Select XML nodes as rows - sql

Question

I am selecting from a table that has an XML column using T-SQL. I would like to select a certain type of node and have a row created for each one.
For instance, suppose I am selecting from a people table. This table has an XML column for addresses. The XML is formated similar to the following:
<address>
<street>Street 1</street>
<city>City 1</city>
<state>State 1</state>
<zipcode>Zip Code 1</zipcode>
</address>
<address>
<street>Street 2</street>
<city>City 2</city>
<state>State 2</state>
<zipcode>Zip Code 2</zipcode>
</address>
How can I get results like this:
Name City State
Joe Baker Seattle WA
Joe Baker Tacoma WA
Fred Jones Vancouver BC

Answer 1

Here is your solution:
/* TEST TABLE */
DECLARE #PEOPLE AS TABLE ([Name] VARCHAR(20), [Address] XML )
INSERT INTO #PEOPLE SELECT
'Joel',
'<address>
<street>Street 1</street>
<city>City 1</city>
<state>State 1</state>
<zipcode>Zip Code 1</zipcode>
</address>
<address>
<street>Street 2</street>
<city>City 2</city>
<state>State 2</state>
<zipcode>Zip Code 2</zipcode>
</address>'
UNION ALL SELECT
'Kim',
'<address>
<street>Street 3</street>
<city>City 3</city>
<state>State 3</state>
<zipcode>Zip Code 3</zipcode>
</address>'
SELECT * FROM #PEOPLE
-- BUILD XML
DECLARE #x XML
SELECT #x =
( SELECT
[Name]
, [Address].query('
for $a in //address
return <address
street="{$a/street}"
city="{$a/city}"
state="{$a/state}"
zipcode="{$a/zipcode}"
/>
')
FROM #PEOPLE AS people
FOR XML AUTO
)
-- RESULTS
SELECT [Name] = T.Item.value('../#Name', 'varchar(20)'),
street = T.Item.value('#street' , 'varchar(20)'),
city = T.Item.value('#city' , 'varchar(20)'),
state = T.Item.value('#state' , 'varchar(20)'),
zipcode = T.Item.value('#zipcode', 'varchar(20)')
FROM #x.nodes('//people/address') AS T(Item)
/* OUTPUT*/
Name | street | city | state | zipcode
~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~
Joel | Street 1 | City 1 | State 1 | Zip Code 1
Joel | Street 2 | City 2 | State 2 | Zip Code 2
Kim | Street 3 | City 3 | State 3 | Zip Code 3

Answer 2

Here's how I do it generically:
I shred the source XML via a call such as
DECLARE #xmlEntityList xml
SET #xmlEntityList =
'
<ArbitrarilyNamedXmlListElement>
<ArbitrarilyNamedXmlItemElement><SomeVeryImportantInteger>1</SomeVeryImportantInteger></ArbitrarilyNamedXmlItemElement>
<ArbitrarilyNamedXmlItemElement><SomeVeryImportantInteger>2</SomeVeryImportantInteger></ArbitrarilyNamedXmlItemElement>
<ArbitrarilyNamedXmlItemElement><SomeVeryImportantInteger>3</SomeVeryImportantInteger></ArbitrarilyNamedXmlItemElement>
</ArbitrarilyNamedXmlListElement>
'
DECLARE #tblEntityList TABLE(
SomeVeryImportantInteger int
)
INSERT #tblEntityList(SomeVeryImportantInteger)
SELECT
XmlItem.query('//SomeVeryImportantInteger[1]').value('.','int') as SomeVeryImportantInteger
FROM
[dbo].[tvfShredGetOneColumnedTableOfXmlItems] (#xmlEntityList)
by utilizing the scalar-valued function
/* Example Inputs */
/*
DECLARE #xmlListFormat xml
SET #xmlListFormat =
'
<ArbitrarilyNamedXmlListElement>
<ArbitrarilyNamedXmlItemElement>004421UB7</ArbitrarilyNamedXmlItemElement>
<ArbitrarilyNamedXmlItemElement>59020UH24</ArbitrarilyNamedXmlItemElement>
<ArbitrarilyNamedXmlItemElement>542514NA8</ArbitrarilyNamedXmlItemElement>
</ArbitrarilyNamedXmlListElement>
'
declare #tblResults TABLE
(
XmlItem xml
)
*/
-- =============================================
-- Author: 6eorge Jetson
-- Create date: 01/02/3003
-- Description: Shreds a list of XML items conforming to
-- the expected generic #xmlListFormat
-- =============================================
CREATE FUNCTION [dbo].[tvfShredGetOneColumnedTableOfXmlItems]
(
-- Add the parameters for the function here
#xmlListFormat xml
)
RETURNS
#tblResults TABLE
(
-- Add the column definitions for the TABLE variable here
XmlItem xml
)
AS
BEGIN
-- Fill the table variable with the rows for your result set
INSERT #tblResults
SELECT
tblShredded.colXmlItem.query('.') as XmlItem
FROM
#xmlListFormat.nodes('/child::*/child::*') as tblShredded(colXmlItem)
RETURN
END
--SELECT * FROM #tblResults

Answer 3

In case this is useful to anyone else out there looking for a "generic" solution, I created a CLR procedure that can take an Xml fragment as above and "shred" it into a tabular resultset, without you providing any additional information about the names or types of the columns, or customizing your call in any way for the given Xml fragment:
http://architectshack.com/ClrXmlShredder.ashx
There are of course some restrictions (the xml must be "tabular" in nature like this sample, the first row needs to contain all the elements/columns that will be supported, etc) - but I do hope it's a few steps ahead of what's available built-in.

Answer 4

Here's an alternate solution:
;with cte as
(
select id, name, addresses, addresses.value('count(/address/city)','int') cnt
from #demo
)
, cte2 as
(
select id, name, addresses, addresses.value('((/address/city)[sql:column("cnt")])[1]','nvarchar(256)') city, cnt-1 idx
from cte
where cnt > 0
union all
select cte.id, cte.name, cte.addresses, cte.addresses.value('((/address/city)[sql:column("cte2.idx")])[1]','nvarchar(256)'), cte2.idx-1
from cte2
inner join cte on cte.id = cte2.id and cte2.idx > 0
)
select id, name, city
from cte2
order by id, city
FYI: I've posted another version of this SQL on the code review site here: https://codereview.stackexchange.com/questions/108805/select-field-in-an-xml-column-where-both-xml-and-table-contain-multiple-matches

Answer 5

If you can use it, the linq api is convenient for XML:
var addresses = dataContext.People.Addresses
.Elements("address")
.Select(address => new {
street = address.Element("street").Value,
city = address.Element("city").Value,
state = address.Element("state").Value,
zipcode = address.Element("zipcode").Value,
});