# Refactor Backlog (Maintainability + System Integrity)

Last updated: 2026-03-12

## Priority 1 — Split `actions/index.js` by concern

- **Problem:** Single high-coupling module mixes API clients, hash/token helpers, logging, file/email/domain operations.
- **Why it matters:** Hard to reason about changes; high regression risk from unrelated edits.
- **Target outcome:** Small modules with explicit boundaries (e.g. `relayClient`, `fileClient`, `notifyClient`, `caseService`, `authService`).
- **Initial scope:** create façade layer first, then migrate call sites incrementally.
- **Detailed plan:** `memory-bank/refactor-plan-actions-index.md`
- **Test gate:** add unit regression tests + coverage reporting as part of this refactor.

## Priority 2 — Centralize relay/hash/token forwarding logic

- **Problem:** `pages/api/endpoint/**` and related routes duplicate request-signing + token + forwarding behavior.
- **Why it matters:** Security-sensitive drift and inconsistent error handling.
- **Target outcome:** Shared relay utility enforcing one signing contract, one timeout/retry policy, one error map, one redacted logging strategy.
- **Initial scope:** pilot on a small endpoint group, then roll out pattern.

## Priority 3 — Extract breadcrumb/back-link route-state rules

- **Problem:** Navigation logic in `components/breadcrumbs.js` and case summary is branch-heavy and query-dependent.
- **Why it matters:** Frequent regressions in search -> case -> myportal paths.
- **Target outcome:** Pure route-state helpers with table-driven tests; UI component mainly renders output.
- **Initial scope:** isolate `va/adv/ads/key` decision matrix first.

## Priority 4 — Standardize security guards in `pages/api/file/**`

- **Problem:** Hash/guard enforcement appears inconsistent across file endpoints.
- **Why it matters:** Uneven protection for document upload/download/delete paths.
- **Target outcome:** Shared guard middleware for hash/auth/input validation + consistent negative-path responses.
- **Initial scope:** enforce a common pre-handler contract on high-risk routes first.

### Priority 4 — Current slice status (2026-03-23)

- Completed on `TASK22224-file-endpoint-contract-bundle`:
    - `pages/api/file/downloadblob.js` hardening + hotfix closure
    - path normalization compatibility (filename-only + prefixed path)
    - encoded/raw hash validation compatibility guard
- Confirmed outcome:
    - user-reported download failures resolved
    - phase21 file + combined contract suites passing

### Priority 4 — Next recommended slice on this branch

- Target routes:
    - `pages/api/file/deleteblob.js`
    - `pages/api/file/deleteblobcase.js`
    - `pages/api/file/deleteblobrep.js`
- Scope:
    - align hash canonicalization strategy with `downloadblob.js`
    - enforce explicit `respondError` contracts (`MISSING_REQUIRED_QUERY`, `INVALID_HASH`, route-specific `*_FAILED`)
    - preserve existing success payload shapes
    - extend `tests/phase21/file-handler-contract.test.cjs` with mixed encoded/raw hash variants and negative paths

## Priority 5 — Establish minimum automated regression baseline

- **Problem:** Limited automated tests for high-risk logic.
- **Why it matters:** Repeated regressions and heavy manual verification burden.
- **Target outcome:** Focused test suite for pure logic and high-risk decisions.
- **Initial scope:**
    1. breadcrumb decision matrix
    2. representation eligibility/date windows
    3. hash utility behavior
    4. locale rewrite mapping sanity checks

## Priority 6 — Decompose `components/elements/index.js` monolith (phased)

- **Problem:** `components/elements/index.js` has grown into a high-coupling UI monolith (~2700+ LOC) combining field primitives, validation/conditional logic, translation helpers, file-upload orchestration, and field-array behavior.
- **Why it matters:** Very high regression surface for new appeal/my portal forms, slower change velocity, and poor testability/isolation.
- **Target outcome:** `components/elements/index.js` reduced to a thin barrel export with responsibility split into focused modules.
- **Context alignment:** Matches architecture direction in `context/architecture.md` (endpoint/component sprawl reduction, bounded slices) and file boundary guidance in `context/coding-conventions.md` (`pages` thin, reusable logic/components split by concern).
- **Guardrail constraints:**
    - Preserve public-service reliability for core flows (`search`, `case`, `myportal`, `newappeal`).
    - No auth/session/security header behavior changes (out of scope).
    - Maintain EN/CY parity for user-facing behavior.
    - Keep accessibility behavior unchanged (labels, focus, keyboard flow, errors).

### Priority 6 — Phase 1 (start here): extract pure helpers only

- **Scope (Phase 1 only):**
    - Move pure/helper logic from `components/elements/index.js` into focused helper modules under `components/elements/` (or `components/elements/helpers/`) without behavior change.
    - Candidate helper extraction set:
        - translation helpers (`FieldsTranslations`, picklist translation helper)
        - file upload helper utilities (icon/doc type naming/pure format helpers)
        - other deterministic pure functions used by field renderers
    - Keep all field renderers/components in place for Phase 1.
- **Non-goals (Phase 1):**
    - No JSX component relocation yet.
    - No upload flow logic rewrites.
    - No validation rule behavior changes.
    - No prop contract changes for existing consumers.

- **Acceptance criteria (Phase 1):**
    - `components/elements/index.js` imports extracted helpers from new helper modules and behavior remains equivalent.
    - No route/API changes.
    - Existing new appeal + myportal form journeys continue to function in EN and CY.
    - Accessibility smoke unchanged for touched form controls (label association, keyboard reachability, inline error visibility).
    - Lint passes for touched files.

- **Validation matrix (minimum):**
    1. `npm run lint`
    2. Manual smoke:
        - new appeal form step rendering + validation messages
        - myportal representation/new appeal editing flow controls
        - file upload field icon/name behavior unchanged
    3. Locale parity checks (EN/CY) for touched user-facing labels/routes.
    4. A11y smoke checks on touched fields (focus, labels, errors).

- **Rollback plan (Phase 1):**
    - Revert helper module extraction commit(s) to restore single-file implementation.
    - No migration/data rollback required.

- **Next phases (for tracking):**
    - **Phase 2:** extract low-risk leaf field renderer components.
    - **Phase 3:** extract `RenderFileUpload` and upload container.
    - **Phase 4:** extract field-array/repeater components.
    - **Phase 5:** finalize `components/elements/index.js` as barrel-only export.

## Sequencing recommendation

1. Priorities 2 + 4 (security/integrity foundation)
2. Priority 3 (high-change regression hotspot)
3. Priority 1 (structural maintainability)
4. Priority 5 (continuous safety net, starts early and expands)