[FBumann]

Prompt: 3. Merging: property maps joined per subject, and a differing set refused.

Note

The following content was generated by AI.

What this changes

merge joins the properties of one subject across fragments, and a property that two fragments declare must be identical in both. A set is shared as a relation is, so two fragments that declare it differently are refused. Stacked on #2.

Why

This is PR 3 of the stack in which sets and properties replace relations.

What each verb does now

merge:

written	result
generator: {bus: bus} in one fragment and generator: {zone: zone} in another	generator: {bus: bus, zone: zone}
bus: bus in one fragment and bus: {into: bus} in another	one property: both forms say the same
one description on a property	carried, because prose is not a claim
site: bus in one fragment and site: zone in another	refused: say different things about the property 'generator.site': {'into': 'bus'} against {'into': 'zone'}. Make the two identical, or give one of them a name of its own.
a property or a set whose missing: differs between fragments	refused, as for a relation
a set with its members in another order	refused, because the two are not identical

override lays a set or a property over the base as it lays a relation:

patch	result
a new property, or a restatement	added, or kept
{missing: absent} on a property written bus: bus	that field changes
bus: null or generator: null	refused: a patch cannot remove a property or a subject
a property with another into:	refused, with the same message as for a relation
a set with other dims:, or cover: null	refused
missing: on a set	changes

How it works: _flat lays the properties out as one entry per property, keyed subject.property and in the long form. The existing rules for a shared entry (_agreed and _shared) then read each property, and _nested puts the result back. _shared now takes the label and the schema class of its entries, and _declared reads the validated fields rather than the serialised form, which writes a property as a bare string.

Departures, and what this PR does not do

• override is in scope too. The ask named merging only. In feat(language): a file may declare sets of labels and the properties of a dimension or a set #1, override laid properties over a subject as owned entries, and a patch could remove them. Because both verbs share one list of sections, adding sets to it changes override as well. So I made properties behave as relations do in both verbs. feat(language): a file may declare sets of labels and the properties of a dimension or a set #1's three tests for that interim behaviour are replaced by the tables above.
• A patch that creates a property without into: is now refused by the schema, properties.generator.zone.into: Field required, as a partial new relation already is. In feat(language): a file may declare sets of labels and the properties of a dimension or a set #1 it was refused with a message of its own.
• No language reference change. The reference page states no composition rule, so only docs/howto/compose.md changed, in both of its tables and in the "redeclared" section.

Gates, and the mutation table

• Passed: pixi run lint, pixi run test (3214 passed, 1 skipped) and pixi run docs-build (strict).
• Not run: compile-tex. Tectonic cannot download its bundle in this container.
• No CI run. GitHub Actions has never run in this fork, and dispatching ci.yml returns 404. Actions must be turned on in the fork's settings.
• Mutation table: the same script as feat(language): a file may declare sets of labels and the properties of a dimension or a set #1 took it, and the tree came back clean.

guard deleted	suite
a set is shared	caught (5)
merge joins the properties	caught (3)
a short form reads as into	caught (6)
a patch cannot remove a subject	caught
a patch changes missing: on anything but a dimension	caught (2)
a restatement reads the fields, not the short form	caught (2)

---

Generated by Claude Code

[coderabbitai]

Important

Draft PR not reviewed

Draft PRs are not automatically reviewed by default.

• Trigger a manual review

To automatically review draft PRs, update your CodeRabbit configuration:

reviews:
  auto_review:
    drafts: true

• Autofix · Keep fixing CodeRabbit findings and required CI, and resolving merge conflicts

_{Comment @coderabbitai help to get the list of available commands.}