PosgtreSQL join tables on json field - sql

Question

Let's say we have 2 tables-
Employee:
id integer (pk)
name char
code char
Status:
id integer (pk)
key char
data jsnob
Now here is the sample data of above tables:
Employee
+----+--------+------+
| id | name | code |
+----+--------+------+
| 1 | Brian | BR1 |
| 2 | Andrew | AN1 |
| 3 | Anil | AN2 |
| 4 | Kethi | KE1 |
| 5 | Smith | SM1 |
+----+--------+------+
Status
+----+---------+---------------------------------------+
| id | key | data |
+----+---------+---------------------------------------+
| 1 | Admin | {'BR1':true, 'AN1':true,'KE1':false} |
| 2 | Staff | {'SM1':true, 'AN2':true,'KE1':false} |
| 3 | Member | {'AN2':false, 'AN1':true,'KE1':false} |
| 4 | Parking | {'BR1':true, 'AN1':true,'KE1':false, |
| | | 'AN2':true,'SM1':true} |
| 5 | System | {'AN2':false, 'AN1':true,'KE1':true} |
| 6 | Ticket | {'AN2':false, 'AN1':true,'KE1':false} |
+----+---------+---------------------------------------+
Now my goal is to get status and name of failure keys, employee code wise. For ex:-
I am not an expert in sql complex queries, so any help is much appreciate.
Note: Above are just sample tables (name and data changed), but design is similar to original tables.

Answer 1

you can use the function jsonb_each_text yo get key and value from jsonb type, and if we mix with sql and ... voilà, the following query is a example for you case :
select employee.code,
case
when dat2.count is null then 'TRUE'
else
'FALSE'
end as status,
case
when dat2.count is null then 0
else
dat2.count
end as failures, string_agg as key from employee left join
(
select key, count(*), string_agg(code,',') from (
select key code , (jsonb_each_text(data)).key,(jsonb_each_text(data)).value
from status) as dat
where value='false'
group by 1 ) dat2 on (employee.code=dat2.key)

Answer 2

As far as I can tell, you don't need the employee table for this (because you only want the code column which is also present in the JSON values of the status table). It's enough to unnest the JSON value and then aggregate on the code from that.
select d.code,
bool_and(d.flag::boolean) as status_flag,
count(*) filter (where not d.flag::boolean) as failures,
coalesce(string_agg(key, ', ') filter (where not d.flag::boolean), 'N/A') as keys
from status st
join lateral jsonb_each_text(st.data) as d(code, flag) on true
group by d.code
order by d.code;
The filter() option is used to only include rows in the aggregate that comply with the where condition. In this case those where the value for the code is false.
bool_and is an aggregate function for boolean values that returns true if all input values are true (and false otherwise)
Online example: https://rextester.com/PEKCZ52605

Answer 3

Use a left-join query between those tables, and apply jsonb_each_text() function for jsonb type column.
The trick is to use conditionals as case when (js).value = 'false' then .. else .. end for the aggregated columns :
select e.id, e.code,
min(case when (js).value = 'false' then 'FALSE' else 'TRUE' end ) as status,
count(case when (js).value = 'false' then 1 end) as failures,
coalesce(
string_agg(case when (js).value = 'false' then s.key end, ',' ORDER BY s.id),'NA'
) as key
from Employee e
left join
(
select *, jsonb_each_text(data) as js
from Status
) s on e.code = (js).key
group by e.id, e.code
order by e.id;
where (js).value is extracted from jsonb type Status.data column
Demo