psf-memo/docs/reviews/like-result-modal-summary.md

like-result-modal — Architect Review

Task: like-result-modal Component: psf-memo-client Base: 3486da3 (last merged architect review); inbound refactorer commit e978516

What was reviewed

Inbound refactorer batch (priority 50), merged onto swarmforge-architect by fast-forwarding 3486da3 -> e978516. The linear chain reviewed:

• efe7a6c — specifier: Record post-options-menu completion in backlog and briefing.
• 3af0d42 — specifier: Add like broadcast result specification. Adds psf-memo-client/specs/like-broadcast-result.feature (3 scenario outlines).
• a0040f9 — coder: Implement like broadcast result modal. Keeps the like/tip modal open after a successful like and shows the success message, the like txid, and a block-explorer link. Adds the pure LikeTipPage result state, the presentational like-result.js, the acceptance render adapter, step handlers, and unit tests.
• e978516 — refactorer: Refactor like result modal and share the block explorer link. Extracts src/services/block-explorer.js as the single source of the bch.loping.net/tx URL shared by the New Post result modal, the post options menu, and the like/tip result; DRYs the new acceptance steps behind renderLikeBroadcastResult; adds property tests.

Architect review commit: e071881ea0 — the mutate4javascript footer manifests, the soft gherkin-mutator acceptance-mutation manifest stamp, and the hardening changes below. The summary and verification record are committed on top, so git diff e071881ea0 HEAD touches only docs/. The record's git_sha is e071881ea0, the commit that contains the verified source state.

Architectural findings and fixes applied

The refactorer's structure is sound: a framework-free controller, a server-renderable presentational component, a pure shared URL module, and a separate acceptance render adapter. The hardening work was mutation- and duplication-driven.

1. UI/Core separation. src/services/block-explorer.js is a pure leaf (if (!txid) return '' + a template literal) with no React, DOM, or IO. LikeTipPage owns the broadcast-result state machine and is exercised with no browser; LikeResult only maps props to markup and is reached by the acceptance adapter through acceptance/lib/render-like-result.js, which server-renders the actual component.
2. Dependency rule. new-post.js, post-options.js, and like-tip-page.js depend inward on block-explorer.js; nothing in the pure modules reaches out to React, the router, or the wallet. No framework or persistence structure leaks across the boundary.
3. Information hiding / DRY. The block-explorer base URL and link shape now have exactly one definition. The old EXPLORER_TX_BASE literals in new-post.js, post-options.js, and like-tip-page.js are gone; the EXPLORER_TX_BASE/explorerUrl/explorerTxUrl names remain as thin aliases for existing acceptance/test callers.
4. Mutation survivors (like-tip-page). Initial run: 13 killed, 8 survived, 0 uncovered across 21 sites. Every survivor was a missing assertion, not a design fault:
   • the false initial tipping/modalOpen, and the deps.postTxid/ deps.authorAddress constructor seeds, were never asserted;
   • open's < dust comparison had no boundary case (balance exactly at 3000);
   • the open success result's true was unchecked;
   • _parseTip's second || (null/undefined) was uncovered;
   • _handleSubmitFailure's err.message || String(err) survived because a broadcast failure is re-derived by the superclass, so the value is only observable on a validation failure. Added unit coverage for each; re-run: 21 killed, 0 survived, 0 uncovered.
5. Mutation survivor (new-post). The NewPostPage constructor's this.showResultModal = false (line 32) was a pre-existing survivor. Added an initial-state assertion; re-run: 10 killed, 0 survived, 0 uncovered.
6. Test duplication. The first DRY pass flagged two structurally identical test pairs in like-tip-page.test.js. Extracted a submitLike() helper and a balance option on makePage(); the remaining pair (dismiss vs. reopen) is semantically distinct and is left as structural test boilerplate.
7. Result-modal controller duplication (documented, not changed). LikeTipPage mirrors the showResultModal/lastResult/submit/ dismissResult shape of NewPostPage, with a different show/dismiss policy. A shared result-modal controller would touch both controllers, their unit tests, and acceptance handlers; per docs/architect-process-notes.md that is a broad cross-module refactor beyond a single review, so it is recorded as a follow-up candidate rather than folded into this task. Likewise, the React like-tip-modal.js keeps its own transient showResult/resultTxid state; it is the environmentally unsuitable shell that node unit and acceptance mutation deliberately do not target, matching the existing component pattern.

Verification results

Language mutation (mutate4javascript, --mutate-all, --max-workers 8)

File	Sites	Killed	Survived	Uncovered
src/services/like-tip-page.js	21	21	0	0
src/services/new-post.js	10	10	0	0
src/services/post-options.js	10	10	0	0
src/services/block-explorer.js	0	0	0	0
src/components/post-feed/like-result.js	0	0	0	0

block-explorer.js and like-result.js contain only !, a ternary, and a template literal — constructs mutate4javascript does not target — so --scan reports 0 sites; this is structural, not a skipped run. The two JSX/ESM components (like-tip-modal.js, post-feed-item.js) are outside the node unit test boundary and are covered end-to-end by acceptance.

DRY (dry4javascript)

Focused run over the changed source, tests, and adapters: the only task-local duplicate left is the semantically distinct dismiss/reopen test pair noted above. A broad run over src test acceptance reports 475 blocks, all pre-existing acceptance/lib/handlers.js step-handler boilerplate and repeated test setup across older suites — none involves the new production modules.

CRAP / cyclomatic complexity (crap4javascript)

All changed functions at or below the 8.0 threshold with 100% coverage, e.g. LikeTipPage._parseTip (CC 5, CRAP 5.0), LikeTipPage.open (CC 4, CRAP 4.0), LikeTipPage.getBroadcastMessage (CC 3, CRAP 3.0), NewPostPage.submit/ dismissResult (CC 3, CRAP 3.0), blockExplorerTxUrl (CC 2, CRAP 2.0), LikeResult (CC 2, CRAP 2.0).

Soft Gherkin acceptance mutation (gherkin-mutator --level soft)

like-broadcast-result.feature: 7 executed, 4 killed, 3 survived, 0 errors. The three survivors are consistent-value intrinsic equivalents — each mutated example value is used consistently on both the setup and assertion sides:

• tip: 600 -> 601 and tip: 25000 -> 25007 (Scenario 2): the tip is entered and then asserted against the same <tip> parameter.
• Scenario 3 liked_txid with an injected x: Scenario 3's purpose is the dismiss-closes-modal behavior; it reuses <liked_txid> on the setup and click steps but makes no txid-validity or broadcast assertion (Scenarios 1 and 2, which do assert the broadcast, killed the same class of mutation).

The tool stamped Scenario 0 (all mutations killed) and left Scenarios 1–2 unstamped, as designed.

Suite status

swarmforge/scripts/verify.sh client --record docs/reviews/like-result-modal-verification.json --task like-result-modal -> pass (5/5): unit 387 pass / 0 fail, property 75 pass / 0 fail, acceptance all 29 suites passed, lint ok, build ok. Record git_sha = e071881ea0.

Handoffs sent

• End-of-chain git_handoff to the specifier (task like-result-modal) with the review commit so it can merge swarmforge-architect into master.
• No coder/refactorer handoff: the review is test hardening plus manifest refresh with no follow-up work for those roles.

By architect.