PostgreSQL, mixing SUM horizontaly and vertically - postgresql

Question

I have a temporary table which is result of previously heavy combined data from which I have to create html document to show.
This table in short illustrates situation:
DROP TABLE IF EXISTS temp11;
CREATE TABLE temp11 (t_idx int PRIMARY KEY, mydate text, myclass int, mypercent double precision, valpercent double precision, valclass double precision);
INSERT INTO temp11
(t_idx, mydate, myclass, mypercent, valpercent, valclass) VALUES
(1, '01.01.2014', 1, 10, 10, 1),
(2, '01.01.2014', 2, 20, 20, 4),
(3, '01.01.2014', 2, 20, 50, 10),
(4, '01.01.2014', 1, 10, 17, 1.7),
(5, '02.01.2014', 2, 20, 40, 8),
(6, '02.01.2014', 1, 10, 18, 1.8),
(7, '02.01.2014', 2, 20, 50, 10),
(8, '03.01.2014', 1, 10, 10, 1),
(9, '03.01.2014', 2, 20, 40, 8),
(10, '03.01.2014', 1, 10, 20, 2),
(11, '03.01.2014', 2, 20, 30, 6);
Now I have a query for grouping and summing that into dates and valclasses:
SELECT mydate, myclass, mypercent,
SUM(valpercent) AS sumvalpercent,
SUM(valclass) AS sumvalclass,
SUM(valpercent+valclass) AS sum_row
FROM temp11
GROUP BY mydate, myclass, mypercent
ORDER BY mydate;
Result of this query is expectable:
"01.01.2014" 2 20 70 14.0 84.0
"01.01.2014" 1 10 27 2.7 29.7
"02.01.2014" 1 10 18 1.8 19.8
"02.01.2014" 2 20 90 18.0 108.0
"03.01.2014" 2 20 70 14.0 84.0
"03.01.2014" 1 10 30 3.0 33.0
But needs are a bit extended.
Is it possible to do with PostgreSQL that in same process after every date I get vertically SUM of data inside that date and after all, at the end, SUM of data from all dates so result will look like this:
"01.01.2014" 2 20 70 14.0 84.0
"01.01.2014" 1 10 27 2.7 29.7
97 16.7 113.7
"02.01.2014" 1 10 18 1.8 19.8
"02.01.2014" 2 20 90 18.0 108.0
108 19.8 127.8
"03.01.2014" 2 20 70 14.0 84.0
"03.01.2014" 1 10 30 3.0 33.0
100 17.0 117.0
305 53.5 358.5
If this is possible such (or similar), how that query should look like with showed data?

Answer 1

The simplest way I can think of is to use UNION ALL to get all the desired output at once.
If you leave out the fact that the dates are shown (needed for the order by clause) this query gives the requested output in the simplest way.
SELECT mydate, myclass, mypercent,
SUM(valpercent) AS sumvalpercent,
SUM(valclass) AS sumvalclass,
SUM(valpercent+valclass) AS sum_row
FROM temp11
GROUP BY mydate, myclass, mypercent
UNION ALL
SELECT mydate || ' total', null, null,
SUM(valpercent) AS sumvalpercent,
SUM(valclass) AS sumvalclass,
SUM(valpercent+valclass) AS sum_row
FROM temp11
GROUP BY mydate
UNION ALL
SELECT 'Total', null, null,
SUM(valpercent) AS sumvalpercent,
SUM(valclass) AS sumvalclass,
SUM(valpercent+valclass) AS sum_row
FROM temp11
ORDER BY mydate;
Here's a fiddle
Perhaps it can be rewritten more elegantly using WITH
EDIT:
This will be more efficient because it only traverses through temp11 table just once. Then it only uses the temporary table temp100 which has much fewer rows for the additional totals (no more than one row per day). The UNIONs still remain and the logic is still the same.
WITH temp100 (mydate,myclass,mypercent, sumvalpercent,sumvalclass,sum_row) as (
SELECT mydate, myclass, mypercent,
SUM(valpercent) AS sumvalpercent,
SUM(valclass) AS sumvalclass,
SUM(valpercent+valclass) AS sum_row
FROM temp11
GROUP BY mydate, myclass, mypercent
)
SELECT mydate,myclass,mypercent, sumvalpercent,sumvalclass,sum_row
FROM temp100
UNION ALL
SELECT mydate || ' total' as mydate, null, null, SUM(sumvalpercent), SUM(sumvalclass), SUM(sum_row)
FROM temp100
GROUP BY mydate
UNION ALL
SELECT 'Total' as mydate, null, null, SUM(sumvalpercent), SUM(sumvalclass), SUM(sum_row)
FROM temp100
ORDER BY mydate;
This is the fiddle