QT4 CG Meeting 098 Minutes 2024-11-12

DN, EP give regrets.

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

Proposal: Accept the agenda, amended to include PR #1569 in the “merge without discussion” group.

Accepted.

Proposal: Accept the minutes of the previous meeting.

Accepted.

This next meeting is planned for 19 November. Any regrets?

DB gives for 19 and 26 November.

(Items marked [X] are believed to have been closed via email before this agenda was posted.)

• [ ] QT4CG-080-07: NW to update the build instructions in the README
• [ ] QT4CG-082-02: DN to work with MK to come to agreement on the fn:ranks proposal
• [ ] QT4CG-088-01: NW to consider how best to add a dedication to MSM.
• [ ] QT4CG-088-04: [Someone] needs to update the processing model diagram needs vis-a-vis the static typing feature
• [ ] QT4CG-089-01: CG to draft a PR that attempts to resolve the operators described in #755 to a smaller number of orthogonal choices.
• [X] QT4CG-097-01: MK to add a note explaining why base-type and friends are functions in the schema-type-record
• [ ] QT4CG-097-02: MK to make the XSD schema component references into links to XSD
• [ ] QT4CG-097-03: DN to proposal an axis for accessing the siblings of a node.

See PR #1565

MK introduces the PR.

• MK: This arose from a bug report in the QT3 test suite; there were tests that were incompatible with the specification.
  • Came to the conclusion that the spec is wrong and the tests are correct.
  • The change is that escaping JSON is false by default.
  • “False” is by far the most usable option; looking at the history of 3.1, we changed the polarity of the option but we forgot to change the default.
• CG: Yeah, that’s fine.

Proposal: Accept this PR.

Accepted.

See PR #1470

• CG: We’ve talked about this before.
  • MK suggested to return a function instead of an item or a string. That’s the change.
• JWL: Can you remind me why we wanted a function?
• CG: The stack trace can be extensive, this defers computation until you need it.
• MK: This drops the free-standing function as well.
• CG: Yes.

Proposal: Accept this PR.

Accepted.

See PR #1562

MK observes that the diff doesn’t show well on this one…

• MK: We publish the schema in two places, both free-standing and copied into an appendix.
• NW: Could we just stop doing that? Is there value in the appendix?
• RD: Are there cross-references into it?
• MK: No.

MK reviews the actual changes in the PR.

• MK: The schema was actually invalid. And some of the stylesheets were invalid against the schema. There are already some caveats that say there are stylesheets that aren’t valid even though they’re correct. That caveat now extends to xsl:note.

MK reviews the changes in the schema.

ACTION: QT4CG-098-01: NW to update the Relax NG grammar for XSLT 4.0

• RD: Would it make sense to generate the schema from the spec. We could put fragments of XSD and RNG into the specification and then pull them all together at the end.
• MK: Are you volunteering?
• RD: I’ll take a look.

ACTION: QT4CG-098-02: NW to look at the XSL stylesheet for XSD, #374.

• SF: Rendering in the specification or in the IDE with a live feed is a problem. If it could be solved, it would be game changing to the coding experience.

Proposal: Accept this PR.

Accepted.

See PR #1560

• MK: The serialization doesn’t define the defaults, that’s up to the host language.
  • I found that XSLT didn’t define a default for indent. This PR fixes that.
  • It also adds a couple of section numbers, which messes with the diff.
  • And in a few places json and adaptive serialization weren’t mentioned.
• MK: All of the prose about defaults for parameters has been moved into a new section.
  • … The only change is to the indent parameter which now says the default is “yes” for HTML and XHTML and “no” everywhere else.

Some discussion of how the diffs are constructed.

Proposal: Accept this PR.

Accepted.

See PR #1558

• MK: This is an ambiguity in the XSLT pattern syntax that JWL identified.
  • The solution was to remove FunctionCallP in RootedPath because it only gives an option you can get to directly (introducing an ambiguity).

Proposal: Accept this PR.

Accepted.

See PR #1546

• MK: This adds json-lines to the XSLT spec, the bulk of the work is in the Serialization specification and we’ve already accepted it.
  • … It adds it as an option in various places and makes a brief mention of what it does.
• JLO: Is it possible to unify the XSLT and XQuery serialization parameters?
• MK: I think moving things into the Serialization specification has refactored about as far as we can.
  • … There are some differences that have nothing to do with XQuery.
• JLO: Is there a way for a user to find out what the current, actual values are? The system that XQuery is running in will override some of those defaults.
• MK: You can override defaults via an API. And the API can also discover what your current settings are.
  • … There’s no way to ask in XSLT or XQuery code what the current settings are. I’m not quite sure why you’d want to know.

Proposal: Accept this PR.

Accepted.

See PR #1544

• MK: This mimics in XSLT a change that was made to XQuery.
  • … The spec had said that a global variable isn’t visible in its own declaration.
  • … It is now visible, but trivial self references are disallowed by circularity rules.
  • … There’s an example where self-reference does make sense, for example, a recursive inline function.
• MK: This removes an unnecessary restriction.
• JWL: Is this only the case with function-valued global variables? Can you do it with any other kind?
• MK: You could do it with maps.
• JWL: That’s effectively a function…
• JLO: If it really only applies to functions, should we say that?

Proposal: Accept this PR.

Accepted.

See PR #1454

• MK: We did take a look at it once before.
  • … In the past, having circular includes was an error.
  • … That meant you couldn’t write stylesheets that simply included everything they depended on because you’d get circularities.
  • … This change says that if you have xsl:includes at the same level, we ignore the duplicates. Only the first one counts.
  • … Duplicates includes the entry point at the stylesheet level.
• MK: There is an edge case where this gives you an incompatibility.
  • … In 3.0 a circularity is an error but a “diamond” isn’t. A diamond where both B and C include D would usually lead to an error because it would include two templates or functions with the same name.
  • … In the rare case where an included module only contains template rules without names then you’d get two copies and next-match would go to the next.
• JLO: Is there some kind of scoping mechanism? Can the output change when xsl:includes are inserted at different levels? Can you have different variables if they’re at different levels?
• MK: No. All the declarations are global to a package. They’re import precedence only depends on the level. The order of tebmplate rules could be different which matters if you haven’t specified priority.
• JLO: Shouldn’t we say that also the last one wins?
• MK: In some ways, yes, but that doesn’t work with static variables because it’s the first one that counts. And you can only reference it in a static declaration that comes after it. So you wouldn’t know if the include you’d already read was the last one.
  • … The evaluation of static expressions is assumed to be serializable. That sort of preprocessing stage has to be sequential and you have to know when it’s the first.
• JLO: Are those includes at the same level turned into a flat list?
• MK: You’re processing your source hierarchically. For static variables and expressions, you work hierarchically. If you reach an xsl:include, you read it immediately, and any variables it contains are avialable anywhere after that include. It’s sort of flattened.
• WP: I think that the edge case that we’re worried about is really out there on the edge. I think the current behavior might really be considered a bug.
• MK: We’ll find out in time!

Some further discussion of includes and nesting and static variables.

Proposal: Accept this PR.

Accepted.