preventing an entity-reference to be resolved (output method XML)

[Joris Gillis]

Hi everyone,

I have this stylesheet:

<?xml version="1.0" encoding="ISO-8859-1" ?>
<xsl:stylesheet version="1.0"
xmlns:xsl="http://www.w3.org/1999/XSL/Transform"> <xslutput method="xml"/> <xsl:template match="/"> <node>&apos;</node> </xsl:template> </xsl:stylesheet>

My goal is to get this output:
<?xml version="1.0" encoding="UTF-8"?><node>&apos;</node>

I've tried '&amp;apos;' , '&apos;'
I tried to do it with custom ENTITY's in the DOCTYPE.
I tried with CDATA sections.
But non of it works.

I know it works if the output method is 'text', but it has to work in 'xml' too.

Any help or comments greatly appreciated...

 

[Martin Honnen]

Joris Gillis wrote:

I have this stylesheet:

<?xml version="1.0" encoding="ISO-8859-1" ?>
<xsl:stylesheet version="1.0"
xmlns:xsl="http://www.w3.org/1999/XSL/Transform"> <xslutput method="xml"/> <xsl:template match="/"> <node>&apos;</node> </xsl:template> </xsl:stylesheet>

My goal is to get this output:
<?xml version="1.0" encoding="UTF-8"?><node>&apos;</node>

I've tried '&amp;apos;' , '&apos;'
I tried to do it with custom ENTITY's in the DOCTYPE.
I tried with CDATA sections.
But non of it works.

I know it works if the output method is 'text', but it has to work in 'xml' too.

I think within the XSLT 1.0 standard there is nothing for you to mandate whether such a character is serialized in UTF-8 as an enity reference or literally. However some processors have extensions to achieve that, for instance Saxon:
http://saxon.sourceforge.net/saxon6.5.3/extensions.html#saxon:entity-ref where the documentation says it is meant for HTML entity references but maybe you can use it for other output methods as well.

 

[Andy Dingley]

I know it works if the output method is 'text', but it has to work in 'xml' too.

As far as I know, there is no way to do this in XSLT. It is an _XML_ problem - in XML the entity reference and the character itself are seen as equivalent and so there is never any need (sic) to control the choice explicitly.

You need to modify the behaviour of the XML serialiser that is used to generate the output document. I've done this, but I had to change the serialiser code to do so. Maybe some other tools will already allow this to be controlled ? I was just working with what I had.

As it's &apos; you're after (and if this is all you're after), an output mode of "HTML" might work for you. &apos; is defined in XML, but not in the HTML DTDs.

 

[Joris Gillis]

Thank you, Martin and Andy, for your replies.

It seems to be there is indeed (not yet) a standard-compliant way to solve this.

Meanwhile, I figured out that
<xsl:text disable-output-escaping="yes">&amp;apos;</xsl:text> returns the returns the result I want.

But the XSLT1.1 specification clearly states that it is illegal to use 'disable-output-escaping' with attribute nodes, which is actually what I'm trying to achieve.

The actual problem I had:
OpenOffice.org (1.1.3) crashes when it finds a single quote (') in font declaration attributes in the content.xml part of a file. It does not accept anything else than '&apos;'. So generating content.xml with XSLT1.1 directly now seems impossible.

About the trick to set the ouptut method to 'HTML' , it doesn't work with me. (I'm using AltovaXslt)

regards,