feat(data-access): auto-normalize enum values in PostgREST query filters

[ekremney]

Summary

• Auto-normalize enum attribute values in PostgREST query filters so that case mismatches (e.g. 'new' vs 'NEW') no longer silently return empty results
• Adds a #normalizeEnumValue helper to BaseCollection that matches input strings case-insensitively against the known enum values array (attribute.type)
• Applied in #applyKeyFilters (covers all synthetic accessors, allByIndexKeys, findByIndexKeys, updateByKeys, removeByIndexKeys) and the batchGetByKeys bulk .in() path
• No schema changes required — automatic for any attribute with type: Object.values(...) (array)

Test plan

• Unit tests: 6 new tests covering lowercase, mixed-case, non-enum passthrough, non-string passthrough, unrecognized value passthrough, and batch normalization
• All 1522 existing unit tests pass
• Integration tests: 2 new tests (Suggestion + Opportunity) verifying case-insensitive enum queries against real Postgres — run with npm run test:it

🤖 Generated with Claude Code

[github-actions]

This PR will trigger a minor release when merged.

[solaris007]

Hey @ekremney

Verdict: Approve (conditional) - right approach, needs a rebase before merge.

Clean, well-scoped fix at the right layer. The core implementation is sound - 11 lines that solve a real class of bugs. One critical branch divergence issue to resolve.

Must fix before merge

#	Issue	Severity
1	caseInsensitive flag regression - The branch predates the caseInsensitive / ilike feature in #applyKeyFilters. The PR's version of #applyKeyFilters drops the if (attr?.caseInsensitive) check entirely, breaking Site.baseURL lookups (which rely on ilike). Rebase on main and merge both behaviors: caseInsensitive -> ilike, else -> eq with #normalizeEnumValue.	Critical
2	Remove incorrect eslint-disable - // eslint-disable-next-line class-methods-use-this on #normalizeEnumValue is wrong. The method uses this.schema.getAttribute(key). Remove it.	Minor (quick fix)
3	Revert package-lock.json noise - 186 lines of unrelated "peer": true additions and sibling package version bumps. Split out or drop.	Minor (hygiene)

Should track as follow-ups

#	Issue	Severity
4	applyWhere gap - The where clause path (postgrest.utils.js) does op.eq() without normalization. where: (attr, op) => op.eq(attr.status, 'new') silently returns empty results. No callers currently use where with enum fields, so this is latent. Document or fix in follow-up.	Important (latent)
5	Write-side asymmetry - Reads now accept 'new' for 'NEW', but #validateItem on the write path uses strict includes() and will reject 'new'. Not a bug per se - reads are forgiving, writes enforce canonical values - but should be documented to avoid confusion.	Low
6	removeByIndexKeys bulk .in() path - Not normalized (unlike batchGetByKeys which is). Low severity since bulk deletes typically use UUIDs, not enum values.	Low
7	Direct PostgREST consumers - Mystique queries PostgREST directly and won't benefit from this fix. If Mystique hits the same case regression, a DB-level solution (citext) will still be needed.	Informational

No concerns

• Security - No vulnerabilities. Bounded match set from schema definitions, safe fallback, parameterized queries. No injection or authorization risk.
• Performance - .find() with .toLowerCase() over 2-8 enum values per key is negligible against PostgREST round-trip latency.
• Schema alignment - All model constants are consistently uppercase and match the schema type arrays.
• FK / ID resolution - #normalizeEnumValue correctly skips non-enum attributes, so UUIDs and FK attributes are untouched.

Recommended merged #applyKeyFilters

After rebase, the method should look like:

#applyKeyFilters(query, keys) {
  if (!isNonEmptyObject(keys)) return query;
  let filtered = query;
  Object.entries(keys).forEach(([key, value]) => {
    const dbField = this.#toDbField(key);
    const attr = this.schema.getAttribute(key);
    if (attr?.caseInsensitive && typeof value === 'string') {
      const escaped = value.replace(/[%_]/g, '\\$&');
      filtered = filtered.ilike(dbField, escaped);
    } else {
      filtered = filtered.eq(dbField, this.#normalizeEnumValue(key, value));
    }
  });
  return filtered;
}

[solaris007]

Correction to review item #1

The caseInsensitive / ilike feature in #applyKeyFilters was from PR #1388, which is closed and never merged to main. So there is no regression - the PR branch is up-to-date with main as-is.

That said, the two features (caseInsensitive ilike for baseURL and enum normalization for status fields) are complementary and should eventually both land. When they do, the merged #applyKeyFilters shown in the review is the right target shape. But this is a coordination item, not a blocker for this PR.

Updated verdict: items #2 and #3 remain (eslint-disable fix, package-lock cleanup). Item #1 is not a blocker.

[ekremney]

Thanks for the thorough review @solaris007!

Addressed the feedback:

Fixed:

• #2: Removed the incorrect eslint-disable-next-line class-methods-use-this — good catch, the method does use this.
• #3: Reverted the package-lock.json noise.
• #6: Added enum normalization to the removeByIndexKeys bulk .in() path as well — easy win.

Acknowledged follow-ups:

• #4 (applyWhere gap): Agreed, no callers currently use where with enum fields so it's latent. Will track for a follow-up.
• #5 (write-side asymmetry): Intentional — reads should be forgiving, writes should enforce canonical values. Worth documenting.
• #7 (direct PostgREST consumers): Noted. If Mystique hits the same issue, a DB-level solution would be the right approach.

And confirmed — #1 is not a blocker since PR #1388 was never merged. When both land, the merged #applyKeyFilters shape you suggested is the right target.

[ekremney]

Re #3 (package-lock.json): turns out the lockfile update is actually required — main's package.json references newer sibling packages (@adobe/spacecat-shared-ims-client@1.11.10, jose@6.1.2, etc.) that aren't reflected in main's lockfile. Without it, npm ci fails in CI. Reverted the revert to keep it in.

[solaris007]

🎉 This PR is included in version @adobe/spacecat-shared-data-access-v3.5.0 🎉

The release is available on:

• npm package (@latest dist-tag)
• GitHub release

Your semantic-release bot 📦🚀