How to pivot table without joining itself multiple times - postgresql

Question

I have a table ctrlr_err_hist that has the following columns:
visit_id
err_cd
err_name
m1_err_cnt
m2_err_cnt
The possible values for err_cd are 0, 1, 3.
Now I want to write a query to have different errors with same visit_id on the same row. For example if the table has following records:
visit_id err_cd err_name m1_err_cnt
1 0 encoder 500
1 3 breakout 212
2 1 obclose 45
2 3 breakout 143
This table will be joined with a tchn_visit_hist table on visit_id. If there is no data for any of the err_cd, it will be null. The end result would look like this:
visit_id encoder_cnt breakout_cnt obclose_cnt
1 500 212 null
2 null 143 45
I wrote the following query which joins same table three times
select t.visit_id, t.door_id, enc.encoder_cnt, brk.breakout_cnt, ob.ob_close_cnt
from tchn_visit_hist t
left join (
select m1_err_cnt as encoder_cnt, visit_id
from ctrlr_err_hist
where err_cd = '0'
) as enc
on t.visit_id = enc.visit_id
left join (
select m1_err_cnt as breakout_cnt, visit_id
from ctrlr_err_hist
where err_cd = '3'
) as brk
on t.visit_id = brk.visit_id
left join (
select m1_err_cnt as ob_close_cnt, visit_id
from ctrlr_err_hist
where err_cd = '1'
) as ob
on t.visit_id = ob.visit_id
I wonder if there is a better, more efficient way to accomplish this.

Answer 1

demo: db<>fiddle
A simple pivot can be achieved by this way:
SELECT
visit_id,
MIN(m1_err_cnt) FILTER (WHERE err_cd = 0) as encoder_cnt,
MIN(m1_err_cnt) FILTER (WHERE err_cd = 1) as ob_close_cnt,
MIN(m1_err_cnt) FILTER (WHERE err_cd = 3) as breakout_cnt
FROM
ctrlr_err_hist
GROUP BY visit_id
Grouping by visit_id and aggregating the columns your are interested in. The FILTER clause filters the element that should be aggregated. In this case the aggregation is only for one special err_cd.
Notice that you have to use an aggregation function. This is because theoretical you could have two or more rows with (visit_id = 1, err_cd = 0). In that case you have to decide what you want to do with several values (SUM, MIN, MAX, AVG, whatever). Because your example contains distinct rows it doesn't matter.
After that you simply join against your tchn_visit_hist table:
SELECT t.visit_id, t.door_id, c.encoder_cnt, c.breakout_cnt, c.ob_close_cnt
FROM (
SELECT
visit_id,
MIN(m1_err_cnt) FILTER (WHERE err_cd = 0) as encoder_cnt,
MIN(m1_err_cnt) FILTER (WHERE err_cd = 1) as ob_close_cnt,
MIN(m1_err_cnt) FILTER (WHERE err_cd = 3) as breakout_cnt
FROM
ctrlr_err_hist
GROUP BY visit_id
) c
JOIN tchn_visit_hist t
ON (t.visit_id = c.visit_id)

Answer 2

You can make Condition in Join
select
a.visit_id,
max(b.encoder_cnt) as encoder_cnt,
max(b.breakout_cnt) as breakout_cnt,
max(b.ob_close_cnt) as ob_close_cnt
from
tchn_visit_hist a
left join(select
case when err_cd = '0' then m1_err_cnt end as encoder_cnt,
case when err_cd = '3' then m1_err_cnt end as breakout_cnt,
case when err_cd = '1' then m1_err_cnt end as ob_close_cnt,
case when err_cd = '0' then visit_id end as encoder_visit,
case when err_cd = '3' then visit_id end as breakout_visit,
case when err_cd = '1' then visit_id end as ob_close_visit
from
tchn_visit_hist) b on case
when a.err_cd = '0' then a.visit_id = encoder_visit
when a.err_cd = '3' then a.visit_id = breakout_visit
when a.err_cd = '1' then a.visit_id = ob_close_visit
end
group by
a.visit_id
order by
a.visit_id asc
Here the Demo