mirror of https://github.com/Permissionless-Software-Foundation/psf-memo.git synced 2026-09-21 16:52:01 -07:00

Files

T

Chris Troutner 30c0949d3f Harden and review post-heights-index refactor

Architectural review of the refactorer's postHeights index work. Extract a
shared assemblePostPage helper and a ListUseCase base class to remove DRY
duplication in the post-list use cases. Build the gherkin-mutator runner
adapter for psf-memo-db and psf-memo-indexer. Add boundary and tie-break
tests to kill all surviving language mutations across the changed files
(post-query, pagination, list use cases, indexer post handler).

Mutation: all changed source files fully killed. DRY: clean. CRAP <= 6.
Soft Gherkin acceptance mutation documented survivors (see review report).

By architect.

2026-08-26 11:39:12 -07:00

4.6 KiB

Raw Blame History

Review summary: post-heights-index

Architect review of the refactorer handoff for task post-heights-index.

Commits reviewed

6a8cf6653e (refactorer): "Refactor postHeights index code and add property tests" — deduplicated postHeights query iteration into a shared generator, extracted shared pagination parsing/enrichment and the idempotent create-if-missing write, added property tests for postHeight key round-trips and ordering in both psf-memo-db and psf-memo-indexer, exposed them via a separate property command.
Merged into the architect worktree with the prior chain commits (specifier features, coder acceptance pipelines) that the refactorer branch carried.

Architectural findings and fixes applied

The refactorer's module structure was sound (shared lib/pagination.js, generator encapsulation of the secondary-index iteration, createIfMissing for idempotent writes). I applied the following structural fixes:

Extracted assemblePostPage into psf-memo-db/src/use-cases/lib/pagination.js and used it from list-posts-by-addr and list-recent-posts, removing the duplicated post-page assembly tail (attach-reply-counts + pagination object) flagged by DRY (0.84).
Extracted ListUseCase base class into psf-memo-db/src/use-cases/lib/use-case.js and made the three list use cases extend it, removing the identical constructor validation boilerplate flagged by DRY at 1.00 across all three files.
Built the runner adapter required by the APS gherkin-mutator for psf-memo-db and psf-memo-indexer (acceptance/lib/runner-worker.js in each). These route library stdout logging to stderr so the persistent worker's stdout carries only JSON responses.

Verification results

Language mutation (`mutate4javascript`, `--max-workers 8`, all covered sites)

Every changed source file is fully killed (0 survivors, 0 uncovered):

psf-memo-db/src/adapters/post-query.js — 22/22 killed
psf-memo-db/src/use-cases/lib/pagination.js — 13/13 killed
psf-memo-db/src/use-cases/lib/use-case.js — 0 sites (constructor only)
psf-memo-db/src/use-cases/list-posts-by-addr.js — killed
psf-memo-db/src/use-cases/list-recent-posts.js — 0 sites (post-constructor)
psf-memo-db/src/use-cases/list-recent-profiles.js — 9/9 killed
psf-memo-indexer/src/use-cases/action-types/post.js — 2/2 killed

Tests added to kill all initially-surviving mutants: parseLimit/parseOffset boundary values (1 and 100), hasMore exact-page boundaries, the sortProfiles seen tie-break with falsy values, scanPostsByAddrTxids/scanRecentPostTxids limit boundary, txidFromPostHeight with absent value, loadPostsByTxids blockHeight fallback, and the MAX_POST_SIZE exact-boundary in the indexer.

DRY (`dry4javascript`)

No duplicate candidates found on any changed source file.

CRAP / cyclomatic complexity (`crap4javascript`)

All changed functions within threshold (max CC 6, CRAP ≤ 6.0). Highest: PostQuery.scanPostsByAddrTxids CC 6 / CRAP 6.0; handlePost CC 4 / CRAP 4.8.

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

psf-memo-db efficient-post-pagination.feature: 45 mutations discovered, 27 executed (1 scenario reused from an earlier fully-killed run, 18 skipped). 23 killed, 4 survived — all limit example-value mutations (2→6, 3→7, 3→11, 3→5). Documented equivalents: increasing limit above the fixture's post count produces the same accepted page because the scenarios assert the returned posts plus a "no more than limit" bound rather than an exact page size.
psf-memo-indexer efficient-post-indexing.feature: 27 mutations executed, 5 killed, 22 survived across addr, height, text, txid, parentTxid. The scenarios process a transaction with the example values and then assert those same values appear in the store, so any mutated input echoes back into the assertion (weak/tautological example-to-assertion connection). These are specifier-side feature-quality improvements.

Suite status

psf-memo-db: unit 58 passing, property 3 passing, acceptance pass.
psf-memo-indexer: unit 29 passing, property 2 passing, acceptance pass.

Handoffs sent

git_handoff priority 00 to coder and refactorer (follow-up review of the architectural changes).
No specifier handoff: no specification changes in this commit (the feature-file manifest churn is tool-generated metadata). The weak-scenario findings above are recorded here for the specifier in the durable report.

By architect.

4.6 KiB Raw Blame History