Top 1 on Left Join SubQuery

Question

I am trying to take a person and display their current insurance along with their former insurance. I guess one could say that I'm trying to flaten my view of customers or people. I'm running into an issue where I'm getting multiple records back due to multiple records existing within my left join subqueries. I had hoped I could solve this by adding "TOP 1" to the subquery, but that actually returns nothing...

Any ideas?

    SELECT 
    p.person_id AS 'MIRID'
    , p.firstname AS 'FIRST'
    , p.lastname AS 'LAST'
    , pg.name AS 'GROUP'
    , e.name AS 'AOR'
    , p.leaddate AS 'CONTACT DATE'
    , [dbo].[GetPICampaignDisp](p.person_id, '2009') AS 'PI - 2009'
    , [dbo].[GetPICampaignDisp](p.person_id, '2008') AS 'PI - 2008'
    , [dbo].[GetPICampaignDisp](p.person_id, '2007') AS 'PI - 2007'
    , a_disp.name AS 'CURR DISP'
    , a_ins.name AS 'CURR INS'
    , a_prodtype.name AS 'CURR INS TYPE'
    , a_t.date AS 'CURR INS APP DATE'
    , a_t.effdate AS 'CURR INS EFF DATE' 
    , b_disp.name AS 'PREV DISP'
    , b_ins.name AS 'PREV INS'
    , b_prodtype.name AS 'PREV INS TYPE'
    , b_t.date AS 'PREV INS APP DATE'
    , b_t.effdate AS 'PREV INS EFF DATE'
    , b_t.termdate AS 'PREV INS TERM DATE'
FROM 
    [person] p
LEFT OUTER JOIN 
    [employee] e
ON 
    e.employee_id = p.agentofrecord_id
INNER JOIN 
    [dbo].[person_physician] pp
ON 
    p.person_id = pp.person_id
INNER JOIN 
    [dbo].[physician] ph
ON
    ph.physician_id = pp.physician_id
INNER JOIN
    [dbo].[clinic] c
ON 
    c.clinic_id = ph.clinic_id
INNER JOIN
    [dbo].[d_Physgroup] pg
ON
    pg.d_physgroup_id = c.physgroup_id
LEFT OUTER JOIN 
    (
        SELECT
            tr1.*
        FROM 
            [transaction] tr1
        LEFT OUTER JOIN 
            [d_vendor] ins1
        ON 
            ins1.d_vendor_id = tr1.d_vendor_id
        LEFT OUTER JOIN 
            [d_product_type] prodtype1
        ON 
            prodtype1.d_product_type_id = tr1.d_product_type_id
        LEFT OUTER JOIN 
            [d_commission_type] ctype1
        ON 
            ctype1.d_commission_type_id = tr1.d_commission_type_id
        WHERE
            prodtype1.name <> 'Medicare Part D'
            AND tr1.termdate IS NULL
    ) AS a_t
ON
    a_t.person_id = p.person_id
LEFT OUTER JOIN 
    [d_vendor] a_ins
ON 
    a_ins.d_vendor_id = a_t.d_vendor_id
LEFT OUTER JOIN 
    [d_product_type] a_prodtype
ON 
    a_prodtype.d_product_type_id = a_t.d_product_type_id
LEFT OUTER JOIN 
    [d_commission_type] a_ctype
ON 
    a_ctype.d_commission_type_id = a_t.d_commission_type_id
LEFT OUTER JOIN
    [d_disposition] a_disp
ON
    a_disp.d_disposition_id = a_t.d_disposition_id
LEFT OUTER JOIN 
    (
        SELECT
            tr2.*
        FROM 
            [transaction] tr2
        LEFT OUTER JOIN 
            [d_vendor] ins2
        ON 
            ins2.d_vendor_id = tr2.d_vendor_id
        LEFT OUTER JOIN 
            [d_product_type] prodtype2
        ON 
            prodtype2.d_product_type_id = tr2.d_product_type_id
        LEFT OUTER JOIN 
            [d_commission_type] ctype2
        ON 
            ctype2.d_commission_type_id = tr2.d_commission_type_id
        WHERE
            prodtype2.name <> 'Medicare Part D'
            AND tr2.termdate IS NOT NULL
    ) AS b_t
ON
    b_t.person_id = p.person_id
LEFT OUTER JOIN 
    [d_vendor] b_ins
ON 
    b_ins.d_vendor_id = b_t.d_vendor_id
LEFT OUTER JOIN 
    [d_product_type] b_prodtype
ON 
    b_prodtype.d_product_type_id = b_t.d_product_type_id
LEFT OUTER JOIN 
    [d_commission_type] b_ctype
ON 
    b_ctype.d_commission_type_id = b_t.d_commission_type_id
LEFT OUTER JOIN
    [d_disposition] b_disp
ON
    b_disp.d_disposition_id = b_t.d_disposition_id
WHERE
    pg.d_physgroup_id = @PhysGroupID

Answer 1

In Sql server 2005 you can use OUTER APPLY

SELECT p.person_id, s.e.employee_id
FROM person p
OUTER APPLY (SELECT TOP 1 *
             FROM Employee
         WHERE /*JOINCONDITION*/
         ORDER BY /*Something*/ DESC) s

http://technet.microsoft.com/en-us/library/ms175156.aspx

Answer 2

Thanks for all of the feedback and ideas...

In the simplest of terms, I have a person table that stores contact information like name, email, etc. I have another table that stores transactions. Each transaction is really an insurance policy that would contain information on the provider, product type, product name, etc.

I want to avoid giving the user duplicate person records since this causes them to look for the duplicates prior to running mail merges, etc. I'm getting duplicates when there is more than 1 transaction that has not been terminated, and when there is more than 1 transaction that has been terminated.

Someone else suggested that I consider a cursor to grab my distinct contact records and then perform the sub selects to get the current and previous insurance information. I don't know if I want to head down that path though.

Answer 3

The pattern I normally use for this is:

SELECT whatever
FROM person
LEFT JOIN subtable AS s1
ON s1.personid = person.personid

...

WHERE NOT EXISTS
( SELECT 1 FROM subtable
WHERE personid = person.personid
AND orderbydate > s1.orderbydate )

Which avoids the TOP 1 clause and maybe makes it a little clearer.

BTW, I like the way you've put this query together in general, except I'd leave out the brackets, assuming you have rationally named tables and columns; and you might even gain some performance (but at least elegance) by listing columns for tr1 and tr2, rather than "tr1.*" and "tr2.*".

Answer 4

Couldn't take the time to dig through all your SQL (what a beast!), here's an idea that might make things easier to handle:

select 
 p.person_id, p.name <and other person columns>,
 (select <current policy columns>
  from pol <and other tables for policy>
  where pol.<columns for join> = p.person_id
  and <restrictions for current policy>),
 (select <previous policy columns>
  from pol <and other tables for policy>
  where pol.<columns for join> = p.person_id
  and <restrictions for previouspolicy>),
 <other columns>
from person p <and "directly related" tables>

This makes the statement easier to read by separating the different parts into their own subselects, and it also makes it easier to add a "Top 1" in without affecting the rest of the statement. Hope that helps.

Answer 5

It's difficult to understand your question so first I'll throw this out there: does changing your SELECT to SELECT DISTINCT do what you want?

Otherwise, let me get this straight, you're trying to get your customers' current insurance and previous insurance, but they may actually have many insurances before that, recorded in the [transactions] table? I looked at your SQL for quite a few minutes but can't figure out what it all means, so could you please reduce it down to only the parts that are necessary? Then I'll think about it some more. It sounds to me like you need a GROUP BY somehow, but I can't work it out exactly yet.