How can I merge several XML-documents into one without namespace accumulation? - sql

Question

I have several SVG-segments in a table stored as xml-document.
Now I need to select all elements from that table, and merge them into one XML-document.
This is the T-SQL code I have:
declare #xml table (xmldocument xml)
insert #xml select '
<svg xmlns="http://www.w3.org/2000/svg" otherattrib="x">
<path id="789" data-objid="0000X2"></path>
</svg>'
insert #xml select '
<svg xmlns="http://www.w3.org/2000/svg" otherattrib="x">
<admin>
<g>
<path></path>
<path data-objid="0000X1"></path>
<path id="123" data-objid="0000X2"></path>
<path id="456" data-objid="0000X3"></path>
</g>
</admin>
<g>
<path></path>
<path data-objid="0000X1"></path>
<path id="789" data-objid="0000X2"></path>
<path id="abc" data-objid="0000X3"></path>
</g>
</svg>'
insert #xml select '
<svg xmlns="http://www.w3.org/2000/svg" otherattrib="x">
<path></path>
</svg>'
insert #xml select '
<svg xmlns="http://www.w3.org/2000/svg" otherattrib="x">
<path id="abc" data-objid="0000X3"></path>
</svg>'
--;WITH XMLNAMESPACES ('http://www.w3.org/2000/svg' AS svg)
;WITH XMLNAMESPACES (default 'http://www.w3.org/2000/svg')
--SELECT
--(
SELECT
--xmldocument
--,c.p.value('.', 'nvarchar(MAX)')
c.p.query('declare default element namespace "http://www.w3.org/2000/svg";.')
FROM #xml AS t
OUTER APPLY t.xmldocument.nodes('/svg//*') AS c(p)
FOR XML PATH(''), root('svg')
-- ) AS merged
But this produces
<svg xmlns="http://www.w3.org/2000/svg">
<path xmlns="http://www.w3.org/2000/svg" id="789" data-objid="0000X2" />
<admin xmlns="http://www.w3.org/2000/svg">
<g>
<path />
<path data-objid="0000X1" />
<path id="123" data-objid="0000X2" />
<path id="456" data-objid="0000X3" />
</g>
</admin>
<g xmlns="http://www.w3.org/2000/svg">
<path />
<path data-objid="0000X1" />
<path id="123" data-objid="0000X2" />
<path id="456" data-objid="0000X3" />
</g>
<path xmlns="http://www.w3.org/2000/svg" />
<path xmlns="http://www.w3.org/2000/svg" data-objid="0000X1" />
<path xmlns="http://www.w3.org/2000/svg" id="123" data-objid="0000X2" />
<path xmlns="http://www.w3.org/2000/svg" id="456" data-objid="0000X3" />
<g xmlns="http://www.w3.org/2000/svg">
<path />
<path data-objid="0000X1" />
<path id="789" data-objid="0000X2" />
<path id="abc" data-objid="0000X3" />
</g>
<path xmlns="http://www.w3.org/2000/svg" />
<path xmlns="http://www.w3.org/2000/svg" data-objid="0000X1" />
<path xmlns="http://www.w3.org/2000/svg" id="789" data-objid="0000X2" />
<path xmlns="http://www.w3.org/2000/svg" id="abc" data-objid="0000X3" />
<path xmlns="http://www.w3.org/2000/svg" />
<path xmlns="http://www.w3.org/2000/svg" id="abc" data-objid="0000X3" />
</svg>
instead of
<svg xmlns="http://www.w3.org/2000/svg">
<path id="789" data-objid="0000X2" />
<admin>
<g>
<path />
<path data-objid="0000X1" />
<path id="123" data-objid="0000X2" />
<path id="456" data-objid="0000X3" />
</g>
</admin>
<g>
<path />
<path data-objid="0000X1" />
<path id="123" data-objid="0000X2" />
<path id="456" data-objid="0000X3" />
</g>
<path />
<path data-objid="0000X1" />
<path id="123" data-objid="0000X2" />
<path id="456" data-objid="0000X3" />
<g>
<path />
<path data-objid="0000X1" />
<path id="789" data-objid="0000X2" />
<path id="abc" data-objid="0000X3" />
</g>
<path />
<path data-objid="0000X1" />
<path id="789" data-objid="0000X2" />
<path id="abc" data-objid="0000X3" />
<path />
<path id="abc" data-objid="0000X3" />
</svg>
What am I missing ? What do I do wrong ?
How to correct this without having to cast to varchar and then do a search-and-replace on "xmlns=..." ?

Answer 1

First of all: The repeated namespaces are not wrong in any way, just annoying and bloating the size of your result...
Unfortunately there is no way to get rid of them in a clean way. Your idea of casting to text and do this on string level is not that bad, if you really need this (but be aware, that casting must go to NVARCHAR and the re-cast to XML can change your XML stucturally (attribute order, CDATA-sections...). Try this:
Short approach
...if you really do not need anything else, than the namespace in the first root <svg>...
DECLARE #NewXML NVARCHAR(MAX)=
(
SELECT t.xmldocument.query('declare default element namespace "http://www.w3.org/2000/svg";svg/*')
FROM #xml AS t
FOR XML PATH('')
);
SELECT CAST(N'<svg xmlns="http://www.w3.org/2000/svg">'
+ REPLACE(#NewXML,' xmlns="http://www.w3.org/2000/svg"','')
+ N'</svg>' AS XML);
More flexible approach
You might try this:
DECLARE #NewXML XML=
(
SELECT t.xmldocument.query('declare default element namespace "http://www.w3.org/2000/svg";svg/*')
FROM #xml AS t
FOR XML PATH(''),TYPE
);
SET #NewXML =CAST(REPLACE(CAST(#NewXML AS NVARCHAR(MAX)),' xmlns="http://www.w3.org/2000/svg"','') AS XML);
--You need to repeat the CAST and REPLACE, otherwise you'd get a lot of xmlns="" defining an empty default namespace for the inner nodes, which was wrong...
WITH XMLNAMESPACES (DEFAULT 'http://www.w3.org/2000/svg')
SELECT #NewXML=CAST(REPLACE(CAST((SELECT #NewXML FOR XML PATH('svg'),TYPE) AS NVARCHAR(MAX)),' xmlns=""','') AS XML);
SELECT #NewXML;
You might use STUFF() to introduce the namespace on string level:
(Don't use WITH XMLNAMESPACES here...
SELECT #NewXML=CAST(STUFF(CAST((SELECT #NewXML FOR XML PATH('svg'),TYPE) AS NVARCHAR(MAX)),5,0,' xmlns="http://www.w3.org/2000/svg" ') AS XML);
SELECT #NewXML;
The result in all cases
<svg xmlns="http://www.w3.org/2000/svg">
<path id="789" data-objid="0000X2" />
<admin>
<g>
<path />
<path data-objid="0000X1" />
<path id="123" data-objid="0000X2" />
<path id="456" data-objid="0000X3" />
</g>
</admin>
<g>
<path />
<path data-objid="0000X1" />
<path id="789" data-objid="0000X2" />
<path id="abc" data-objid="0000X3" />
</g>
<path />
<path id="abc" data-objid="0000X3" />
</svg>