, Fred has an association to both a HOMEPHONE and a WORKPHONE.

Answer 1

I would solve this using a Common Table Expression – This should be more efficient than using APPLY which would be evaluated for each row. Using a CTE, you would just get a single result set and then do a left outer join on that.

I would use RANK to make sure you only get one contact back … by ordering the CTE and joining the CTE by the first rank you will just get a single contact.

The SQL would look something like this:

WITH [CTE] AS (
SELECT
        pc.PersonId
    ,   pc.ContactValue
    ,   RANK() OVER (PARTITION BY PersonId ORDER BY ContactValue DESC) AS [Seed]
FROM
    PeopleContacts pc
    INNER JOIN  Contacts c ON (pc.ContactId = c.ContactId AND c.ContactType='HOME')
    WHERE pc.StartValidDate <= GETDATE()
    AND (pc.EndValidDate > GETDATE() OR pc.EndValidDate IS NULL)
)

SELECT      FullName
        ,   ISNULL(CTE.ContactValue, '') 
FROM People p
LEFT OUTER JOIN CTE ON (CTE.PersonId = p.PersonID AND CTE.Seed = 1)

(This code is untested, if you set up a sql fiddle then I’d be happy to get it working)