Question

If I need to get five+ different SUM values with similar WHERE conditions, is the best (maybe only) way to do it through one query with five subqueries? For example, my query is:

SELECT user, 
(SELECT SUM(amount) FROM table t, table2 t2 WHERE t.table2_id = t2.id AND table2.type = 1 AND user = 1 AND date = x) AS one,
(SELECT SUM(amount) FROM table t, table2 t2 WHERE t.table2_id = t2.id AND table2.type = 2 AND user = 1 AND date = x) AS two, 
(SELECT SUM(amount) FROM table t, table2 t2 WHERE t.table2_id = t2.id AND table2.type = 3 AND user = 1 AND date = x) AS three, 
(SELECT SUM(amount) FROM table t, table2 t2 WHERE t.table2_id = t2.id AND table2.type = 4 AND user = 1 AND date = x) AS four, 
(SELECT SUM(amount) FROM table t, table2 t2 WHERE t.table2_id = t2.id AND table2.type = 5 AND user = 1 AND date = x) AS five 
FROM table 
WHERE user = 1

So far my queries have been very quick but I’m worried about the implication of five+ sub-queries when there’s many more records.

And out of curiosity does one query like this with five sub-queries have a performance advantage / dis-advantage over five individual queries? I was assuming it’s faster because it’s only one trip to the database vs 5?

EDIT: There is a table 2 of about 60 records that are used to classify the thousands of records in table 1. Each record in table 2 is categorized by a type value of 1 – 5 (which is what I’m using to group items I want the SUM of).

Answer 1

1. Use ANSI JOINS
2. You are better writing your query in PIVOT form like the query below

Rewritten:

SELECT t.user, 
SUM(CASE table2.type WHEN 1 then amount else 0 end) As one,
SUM(CASE table2.type WHEN 2 then amount else 0 end) As two,
SUM(CASE table2.type WHEN 3 then amount else 0 end) As three,
SUM(CASE table2.type WHEN 4 then amount else 0 end) As four,
SUM(CASE table2.type WHEN 5 then amount else 0 end) As five
FROM table t
INNER JOIN table2 t2 ON t.table2_id = t2.id 
WHERE t.user = 1
  AND table2.type in (1,2,3,4,5)

I understand your query is pseudo but it has a hanging WHERE user without a FROM clause. As for performance, the above goes through the table in one pass. The 5x subquery form would have taken 5 passes, but should not be significantly slower if proper indexes are available.