There are a couple of non-trivial things in this request. First is to order the resulting rows by the document position, which is not visible when you use OPENJSON … WITH to project the columns. And the second one is that you need a hierarchical query (assuming there could be multiple levels).
Anyway, something like this:
declare @doc nvarchar(max) = N'[{"id":1},{"id":3},{"id":2,"children":[{"id":4},{"id":5}]}]';
with q as
(
select [key] nodePath,
cast(json_value(d.[value],'$.id') as int) Id,
cast(null as int) ParentId,
cast(json_query(d.[value],'$.children') as nvarchar(max)) children
from openjson(@doc) d
union all
select q.nodePath + '.' + d.[key] nodePath,
cast(json_value(d.[value],'$.id') as int) Id,
q.id ParentId,
cast(json_query(d.[value],'$.children') as nvarchar(max)) children
from q
outer apply openjson(q.children) d
where q.children is not null
)
select Id, row_number() over (order by nodePath) [Order/Index], ParentId
from q
order by [Order/Index]
outputs
Id Order/Index ParentId
----------- -------------------- -----------
1 1 NULL
3 2 NULL
2 3 NULL
4 4 2
5 5 2
(5 rows affected)
childrenon the first level only or might there be more and deeper nestedchildrentoo? In other words: Couldid=5havechildrenitself?