MS Access 2010 SQL Top N query by group performance issue (continued2)

问题

I have rephrased a previous question MS Access 2010 SQL Top N query by group performance issue (continued) as I believe the context was not clearly described before. The anwwer to my previous question did not provide the top n by group result. The rephrased question is more generic. I have now all data in one table.

Here is my situation: I have a table (Analysis) that contains products (Partnumber) of various categories (Category). Every product has a price (Value). The objective of the query is to show the 10 products with the highest price of each category The table contains 15000 records and will continue to grow.

This is the query:

  SELECT 
    a.Location,
    a.Category,
    a.Partnumber
    a.Value

  FROM Analysis a

  WHERE a.Partnumber IN (
    SELECT TOP 10 aa.Partnumber
    FROM Analysis aa
    WHERE aa.Category = a.Category
    ORDER BY aa.Value DESC
  )


  ORDER BY 
    a.Category;

Here is my question: My current query works with 1000 records in the table (respond time 3 seconds). With 15000 records the query runs endlessly long. How can I rebuild the query to significantly improve performance?

The answer to my previous question was to not use the in-list operation. But this eliminated function to give the top n records by group. The query gave the top n of all records.

回答1:

For sample data in a table called [Analysis]

ID  Location   Category  Partnumber  Value
--  ---------  --------  ----------  -----
 1  here       cat1      part001         1
 2  there      cat1      part002         2
 3  wherever   cat1      part003         3
 4  someplace  cat2      part004         4
 5  nowhere    cat2      part005         5
 6  unknown    cat2      part006         6

the "ranking query"

SELECT 
    a1.ID,
    a1.Location,
    a1.Category,
    a1.Partnumber,
    a1.Value,
    COUNT(*) AS CategoryRank
FROM
    Analysis a1
    INNER JOIN
    Analysis a2
        ON a1.Category = a2.Category
            AND a1.Value <= a2.Value
GROUP BY
    a1.ID,
    a1.Location,
    a1.Category,
    a1.Partnumber,
    a1.Value

returns

ID  Location   Category  Partnumber  Value  CategoryRank
--  ---------  --------  ----------  -----  ------------
 1  here       cat1      part001         1             3
 2  there      cat1      part002         2             2
 3  wherever   cat1      part003         3             1
 4  someplace  cat2      part004         4             3
 5  nowhere    cat2      part005         5             2
 6  unknown    cat2      part006         6             1

so if you only want the top 2 items in each category just wrap the above query in a SELECT ... WHERE

SELECT *
FROM
(
        SELECT 
            a1.ID,
            a1.Location,
            a1.Category,
            a1.Partnumber,
            a1.Value,
            COUNT(*) AS CategoryRank
        FROM
            Analysis a1
            INNER JOIN
            Analysis a2
                ON a1.Category = a2.Category
                    AND a1.Value <= a2.Value
        GROUP BY
            a1.ID,
            a1.Location,
            a1.Category,
            a1.Partnumber,
            a1.Value
) AS RankingQuery
WHERE CategoryRank <= 2
ORDER BY Category, CategoryRank

to give you

ID  Location  Category  Partnumber  Value  CategoryRank
--  --------  --------  ----------  -----  ------------
 3  wherever  cat1      part003         3             1
 2  there     cat1      part002         2             2
 6  unknown   cat2      part006         6             1
 5  nowhere   cat2      part005         5             2

Note: Ensure that the [Category] and [Value] fields are indexed for best performance.

来源：https://stackoverflow.com/questions/24405074/ms-access-2010-sql-top-n-query-by-group-performance-issue-continued2