pgsql 1 to n relation into json - sql

Question

Long story short, how can i use 1 to n select data to build json like shown in example:
SELECT table1.id AS id1,table2.id AS id2,t_id,label
FROM table1 LEFT JOIN table2 ON table2.t_id = table1.id
result
|id1|id2|t_id|label|
+---+---+----+-----+
|1 | 1 | 1 | a |
| | 2 | 1 | b |
| | 3 | 1 | c |
| | 4 | 1 | d |
|2 | 5 | 2 | x |
| | 6 | 2 | y |
turn into this
SELECT table1.id, build_json(table2.id,table2.label) AS json_data
FROM table1 JOIN table2 ON table2.t_id = table1.id
GROUP BY table1.id
|id1|json_data
+--+-----------------
|1 |{"1":"a","2":"b","3":"c","4":"d"}
|2 |{"5":"x","6":"y"}
My guess the best start woulb be building an array from columns
Hstore instead of json would be ok too

Answer 1

well your table structure is a bit strange (is looks more like report than table), so I see two tasks here:
Replace nulls with correct id1. You can do it like this
with cte1 as (
select
sum(case when id1 is null then 0 else 1 end) over (order by t_id) as id1_partition,
id1, id2, label
from Table1
), cte2 as (
select
first_value(id1) over(partition by id1_partition) as id1,
id2, label
from cte1
)
select *
from cte2
Now you have to aggregate data into json. As far as I remember, there's no such a function in PostgreSQL, so you have to concatenate data manually:
with cte1 as (
select
sum(case when id1 is null then 0 else 1 end) over (order by t_id) as id1_partition,
id1, id2, label
from Table1
), cte2 as (
select
first_value(id1) over(partition by id1_partition) as id1,
id2, label
from cte1
)
select
id1,
('{' || string_agg('"' || id2 || '":' || to_json(label), ',') || '}')::json as json_data
from cte2
group by id1
sql fiddle demo
And if you want to convert into hstore:
with cte1 as (
select
sum(case when id1 is null then 0 else 1 end) over (order by t_id) as id1_partition,
id1, id2, label
from Table1
), cte2 as (
select
first_value(id1) over(partition by id1_partition) as id1,
id2, label
from cte1
)
select
c.id1, hstore(array_agg(c.id2)::text[], array_agg(c.label)::text[])
from cte2 as c
group by c.id1
sql fiddle demo