QT4 CG Meeting 189 Minutes 2026-11-17
Meeting index / QT4CG.org / Dashboard / GH Issues / GH Pull Requests
Table of Contents
- Summary of new and continuing actions
[0/7] - Draft Minutes
- 1. Administrivia
- 2. Technical agenda
- 2.1. PR #3069: 3068 `item` axis: document order
- 2.2. PR #3056: 3053 Guarded expressions: runtime type errors
- 2.3. PR #3073: 3072 Switch, typeswitch: optional default clause
- 2.4. PR #3067: 3066 Node constructors: JNodes
- 2.5. PR #3076: 2935A: Default function options in the static context
- 2.6. PR #2981: 2980 Adaptive serialization: modes
- 2.7. PR #3054: 3050 Clarifications regarding errors and type inferences
- 3. Any other business
Summary of new and continuing actions [0/7]
[ ]QT4CG-143-02: MK to try to recover the ability to extract formal equivalences into tests[ ]QT4CG-167-07: NW to review tests for interpolated strings with edge cases in mind[ ]QT4CG-170-01: RD to draft a proposal that attempts to address the lexical issues differently.[ ]QT4CG-177-01: NW to put the progress graphs back in the minutes.[ ]QT4CG-182-01: NW to review CG’s diff script[ ]QT4CG-187-01: MK to elaborate on the proposal in PR #3031[ ]QT4CG-189-01: NW to coordinate Adam joining next week.
Draft Minutes
1. Administrivia
1.1. Roll call [7/11]
Regrets: WP, AP, BTW.
[ ]David J Birnbaum (DB)[X]Reece Dunn (RD)[X]Christian Grün (CG)[X]Joel Kalvesmaki (JK)[X]Michael Kay (MK)[X]Juri Leino (JLO)[X]John Lumley (JWL)[ ]Alan Painter (AP)[ ]Wendell Piez (WP)[ ]Bethan Tovey-Walsh (BTW)[X]Norm Tovey-Walsh (NW), Scribe, Chair.
1.2. Accept the agenda
Proposal: Accept the agenda.
Accepted with AOB for Prague and EXPath CG.
1.3. Approve minutes of the previous meeting
Proposal: Accept the minutes of the previous meeting.
Accepted.
1.4. Next meeting
The next meeting is planned for 24 November.
Regrets: JLO.
1.5. Review of open action items [0/6]
[ ]QT4CG-143-02: MK to try to recover the ability to extract formal equivalences into tests[ ]QT4CG-167-07: NW to review tests for interpolated strings with edge cases in mind[ ]QT4CG-170-01: RD to draft a proposal that attempts to address the lexical issues differently.[ ]QT4CG-177-01: NW to put the progress graphs back in the minutes.[ ]QT4CG-182-01: NW to review CG’s diff script[ ]QT4CG-187-01: MK to elaborate on the proposal in PR #3031[X]QT4CG-188-01: NW to assist in the merge of PR #2998- Overtaken by events, it was already merged
1.6. Review of open pull requests and issues
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.
1.6.1. Blocked
The following PRs are open but have merge conflicts or comments which suggest they aren’t ready for action.
- PR #2247: 716 Deferred Evaluation in XPath - the f:generator record
1.6.2. Merge without discussion
The following PRs are editorial, small, or otherwise appeared to be uncontroversial when the agenda was prepared. The chairs propose that these can be merged without discussion. If you think discussion is necessary, please say so.
- PR #3080: 3079 Split %fixed into two terminal symbols
PR #3078: 3077 Function assertions in reserved namespaces- Withdrawn
- PR #3061: 3049 err:map becomes a labeled record
- PR #3060: 3059 fn:char: FOCH0001 instead of FOCH0005
A few moved to technical agenda.
Accepted.
1.6.3. Close without action
1.6.4. Substantive PRs
The following substantive PRs were open when this agenda was prepared.
- PR #3076: 2935A: Default function options in the static context
- PR #2981: 2980 Adaptive serialization: modes
- PR #3054: 3050 Clarifications regarding errors and type inferences
- PR #3058: 3057 Casting to records: use target %fixed values
- PR #3063: 3062 fn:resolve-uri: non-hierarchical bases
- PR #3065: 3064 `fn:jvalue`: accept any item
- PR #3071: 3070 Guarded expressions: missing constructs
- PR #3075: 3074 Regular expressions: Unicode scripts
- PR #3081: 2935B Arithmetic Mode Decimal
2. Technical agenda
2.1. PR #3069: 3068 `item` axis: document order
See PR #3069
CG explains the PR; it establishes document order for item axis. This avoids having an implementation-defined behavior for the following axis, for example.
- MK: I think I need to see some examples, but it fills a gap.
Proposal: accept this PR.
Accepted.
2.2. PR #3056: 3053 Guarded expressions: runtime type errors
See PR #3056
CG explains the PR. Testing revealed that some expressions needed to be guarded.
- CG: The definition of a guarded has been updated to add a type error detected during dynamic evaluation.
- MK: I think what worries me here slightly is just-in-time optimization.
- … If we’re allowed to do something statically, should we be allowed to do it in a runtime JIT optimization?
- … For example, once you know the values of global parameters, some optimizations are easier.
- … This says if you delay the optimization, you can’t report it.
- CG: Let’s keep it open and have another look.
- MK: I do know that there are problems here; there are users who complain about type errors on a conditional branch when you can see clearly that it’ll never be executed.
- RD: Am I right in think that a type error is a subset of a dynamic error?
- MK: No, it’s a third category.
- CG: It can happen during the static or dynamic phases.
- RD: It would be useful to see some examples.
- MK: It’s muddled in our case by the fact that we don’t make a completely clear
distinction between static and dynamic phases anyway.
- … We allow JIT compilation of template rules; we think of that as a delayed part of the static process.
RD: What this section is saying is that if the processor can guarantee that given the current static context and values of variables and things that a guarded expression won’t trigger therefore they can do optimizations
- … Like for example in the
(1 to $n)case they can short circuit.
- JLO: This looks very similar to a bug that I found recently.
- … Like for example in the
Revisit in the future.
2.3. PR #3073: 3072 Switch, typeswitch: optional default clause
See PR #3073
- CG: It’s a bit more than editorial, but it should be easy to explain.
- … The switch and typeswitch expressions require a default case.
- … But we now have enums and choice item types where we know that the list is exhaustive.
- … My proposal says that the default clause can be omitted and there’s a new error if the processor falls of the end of the list.
- MK: A sensible thing. My only question is, XSLT defaults otherwise to returning an empty sequence rather than raising an error.
- … I wonder what the relative merits are.
- CG: An error gives you better control.
- MK: In terms of handling semi-structured content where the typeswitch is over
a collection of elements that’s over a collection of elements that might
appear in the input…well, I suppose you can always have an explicit
defaultclause that returns empty sequence.- … There are probably use cases for both.
- … It just seems slightly odd to make them different.
- CG: Yes, I can see that.
- JWL: Is
typeswitchin XPath? - RD: No, it’s only in XQuery.
- JWL: Is there any reason not to put it in XPath?
- MK: I don’t think there’s a strong justification anymore for not having it in XPath.
- JWL: I’d suggest putting in in XPath.
Proposal: CG to rewrite for returning the empty sequence instead of error; mark as “merge without discussion” for next week.
2.4. PR #3067: 3066 Node constructors: JNodes
See PR #3067
- CG: This is about JNodes that are supplied within node constructors.
- … I haven’t added an example, but there are some in the pull request and the issue.
CG reviews the examples:
<a>{ { 'x': 'y' }/x }</a>
<a>{ [ <b/> ]/1 }</a>
<a>{ [ [ 1, 2 ] ]/1 }</a>
+CG: I think just using the jvalue is the right choice.
- MK: My only comment is that we should do the same in XSLT.
- RD: As written, does this handle things like atomization? If you’re putting the number 1 into an element, if I’m correct that gets stringifiied into a string literal 1.
- CG: Atomization should be a subsequent step.
- CG: There’s no reason to atomize it.
- RD: If you select the first item of an array as a jnode, that would be the same as an enclosed expression with the same value.
- CG: It happens first.
- MK: I’m not sure what part of the document I’m looking at. We’re talking about
content of element constructors. What happens about constructors of other
kinds of nodes.
- CG: … good question! … I should look at that separately.
Proposal: merge this PR?
Accepted.
2.5. PR #3076: 2935A: Default function options in the static context
See PR #3076
- MK: This came about by a slightly devious route. In a very detailed issue
about arithmetic modes and defaults. One item in that was the ability to
change the default rounding mode globally to handle accounting standards.
- … Thinking laterally from that, I wonder if we need a more general mechanism to set options from the static context.
- … What I forgot when proposing that is that the rounding mode of round is not an options parameter, so that ended up being treated specially anyway!
- MK: What the proposal does is add default options to the static context and I’ve defined syntax in XSLT for setting them.
- … I haven’t designed syntax for XQuery, but I thought I’d get input first.
MK reviews the XSLT specification.
- MK: The substance of it is in 5.6.
- … There’s a
xsl:default-function-optionsinstruction to set default options for a particular function. - … It’s a bit difficult to define in XQuery because I’m exploiting static expressions in XSLT.
- … There’s detail about when you have to check the options.
- … It got too complicated to try to define them for dynamic function calls, only static ones.
- … And there are a number of places where they don’t apply.
- … They do apply when you pass options to a static function call.
- … There’s a
- RD: In XQuery, there’s a declare option syntax, the options declarations in the prolog. We could extend that.
- MK: Yes, that’s probably what I’d do. But it’s a bit tricky because we don’t have static options.
- JWL: In one sense, this is effectively a record but it’s one that gets put by default on the appropriate function calls.
- … You could do that by defining a record for the function options.
- MK: Sort of.
- JWL: What happens for curried functions.
- MK: At the moment, it only applies to static function calls, not partial applications. That’s to make things simpler.
- JLO: I like this, I would like to see this in XQuery. We do already have that in part as we can do that in global configurations.
- … The context can take a lot of defaults from there.
Some discussion of how the options are applied, in brief: they’re merged.
- CG: I can see the advantages of having XQuery syntax. But I’m a bit hesitant because it’s powerful machinery but it can’t do everything. Lots of applications have dynamic options and we put them on a variable that is referenced.
- JK: What impact does this have on user-written XSL function instructions? Can be exploited by a user?
- MK: At the moment, options parameters are only defined for system functions. I guess vendor extension functions could use them. There’s no current way to do it for user defined functions to exploit it.
- JK: I think the prose needs to be made more clear on that point.
- MK: I think it’s worth persuing. I’ll revise it.
2.6. PR #2981: 2980 Adaptive serialization: modes
See PR #2981
Discuss next time.
2.7. PR #3054: 3050 Clarifications regarding errors and type inferences
See PR #3054
- MK: I wrote this partly because I found I was confused.
- MK: There’s a bit of clarification on dynamic errors.
- … Disjoint value spaces aren’t a sufficent condition to raise a type error (consider fn:error)
MK reviews a number of examples.
- MK: There’s no change to the status quo, just new notes and examples.
- RD: In the match example, is the reason you can avoid that call because it’s a built-in function.
- MK: No, it’s because it has a declared type of
xs:boolean, so the type will always bexs:boolean. - RD: The XQuery type system is more of a hint; I’m thinking of the case where a
user-defined function is specified to return an
xs:booleanbut returns"hello".- … In that case it would be a dynamic error when evaluating the function.
- MK: It might be a dynamically detected type error and if you avoid evaluating the function, you might miss it.
- … I think that’s the status quo.
Proposal: accept this PR.
Accepted.
3. Any other business
3.1. Meetup at XML Prague in February
- NW: Jirka has announced that there will be a QT4-themed meetup in February in Prague.
- JWL: It seems very short notice to me.
3.2. EXPath CG
- NW: Adam is spinning the EXPath CG back up.
Some discussion of where the development should continue.
ACTION QT4CG-189-01: NW to coordinate Adam joining next week.