QT4 CG Meeting 147 Minutes 2026-01-06

Regrets: CG.

• [X] David J Birnbaum (DB)
• [X] Reece Dunn (RD)
• [ ] Sasha Frisov (SF)
• [ ] Christian Grün (CG)
• [ ] Joel Kalvesmaki (JK)
• [X] Michael Kay (MK)
• [ ] Juri Leino (JLO)
• [X] John Lumley (JWL)
• [X] Alan Painter (AP
• [X] Wendell Piez (WP)
• [ ] Ruvim Pinka (RP)
• [ ] Ed Porter (EP)
• [X] Norm Tovey-Walsh (NW) Scribe. Chair.

Proposal: Accept the agenda.

MK observes that #2336 should have been at the top of the agenda.

Accepted with that change.

Proposal: Accept the minutes of the previous meeting.

Accepted.

The next meeting is planned for 13 January 2026.

No regrets heard.

JWL gives regrets for 20, 27 January.

• [ ] QT4CG-143-02: MK to try to recover the ability to extract formal equivalences into tests
• [ ] QT4CG-143-03: JK to look for C14N test suites.
• [ ] QT4CG-144-01: MK to consider if any now lost value comparisons should be added as examples.
• [X] QT4CG-146-01: NW to attempt to provide a markup solution for argument defaults

This section summarizes all of the issues and pull requests that need to be resolved before we can finish. See Technical Agenda below for the focus of this meeting.

See PR #2336.

MK introduces the PR.

• MK: Primarily about XSLT, but there are one or two knock-on effects on the XPath grammar.
  • … One is completely trivial; the pattern syntax now reuses the fragment for constants so it’s been rearranged a bit.
    • … 2.6.6 Constants has been drawn out into a section of their own.
  • … The other is not so trivial; in the interest of alignment, I’ve changed the syntax for JNode types to work like patterns. You can specify first the selector, then the type. That brings pattersn and types into alignment for JNodes. Let’s look at that first.
• MK: The JNodeType production is changed; it now has two parts, a selector property followed by a content type. That aligns with JNode patterns and element tests.
  • … The selector has to be a constant, that again aligns with elements.
  • … If you want to use the selector with a type, that implies that it’s statically known.
• MK: The other changes are to XSLT.
  • … The way patterns work has been restructured a bit.
  • … Predicate patterns aren’t new, but we’ve changed the way the semantics are described.
  • … I’ve changed the syntax for type patterns, it was getting too messy.
  • … I gave up trying to write some types in type() and some outside. They’re now all uniformly a tilde followed by an item type.
• MK: Type patterns are now simple conceptually; you can’t combine them with union and intersect.
• MK: The third kind of pattern are ~GNodePattern~s. You can now use paths for both JTrees and XTrees and the semantics are defined in the same way.
  • … This removes the ambiguity of parenthesis in JNode patterns.
• MK: The meaning of GNode pattern has been generalized to handle JNodes.
• MK: Changing the intersect and excepts back to the way it was also removes the backwards incompatibility that we introduced. But also reintroduces the problem that except doesn’t always mean what you might think it means.
• MK: There are no longer special rules for intersect and except in priorities.

MK reviews the priorities outlined in the spec.

• JWL: In 6.3.2.2, the fifth example in type patterns, the jnode(address… example is missing a closing ).
  • … I’ve run some tests on the grammar and I don’t see any ambiguities.
• JWL: We don’t use tilde anywhere else, right?
• MK: I think that’s right. At one stage we had a separate syntax for JTrees but we abandoned that.
• JWL: The * in the selectors and content things, is there an argument for saying that that is a wildcard production? Or is that too complex?
  • … When I’m doing the grammar conversions, there isn’t a production for the *, it’s just text.
  • … If we made it a specific production, then it would be clearer in the grammar.
• MK: There’s a related problem when you just have a name there; you have to remember that it was a bare name because you don’t know until runtime if it’s being used as an element name or not. (Is it a QName in the default namespace or a plain name?)
  • … The semantics get a little bit complex in that case.
• JWL: It’s just a suggestion.
• RD: We do something similar for element and attribute tests, because they were missing the *:name production.
• JWL: That might be what I’m thinking wildcard is set to. I’ll review and send some mail.

Proposal: Accept this PR.

Accepted.

• JWL: In the example that has a record with a * at the end is a production for an extensibility key. I think we’ve got reasonable stuff about trying to make things grammar productions.
• RD: See also:
  • https://qt4cg.org/specifications/xquery-40/xpath-40.html#doc-xpath40-ItemType-NameTest
  • https://qt4cg.org/specifications/xquery-40/xpath-40.html#prod-xpath40-Wildcard
• JWL: You need to know that a star was present which is unlike tilde.

See PR #2348.

• MK: I think the issue here describes a wider range of things; I just picked some low-hanging fruit.
• MK: The first thing is adding a source-location option.
  • … We needed this in Saxon to allow fn:transform to do streaming.
• MK: … We add the option to write outputs to persistent storage (the filesystem)
  • … We resisted this previously because it’s a side effect; let’s just bite the bullet.

Proposal: Accept this PR.

Accepted.

See PR #2266.

Some discussion of whether or not this is awaiting further work…

See PR #2274.

MK introduces the PR. Essentially, it handles the XSLT bits of the context the same way we handle the XPath bits. When you bind a function item to something like regex groups, then it explicitly captures the regex groups as part of the bound context in that function item rather than throwing some sort of error. It’s just an alignment thing with the XPath context functions.

• MK: The detail mainly involves removing rules.

Proposal: Accept this PR.

Accepted.

ACTION QT4CG-147-01: MK to fix the change markup entry for fn:curent-merge-group; it refers to the wrong function name.

• JWL: Inside a function that captures the regex group, is it available for the whole function?

Some discussion of a complex case with a partial function bound to an XSLT function with the capturing group.

• MK: None of this is especially useful, it just follows the general rule and removes some special cases.

Proposal: Accept this PR.

Accepted.