XSLT 2.0 implementation
There was a restriction that the characters used for digits and other symbols in
xsl:decimal-format and in the
format-number function must be
Unicode BMP characters (that is, codepoints above 65535 were not allowed). This restriction
has been removed. This means, for example, that you can now format numbers using Osmanya digits.
As a consequence of this change, the old version of the
implementation, which relied on Java decimal formatting and was retained under the name
format-number-1.0, is no longer available.
The output of
xsl:number has been changed, reflecting a discussion on the
xsl-list about the correct interpretation of the specification (which in this respect is unchanged from
XSLT 1.0). It appears that when formatting the number sequence (1 to 4) using the format token
, the correct output is
[220.127.116.11], whereas Saxon (in all previous
releases including 6.5.x) was outputting
[1[2[3. The distinction hinges on the
interpretation of the term "separator token" in the specification.
The documentation for previous releases described a restriction:
key() is called in a match pattern, the argument must be a string literal."
This restriction was in fact removed many releases ago.
Error XTSE1560 is now reported. A static error is reported if there are two conflicting
A number of changes have been made to the
- The function now reports an error if the content of the file contains characters that are not valid in XML.
- The version of the function with a single argument is available. The encoding is inferred first by looking the HTTP headers if available; then by examining the file to see if it starts with a byte order mark or an XML declaration. If neither strategy succeeds, UTF-8 is assumed.
- If the encoding is UTF-8 and the file starts with a byte order mark, this is not included in the string returned as the result of the function.
unparsed-text-available()function is implemented (both forms, with one argument or two arguments). However, note that the current implementation is not efficient: if a call on
unparsed-text-available()is followed by a call on
unparsed-text()to read the file, the file is actually read twice.
- UTF-16 little-endian encoding, when signalled by a byte-order-mark, is now handled correctly.
It is no longer an error to disable output escaping while writing a temporary tree. Instead, the request to disable output escaping is simply ignored.
An error is now reported when no value is supplied for a required stylesheet parameter, even if the parameter is never referenced. Previously an error was reported only when the parameter was actually referenced.
Minor changes have been made to the way that namespace prefixes are allocated to constructed
elements and attributes in cases where a prefix needs to be invented. As required by the specification,
in the event of a conflict the prefix of a namespace node that is explicitly attached to an element takes
precedence over the prefix used in the
name attribute of the
In general it is no longer an error to use a relative URI in the
xsl:result-document when no base URI is known for the principal output file. Instead,
the relative URI is interpreted as being relative to the current working directory in the calling environment.
However, the error is retained if the configuration option to disallow extension function calls is set: this
option indicates that the stylesheet is not trusted to access external resources, and it must
therefore not be trusted to write files relative to the current directory.
When a stylesheet selects forwards-compatible mode (that is, when the
attribute is greater than 2.0), unrecognized top-level elements are now ignored together with all their
content, as required by the specification. Previously errors were reported if such elements contained
XSLT instructions with invalid attributes or content.
The extension attribute
saxon:explain="yes" now causes display of the optimized
expression tree for the entire template or function in which it appears, not only for XPath expressions
appearing on that specific element. This change has been made because optimization is now carried
out on the template or function as a whole.