SITES-49202: remove deprecated brand semrushWorkspaceId mirror (§5 PR B, spacecat-shared)

[aliciadriani]

What & why

PR B of the phased §5 Serenity post-GA cleanup — removes the deprecated, read-only brand semrushWorkspaceId mirror from @adobe/spacecat-shared-data-access.

This mirror shadowed the canonical semrushSubWorkspaceId (write-of-record) and was maintained by the mysticat-data-service brands_sync_semrush_workspace_id DB trigger. Now that PR A (project-elmo-ui #2730, merged) migrated the last direct reader, the mirror can go.

Precondition (verified before removal)

• Prod DB: 0 rows where the brand mirror is the only value (carries no unique data).
• No backend caller of the brand finder/setter. Grepped origin/main of spacecat-shared, spacecat-api-service, llmo-data-retrieval-service, project-elmo-ui, mysticat-data-service for findBySemrushWorkspaceId / allBySemrushWorkspaceId / setSemrushWorkspaceId / brand-context getSemrushWorkspaceId( — every hit is org-context (Organization / org. / organization.) or design-spec prose. No genuine brand-context reader.

Datastore / index note — NO infra op (corrected)

An earlier version of this description flagged a "DynamoDB GSI infra review". That was wrong and is retracted.

This data-access layer is PostgREST/Postgres-backed (base.collection.js → static DATASTORE_TYPE = DATASTORE_TYPE.POSTGREST, @supabase/postgrest-js, POSTGREST_URL). Despite the legacy electrodb-style gsi{N}pk / addAllIndex naming in schema.builder.js, there is no DynamoDB table or GSI here — verified directly against spacecat-infrastructure (which provisions the Mysticat Postgres service + one unrelated github-service lock table, no spacecat data table with brand GSIs).

Removing addAllIndex(['semrushWorkspaceId']) just stops the data-access layer generating the semrushWorkspaceId finder — no infra/table change, and none needed. The one physical artifact is the Postgres unique index brands_semrush_workspace_id_key ON brands(semrush_workspace_id) (confirmed live on prod), which drops automatically with the column in PR C (mysticat-data-service migration). So this PR is code-only with no infra dependency.

org-level field deliberately KEPT

Organization.semrushWorkspaceId (correctly-named, distinct entity) is untouched — this is a brand-only removal. The symbol collision is the trap here; org getter/finder/setter all stay.

Changes (brand model only)

File	Change
src/models/brand/brand.schema.js	Removed the semrushWorkspaceId readOnly mirror attribute and addAllIndex(['semrushWorkspaceId']); tidied surrounding comments.
src/models/brand/brand.model.js	Removed the @deprecated setSemrushWorkspaceId() BC delegate + docblock; updated the class docblock to describe semrushSubWorkspaceId as the dual-mode switch.
src/models/brand/brand.collection.js	Updated docblock: pointer-repair path now references findBySemrushSubWorkspaceId.
src/models/brand/index.d.ts	Removed the mirror type decls: getSemrushWorkspaceId, setSemrushWorkspaceId, allBySemrushWorkspaceId, findBySemrushWorkspaceId.
test/unit/models/brand/brand.schema.test.js	Removed the deprecated-mirror attribute describe block.
test/unit/models/brand/brand.model.test.js	Removed the deprecated-mirror getter/setter describe block and the fixture field.
test/unit/models/brand/brand.collection.test.js	Removed the mirror index-accessor assertion and the fixture field.

Next step — PR C (do NOT land before this releases)

PR C = mysticat-data-service migration dropping the brands.semrush_workspace_id column and the sync_semrush_workspace_id_from_sub trigger (the Postgres unique index brands_semrush_workspace_id_key drops with the column). It must land only AFTER this PR is released/deployed, so no deployed consumer still references the mirror attribute when the column/trigger disappear.

Semver

Published package — this removes public model members, so a semver bump is expected and handled by semantic-release from the conventional-commit message. No manual version edits.

Validation

• npm run lint (data-access): clean.
• npm test (data-access, full unit suite): 2774 passing, 0 failing.

Links

• Jira: https://jira.corp.adobe.com/browse/SITES-49202
• Serenity cleanup spec (§5): https://github.com/adobe/serenity-docs/pull/62

[aliciadriani]

Review — §5 PR B (manual)

Summary

Removes the deprecated brand semrushWorkspaceId mirror (readOnly attribute, its addAllIndex GSI, the @deprecated setSemrushWorkspaceId() delegate, and the finder/getter type decls) from the electrodb Brand model. Consistent and well-scoped; the org-level field is correctly left alone.

Must Fix

None.

Should Fix

None.

Nits

None material. The brand.collection.js pointer-repair docblock now points at findBySemrushSubWorkspaceId — verified that finder is generated from the retained addAllIndex(['semrushSubWorkspaceId']), so the doc is accurate.

What's Good

• Pre-removal grep across spacecat-shared / spacecat-api-service / llmo-data-retrieval-service / project-elmo-ui / mysticat-data-service found no brand-context reader — every hit was the org field or design-spec prose. Precondition solid.
• Reads of existing DynamoDB items are unaffected (electrodb ignores stored fields absent from the schema); the attribute was readOnly, so nothing wrote through it.
• 2774 data-access unit tests pass; lint clean.
• PR body correctly flags the DynamoDB GSI infra-review gate, the PR C ordering (must land after this releases), and the semver bump.

[aliciadriani]

DynamoDB GSI infra review — CLEARED (safe to merge; no infra change needed)

TL;DR: removing addAllIndex(['semrushWorkspaceId']) is safe with no Terraform/GSI teardown required, because the electrodb Brand entity is not populated — nothing anywhere writes or reads it.

What I checked

1. Is the entity used? Grepped all 5 repos (spacecat-shared, spacecat-api-service, llmo-data-retrieval-service, project-elmo-ui, mysticat-data-service) for dataAccess.Brand reads/writes/queries → zero. Serenity brands live in mysticat-data-service Postgres via PostgREST (brands-storage.js), never the electrodb Brand model. So there are no Brand items in DynamoDB.
2. Does removal renumber the GSIs? Yes. schema.builder.js assigns physical gsi{N}pk/sk by index insertion order. On main Brand has two all-indexes (semrushWorkspaceId=gsi1, semrushSubWorkspaceId=gsi2); after removal semrushSubWorkspaceId shifts to gsi1 (empirically confirmed: PR branch builds gsi1 only). On a populated entity this would orphan existing items' GSI-projected data and break findBySemrushSubWorkspaceId — but per (1) the entity is empty, so zero runtime impact.
3. Physical GSI teardown? Not applicable. This is a shared single-table design — physical gsi1..gsi5 are shared across all entities (Site/Org/…) and stay provisioned. Removing Brand's logical index doesn't delete any physical GSI. There is no dedicated brand GSI to tear down.

Verdict

Infra-safe to merge. The PR body's "physical GSI teardown is a separate infra op" caveat can be downgraded — no such op exists here.

One caveat + a follow-up

• I could not inspect spacecat-infrastructure directly (not checked out). Evidence strongly indicates standard shared gsi1-5; a 30-sec confirm from the table owner that no dedicated brand GSI is provisioned would make this airtight.
• Bigger picture: the electrodb Brand entity appears entirely dead (unused everywhere). Worth a separate follow-up to consider removing the whole model — out of scope for §5.

[github-actions]

This PR will trigger a patch release when merged.

[aliciadriani]

• spacecat-shared-data-access is PostgREST/Postgres-backed, not DynamoDB. base.collection.js:53 is static DATASTORE_TYPE = DATASTORE_TYPE.POSTGREST; the layer requires POSTGREST_URL/POSTGREST_SCHEMA/POSTGREST_API_KEY and queries via @supabase/postgrest-js (postgrestService.from(tableName)). The gsi{N}pk / addAllIndex naming in schema.builder.js is a legacy electrodb vestige — no DynamoDB table or GSI is involved.
• spacecat-infrastructure has no spacecat data table with brand GSIs — it provisions the Mysticat Postgres data service and one unrelated mysticat_github_service idempotency lock table. There is no DynamoDB GSI to review or tear down.
• The real physical artifact behind findBySemrushWorkspaceId is a Postgres UNIQUE index — brands_semrush_workspace_id_key ON brands (semrush_workspace_id), confirmed present on prod right now. It is dropped automatically when §5 PR C drops the brands.semrush_workspace_id column (mysticat-data-service migration). Nothing to do in Terraform/DynamoDB.

Net

This PR is a code-only schema/finder removal with NO infra dependency — it just stops the data-access layer generating the semrushWorkspaceId finder. The prior "DynamoDB GSI teardown is a separate infra op" flag in the PR body is retracted — no such op exists. The only physical cleanup (the Postgres index) rides along with §5 PR C.

The renumbering note from my earlier comment is also moot (and was DynamoDB-specific); disregard it.