SELECT dynamic json stored in SQL as part the select query - sql

Question

I know similar questions has been asked multiple times, however my scenario seems to be a bit different.
My database table is like this:
App ID | ID | JSONData | URL | CreatedOn
----------+-----------+-------------------+--------------+-----------------
5b5cd8 | 1 | {"F":"B", "S":"D"}| http://local | Mar 19 2018 13:04
5b5cd8 | 2 | {"F":"C", "S":"K"}| http://remote| Mar 29 2018 09:34
6b9df0 | 3 | {"T":"N", "D":"S"}| http://site | Apr 04 2018 16:12
App ID column can have varying values, however the structure of JSONData is (*supposed to be) same for the same App ID.
Is there anyway I can split the JSONData data and get a result like this?
App ID | ID | F | S | URL | CreatedOn
----------+-----------+-----+-----+--------------+-------------------------
5b5cd8 | 1 | B | D | http://local | Mar 19 2018 13:04
5b5cd8 | 2 | C | K | http://remote| Mar 29 2018 09:34
For the next App ID it is like this
App ID | ID | T | D | URL | CreatedOn
----------+-----------+-----+-----+--------------+-------------------------
6b9df0 | 3 | N | S | http://site | Apr 04 2018 16:12
Note: The data in JSONData field will mostly be one level deep i.e. all the data will be string and no further objects.
The solution I found most of the times like this, was either using static JSON key names to split, or creating temp table that is going to cause performance issues.

Answer 1

You were told already, that the column names of a result set must be known in advance.
The only workaround was dynamic SQL (creation of the statement as string and EXEC() to get its result). But this has some major draw-backs (and some advantages)...
You might go with something along this (needs SQL-Server 2016+):
A mockup-table
DECLARE #tbl TABLE(AppID VARCHAR(100),ID INT,JSONData NVARCHAR(MAX));
INSERT INTO #tbl VALUES
('5b5cd8',1,N'{"F":"B", "S":"D"}')
,('5b5cd8',2,N'{"F":"C", "S":"K"}')
,('6b9df0',3,N'{"T":"N", "D":"S"}');
--This query fetches the values using JSON_VALUE
--You'd need to create one statement for each possible list of columns
--Apply a WHERE to filter for appropriate rows
SELECT t.AppID
,t.ID
,JSON_VALUE(t.JSONData,'$.F') AS F
,JSON_VALUE(t.JSONData,'$.S') AS S
FROM #tbl t
WHERE t.AppID='5b5cd8'
--You might include all possible columns
--This works without a filter, but will return a lot of NULLs
SELECT t.AppID
,t.ID
,JSON_VALUE(t.JSONData,'$.F') AS F
,JSON_VALUE(t.JSONData,'$.S') AS S
,JSON_VALUE(t.JSONData,'$.T') AS T
,JSON_VALUE(t.JSONData,'$.D') AS D
FROM #tbl t
--A bit cleaner / better to read was OPENJSON() in connection with a WITH-clause
SELECT t.AppID
,t.ID
,JsonColumns.*
FROM #tbl t
CROSS APPLY OPENJSON(t.JSONData) WITH(F CHAR(1)
,S CHAR(1)
,T CHAR(1)
,D CHAR(1)) JsonColumns
My suggestion: Create the last one as a VIEW or (probably better) an iTVF and use dedicated statements against this, one per each type of structure.