SQL中SUM的求和

Question

I have two tables, one representing securities ownership (where there may be more than one holding in each security) and one representing distributions on each security (where there may be more than one distribution on each). I would like to compute total distribution paid on each security.

Thus, given input like

Security ID    shares held
 44                100
 44                100
 45                200
 55                300


Security ID    distribution
 44               0.05
 45               0.06
 55               0.07
 55               0.03
 44               0.05

The output should be

Security ID    total distribution
 44               20
 45               12
 55               30

The problem is, SUM(A.sharecount)*SUM(B.distribution) ends up double counting some of the positions and I get answers that are too big [(80, 12, 60) instead of (20, 12, 30) in the example above].

SQLFiddle with my sample input and failing SQL can be seen at http://sqlfiddle.com/#!9/e3caa/3/0

All advice appreciated.

Answer 1

This works as requested:

SELECT
  S.instrumentID,
  S.shares,
  D.distribution,
  S.shares * D.distribution as total
FROM (
  SELECT instrumentID, SUM(sharecount) shares
  FROM samplePortfolio
  GROUP BY instrumentID
) S
JOIN (
  SELECT instrumentID, SUM(distribution) distribution
  FROM sampleDistributions
  GROUP BY instrumentID
) D USING(instrumentID)

SQL Fiddle

It works by pre-computing the total number of shares and the total distribution in separate subqueries, and then joins them together to make the product, giving the correct value.

Answer 2

Simplest fix:

SELECT
  A.instrumentID,
  SUM(A.sharecount * B.distribution)
FROM
    samplePortfolio A
INNER JOIN
    sampleDistributions B
ON
    A.instrumentID=B.instrumentID
GROUP BY
    A.instrumentID

The reason this is a fix is that the join matches every row from A with every row from B. If you sum across A, then sum across B, then multiply, you double-count for the rows that got duplicated during the join. If you multiply, then sum, you are multiplying rows only with their "partner" rows.

Problems coming from double-counting join results are one of the most common errors I see in SQL, so be on the lookout!

SQL Fiddle: http://sqlfiddle.com/#!9/e3caa/14