Question

I have two tables:

• PERSON with columns person_id and total
• DATA with columns data_a, data_b, data_c, and data_person_id

Each “person” can have zero or more entries in DATA – your standard one-to-many relationship. PERSON has a total column that is the sum of values in DATA. There are currently some discrepancies between total and the actual entries in DATA where DATA is correct but total is wrong.

This is the query I’m using to find the discrepancies:

SELECT
  person_id
FROM PERSON JOIN (
  SELECT
    data_person_id,
    SUM( data_a + data_b + data_c ) as data_total
  FROM
    DATA
  GROUP BY
    data_person_id
  ) x ON data_person_id = person_id
WHERE
  person_total != data_total

I plan on doing this through Hibernate as a query where the backend will be Postgres 9.x.

The incorrect query that I’m trying to understand/fix is:

UPDATE
  ONLY PERSON
SET
  total = data_info.calc_total
FROM (
  SELECT
    SUM( data_a + data_b + data_c ) as calc_total
  FROM
    DATA
  WHERE
    DATA.data_person_id = person_id
  GROUP BY
    DATA.data_person_id
) as data_info
WHERE
  PERSON.person_id IN (
    SELECT
      data_person_id
    FROM PERSON JOIN (
      SELECT
        data_person_id,
        SUM( data_a + data_b + data_c ) as data_total
      FROM
        DATA
      GROUP BY
        data_person_id
      ) x ON person_id = data_person_id
    WHERE
      total != data_total
  )

Right now, it won’t run because of WHERE DATA.data_person_id = person_id. But if I take that out, the wrong values get used.

The following seems to work but I’m confused as to why:

UPDATE
  ONLY PERSON
SET
  total = data_info.calc_total
FROM 
  PERSON P JOIN (
    SELECT
      data_person_id,
      SUM( data_a + data_b + data_c ) as calc_total
    FROM
      DATA
    WHERE
      DATA.data_person_id = person_id
    GROUP BY
      DATA.data_person_id
  ) as data_info ON P.person_id = data_person_id
WHERE
  PERSON.person_id IN (
    SELECT
      data_person_id
    FROM PERSON JOIN (
      SELECT
        data_person_id,
        SUM( data_a + data_b + data_c ) as data_total
      FROM
        DATA
      GROUP BY
        data_person_id
      ) x ON person_id = data_person_id
    WHERE
      total != data_total
  )

I believe my problem lies in my misunderstanding of the doc (I’m guessing the part about the self-join).

Also, any ways to improve this query is appreciated!

Answer 1

Seems your queries are way to complex. The task should be as simple as:

UPDATE person p
SET    total = d.calc_total
FROM (
  SELECT data_person_id, sum(data_a + data_b + data_c) as calc_total
  FROM   data
  GROUP  BY 1
) d
WHERE  p.person_id = d.data_person_id
AND    p.total IS DISTINCT FROM d.calc_total;

• First aggregate calc_total from the data table, grouped by data_person_id.

• Then use this subquery in the FROM clause of the UPDATE.

• I use IS DISTINCT FROM to make sure NULL values are covered, while only rows that would change are actually updated.
  If all involved columns are defined NOT NULL, you can use = instead.

-> sqlfiddle demo.