Re: [xsl] xs:anyURI allows a space

["Andrew Welch" <andrew.j.welch@xxxxxxxxx>]

Thanks for a response that is a model of clarity...

I've copied the technique you've used in the StandardURIResolver, and
the tests are passing so I'm happy.

cheers
andrew

On 7/16/07, Michael Kay <mike@xxxxxxxxxxxx> wrote:
> The rules in XML Schema for xs:anyURI are not exactly a model of clarity,
> and I have whinged about them loudly and often.
>
> The sentence "The mapping from anyURI values to URIs is as defined by the
> URI reference escaping procedure..." means (I am told) that to get from an
> xs:anyURI value to a real URI, you percent-encode the characters that are
> not allowed in a URI. This includes space.
>
> Later on the spec says "Note:  Spaces are, in principle, allowed in the
> .lexical space. of anyURI, however, their use is highly discouraged (unless
> they are encoded by %20)." Which has caused no end of confusion, because in
> fact spaces are also allowed in the value space (which is the same as the
> lexical space). And of course, discouraging something in a Note doesn't make
> it invalid.
>
> The whitespace facet for xs:anyURI, which defines what happens to whitespace
> as written in the source XML before creating the "lexical" value, is
> "collapse" (see the schema-for-schemas), which is essentially the same as
> XPath normalize-space(), so it can leave single spaces in the middle of the
> value. There's a case to be made for collapsing whitespace before calling
> the URIResolver, but I don't think it's a strong one on either theoretical
> or practical grounds.
>
> The JAXP spec for the URIResolver interface is not exactly crystal-clear
> either. However, it doesn't mandate that the XSLT processor should do
> anything to the values supplied in the stylesheet source before calling the
> URIResolver, and it doesn't mandate that the values passed across should be
> valid URIs (or relative references) as per the RFC. My interpretation of the
> spec (and yours has just as much claim to validity...) is that it's the
> responsibility of the URIResolver to do any percent-encoding that is needed,
> and if you look at Saxon's StandardURIResolver you will see that it does
> this. The advantage of giving the URIResolver the raw string as supplied to
> xsl:include or document() is that it gives the user maximum control over how
> to interpret it.
>
> Michael Kay
> http://www.saxonica.com/
>
>
>
> > -----Original Message-----
> > From: Andrew Welch [mailto:andrew.j.welch@xxxxxxxxx]
> > Sent: 16 July 2007 15:11
> > To: xsl-list@xxxxxxxxxxxxxxxxxxxxxx
> > Subject: [xsl] xs:anyURI allows a space
> >
> > What is it that xs:anyURI allows a space, but Java's URI
> > class does not (when using the single argument constructor) ?
> >
> > If makes it difficult when resolving relative paths in the
> > URIResolver as something that's a valid xs:anyURI in a call
> > to doc() causes a URISyntaxException in Java when it hits the
> > resolver - don't they both sing from the same rec?
> >
> > The solution it seems is to manually %HH escape the spaces
> > (and use the single arg constructor), or manually deconstruct
> > the href and then use the appropriate constructor, eg for the
> > relative href "the
> > file.xml":
> >
> > new URI(null, null, href, null)
> >
> > Neither seems like the right way.
> >
> > Related-but-possibly-Saxon-specific-question: Should the href
> > argument that's passed to the resolve(href, base) method in
> > the URIResolver be %HH encoded - the base argument is, but
> > the href argument comes through exactly as it's written in
> > the stylesheet ?
> >
> >
> > --
> > http://andrewjwelch.com


--
http://andrewjwelch.com