Pivot TSQL table with dynamic columns - sql

Question

I am passing up JSON as a parameter to a SQL stored proc. I use a function that takes a JSON dictionary and creates a table with key value pairs (two columns) that I then use with a COALESCE to create dynamic sql for an INSERT statement. This works fine for a single dictionary, but I need to also be able to send up JSON string that contains an array of dictionaries. Right now, my parse function gives me this table variable:
What I need is a table variable like this:
I can get the column names from the first table with this:
SELECT DISTINCT element name from #JSONTable
I should mention that these elementname's can and will change. I will not know the number of distinct elementname values.
UPDATE - Using Umair's answer, I am getting close:
DECLARE #JSONString AS VARCHAR(MAX)
SET #JSONString = '[{"ravid":3,"ravversion":2,"taskid":3},{"ravid":4,"ravversion":7,"taskid":99}]'
IF OBJECT_ID('tempdb..#JSONTable') IS NOT NULL
DROP TABLE #JSONTable
CREATE TABLE #JSONTable
(
elementname VARCHAR(500) ,
elementvalue VARCHAR(500)
)
INSERT INTO #JSONTable
( elementname ,
elementvalue
)
SELECT NAME ,
StringValue
FROM dbo.parseJSON(#JSONString)
WHERE LEN(StringValue) > 0 AND NAME IS NOT NULL
--declare a csv variable with all the distinct elements
DECLARE #csv NVARCHAR(max) = STUFF(
(
SELECT ',' + elementname
FROM (
SELECT DISTINCT elementname
FROM #JSONTable
) AS e
FOR XML PATH(''), TYPE
).value('.', 'nvarchar(max)'),
1,
1,
''
);
DECLARE #sql NVARCHAR(MAX) = '
SELECT *
FROM (
SELECT *, Row = ROW_NUMBER() OVER (PARTITION BY elementname ORDER BY elementname)
FROM #JSONTable
) AS t
PIVOT (
MAX(elementvalue)
FOR elementname IN (' + #csv + ')
) AS p
'
EXEC sp_executesql #sql
But the dictionary values don't correspond to the key. Here is the results of Umair's current answer:

Answer 1

How about
--declare a csv variable with all the distinct elements
DECLARE #csv NVARCHAR(max) = STUFF(
(
SELECT ',' + elementname
FROM (
SELECT DISTINCT elementname
FROM #JSONTable
) AS e
FOR XML PATH(''), TYPE
).value('.', 'nvarchar(max)'),
1,
1,
''
);
DECLARE #sql NVARCHAR(MAX) = '
SELECT *
FROM (
SELECT *, Row = ROW_NUMBER() OVER (PARTITION BY elementname ORDER BY elementvalue)
FROM #JSONTable
) AS t
PIVOT (
MAX(elementvalue)
FOR elementname IN (' + #csv + ')
) AS p
'
EXEC sp_executesql #sql
i.e dynamic pivoting :)
Edit:
Since you do not want to aggregate by anything in the pivot, I added a row number function to assign each distinct elementname a sequentially increasing id (based on element value). This will essentially group the pivot by this row column, producing all the required rows.