Shred JSON array into child tables using SQL Server functions - arrays

Question

Here is JSON I would like to shred into three tables using SQL Server JSON functions:
{
"school" : "Ecole",
"classes": [
{
"className": "Math",
"Students": ["LaPlace", "Fourier","Euler","Pascal"]
}
{
"className": "Science",
"Students": ["Newton", "Einstein","Al-Biruni", "Cai"]
},
]
}
Table 1
+-------+--------+
| ID | school |
+-------+--------+
Table 2
+-------+---------------+-----------+
| ID | schoolID (FK) | className |
+-------+---------------+-----------+
Table 3
+-------+---------------+-----------+
| ID | classID (FK) | student |
+-------+---------------+-----------+
My queries so far:
SELECT * FROM OPENJSON(#json, '$.school') --Returns the name of the school
SELECT
ClassName = JSON_VALUE(c.value, '$.className'),
Students = JSON_QUERY(c.value, '$.Students')
FROM
OPENJSON(#json, '$.classes') c
-- Returns the name of the class and a JSON array of students.
I am wondering how use SQL to shred the JSON array to extract the data for the third table so that it looks like this:
Math class Id = 1
Science class Id =2
Id ClassId Student
+-------+--------+-----------+
| 1 | 1 | LaPlace |
+-------+--------+-----------+
| 2 | 1 | Fourier |
+-------+--------+-----------+
| 3 | 1 | Euler |
+-------+--------+-----------+
| 4 | 1 | Pascal |
+-------+--------+-----------+
| 5 | 2 | Newton |
+-------+--------+-----------+
| 6 | 2 | Einstein |
+-------+--------+-----------+
| 7 | 2 | Al-Biruni |
+-------+--------+-----------+
| 8 | 2 | Cai |
+-------+--------+-----------+
I can get the Ids from the other tables, but I don't know how to write a query to extract the students from the JSON arrays.
I do have the ability to restructure the JSON schema so that instead of arrays of strings, I could make arrays of objects:
"Students": [{"StudentName"}:"Newton", {"StudentName":"Einstein"},{"StudentName":"Al-Biruni"}, {"StudentName":"Cai"}]
But I am not certain that makes it any easier. Either way, I would still like to know how to write a query to accomplish the first case.

Answer 1

JSON is supported starting with SQL-Server 2016.
As your JSON is deeper nested (array of classes contains an array of students) I'd solve this with a combination of OPENJSON and a WITH-clause. Please look a bit closer to the AS JSON in the WITH-clause. This will allow for another CROSS APPLY OPENJSON(), hence moving deeper and deeper into your JSON-structure.
DECLARE #json NVARCHAR(MAX) =
N'{
"school" : "Ecole",
"classes": [
{
"className": "Math",
"Students": ["LaPlace", "Fourier","Euler","Pascal"]
},
{
"className": "Science",
"Students": ["Newton", "Einstein","Al-Biruni", "Cai"]
}
]
}';
--The query
SELECT ROW_NUMBER() OVER(ORDER BY B.className,C.[key]) AS RowId
,A.school
,B.className
,CASE B.className WHEN 'Math' THEN 1 WHEN 'Science' THEN 2 ELSE 0 END AS ClassId
,C.[key] AS StudentIndex
,C.[value] AS Student
FROM OPENJSON(#json)
WITH(school NVARCHAR(MAX)
,classes NVARCHAR(MAX) AS JSON) A
CROSS APPLY OPENJSON(A.classes)
WITH(className NVARCHAR(MAX)
,Students NVARCHAR(MAX) AS JSON) B
CROSS APPLY OPENJSON(B.Students) C
The result
+-------+--------+-----------+---------+--------------+-----------+
| RowId | school | className | ClassId | StudentIndex | Student |
+-------+--------+-----------+---------+--------------+-----------+
| 1 | Ecole | Math | 1 | 0 | LaPlace |
+-------+--------+-----------+---------+--------------+-----------+
| 2 | Ecole | Math | 1 | 1 | Fourier |
+-------+--------+-----------+---------+--------------+-----------+
| 3 | Ecole | Math | 1 | 2 | Euler |
+-------+--------+-----------+---------+--------------+-----------+
| 4 | Ecole | Math | 1 | 3 | Pascal |
+-------+--------+-----------+---------+--------------+-----------+
| 5 | Ecole | Science | 2 | 0 | Newton |
+-------+--------+-----------+---------+--------------+-----------+
| 6 | Ecole | Science | 2 | 1 | Einstein |
+-------+--------+-----------+---------+--------------+-----------+
| 7 | Ecole | Science | 2 | 2 | Al-Biruni |
+-------+--------+-----------+---------+--------------+-----------+
| 8 | Ecole | Science | 2 | 3 | Cai |
+-------+--------+-----------+---------+--------------+-----------+

Answer 2

Something like this:
declare #json nvarchar(max) = N'
{
"school" : "Ecole",
"classes": [
{
"className": "Math",
"Students": ["LaPlace", "Fourier","Euler","Pascal"]
},
{
"className": "Science",
"Students": ["Newton", "Einstein","Al-Biruni", "Cai"]
}
]
}
';
with q as
(
SELECT
ClassID = c.[key]+1,
ClassName = JSON_VALUE(c.value, '$.className'),
Id = row_number() over (order by c.[Key], students.[key] ),
Student = students.value
FROM
OPENJSON(#json, '$.classes') c
cross apply openjson(c.value,'$.Students') students
)
select Id, ClassId, Student
from q
/*
Id ClassId Student
----------- ----------- -----------
1 1 LaPlace
2 1 Fourier
3 1 Euler
4 1 Pascal
5 2 Newton
6 2 Einstein
7 2 Al-Biruni
8 2 Cai
*/