如何在PostgreSQL中使用交叉表功能创建具有四列的数据透视表

Question

Unfortunately I couldn't find another questions regarding my issue, so I really hope someone can help me out here.

I am trying to create a pivot table using the crosstab function in PostgreSQL. My problem is that I only can pivot three columns from my original table. But I need to use four.

The Table consists of four columns: kgs12, Date, V4, Hach. I want to count the number of records of every category of V4 for every kgs12 and Date. But with crosstab I only achieve to count the categories for either kgs12 or Date.

My table looks like this:

kgs12   Date   V4   Hach
5158020050   05.07.2016   IAB14   1
5158020050   05.07.2016   IAB14   2
5158020050   06.07.2016   IAB14   3
5158020050   06.07.2016   IAB12   4
5158020060   05.07.2016   IAB14   5
5158020060   05.07.2016   IAB12   6
5158020060   06.07.2016   IAB12   7
5158020060   06.07.2016   IAB12   8

I want a pivot table that looks the following way:

kgs12                Date         IAB12_count       IAB14_count
05158020050       2016-07-05      null               2
05158020050       2016-07-06       1                 1
05158020060       2016-07-05       1                 1
05158020060       2016-07-06       2                null

select * from crosstab
('select  kgs12, "V4", count(*) 
from pivot_test 
group by kgs12, "V4" order by kgs12')
as 
ct(kgs12 character varying, iab12 bigint, iab14 bigint)

But as stated I only can create a table with three columns like:

kgs12           IAB12_count  IAB14_count  
05158020050        1            3
05158020060        3            1

or

Date             IAB12_count  IAB14_count
2016-07-05          1              3
2016-07-06          3              1

I am sure the solution is very easy but I am unable to find any. So if someone could help me on this issue, that would be fantastic.

Answer 1

I've never used the crosstab function to do pivots, but if it is as you say and limited to one column for the keycolumn, why not merge the data going into it?

select * from crosstab
('select  concat(kgs12, '--', Date) as key, "V4", count(*) 
from pivot_test 
group by kgs12, Date, "V4" order by kgs12')
as 
ct(key character varying, iab12 bigint, iab14 bigint)

(or maybe this version:)

select * from crosstab
('select  concat(kgs12, '--', Date) as key, "V4", count(*) 
from pivot_test 
group by concat(kgs12, '--', Date), "V4" order by kgs12')
as 
ct(key character varying, iab12 bigint, iab14 bigint)

Maybe even split it apart again later:

select split_part(key, '--', 1), split_part(key, '--', 2), iab12_count, iab14_count from crosstab
(...)

I've always done my crosstabs in a database agnostic way:

select 
  kgs12,
  date,
  count(CASE WHEN "V4" = 'IAB12' THEN 1 END) as iab12_count,
  count(CASE WHEN "V4" = 'IAB14' THEN 1 END) as iab14_count
from pivot_test 
group by kgs12, date

But it does require you to code for the variations in advance, you can't just point it at a table with 10k different values of v4 and automatically have it produce a 10002 column output