Bug 25185 – XSLT Text Transform: contains(TAB) matches SPACE

• Status: NEW
• Component: XML

[Doeke Zanstra]

I'm transforming an XML document with an XSLT document to a text string (output method=text) with the XSLTProcessor. It seems the selector text()[contains(.,'	')] matches a space ( ) and instead of the specified tab (	).

I'm not to sure if the conclusion is correct, but Gecko (Firefox, Camino) is matching correctly. 

The bug appears on the nightly build, but also on Safari 4 latest beta windows and Google Chrome 2 latest beta.

[Doeke Zanstra]

Testcase in the URL.

When transforming the XML without script, it's working as expected (no bug is showing). 

Case:http://mcbak.local/~doekman/bugs/xslt-tab.xml

[Alexey Proskuryakov]

This is a difference in what steps are taken for XSLT processing.

When the HTML code is parsed, 	 is replaced with 	. DOMParser replaces that with a tab character. These steps work identically in WebKit and in Gecko.

However, WebKit's XSLTProcessor serializes both documents to strings to pass them to libxslt. When libxml2/lixslt parse the string containing a tab character, they correctly convert it to a space (see section 3.3.3 "Attribute-Value Normalization" of XML 1.0). On the other hand, Gecko operates on the existing DOM tree.

In contrast with bug 14530 comment 3, I think that Gecko is correct here.

See also: bug 14101.

[Alex Milowski]

This seems like one of those "under specified" edge cases.  There is nothing wrong with serialization and, as such, you should expect to be able to match tab characters.  You cannot parse an XSLT stylesheet document with an XML parser and expect a tab character in your match pattern.

Via Javascript, you can create a XML DOM with a tab character in the match attribute.  As such, you can give the XSLT compiler an instance that you cannot receive from the XML parser.

I'm not sure I would classify this as a bug.