Select top 1 for each group

Question

I have an Access database that contains a table with information about parts we sort. This table has an autonumber ID field and a 110ID that links to another table with the part information. It also contains a sortDate, sortShift, sorted, scrapped, and repaired. I need to find how many parts have been sorted since the last defect (none scrapped or repaired) was found for each 110ID.

The problem is that I cannot guarantee that the information will be entered into the database in chronological order. So I need sum the 'sorted' field for any records that have a 'sortDate' greater than the last defect, or a 'sortDate' the same as the last defect but greater 'sortShift', or use the autonumber id as a last resort if both the 'sortDate' and 'sortShift' match.

This is the query I am currently using:

SELECT SortInfo.[110ID], Sum(SortInfo.Sorted) AS SumOfSorted
FROM SortInfo
WHERE (
    ((SortInfo.sortdate)>(select top 1 dupe.sortdate from sortinfo as dupe where     (((dupe.[110id])=(sortinfo.[110id])) and (((dupe.repaired)<>0) or ((dupe.scrapped)<>0))) order by dupe.sortdate desc, dupe.sortshift desc, dupe.id desc)))
    OR (((SortInfo.sortdate)=(select top 1 dupe.sortdate from sortinfo as dupe where (((dupe.[110id])=(sortinfo.[110id])) and (((dupe.repaired)<>0) or ((dupe.scrapped)<>0))) order by dupe.sortdate desc, dupe.sortshift desc, dupe.id desc)) 
        AND ((SortInfo.sortshift)>(select top 1 dupe.sortshift from sortinfo as dupe where (((dupe.[110id])=(sortinfo.[110id])) and (((dupe.repaired)<>0) or ((dupe.scrapped)<>0))) order by dupe.sortdate desc, dupe.sortshift desc, dupe.id desc))) 
    OR (((SortInfo.sortdate)=(select top 1 dupe.sortdate from sortinfo as dupe where (((dupe.[110id])=(sortinfo.[110id])) and (((dupe.repaired)<>0) or ((dupe.scrapped)<>0))) order by dupe.sortdate desc, dupe.sortshift desc, dupe.id desc)) 
        AND ((SortInfo.sortshift)=(select top 1 dupe.sortshift from sortinfo as dupe where (((dupe.[110id])=(sortinfo.[110id])) and (((dupe.repaired)<>0) or ((dupe.scrapped)<>0))) order by dupe.sortdate desc, dupe.sortshift desc, dupe.id desc)) 
        AND ((SortInfo.ID)>(select top 1 dupe.id from sortinfo as dupe where (((dupe.[110id])=(sortinfo.[110id])) and (((dupe.repaired)<>0) or ((dupe.scrapped)<>0))) order by dupe.sortdate desc, dupe.sortshift desc, dupe.id desc))
)
GROUP BY SortInfo.[110ID];

The problem is that this is extremely slow. Is there a better way to accomplish this that will yield better performance?

Answer 1

Instead of using all of those subqueries, you can do this with either a LEFT OUTER JOIN or a NOT EXISTS subquery. I don't use Access much, so you may need to tweak these if it's one of those areas where Access falls short of following the ANSI standard.

SELECT
    SI.[110ID],
    SUM(SI.Sorted) AS SumOfSorted
FROM
    SortInfo SI
LEFT OUTER JOIN SortInfo SI2 ON
    SI2.Repaired <> 0 AND
    SI2.Scrapped <> 0 AND
    (
        SI2.SortDate > SI.SortDate OR
        (SI2.SortDate = SI.SortDate AND SI2.SortShift > SI.SortShift) OR
        (SI2.SortDate = SI.SortDate AND SI2.SortShift = SI.SortShift AND SI2.ID > SI.ID)
    )
WHERE
    SI2.ID IS NULL
GROUP BY
    SI.[110ID]

---

SELECT
    SI.[110ID],
    SUM(SI.Sorted) AS SumOfSorted
FROM
    SortInfo SI
WHERE
    NOT EXISTS
    (
        SELECT *
        FROM
            SortInfo SI2
        WHERE
            SI2.Repaired <> 0 AND
            SI2.Scrapped <> 0 AND
            (
                SI2.SortDate > SI.SortDate OR
                (SI2.SortDate = SI.SortDate AND SI2.SortShift > SI.SortShift) OR
                (SI2.SortDate = SI.SortDate AND SI2.SortShift = SI.SortShift AND SI2.ID > SI.ID)
            )    
    GROUP BY
        SI.[110ID]
    )