W3C home > Mailing lists > Public > public-qt-comments@w3.org > May 2015

[Bug 28654] [FO31] load-xquery-module error codes

From: <bugzilla@jessica.w3.org>
Date: Fri, 29 May 2015 10:55:25 +0000
To: public-qt-comments@w3.org
Message-ID: <bug-28654-523-lgM09Kuo68@http.www.w3.org/Bugs/Public/>
https://www.w3.org/Bugs/Public/show_bug.cgi?id=28654

--- Comment #3 from Michael Kay <mike@saxonica.com> ---
I'd like to "reopen" this (though it hasn't actually been closed) because I
think Debbie's suggestions for extra error conditions actually raise a couple
of questions about the spec.

(a) fn:transform defines an option saying what version of XSLT needs to be
supported.

Specifically:

xslt-version    xs:decimal    The minimum level of the XSLT language that the
processor must support. Defaults to the [xsl:]version attribute at the
outermost level of the stylesheet.

I think fn:load-xquery-module should have a similar option? 

Proposal: have an xquery-version option:

xquery-version    xs:decimal    The minimum level of the XQuery language that
the processor must support. Defaults to the value appearing in the version
declaration of the loaded module, or implementation-defined in the absence of
such a declaration.

Then add the error condition:

FOQM0006 is raised if no suitable XQuery processor is available.

(b) For load-xquery-module we say:

Nodes that are passed to or from the dynamically loaded module should retain
their node identity. However, if this is impossible to achieve (for example,
because nodes need to be reconstructed into the representation used by a
different software product) then the tree containing a node may be copied,
causing a loss of node identity.

which is not quite the same as what we say for fn:transform:

Where nodes are passed to the transformation, for example as the value of the
context item or of a stylesheet parameter, they should if possible retain their
node identity, their base URI, and their relationships to all other nodes in
the containing tree (including ancestors and siblings). If this is not
possible, for example because the only way of passing nodes to the chosen XSLT
implementation is by serializing and re-parsing, then a node may be passed in
the form of a deep copy, which may lose information about the ancestors and
siblings of the node, and its relationships to other nodes passed across the
interface.

(also, it's a note in one case and a rule in the other).

I suggest unifying these to avoid any unintended differences:

Where nodes are passed to (X), for example as the value of (Y), they should if
possible retain their node identity, their base URI, and their relationships to
all other nodes in the containing tree (including ancestors and siblings). If
this is not possible, for example because the only way of passing nodes to the
chosen (XSLT|XQuery) implementation is by serializing and re-parsing, then a
node may be passed in the form of a deep copy, which may lose information about
the identity of the node, about its ancestors and sibling, and about its
relationships to other nodes passed across the interface.

-- 
You are receiving this mail because:
You are the QA Contact for the bug.
Received on Friday, 29 May 2015 10:55:27 UTC

This archive was generated by hypermail 2.3.1 : Friday, 29 May 2015 10:55:27 UTC