feat(objectql): materialize seed rows on EVERY publish path (seedApplied)#1686
Merged
Conversation
…ied) Publishing a `seed` draft only promoted its metadata — whether rows actually landed depended on WHICH route you published through. The package route (/packages/:id/publish-drafts) applied seeds at the dispatcher level and swallowed failures silently; the per-ref publish (/meta/:type/:name/publish — what the home banner uses) never applied them at all. Result: "Published!" with an app full of empty tables. Move SeedLoaderService into @objectstack/objectql (runtime keeps a re-export shim) and apply seeds where every path converges — the protocol itself: - publishMetaItem: publishing a single `seed` materializes its rows from the just-promoted body (no read-back, no org-scope resolution pitfalls) and reports under `seedApplied`. Best-effort: a seed problem NEVER fails the publish, but it is now LOUD in the response instead of invisible. - publishPackageDrafts: structure publishes first, seeds LAST (their objects' tables must exist), with every seed body batch-applied in ONE loader pass so cross-seed references resolve regardless of draft order. Per-item self-apply is suppressed (_skipSeedApply) so nothing double-inserts. - http-dispatcher publish-drafts route: falls back to its route-level apply ONLY when the protocol didn't self-apply (custom protocol back-compat). objectql 546 + runtime 381 tests green; new coverage for ordering, batch single-pass, _skipSeedApply, non-seed publishes, and a real-loader smoke test. Co-Authored-By: Claude Opus 4.8 <noreply@anthropic.com>
|
The latest updates on your projects. Learn more about Vercel for GitHub.
|
xuyushun441-sys
added a commit
that referenced
this pull request
Jun 13, 2026
* chore: bump objectui to 893e5302174c feat(ADR-0046): package documentation portal + nav entry (#1686) objectui@893e5302174c7cbf75a7bed9e8e6dcb935273363 * feat(objectql): ADR-0048 Phase 1+2 — namespace install gate + prefer-local resolution Phase 1 — install-time namespace gate. `SchemaRegistry.installPackage` refuses a package whose `manifest.namespace` is already owned by a DIFFERENT installed package (new `NamespaceConflictError`), making explicit and early the constraint the object/table layer already enforces implicitly (a duplicate `CREATE TABLE <ns>_<obj>` fails at the DB). Same-package reinstall and shareable platform namespaces (base/system/sys) are exempt; OS_METADATA_COLLISION=warn downgrades to a warning. Phase 2 — prefer-local (container-scoped) resolution. `getItem(type, name, ns?)` resolves a bare name to the item owned by `ns`'s package before any cross-package fallback, preserving ADR-0005 overlay precedence and remaining backward compatible (param optional). `getApp` resolves prefer-local against its own name (app.name ≡ namespace in v1). The per-item collision guard now narrows to the cases prefer-local CANNOT disambiguate (shared/missing namespace), so two DIFFERENT-namespace packages legitimately coexist on the same bare name — the marketplace coexistence the revised ADR-0048 targets. Every existing collision test still passes (namespace-less fixtures remain a hard error). Tests: registry-namespace-install-gate.test.ts (7), registry-prefer-local- resolution.test.ts (7). Full objectql suite green (600). Co-Authored-By: Claude Opus 4.8 (1M context) <noreply@anthropic.com> * feat(ADR-0046): add optional `description` to package docs A doc can now carry a one-line `description` (frontmatter `description:`) — title / summary / body, the natural minimal model. - spec: DocSchema gains optional `description`. - cli: collect-docs reads `description:` from frontmatter (generalized the scalar extractor; title + description share it). Regression test added. - It travels in GET /meta/doc list (content is omitted there), so the docs portal shows summaries without fetching each body — verified live: showcase /meta/doc returns description, hasContent:false. - Example docs (app-showcase, app-todo) carry descriptions. - ADR-0046: schema + a P3 design note for doc TAGS (deferred) — tags are i18n-resolved keys, core protocol vocabulary + namespace-prefixed package tags; not a field to bolt on early. Co-Authored-By: Claude Opus 4.8 (1M context) <noreply@anthropic.com> --------- Co-authored-by: Jack Zhuang <277994282+os-zhuang@users.noreply.github.com> Co-authored-by: Claude Opus 4.8 (1M context) <noreply@anthropic.com>
xuyushun441-sys
added a commit
that referenced
this pull request
Jun 13, 2026
…local resolution (#1810) * chore: bump objectui to 893e5302174c feat(ADR-0046): package documentation portal + nav entry (#1686) objectui@893e5302174c7cbf75a7bed9e8e6dcb935273363 * feat(objectql): ADR-0048 Phase 1+2 — namespace install gate + prefer-local resolution Phase 1 — install-time namespace gate. `SchemaRegistry.installPackage` refuses a package whose `manifest.namespace` is already owned by a DIFFERENT installed package (new `NamespaceConflictError`), making explicit and early the constraint the object/table layer already enforces implicitly (a duplicate `CREATE TABLE <ns>_<obj>` fails at the DB). Same-package reinstall and shareable platform namespaces (base/system/sys) are exempt; OS_METADATA_COLLISION=warn downgrades to a warning. Phase 2 — prefer-local (container-scoped) resolution. `getItem(type, name, ns?)` resolves a bare name to the item owned by `ns`'s package before any cross-package fallback, preserving ADR-0005 overlay precedence and remaining backward compatible (param optional). `getApp` resolves prefer-local against its own name (app.name ≡ namespace in v1). The per-item collision guard now narrows to the cases prefer-local CANNOT disambiguate (shared/missing namespace), so two DIFFERENT-namespace packages legitimately coexist on the same bare name — the marketplace coexistence the revised ADR-0048 targets. Every existing collision test still passes (namespace-less fixtures remain a hard error). Tests: registry-namespace-install-gate.test.ts (7), registry-prefer-local- resolution.test.ts (7). Full objectql suite green (600). Co-Authored-By: Claude Opus 4.8 (1M context) <noreply@anthropic.com> * refactor(objectql): pivot ADR-0048 to option A — package-id-scoped resolution Per the sharpened ADR-0048 decision: the URL/route key is the package id (A), not the namespace. Migration cost is symmetric between A and B (both change one thing), so A wins on the dimensions that matter for an open marketplace — coordination-free global uniqueness, a single identity coordinate, no land-grab, and a self-describing URL. Backend: - getItem(type, name, currentPackageId?) prefers `${currentPackageId}:${name}` directly (was namespace→owner indirection). ADR-0005 overlay precedence and backward compatibility unchanged. - getApp(name, currentPackageId?) resolves prefer-local by package id. - Retire the per-item CROSS-package throw: package ids are globally unique, so package-scoped resolution always disambiguates two distinct packages — what the old guard flagged is now the supported coexistence case. Remove the now- dead MetadataCollisionError, findOtherPackageOwner, isPreferLocalDisambiguable, isRealPackage, SYS_METADATA_OWNER. collisionPolicy now governs only the Phase 1 namespace gate. Tests: rewrite *-cross-package-collision.test.ts from "throws" to "coexists + package-scoped resolves"; prefer-local test keys on package id. Full objectql suite green (598). ADR: §3.3 package-scoped (package-id) resolution; §3.4 cross-package throw retired; §3.6 transparent package-id URL + optional per-tenant namespace alias; phasing marks Phase 1 + Phase 2 backend/frontend done, package-id routing as the remaining frontend step. Co-Authored-By: Claude Opus 4.8 (1M context) <noreply@anthropic.com> * ci: re-trigger checks for the option-A pivot commit Co-Authored-By: Claude Opus 4.8 (1M context) <noreply@anthropic.com> --------- Co-authored-by: Jack Zhuang <277994282+os-zhuang@users.noreply.github.com> Co-authored-by: Claude Opus 4.8 (1M context) <noreply@anthropic.com>
This file contains hidden or bidirectional Unicode text that may be interpreted or compiled differently than what appears below. To review, open the file in an editor that reveals hidden Unicode characters.
Learn more about bidirectional Unicode characters
Sign up for free
to join this conversation on GitHub.
Already have an account?
Sign in to comment
1 participant
Add this suggestion to a batch that can be applied as a single commit.This suggestion is invalid because no changes were made to the code.Suggestions cannot be applied while the pull request is closed.Suggestions cannot be applied while viewing a subset of changes.Only one suggestion per line can be applied in a batch.Add this suggestion to a batch that can be applied as a single commit.Applying suggestions on deleted lines is not supported.You must change the existing code in this line in order to create a valid suggestion.Outdated suggestions cannot be applied.This suggestion has been applied or marked resolved.Suggestions cannot be applied from pending reviews.Suggestions cannot be applied on multi-line comments.Suggestions cannot be applied while the pull request is queued to merge.Suggestion cannot be applied right now. Please check back later.
Problem
Publishing a
seeddraft only promoted its metadata — whether the rows actually landed depended on which route you published through:/packages/:id/publish-draftsapplied seeds at the dispatcher level, swallowing failures silently (the chat says "Published!", the app's tables are empty — exactly what we hit live on staging with an AI-built recruiting app)./meta/:type/:name/publish(per-ref — what the home Publish banner uses since objectui#1621) never applied seeds at all.Fix
Move
SeedLoaderServiceinto@objectstack/objectql(runtime keeps a re-export shim —restcan't depend onruntime, and objectql owns both the engine and the publish primitive), then apply seeds where every path converges — the protocol itself:publishMetaItem: publishing a singleseedmaterializes its rows from the just-promoted body (no read-back → no org-scope resolution pitfalls) and reports underseedApplied. Best-effort: a seed problem never fails the publish, but it's now loud in the response instead of invisible.publishPackageDrafts: structure first, seeds last (their objects' tables must exist), all seed bodies batch-applied in one loader pass so cross-seed references resolve regardless of draft order. Per-item self-apply suppressed (_skipSeedApply) so nothing double-inserts.The per-ref REST route passes the protocol result through unchanged, so the banner path gets
seedAppliedwith zero rest-server changes.Test
_skipSeedApply; non-seed publishes untouched; per-ref self-apply; loud failure for unreadable bodies; real-loader smoke test (rows actually inserted via the engine).🤖 Generated with Claude Code