Replace multiple values in XML column based on a mapping table - sql

Question

Suppose I have a mapping table called tblMap that just maps an old attribute ID to a new attribute ID (oldID -> newID). NOTEWORTHY: newID is not contained in the list of oldID's.
I then have a table tblData which contains an xml string that has a number of attribute ids. I would like to replace all the current attribute ids with the newIDs that are found in tblMap. If an id mapping is not found in tblMap then it should stay as is. Any hints on how I can achieve this?
WHAT I TRIED:
I was trying to coerce something using XMLText.modify('replace value of ...') as described in: This StackOverflow Article but haven't been successful at getting it to work.
CREATE TABLE tblmap (
oldid INT,
newid INT
)
GO
INSERT INTO tblMap
VALUES
( 58, 1002),
( 85, 5002),
( 70, 3202),
(2, 2340),
(5, 7432)
GO
CREATE TABLE tblData ( [SourceID] int, [SourceRecID] bigint, [Value] xml )
GO
INSERT INTO tblData
VALUES
( 1, 0, N'<attributes><attribute id="58" value="0" /><attribute id="86" value="1" /><attribute id="85" value="1" /><attribute id="70" value="0" /><attribute id="38" value="0" /><attribute id="68" value="0" /><attribute id="42" value="1" /><attribute id="67" value="1" /><attribute id="62" value="1" /></attributes>' ),
( 1, 686, N'<attributes><attribute id="1" value="0.25" /><attribute id="4" value="1" /><attribute id="10" value="3" /><attribute id="11" value="1" /><attribute id="12" value="6" /></attributes>' ),
( 1, 687, N'<attributes><attribute id="1" value="2.00" /><attribute id="2" value="60.00" /><attribute id="3" value="-1" /><attribute id="5" value="252.00" /><attribute id="6" value="0" /><attribute id="7" value="1" /><attribute id="9" value="1" /><attribute id="10" value="1" /><attribute id="11" value="2" /><attribute id="12" value="10" /></attributes>' ),
( 1, 688, N'<attributes><attribute id="1" value="2.00" /><attribute id="2" value="60.00" /><attribute id="3" value="-1" /><attribute id="5" value="252.00" /><attribute id="6" value="0" /><attribute id="7" value="1" /><attribute id="11" value="2" /><attribute id="12" value="10" /></attributes>' )
SELECT *
FROM tblMap
GO
SELECT *
FROM tblData
GO
I have constructed all the schema/sample data here for your convenience:
https://rextester.com/MUMI61854

Answer 1

I would try to completely recreate the entire XML (or rather the /attributes node) and update the table with new value:
declare #tblmap table (oldid INT, newid INT);
INSERT INTO #tblMap
VALUES
( 58, 1002),
( 85, 5002),
( 70, 3202),
(2, 2340),
(5, 7432);
declare #tblData table ([SourceID] int, [SourceRecID] bigint, [Value] xml);
INSERT INTO #tblData
VALUES
( 1, 0, N'<attributes><attribute id="58" value="0" /><attribute id="86" value="1" /><attribute id="85" value="1" /><attribute id="70" value="0" /><attribute id="38" value="0" /><attribute id="68" value="0" /><attribute id="42" value="1" /><attribute id="67" value="1" /><attribute id="62" value="1" /></attributes>' ),
( 1, 686, N'<attributes><attribute id="1" value="0.25" /><attribute id="4" value="1" /><attribute id="10" value="3" /><attribute id="11" value="1" /><attribute id="12" value="6" /></attributes>' ),
( 1, 687, N'<attributes><attribute id="1" value="2.00" /><attribute id="2" value="60.00" /><attribute id="3" value="-1" /><attribute id="5" value="252.00" /><attribute id="6" value="0" /><attribute id="7" value="1" /><attribute id="9" value="1" /><attribute id="10" value="1" /><attribute id="11" value="2" /><attribute id="12" value="10" /></attributes>' ),
( 1, 688, N'<attributes><attribute id="1" value="2.00" /><attribute id="2" value="60.00" /><attribute id="3" value="-1" /><attribute id="5" value="252.00" /><attribute id="6" value="0" /><attribute id="7" value="1" /><attribute id="11" value="2" /><attribute id="12" value="10" /></attributes>' );
SELECT * FROM #tblMap;
SELECT * FROM #tblData;
-- Update table with new XML
with cte as (
select d.*, (
select isnull(m.newid, a.c.value('./#id', 'int')) as [#id], a.c.value('./#value', 'nvarchar(max)') as [#value]
from d.Value.nodes('/attributes[1]/attribute') a(c)
left join #tblmap m on m.oldid = a.c.value('./#id', 'int')
for xml path('attribute'), type, root('attributes')
) as [NewValue]
from #tblData d
)
update c set Value = NewValue
from cte c;
-- New version
select * from #tblData;
(I have turned your tables into table variables, as it leaves zero footprint on the instance. Everything else is the same.)
Unfortunately, this approach can become prohibitively difficult to implement if your actual XML schema is more complex than your example shows, and involves additional unpredictable elements and / or attributes under the /attributes node. In that case, I would recommend either a FLWOR (which is slow and quite difficult to write, at least for me) or cursored update.
To DEBUG:
-- Update table with new XML
with cte as (
select d.*, (
select isnull(m.newid, a.c.value('./#id', 'int')) as [#id], a.c.value('./#value', 'nvarchar(max)') as [#value]
from d.Value.nodes('/attributes[1]/attribute') a(c)
left join #tblmap m on m.oldid = a.c.value('./#id', 'int')
for xml path('attribute'), type, root('attributes')
) as [NewValue]
from #tblData d
)
SELECT c.SourceID,
c.SourceRecID,
c.Value,
c.NewValue
from cte c;

Answer 2

My suggestion calls XQuery to the rescue (txh Roger Wolf for the declared table variables, used them too...):
declare #tblmap table (oldid INT, newid INT);
INSERT INTO #tblMap
VALUES
( 58, 1002),
( 85, 5002),
( 70, 3202),
(2, 2340),
(5, 7432);
declare #tblData table ([SourceID] int, [SourceRecID] bigint, [Value] xml);
INSERT INTO #tblData
VALUES
( 1, 0, N'<attributes><attribute id="58" value="0" /><attribute id="86" value="1" /><attribute id="85" value="1" /><attribute id="70" value="0" /><attribute id="38" value="0" /><attribute id="68" value="0" /><attribute id="42" value="1" /><attribute id="67" value="1" /><attribute id="62" value="1" /></attributes>' ),
( 1, 686, N'<attributes><attribute id="1" value="0.25" /><attribute id="4" value="1" /><attribute id="10" value="3" /><attribute id="11" value="1" /><attribute id="12" value="6" /></attributes>' ),
( 1, 687, N'<attributes><attribute id="1" value="2.00" /><attribute id="2" value="60.00" /><attribute id="3" value="-1" /><attribute id="5" value="252.00" /><attribute id="6" value="0" /><attribute id="7" value="1" /><attribute id="9" value="1" /><attribute id="10" value="1" /><attribute id="11" value="2" /><attribute id="12" value="10" /></attributes>' ),
( 1, 688, N'<attributes><attribute id="1" value="2.00" /><attribute id="2" value="60.00" /><attribute id="3" value="-1" /><attribute id="5" value="252.00" /><attribute id="6" value="0" /><attribute id="7" value="1" /><attribute id="11" value="2" /><attribute id="12" value="10" /></attributes>' );
--The query will do the whole process in one single pass
WITH CombineThem AS
(
SELECT d.SourceID
,d.SourceRecID
,d.[Value]
,(SELECT
(SELECT *
FROM #tblMap
FOR XML PATH('map'),ROOT('maps'),TYPE)
,[Value] AS [*]
FOR XML PATH('Combined'),TYPE) AS Combined
FROM #tblData d
)
,updateableCTE AS
(
SELECT ct.[Value]
,ct.Combined
.query('<attributes>
{
for $attr in /Combined/attributes/attribute
return <attribute id="{
(
/Combined/maps/map[oldid[1]=$attr/#id]/newid
,$attr/#id
)[1]
}"
value="{$attr/#value}"/>
}
</attributes>') NewValue
FROM CombineThem ct
)
UPDATE updateableCTE SET [Value]=NewValue;
--check the result
SELECT * FROM #tblData;
Some Explanation
In order to use the mapping and the data in XQuery I create a combined XML in the first CTE. This will include the full <attributes> element together with a <maps> element.
The .query() will run through the attributes and search the <maps> for a fitting re-mapping. The magic happens in (val1,val2)[1]. This behaves like COALESCE(). It will pick the first non-null-value, which is either the fitting new id or the existing value.
Instead of updating the XML using .modify() the final step will set the [Value] column to the newly created XML in one go.

Answer 3

Honestly, I'm not 100% the reliability on ORDER BY (SELECT NULL) here, however, I don't have many options other than to hope the order is that of the nodes.
Anyway, the solution involves dynamic SQL; there may be a "better" way of doing this, but if there is I don't know it. I suggest doing some decent testing first, however, this appears to get the results you're after:
DECLARE #SQL nvarchar(MAX);
SET #SQL = STUFF((SELECT NCHAR(10) +
N'UPDATE tblData' + NCHAR(10) +
N'SET [Value].modify(''replace value of (/attributes/attribute/#id)[' + CONVERT(varchar(4),ROW_NUMBER() OVER (PARTITION BY D.SourceID, D.SourceRecID ORDER BY (SELECT NULL))) + N'] with "' + CONVERT(varchar(4),ISNULL(M.newid,V.AA.value('#id','int'))) + N'"'')' + NCHAR(10) +
N'WHERE SourceID = ' + CONVERT(varchar(4),D.SourceID) + NCHAR(10) +
N' AND SourceRecID = ' + CONVERT(varchar(4),D.SourceRecID) + N';'
FROM tblData D
CROSS APPLY D.[Value].nodes('attributes/attribute') V(AA)
LEFT JOIN tblmap M ON V.AA.value('#id','int') = M.oldid
FOR XML PATH(N'')),1,1,N'');
EXEC sp_executesql #SQL;