XPath data model and the XML Declaration

[David Carlisle]

Does the XML Declaration map to a Xpath node type or not.
If yes - what type of node?

No, it affects the parser (eg telling it what encoding has been used), and so affects the construction of the node tree, but no record of the declaration is left in the node set.

That's (normally) what you want. If you have a file in latin1 that starts <?xml version="1.0" encoding="iso-8859-1"?> ...

and the same data encoded in utf-8 that starts

<?xml version="1.0" encoding="utf-8"?>

Then you want these two to be equivalent. The actual data in the nodes are just logical unicode characters, the encoding used to store them originally in a file shouldn't be relevant to the logical view of the XML tree (and isn't available to XPath/XSLT).

Of course in many cases it's not unrreasonable to want to output a file with the same encoding as was used for input, but that information has gone, along with information about whether " or ' was used for attributes, or where CDATA sections were, etc.

David

 

[Helmut Dirtinger]

Hi

Compontents of an xml file are mapped to the different node types of the xpath data model. An element is mapped to an element node, an attribute node represents an attribute and its value ... and so on.

so far so good

But what happens to the xml declaration (<?xml version="1.0" ......?>). Though it looks like a processing instruction it isn't.

The XPath specification [1] only says that the XML Declaration does not map to a the node type processing instruction node (5.5).

Does the XML Declaration map to a Xpath node type or not.
If yes - what type of node?

thx

Helmut

[1] XPath Specification http://www.w3c.org/TR/xpath

 

[Bjoern Hoehrmann]

* Helmut Dirtinger wrote in comp.text.xml:

Does the XML Declaration map to a Xpath node type or not.

No, it does not. The XML declaration does not contain information that would make sense to be available in the data model.

 

[Richard Tobin]

No, it does not.
True.

The XML declaration does not contain information that
would make sense to be available in the data model.

But this is an exaggeration. The encoding is a useful hint when serializing the data model, and the XML version number affects what assumptions you can make about the data (e.g. whether it might have C0 control characters in it).

-- Richard

 

[Bjoern Hoehrmann]

* Richard Tobin wrote in comp.text.xml:

But this is an exaggeration. The encoding is a useful hint when serializing the data model, and the XML version number affects what assumptions you can make about the data (e.g. whether it might have C0 control characters in it).

The problem is however that the value of the encoding pseudo-attribute might not be the actual character encoding of the document, for example if the document is delivered via HTTP the acutal encoding might be specified in the HTTP header which would then override the value in the XML declaration (if any). So it would make more sense to use other means to get that information, for example using DOM Level 3 Core methods.

Regarding C0 control characters, I do not think these may appear in the XPath 1.0 data model which is only defined for XML 1.0 (they are not allowed to appear in expressions either, strings are restricted to chars as defined in XML 1.0 which excludes C0 controls)... and if they are in the data model, you can infer that the XML version is 1.1 so this is of limited use, too.

Other uses would be possible too, for example, if you want to write an XSLT document that discovers meta-data from XML documents like

XML Declaration: yes
Encoding Declaration: ISO-8859-1
Standalone: no

Elements:

1 <html>
1 <head>
1 <body>
32 <div>
...

or whatever, so maybe I should rephrase to, the value such functionality would add is not considered worth the additional complication and such functionality might contribute to making false assumptions such as that the encoding in the XML declaration is the actual document encoding.