postgres – aggregate items in text array

I have an array of text items, each text item in the array is a delimited string, here is a sample:

drop table if exists tmp;
create table tmp(x text);
insert into tmp select ('1~41.5~50~1|2~43.72~1000~1|3~52.0~1~1|4~57.5~7500~1|5~68.0~8~1|6~80.95~6~1|7~84.25~300~1');

with t as (
    select string_to_array(x , '|') arr 
    from  tmp
)
select * , cardinality(arr) arr_len , split_part(arr[1],'~',2)
from t

sample output:

"{1~41.5~50~1,2~43.72~1000~1,3~52.0~1~1,4~57.5~7500~1,5~68.0~8~1,6~80.95~6~1,7~84.25~300~1}";7;"41.5"

I’m interested in extracting the price, which is represented as the second item of each array entry (when delimited by ~), so that the output would be:

41.5,43.72,52.0,...

when using split_part, that return only one result of a specific array index, and I do not want to hard-code all indexes, since it varies.

suggestions are appreciated.

Answer

Here is one approach:

select t.*, w.prices
from tmp t
cross join lateral (
    select array_agg(split_part(v.z, '~', 2)) prices
    from regexp_split_to_table(t.x, '|') as v(z)
) w

Within the subquery, we split the string to rows, using separator |; then, we aggregate the second element of each row back into an array.

You can be explicit that you want to preserve the ordering of the elements with with ordinality:

select t.*, w.prices
from tmp t
cross join lateral (
    select array_agg(split_part(v.z, '~', 2) order by v.n) prices
    from regexp_split_to_table(t.x, '|') with ordinality as v(z, n)
) w