I have following table in Postgres
| phone | group | spec |
| 1 | 1 | 'Lock' |
| 1 | 2 | 'Full' |
| 1 | 3 | 'Face' |
| 2 | 1 | 'Lock' |
| 2 | 3 | 'Face' |
| 3 | 2 | 'Scan' |
Tried this
SELECT phone, string_agg(spec, ', ')
FROM mytable
GROUP BY phone;
Need this ouput for each phone where there is empty string for missing group.
| phone | spec
| 1 | Lock, Full, Face
| 2 | Lock, '' , Face
| 3 | '', Scan ,''
You need a CTE which returns all possible combinations of phone and group and a left join to the table so you can group by phone:
with cte as (
select *
from (
select distinct phone from mytable
) m cross join (
select distinct "group" from mytable
) g
)
select c.phone, string_agg(coalesce(t.spec, ''''''), ',') spec
from cte c left join mytable t
on t.phone = c.phone and t."group" = c."group"
group by c.phone
See the demo.
Results:
| phone | spec |
| ----- | -------------- |
| 1 | Lock,Full,Face |
| 2 | Lock,'',Face |
| 3 | '',Scan,'' |
You can use conditional aggregation:
select phone,
(max(case when group = 1 then spec else '''''' end) || ', ' ||
max(case when group = 2 then spec else '''''' end) || ', ' ||
max(case when group = 3 then spec else '''''' end)
) as specs
from mytable t
group by phone;
Alternatively, you can general all the groups using generate_series() and then aggregation:
select p.phone,
string_agg(coalesce(t.spec, ''''''), ', ') as specs
from (select distinct phone from mytable) p cross join
generate_series(1, 3, 1) gs(grp) left join
mytable t
on t.phone = p.phone and t.group = gs.grp
group by p.phone
You can consider using a self - (RIGHT/LEFT)JOIN with all three distinct groups (which's stated within the subquery just after RIGHT JOIN keywords ) and a correlated query for your table :
WITH mytable1 AS
(
SELECT distinct t1.phone, t2."group",
( SELECT spec FROM mytable WHERE phone = t1.phone AND "group"=t2."group" )
FROM mytable t1
RIGHT JOIN ( SELECT distinct "group" FROM mytable ) t2
ON t2."group" = coalesce(t2."group",t1."group")
)
SELECT phone, string_agg(coalesce(spec,''''''), ', ') as spec
FROM mytable1
GROUP BY phone;
