Is it possible to use an XML DML statement to rename an element in an untyped XML column?
I am in the process of updating an XML Schema Collection on an XML column and need to patch the existing XML instances by renaming one element before I can apply the latest schema.
As far as I can tell from the docs you can only insert / delete nodes or replace their value.
On
Yes you can use DML to rename an element by snipping it at the node you want renamed, injecting a new node at that element and then pasting the snipped elements back into the xml at that node. Ive done a SQL fiddle to demo. http://sqlfiddle.com/#!3/dc64d/1 This will change
<animal species="Mouse">
<legs>
<leg>Front Right</leg>
<leg>Front Left</leg>
<leg>Back Right</leg>
<leg>Back Left</leg>
</legs>
</animal>
into
<animal species="Mouse">
<armsandlegs>
<leg>Front Right</leg>
<leg>Front Left</leg>
<leg>Back Right</leg>
<leg>Back Left</leg>
</armsandlegs>
</animal>
SqlFiddle looks to have long since broken my solution. From memory ive pasted the basis of my solution below...
DECLARE @XML2 xml
DECLARE @XML3 xml = '<limbs></limbs>'
DECLARE @XML xml =
'<animal species="Mouse">
<legs>
<leg>Front Right</leg>
<leg>Front Left</leg>
<leg>Back Right</leg>
<leg>Back Left</leg>
</legs>
</animal>'
SET @XML2 = @XML.query('animal/legs/*')
SET @XML.modify('
insert
(sql:variable("@XML3"))
after
(/animal/legs)[1]
')
SET @XML.modify('
delete (/animal/legs[1])
')
SET @XML.modify('
insert
(sql:variable("@XML2"))
as last into
(/animal/limbs)[1]
')
select @XML
During development of SQL Server Unit Test (ssut - see related blog post) I wanted to standardize an xml set coming from a tested object. As I will call the tested object multiple times, each time the set and record names will be the same. For reading ease, I want the record set from the original records to be named similar to <original_record_set><original_record /></original_record_set> and the record set for
test records to be named similar to <test_record_set><test_record /></ test_record_set >.
Obviously this is trivial to do if you can modify the call in the tested object as first:
SET @output = (SELECT col1, col2
FROM @test_object_result
FOR xml path ( test_record '), root( test_record_set '));
and then:
SET @output = (SELECT col1, col2
FROM @test_object_result
FOR xml path ( original_record'), root( original_record_set '));
However, since I'm calling the SAME object multiple times, and "for xml path" does NOT allow variables in the path('...') and root('...') methods, I had to come up with a different method.
This function accepts an xml tree and builds a new tree, replacing the root node with the value of @relation_name and the name of each record with @tuple_name. The new tree is built with all the attributes of the original, even if there are different numbers per record.
EXCEPTIONS
Obviously this does NOT work with multiple element levels! I have built it specifically to handle a single level attribute based tree as shown in the example below. I may build it out for a multi-level mixed attribute/element tree in the future, but I think that the method to do so becomes obvious now that I've solved the basic problem as below, and will leave that exercise to the reader pending that time.
USE [unit_test];
GO
IF EXISTS (SELECT * FROM sys.objects WHERE object_id = OBJECT_ID(N'[dbo].[standardize_record_set]') AND type IN ( N'FN', N'IF', N'TF', N'FS', N'FT' ))
DROP FUNCTION [dbo].[standardize_record_set];
GO
SET ANSI_NULLS ON;
GO
SET QUOTED_IDENTIFIER ON;
GO
SET nocount ON;
GO
/*
DECLARE
@relation_name nvarchar(150)= N'standardized_record_set',
@tuple_name nvarchar(150)= N'standardized_record',
@xml xml,
@standardized_result xml;
SET @xml='<Root>
<row id="12" two="now1" three="thr1" four="four1" />
<row id="232" two="now22" three="thr22" />
<row id="233" two="now23" three="thr23" threeextra="extraattrinthree" />
<row id="234" two="now24" three="thr24" fourextra="mealsoin four rwo big mone" />
<row id="235" two="now25" three="thr25" />
</Root>';
execute @standardized_result = [dbo].[standardize_record_set] @relation_name=@relation_name, @tuple_name=@tuple_name, @xml=@xml;
select @standardized_result;
*/
CREATE FUNCTION [dbo].[standardize_record_set] (@relation_name nvarchar(150)= N'record_set',
@tuple_name nvarchar(150)= N'record', @xml xml )
returns XML
AS
BEGIN
DECLARE
@attribute_index int = 1,
@attribute_count int = 0,
@record_set xml = N'<' + @relation_name + ' />',
@record_name nvarchar(50) = @tuple_name,
@builder nvarchar(max),
@record xml,
@next_record xml;
DECLARE @record_table TABLE (
record xml );
INSERT INTO @record_table
SELECT t.c.query('.') AS record
FROM @xml.nodes('/*/*') T(c);
DECLARE record_table_cursor CURSOR FOR
SELECT cast([record] AS xml)
FROM @record_table
OPEN record_table_cursor
FETCH NEXT FROM record_table_cursor INTO @next_record
WHILE @@FETCH_STATUS = 0
BEGIN
SET @attribute_index=1;
SET @attribute_count = @next_record.query('count(/*[1]/@*)').value('.', 'int');
SET @builder = N'<' + @record_name + N' ';
-- build up attribute string
WHILE @attribute_index <= @attribute_count
BEGIN
SET @builder = @builder + @next_record.value('local-name((/*/@*[sql:variable("@attribute_index")])[1])',
'varchar(max)') + '="' + @next_record.value('((/*/@*[sql:variable("@attribute_index")])[1])',
'varchar(max)') + '" ';
SET @attribute_index = @attribute_index + 1
END
-- build record and add to record_set
SET @record = @builder + ' />';
SET @record_set.modify('insert sql:variable("@record") into (/*)[1]');
FETCH NEXT FROM record_table_cursor INTO @next_record
END
CLOSE record_table_cursor;
DEALLOCATE record_table_cursor;
RETURN @record_set;
END;
GO
As the saying goes, "Where there's a will there's a way"
Here's two methods: the first is to simply replace the previous xml with a new xml constructed from the original with the new element name. In my example I've changed Legs/Leg to Limbs/Limb this could get very complicated for anything but the simplest schema
And secondly, a more appropriate approach of combining insert and delete.
I've combined them into one simple example: