- From: <bugzilla@jessica.w3.org>
- Date: Sun, 06 Nov 2016 14:33:53 +0000
- To: public-qt-comments@w3.org
https://www.w3.org/Bugs/Public/show_bug.cgi?id=29983 --- Comment #1 from Michael Kay <mike@saxonica.com> --- My mental model for scanning expressions is that they are expressions that can be evaluated by considering each descendant node individually and testing it against a pattern. Anything that involves evaluating a user-defined function that selects which nodes should match the pattern doesn't fit into that model at all. Remember that all this started with the //section/footer problem, with a a source tree like <section id="A"> <section id="B"> <footer of="B"/> </section> <footer of="A"/> </section> Here the nested-loop evaluation strategy for the path expression yields nodes that are not in document order (the child of the second section PRECEDES the child of the first section). The only streamable strategy for evaluating this is to consider the footer nodes in document order and test each one against the pattern //section/footer. If we replaced "//section" here by a function that selects the section elements then I would have no idea how to devise a streamable execution strategy. -- You are receiving this mail because: You are the QA Contact for the bug.
Received on Sunday, 6 November 2016 14:34:00 UTC