SQL pivoted table is read-only and cells can't be edited? - sql

Question

If I create a VIEW using this pivot table query, it isn't editable. The cells are read-only and give me the SQL2005 error: "No row was updated. The data in row 2 was not committed. Update or insert of view or function 'VIEWNAME' failed because it contains a derived or constant field."
Any ideas on how this could be solved OR is a pivot like this just never editable?
SELECT n_id,
MAX(CASE field WHEN 'fId' THEN c_metadata_value ELSE ' ' END) AS fId,
MAX(CASE field WHEN 'sID' THEN c_metadata_value ELSE ' ' END) AS sID,
MAX(CASE field WHEN 'NUMBER' THEN c_metadata_value ELSE ' ' END) AS NUMBER
FROM metadata
GROUP BY n_id

Answer 1

Assuming you have a unique constraint on n_id, field which means that at most one row can match you can (in theory at least) use an INSTEAD OF trigger.
This would be easier with MERGE (but that is not available until SQL Server 2008) as you need to cover UPDATES of existing data, INSERTS (Where a NULL value is set to a NON NULL one) and DELETES where a NON NULL value is set to NULL.
One thing you would need to consider here is how to cope with UPDATES that set all of the columns in a row to NULL I did this during testing the code below and was quite confused for a minute or two until I realised that this had deleted all the rows in the base table for an n_id (which meant the operation was not reversible via another UPDATE statement). This issue could be avoided by having the VIEW definition OUTER JOIN onto what ever table n_id is the PK of.
An example of the type of thing is below. You would also need to consider potential race conditions in the INSERT/DELETE code indicated and whether you need some additional locking hints in there.
CREATE TRIGGER trig
ON pivoted
INSTEAD OF UPDATE
AS
BEGIN
SET nocount ON;
DECLARE #unpivoted TABLE (
n_id INT,
field VARCHAR(10),
c_metadata_value VARCHAR(10))
INSERT INTO #unpivoted
SELECT *
FROM inserted UNPIVOT (data FOR col IN (fid, sid, NUMBER) ) AS unpvt
WHERE data IS NOT NULL
UPDATE m
SET m.c_metadata_value = u.c_metadata_value
FROM metadata m
JOIN #unpivoted u
ON u.n_id = m.n_id
AND u.c_metadata_value = m.field;
/*You need to consider race conditions below*/
DELETE FROM metadata
WHERE NOT EXISTS(SELECT *
FROM #unpivoted u
WHERE metadata.n_id = u.n_id
AND u.field = metadata.field)
INSERT INTO metadata
SELECT u.n_id,
u.field,
u.c_metadata_value
FROM #unpivoted u
WHERE NOT EXISTS (SELECT *
FROM metadata m
WHERE m.n_id = u.n_id
AND u.field = m.field)
END

Answer 2

You'll have to create trigger on view, because direct update is not possible:
CREATE TRIGGER TrMyViewUpdate on MyView
INSTEAD OF UPDATE
AS
BEGIN
SET NOCOUNT ON;
UPDATE MyTable
SET ...
FROM INSERTED...
END