I'm trying to use the crosstab function in postgresql
to create a pivot table
. However, I'm having difficulty understanding how to structure my SQL within the query. My data consists of four columns and looks like this:
I create this table using the following code:
CREATE TABLE ct(id SERIAL, zone_id int, group_id int, area double precision);
INSERT INTO ct(zone_id, group_id, area) VALUES(1,2,6798.50754160784);
INSERT INTO ct(zone_id, group_id, area) VALUES(1,3,10197.7613124118);
INSERT INTO ct(zone_id, group_id, area) VALUES(2,1,85708.8676744647);
INSERT INTO ct(zone_id, group_id, area) VALUES(2,2,56006.5971338327);
INSERT INTO ct(zone_id, group_id, area) VALUES(2,3,5584.33145616642);
INSERT INTO ct(zone_id, group_id, area) VALUES(2,5,8611.99732832252);
INSERT INTO ct(zone_id, group_id, area) VALUES(2,6,36103.5509183704);
INSERT INTO ct(zone_id, group_id, area) VALUES(2,8,9801.14541428806);
INSERT INTO ct(zone_id, group_id, area) VALUES(5,1,45796.0020793546);
And following the postgresql
documentation closely, I use the following code in my crosstab
query:
SELECT *
FROM crosstab(
'select zone_id, group_id, area
from ct
')
AS ct(row_name integer,
g_1 double precision,
g_2 double precision,
g_3 double precision,
g_4 double precision,
g_5 double precision,
g_6 double precision,
g_7 double precision,
g_8 double precision);
This results in the following table which is not what I want them to be:
For example, in row two, I want the following values:
85708.8676744647, 56006.5971338327, 5584.33145616642, NULL, 8611.99732832252, 36103.5509183704, NULL, 9801.14541428806
Instead the values are:
85708.8676744647, 56006.5971338327, 5584.33145616642, 8611.99732832252, 36103.5509183704, 9801.14541428806
However, it seems that the null
values are ignored, so that my column names g1
to g8
, do not correspond to the original groups.
Use the crosstab()
variant with two parameters:
SELECT * FROM crosstab(
'SELECT zone_id, group_id, area
FROM ct
ORDER BY 1,2'
,'SELECT g FROM generate_series(1,8) g' -- ! Provide values explicitly
)
AS ct(
row_name integer
, g_1 float8, g_2 float8
, g_3 float8, g_4 float8
, g_5 float8, g_6 float8
, g_7 float8, g_8 float8);
Thereby declaring explicitly which value goes in which output column. So the function knows where to fill in NULL
values. In this case generate_series()
comes in handy to provide 8 rows with the numbers 1-8. A VALUES
expression would be an alternative:
'VALUES (1), (2), (3), (4), (5), (6), (7), (8)'
Also, don't forget the ORDER BY
clause in the first parameter query.
I provided a detailed explanation in this related answer.
I can't make it work in the Demo as I can't create the tablefunc extension but it works in my desktop running 9.2.1:
SELECT *
FROM crosstab('
select
s.zone_id, s.group_id, area
from
ct
right join (
(select distinct zone_id from ct) z(zone_id)
cross join
generate_series(1, 8) g(group_id)
) s on s.group_id = ct.group_id and s.zone_id = ct.zone_id
order by s.zone_id, s.group_id
') AS ct (
row_name integer,
g_1 double precision,
g_2 double precision,
g_3 double precision,
g_4 double precision,
g_5 double precision,
g_6 double precision,
g_7 double precision,
g_8 double precision
);
来源:https://stackoverflow.com/questions/12885572/how-to-include-null-values-in-tablefunc-query-in-postgresql