Sales data contains dynamic product names which can contian any characters.
Dynamic pivot table is created based on sample from Crosstab with a large or undefined number of categories
translate() is used to remove bad characters.
In result pivot table column names are corrupted: missing characters and spaces are removed. How to return data with same column names as in source data ? I tried to use
quote_ident(productname) as tootjakood,
instead of
'C'||upper(Translate(productname,'Ø. &/+-,%','O')) as tootjakood,
but it returns error
ERROR: column "Ø, 12.3/3mm" does not exist
testcase:
create temp table sales ( saledate date, productname char(20), quantity int );
insert into sales values ( '2016-1-1', 'Ø 12.3/3mm', 2);
insert into sales values ( '2016-1-1', '+-3,4%/3mm', 52);
insert into sales values ( '2016-1-3', '/3,2m-', 246);
do $do$
declare
voter_list text;
begin
create temp table myyk on commit drop as
select saledate as kuupaev,
'C'||upper(Translate(productname,'Ø. &/+-,%','O')) as tootjakood,
sum(quantity)::int as kogus
from sales
group by 1,2
;
drop table if exists pivot;
voter_list := (
select string_agg(distinct tootjakood, ' ' order by tootjakood) from myyk
);
execute(format('
create table pivot (
kuupaev date,
%1$s
)', (replace(voter_list, ' ', ' integer, ') || ' integer')
));
execute (format($f$
insert into pivot
select
kuupaev,
%2$s
from crosstab($ct$
select
kuupaev,tootjakood,kogus
from myyk
order by 1
$ct$,$ct$
select distinct tootjakood
from myyk
order by 1
$ct$
) as (
kuupaev date,
%4$s
);$f$,
replace(voter_list, ' ', ' + '),
replace(voter_list, ' ', ', '),
'',
replace(voter_list, ' ', ' integer, ') || ' integer' -- 4.
));
end; $do$;
select * from pivot;
Postgres 9.1 is used.