Comma separated string to JSON object - sql

Question

I need to update/migrate a table IdsTable in my SQL Server database which has the following format:
+----+------------------+---------+
| id | ids | idType |
+----+------------------+---------+
| 1 | id11, id12, id13 | idType1 |
| 2 | id20 | idType2 |
+----+------------------+---------+
The ids column is a comma separate list of ids. I need to combine the ids and idType column to form a single JSON string for each row and update the ids column with that object.
The JSON object has the following format:
{
"idType": string,
"ids": string[]
}
Final table after transforming/migrating data should be:
+----+-----------------------------------------------------+---------+
| id | ids | idType |
+----+-----------------------------------------------------+---------+
| 1 | {"idType": "idType1","ids": ["id11","id12","id13"]} | idType1 |
| 2 | {"idType": "idType2","ids": ["id20"]} | idType2 |
+----+-----------------------------------------------------+---------+
The best I've figured out so far is to get the results into a format where I could GROUP BY id to try and get the correct JSON format:
SELECT X.id, Y.value, X.idType
FROM
IdsTable AS X
CROSS APPLY STRING_SPLIT(X.ids, ',') AS Y
Which gives me the results:
+----+------+---------+
| id | ids | idType |
+----+------+---------+
| 1 | id11 | idType1 |
| 1 | id12 | idType1 |
| 1 | id13 | idType1 |
| 2 | id20 | idType2 |
+----+------+---------+
But I'm not familiar enough with SQL Server JSON to move forward.

Answer 1

If it's a one-off op I think I'd just do it the dirty way:
UPDATE table SET ids =
CONCAT('{"idType": "', idType, '","ids": ["', REPLACE(ids, ', ', '","'), '"]}'
You might need to do some prep first, like if your ids column can look like:
id1,id2,id3
id4, id5, id6
id7 ,id8 , id9
etc, a series of replacements like:
UPDATE table SET ids = REPLACE(ids, ' ,', ',') WHERE ids LIKE '% ,%'
UPDATE table SET ids = REPLACE(ids, ', ', ',') WHERE ids LIKE '%, %'
Keep running those until they don't update any more records
ps; if you've removed all the spaces from around the commas, you'll need to tweak the REPLACE in the original query - I specified ', ' as the needle

Answer 2

I found this blog post that helped me construct my answer:
-- Create Temporary Table
SELECT
[TAB].[id], [TAB].[ids],
(
SELECT [STRING_SPLIT_RESULTS].value as [ids], [TAB].[idType] as [idType]
FROM [IdsTable] AS [REQ]
CROSS APPLY STRING_SPLIT([REQ].[ids],',') AS [STRING_SPLIT_RESULTS]
FOR JSON PATH
) as [newIds]
INTO [#TEMP_RESULTS]
FROM [IdsTable] AS [TAB]
-- Update rows
UPDATE [IdsTable]
SET [ids] = [#TEMP_RESULTS].[newIds]
FROM [#TEMP_RESULTS]
WHERE [IdsTable].[Id] = [#TEMP_RESULTS].[Id]
-- Delete Temporary Table
DROP TABLE [#TEMP_RESULTS]
Which replaces those ids column (not replaced below for comparison):
+----+----------------+---------+------------------------------------------------------------------------------------------------------+
| id | ids | idType | newIds |
+----+----------------+---------+------------------------------------------------------------------------------------------------------+
| 1 | id11,id12,id13 | idType1 | [{"id":"id11","idType":"idType1"},{"id":"id12","idType":"idType1"},{"id":"id13","idType":"idType1"}] |
| 2 | id20 | idType2 | [{"id":"id20","idType":"idType2"}] |
+----+----------------+---------+------------------------------------------------------------------------------------------------------+
This is more verbose that I wanted but considering the table size and the number of ids stored in the ids column which translates to the size of the JSON object, this is fine for me.