Pivots skip frontend emission
Auto-generated from the source archive. Do not edit by hand — rerun
catalyst-changelog-syncinstead.
What changed
Section titled “What changed”- BREAKING — pivot aggregates (auto-derived from m2m relationships) no longer emit
list / detail / form / form-embed / list-actions-cell / columns / indexfiles. Their UI representation lives in the m2m widget on the parent aggregate. - Data-access stays: pivots keep emitting
list-config,detail-config,graphql,resolvers, and their type entries — these feed the manager widget’s orchestrator. - New
[ORPHAN PIVOT FILES]warning lists pre-existing pivot UI files marked safe-to-delete on regen. The codegen never deletes user files; the developer runsrmmanually. - New helper
isPivotAggregate(schema)detects pivots structurally (every non-system property is many-to-one).
Why it matters
Section titled “Why it matters”This encodes a longstanding Aurora rule at the codegen level: pivots are never first-class frontend subjects. They exist as backend storage for an m2m; their UI is the widget on the m2m field on the parent aggregate. Existing projects with previously-emitted pivot directories (iam/permission-role/, iam/role-account/, iam/tenant-account/) need a manual cleanup pass after regen — the codegen warns but does not delete. After cleanup, the regen output is consistent: no orphan pivot files, no broken types, no TS2339 row.name on aggregates without a name.