Converting Parent child rows to JSON in oracle - sql

Question

Is there a way to create JSON object in Oracle, for parent child relationship data? For example an organizational structure. Table contains
EmpId Name Title ManagerId
1 John GM 0
2 Smith Manager 1
3 Jason Manager 1
4 Will IP1 3
5 Jade AM 3
6 Mark IP2 5
7 Jane AM2 5
8 Tamara M1 1
9 Dory M2 1
Something like below JSON object is expected.
{
'name': 'John',
'title': 'GM',
'children': [
{ 'name': 'Smith', 'title': 'manager' },
{ 'name': 'Jason', 'title': 'manager',
'children': [
{ 'name': 'Will', 'title': 'IP1' },
{ 'name': 'Jade', 'title': 'AM',
'children': [
{ 'name': 'Mark', 'title': 'IP2' },
{ 'name': 'Jane', 'title': 'AM2' }
]
}
]
},
{ 'name': 'Tamara', 'title': 'M1' },
{ 'name': 'Dory', 'title': 'M2' }
]
}

Answer 1

Oracle Database 12.2 does have a number of JSON generation functions. But these are of limited use. You need to build up the document recursively.
Which I believe requires a bit of hand-crafting.
First use a recursive query to create the org chart, adding which level each person is in the hierarchy.
Then build the JSON by:
If level for the next row is greater than the current, the employee is a manager. And you need to start a child array. Otherwise return a JSON object for the current row
If the current row is the last in the tree, you need to close N arrays and objects. N is how deep the row is in the tree minus one.
Otherwise if the next row is a lower level than the current, you need to close ( current level - next level ) arrays and objects
Then if the next level equals or is less than the current, add a comma
Which gives something like:
create table t (
EmpId int,
Name varchar2(10),
Title varchar2(10),
ManagerId int
);
insert into t values (1, 'John', 'GM' , 0 );
insert into t values (2, 'Smith', 'Manager' , 1 );
insert into t values (3, 'Jason', 'Manager' , 1 );
insert into t values (4, 'Will', 'IP1' , 3 );
insert into t values (5, 'Jade', 'AM' , 3 );
insert into t values (6, 'Mark', 'IP2' , 5 );
insert into t values (7, 'Jane', 'AM2' , 5 );
insert into t values (8, 'Tamar', 'M1' , 1 );
insert into t values (9, 'Dory', 'M2' , 1 );
commit;
with chart (
empid, managerid, name, title, lvl
) as (
select empid, managerid,
name, title, 1 lvl
from t
where empid = 1
union all
select t.empid, t.managerid,
t.name, t.title,
lvl + 1 lvl
from chart c
join t
on c.empid = t.managerid
) search depth first by empid set seq,
jdata as (
select case
/* The employee has reports */
when lead ( lvl ) over ( order by seq ) > lvl then
'{"name": "' || name ||
'", "title": "' || title ||
'", "children": ['
else
json_object ( 'name' value name, 'title' value title )
end ||
case
/* Close arrays & objects */
when lead ( lvl ) over ( order by seq ) is null then
lpad ( ']}', ( lvl - 1 ) * 2, ']}' )
when lead ( lvl ) over ( order by seq ) < lvl then
lpad ( ']}', ( lvl - lead ( lvl ) over ( order by seq ) ) * 2, ']}' )
end ||
case
/* Add closing commas */
when lead ( lvl ) over ( order by seq ) <= lvl then
','
end j,
lead ( lvl ) over ( order by seq ) nlvl,
seq, lvl
from chart
)
select json_query (
listagg ( j )
within group ( order by seq ),
'$' returning varchar2 pretty
) chart_json
from jdata;
CHART_JSON
{
"name" : "John",
"title" : "GM",
"children" :
[
{
"name" : "Smith",
"title" : "Manager"
},
{
"name" : "Jason",
"title" : "Manager",
"children" :
[
{
"name" : "Will",
"title" : "IP1"
},
{
"name" : "Jade",
"title" : "AM",
"children" :
[
{
"name" : "Mark",
"title" : "IP2"
},
{
"name" : "Jane",
"title" : "AM2"
}
]
}
]
},
{
"name" : "Tamar",
"title" : "M1"
},
{
"name" : "Dory",
"title" : "M2"
}
]
}