- From: James Clark <jjc@jclark.com>
- Date: Tue, 13 Jul 1999 23:52:50 +0700
- To: "Weichel Bernhard (K3/EMW4) *" <Bernhard.Weichel@de.bosch.com>
- CC: "'www-xpath-comments@w3.org'" <www-xpath-comments@w3.org>
Thanks for your comment. I think it's unlikely that the WG would want to consider a major redesign at this stage. Also this suggestion seems to lose the convenience of the abbreviations. For example, being able to use simply foo to select all foo children, or foo/bar to select all the bar children of the foo children. These abbreviations account for a large percentage of the uses in XSLT and XPointer. "Weichel Bernhard (K3/EMW4) *" wrote: > > I think it would be better to define the step as it is done in MetaMorphosis > (www.ovidius.com): > > basis predicate rank > > where basis is simply the axis (such as child, parent, descendant, left, > right, preceeding etc.) with no NodeTest. > > The NodeTest should be part of the predicate using a special operator such > as (taken from MetaMorphosis language) > > child[?para] > > The ? operator can be applied to multiple properties such as types, > attributes contents. If no property is specified, then the element name is > taken: > > child[data?introduction] all children containing the text > "introduction) > child[@id?foo] all children with Attribute id equals foo > child[gid?para] all children with Generic Identifier "para" > > predicate should be a location path which is treated as true as long as the > adressed ressource exists. > > rank is separated by ":" and allows to specify a subset of the result: > > child[?para]:2~-2 returns the child paragraph except the first and the > last one > > This solves the issues (node-ordering) and (bracket-overload).
Received on Tuesday, 13 July 1999 15:58:13 UTC