Join 2 records in 1 table with 2 records in another table to produce 2 records

Question

I have 2 tables with lots of data that I need to join. The problem is that the 2 tables hold mostly the same data, and the join sometimes produces undesired, though not unexpected. results. Here is an example:

week_end_date  nugly   payroll_code  rate      hours     check_number
--------------------------------------------------------------
2010-01-17     AU9T8K  HRLY-W        13.00000  40.00000  530957               
2010-01-17     AU9T8K  HRLY-W        13.00000  40.00000  DD00000105382 

week_end_date   nugly   trx_number  pay_code    hours   rate
2010-01-17  AU9T8K  ETS00000010771815   HRLY-W  40.00000    13.00000
2010-01-17  AU9T8K  ETS00000010771684   HRLY-W  40.00000    13.00000

I'm looking to the the check # and the trx_number combined in the join, but I end up with a cross join because everything is the same that I'm joining on. For every case I have like this, I really don't care with trx_number ends up with which check #.

Any thoughts?

Here are the current results:

week_end_date   nugly   payroll_code    rate    hours   check_number    trx_number
2010-01-17  AU9T8K  HRLY-W  13.00000    40.00000    DD00000105382           ETS00000010771815
2010-01-17  AU9T8K  HRLY-W  13.00000    40.00000    530957                  ETS00000010771815
2010-01-17  AU9T8K  HRLY-W  13.00000    40.00000    DD00000105382           ETS00000010771684
2010-01-17  AU9T8K  HRLY-W  13.00000    40.00000    530957                  ETS00000010771684

What I'd like is:

week_end_date   nugly   payroll_code    rate    hours   check_number    trx_number
2010-01-17  AU9T8K  HRLY-W  13.00000    40.00000    DD00000105382           ETS00000010771815
2010-01-17  AU9T8K  HRLY-W  13.00000    40.00000    530957                  ETS00000010771684

Where I don't really care which trx_number is with which check_number.

Here is my current query:

SELECT c.week_end_date, c.nugly, c.payroll_code, c.rate, c.hours, c.check_number, t.trx_number
    FROM checksByNuglyPaycode c 
    LEFT OUTER JOIN trxNumByNuglyPaycode t ON c.db_id = t.db_id AND c.fridate = t.fridate
        AND c.nugly = t.nugly AND c.trx_type = t.trx_type AND c.payroll_code = t.pay_code
        AND c.hours = t.hours AND c.rate = t.rate AND c.week_end_date = t.week_end_date
    WHERE t.db_id = 'lal' AND c.nugly = 'AU9T8K' AND c.payroll_code = 'HRLY-W' 
        AND c.fridate = '2010-01-22' AND c.week_end_date = '2010-01-17'
ORDER BY c.fridate, c.nugly, payroll_code

The where clause is obviously specifically for this case, in the final query, there will not be a where clause.

Answer 1

It's a guess, but could you use something like ROW_NUMBER to make a sort of identity field for each table and then join on them?

Something like

CREATE VIEW vwOrderedTable1
AS
SELECT ROW_NUMBER() OVER(ORDER BY week_end_date) AS 'RowNumber', 
    week_end_date, 
    nugly, 
    payroll_code...
FROM Table1
GO

CREATE VIEW vwOrderedTable2
AS
SELECT ROW_NUMBER() OVER(ORDER BY week_end_date) AS 'RowNumber', 
    week_end_date, 
    nugly, 
    'payroll_code' = pay_code...
FROM Table2
GO

SELECT *
FROM vwOrderedTable1 
INNER JOIN vwOrderedTable2 ON vwOrderedTable1.RowNumber = vwOrderedTable2.RowNumber

Answer 2

Could you provide a bit more information? What's your current query look like? How are the tables structured? Are you trying to join on keys?

You mention you're "looking" to the check # and trx_number... are these the results?

Do you have an example of what would be returned for the lines you've provided?

---

I like Alex's answer, it seems solid.

Perhaps a silly comment, but if you don't care what trx_number you get, and in your sample query you're using a left join that implies you only want "all" the data from checksByNuglyPaycode, why are you joining the two tables? do you need all the data for both tables? More like a full outer join?

So if there is a record in checksByNuglyPaycode that is not in trxNumByNuglyPaycode you would need it?

And if there is a record in trxNumByNuglyPaycode and not in checksByNuglyPaycode you would need that?

---

Interesting aside, if you do want all the data from both tables you could use the select below. Note: it uses distinct which would likely be slow, and I think full outer joins can also be slow... it's been quite a while...

SELECT DISTINCT
    ISNULL(c.week_end_date, t.week_end_date), ISNULL(c.nugly,t.nugly), 
    ISNULL(c.payroll_code,t.payroll_code), ISNULL(c.rate,t.rate), 
    ISNULL(c.hours, t.hours), c.check_number, t.trx_number
FROM 
    checksByNuglyPaycode c 
FULL OUTER JOIN 
    trxNumByNuglyPaycode t ON c.db_id = t.db_id 
                           AND c.fridate = t.fridate
                           AND c.nugly = t.nugly 
                           AND c.trx_type = t.trx_type 
                           AND c.payroll_code = t.pay_code
                           AND c.hours = t.hours 
                           AND c.rate = t.rate 
                           AND c.week_end_date = t.week_end_date
ORDER BY 
    c.fridate, c.nugly, payroll_code

Answer 3

What about something like the query you've already got as a sub-query wrapped up with:

SELECT MAX(trx_number), week_end_date, nugly, payroll_code, rate, hours, check_number
FROM
(
    QUERY_YOU_ALREADY_HAVE
) tbl
GROUP BY week_end_date, nugly, payroll_code, rate, hours, check_number