Generating a ordered sequence rank on complex query

问题

I am trying to create an implicit rank in the output from the query. The problem seems to be that row_number() is running before worth is calculated.

SELECT 
    firstname,
    lastname,
    personid,
    year,
    (
        SELECT COALESCE(SUM(thevalue),0) 
        FROM assets WHERE personidref = w.personid AND year = w.year
    ) AS assets ,
    (
        SELECT COALESCE(SUM(amount),0) 
        FROM liabilities WHERE personidref = w.personid AND year = w.year
    ) AS liabilities,
    (
        (SELECT COALESCE(SUM(thevalue),0) 
         FROM assets 
         WHERE personidref = w.personid AND year = w.year) 
        - 
        (SELECT COALESCE(SUM(amount),0) 
         FROM liabilities 
         WHERE personidref = w.personid AND year = w.year)
    ) as worth,
    row_number() over(ORDER BY w.worth) as rank 
FROM members w 
WHERE year = 2012 
ORDER BY worth DESC LIMIT 2;

The result is I am getting this:

| firstname | lastname | personid | year | assets | liabilities | worth | rank |
+-----------+----------+----------+------+--------+-------------+-------+------+
| foo       | bar      | 234      | 2012 | 30000  | 20          | 29980 | 32   |
| foo2      | bar2     | 5234     | 2012 | 30000  | 100         | 29900 | 69   |

Instead of this desired output:

| firstname | lastname | personid | year | assets | liabilities | worth | rank |
+-----------+----------+----------+------+--------+-------------+-------+------+
| foo       | bar      | 234      | 2012 | 30000  | 20          | 29980 | 1    |
| foo2      | bar2     | 5234     | 2012 | 30000  | 100         | 29900 | 2    |

Is there a way to pre-run this query and pre-sort by rank first?

回答1:

How about doubling up on the row_number()?

with enchilada as (
SELECT firstname,lastname,personid,year,(SELECT COALESCE(SUM(thevalue),0)
FROM assets WHERE personidref = w.personid) AS assets ,
(SELECT COALESCE(SUM(amount),0) FROM liabilities WHERE personidref = w.personid AND year = w.year) AS liabilities,
((SELECT COALESCE(SUM(thevalue),0) FROM assets WHERE personidref = w.personid AND year = w.year) - (SELECT COALESCE(SUM(amount),0) FROM liabilities WHERE personidref = w.personid AND year = w.year)) as worth,
row_number() over(ORDER BY w.worth) as rank
FROM members w
WHERE year = 2012
ORDER BY worth DESC LIMIT 2 )
select row_number() over (order by rank) as new_rank, * from enchilada;

回答2:

Without sample data it's hard to be sure, but if you want a ranking, why not use rank or dense_rank?

rank() over(ORDER BY w.worth) as rank

dense_rank() over(ORDER BY w.worth) as rank

Here's what I think you might be trying to do. Untested, as there's no sample data or http://sqlfiddle.com/ to test with, but:

SELECT
    *,
    dense_rank() OVER (ORDER BY worth)
FROM
(
    SELECT
        *,
        assets - liabilities AS worth
    FROM
    (
        SELECT 
            firstname,
            lastname,
            personid,
            year,
            (
                SELECT COALESCE(SUM(a.thevalue),0) 
                FROM assets a WHERE a.personidref = w.personid AND a.year = w.year
            ) AS assets ,
            (
                SELECT COALESCE(SUM(l.amount),0) 
                FROM liabilities l WHERE l.personidref = w.personid AND l.year = w.year
            ) AS liabilities
        FROM members w
        WHERE year = 2012
    ) AS a_and_l
) AS net_worths
ORDER BY worth DESC
LIMIT 2;

来源：https://stackoverflow.com/questions/25732309/generating-a-ordered-sequence-rank-on-complex-query

标签

postgresql

subquery