psf-memo/docs/reviews/page-size-50-summary.md

Review: page-size-50

By architect.

Task and commits reviewed

• Task: page-size-50 — reduce the client page size from 100 to 50 across all paginated pages (recent feed, following feed, topic feed, notifications, search, profile, recent profiles) to cut payload size and reduce page load times, plus add pagination controls and full acceptance coverage.
• Inbound handoff: refactorer a9488d9ae2 (fast-forward merge onto swarmforge-architect).
• Reviewed commits: 81cedab (spec), 59cce442/merge 38f7a42 (implementation by coder), a9488d9 (refactorer — PaginatedPage base + property coverage), and 0dba837 (briefing/ backlog update after the youtube-embed merge).
• Scope note: the fast-forward also carried the prior 0dba837 specifier/backlog update; it is informational only, no code.

Architectural findings and fixes applied

• Good UI/Core separation: page services remain pure, testable controllers with the MemoDb client injected. The coder's change was a one-line default (limit = 100 → limit = 50) in each load/submit plus new canLoadMore() methods on the four pages that lacked them.
• Refactorer DRY extraction: extracted a shared PaginatedPage base (new src/services/paginated-page.js) and subclassed RecentFeedPage and RecentProfilesPage onto it, removing the duplicated load/canLoadMore pattern. Dependency direction is correct: the concrete pages depend on the pure base; the base has no IO. Each subclass keeps a thin custom item-finder (getPost/getProfile). Clean, cohesive, well-hidden. No structural changes required.
• Hardening (this review): killed the surviving canLoadMore mutation in the base and the three pages that added canLoadMore without tests (profile, search, topic-feed), plus notifications-page, all of which had the same uncovered ?? false default-on-null edge:
  • Added new test/unit/paginated-page.test.js (6 tests) covering load storage, the missing-client guard, the 50/0 defaults, and canLoadMore under null / missing-hasMore pagination.
  • Added canLoadMore null-pagination tests to notifications, profile, search, and topic-feed unit test files.
• Documented equivalent (not chased): the notifications-page.js constructor this.empty = false → true mutation survives because load() always overwrites empty before it is read (component and acceptance both read it after load). Genuine equivalent.
• Noted, not refactored: the pagination Previous/Next UI is duplicated across the profile, recent-profiles, and search components. Extracting a shared pagination component is a broad cross-module UI refactor beyond this handoff; the controller-side dedup via PaginatedPage is the appropriate scope here.

Verification results

• Language mutation (mutate4javascript <file> --max-workers 8 --mutate-all, sequential):
  • paginated-page.js Killed 5 / 0 survived / 0 uncovered
  • recent-feed-page.js Killed 1 / 0 / 0
  • recent-profiles-page.js Killed 1 / 0 / 0
  • following-feed-page.js Killed 12 / 0 / 0
  • notifications-page.js Killed 11 / 1 survived / 0 uncovered (the constructor empty equivalent documented above)
  • profile-page.js Killed 22 / 0 / 0
  • search-page.js Killed 9 / 0 / 0
  • topic-feed-page.js Killed 15 / 0 / 0
  • memo-db.js is excluded from mutation testing (ESM + ../config directory import, standing precedent).
• DRY (dry4javascript) on all touched services: no duplicate candidates.
• Soft Gherkin acceptance mutation (gherkin-mutator --level soft on page-size.feature): 0 killed, 24 survived. All 24 are single-character case/value mutations of setup example values — counts that stay above 50 and addresses/topics/queries/txids used consistently on both sides of their scenario. Intrinsic equivalents for a read-only feature; not chased.
• Unit tests (npm test): 280 passing (267 prior + 13 hardening), 0 fail.
• Property tests (npm run test:property): 40 passing (37 prior + 3 new paginated-page properties), 0 fail.
• Acceptance (npm run test:acceptance): all 22 generated suites PASS, including the new page-size scenarios.
• Lint: clean. Build: OK.

Suite status

• psf-memo-client: 280 unit, 40 property, acceptance PASS, lint clean, build OK.

Handoffs sent

• git_handoff to coder and refactorer (priority: 00) with the review commit for follow-up review.
• No functional commit for the specifier (the task produced a functional client change, but the specifier-authored spec 81cedab is already merged and there is no new feature to specify; the handoff chain ends at architect + coder/refactorer review).