105

If you have a varchar field you can easily do SELECT * FROM TABLE WHERE ColumnA LIKE '%Test%' to see if that column contains a certain string.

How do you do that for XML Type?

I have the following which returns only rows that have a 'Text' node but I need to search within that node

select * from WebPageContent where data.exist('/PageContent/Text') = 1
Cœur
  • 37,241
  • 25
  • 195
  • 267
Jon
  • 38,814
  • 81
  • 233
  • 382

5 Answers5

125

Yet another option is to cast the XML as nvarchar, and then search for the given string as if the XML vas a nvarchar field.

SELECT * 
FROM Table
WHERE CAST(Column as nvarchar(max)) LIKE '%TEST%'

I love this solution as it is clean, easy to remember, hard to mess up, and can be used as a part of a where clause.

This might not be the best performing solution, so think twice before deploying it to production. It is however very useful for a quick debug session, which is where I mostly use it.

EDIT: As Cliff mentions it, you could use:

...nvarchar if there's characters that don't convert to varchar

Squazz
  • 3,912
  • 7
  • 38
  • 62
75

You should be able to do this quite easily:

SELECT * 
FROM WebPageContent 
WHERE data.value('(/PageContent/Text)[1]', 'varchar(100)') LIKE 'XYZ%'

The .value method gives you the actual value, and you can define that to be returned as a VARCHAR(), which you can then check with a LIKE statement.

Mind you, this isn't going to be awfully fast. So if you have certain fields in your XML that you need to inspect a lot, you could:

  • create a stored function which gets the XML and returns the value you're looking for as a VARCHAR()
  • define a new computed field on your table which calls this function, and make it a PERSISTED column

With this, you'd basically "extract" a certain portion of the XML into a computed field, make it persisted, and then you can search very efficiently on it (heck: you can even INDEX that field!).

Marc

marc_s
  • 732,580
  • 175
  • 1,330
  • 1,459
  • 1
    I'm basically implemeting a search feature so I want to search the XML column only on the 'Text' nodes and then return a substring to indicate that the search has found a match. For example search on 'hi there' instead of returning the whole xml column I'd just return a substring such as 'the chap said hi there and carried...' – Jon Dec 02 '09 at 13:51
  • 1
    Beat me to it by 5 seconds. Another possibility is to consider using free text search, if your data is amenable... – RickNZ Dec 02 '09 at 13:51
  • 11
    to search the whole filed: WHERE xmlField.value('.', 'varchar(max)') LIKE '%FOO%' – jhilden Apr 25 '13 at 19:03
  • watch out for pesky Xml namespaces if you get NULL back – Simon_Weaver Mar 15 '18 at 23:32
12

Another option is to search the XML as a string by converting it to a string and then using LIKE. However as a computed column can't be part of a WHERE clause you need to wrap it in another SELECT like this:

SELECT * FROM
    (SELECT *, CONVERT(varchar(MAX), [COLUMNA]) as [XMLDataString] FROM TABLE) x
WHERE [XMLDataString] like '%Test%'
Carl Onager
  • 4,112
  • 2
  • 38
  • 66
0

This is what I am going to use based on marc_s answer:

SELECT 
SUBSTRING(DATA.value('(/PAGECONTENT/TEXT)[1]', 'VARCHAR(100)'),PATINDEX('%NORTH%',DATA.value('(/PAGECONTENT/TEXT)[1]', 'VARCHAR(100)')) - 20,999)

FROM WEBPAGECONTENT 
WHERE COALESCE(PATINDEX('%NORTH%',DATA.value('(/PAGECONTENT/TEXT)[1]', 'VARCHAR(100)')),0) > 0

Return a substring on the search where the search criteria exists

Squazz
  • 3,912
  • 7
  • 38
  • 62
Jon
  • 38,814
  • 81
  • 233
  • 382
0

The correct way to do this is to use the contains() XQuery predicate

SELECT *
FROM WebPageContent wpc
WHERE wpc.data.exist('/PageContent/Text/text()[contains(., "Test")]') = 1;
Charlieface
  • 52,284
  • 6
  • 19
  • 43