A typed IHttpRequest test builder — 4 ledgered TS2345 in packages/rest are hand-built request literals, and the only in-repo alternative is `as any`

[os-trump]

Filed unassigned and ungraded by the domain:cli dev seat, session session_01TvqBFLRzXdSPcbusDoED9k, while paying down packages/rest's test-typecheck ledger under #12573. Not graded, not routed.

This card exists because #12573's 08-27 record said the durable answer to this class "is its own card, not a rider" — and a dedup search finds no such card. Measured, the class is twice the size that record assumed.

The measurement

On origin/main at 56c5b1dbe7, packages/rest's test layer holds 6 errors. Four of the six are this one class:

npx tsc --noEmit --pretty false -p tsconfig.test.json    # in packages/rest

site	missing members of IHttpRequest
src/meta-public-book-grant.test.ts(138,25)	headers, path
src/rest-batch-size-cap.test.ts(152,26)	headers, path
src/rest.test.ts(2064,7)	headers, method, path
src/rest.test.ts(2089,7)	headers, method, path

All four are TS2345: an object literal is handed to a route handler whose parameter is IHttpRequest, and the literal omits members the interface declares required. The 08-27 record on #12573 knew of the first two; the two in rest.test.ts were behind PR #12421's fence at the time and were never classified. They are the same class, and they are missing method on top of headers and path.

Why no annotation repairs them

Adding the missing members changes what the handler receives — the fixture's data, not its annotation. A handler that reads req.path or req.method behaves differently once they are present, so inventing values silently changes what each test measures. That is the reason both #12573 dev rounds stopped here rather than forcing it.

The surrounding convention is as any, which #12573 forbids: of 281 .handler( call sites in packages/rest/src/*.test.ts, 151 carry as any within the call window. So the package already answers this question 151 times, in a form the paydown card rules out — and each of those sites invents its own header and path values, or omits them.

The shape of the durable answer

One typed builder that returns a complete IHttpRequest, so the required members are stated once instead of at every call site:

• required members get a single documented default (an empty headers object, a path derived from the route under test, an explicit method);
• each site overrides only what its own assertion is about;
• migrating a site becomes a mechanical, reviewable change instead of an invention.

⚠️ Deciding the defaults is the real content here, and it is a test-semantics question: a default path that does not match the route under test would make a passing test measure something other than what it names. ⛔ That is why this is a card and not a rider on a paydown.

Re-check

cd packages/rest && npx tsc --noEmit --pretty false -p tsconfig.test.json
cat packages/rest/test-typecheck-debt.json

Positive control for the as any count: the same scan reports 281 .handler( sites in the package's tests, so a zero would be visible as an instrument failure rather than a finding.

Refs

• Pay down packages/rest's new 37-error test-typecheck ledger — 13 of them are one class in one file, and 14 more look like one call-signature story across five #12573 — the ledger paydown these four block. Not addressed by that card; its remaining scope excludes them by ruling.
• State SaveReportInput's requirements at the reports.save door #12421 — merged as 3519f8d11f, which released the fence over src/rest.test.ts and made its two sites measurable.

Severity not judged; no urgency claimed — the debt is pinned by an EXACT shrink-only ratchet and reds on regression either way.

Generated by Claude Code

[os-trump]

Claim: domain:cli execution seat (#6024), session session_01TvqBFLRzXdSPcbusDoED9k, branch claude/issue-13377-typed-ihttprequest-builder.

All agents share one GitHub identity, so the assignee field cannot say whose claim this is — this comment can. An earlier Claim: comment here with a different session ID outranks this one; I re-read the thread after posting and found none.

Serial released by measurement, not by assumption

This card was held behind PR #13425 on a same-file hard serial (packages/rest/src/rest.test.ts). The release condition is the merge, not the arm and not the PR opening. Measured just now:

• origin/main = 82faea03f5 — fix(rest): state the exceljs Buffer-shim assertion once, in a typed test loader (#13425), judged by non-empty git log output, never by exit code (git log --grep exits 0 on zero matches — the [finding] ${PIPESTATUS[0]} is documented as a SAFE form but reads 0 for a red gate when the downstream reader closes the pipe early (| head) #12979 family).
• Cross-checked against the merge-queue refs: pr-13425-… is gone from refs/heads/gh-readonly-queue/main/* and its commit is in main. A vanished queue ref alone would be ambiguous — it means merged or ejected — so both readings were taken.

⇒ released. Dispatching now.

Line numbers re-measured, not carried forward on faith

I predicted 2063/2088 when #13425 was still open. Read off origin/main after the merge, that prediction holds:

site	literal	supplies
src/rest.test.ts:2063	{ params: { environmentId: 'proj-123', object: 'task' }, query: {} }	params, query
src/rest.test.ts:2088	{ params: { object: 'task' }, query: {} }	params, query

IHttpRequest (packages/spec/src/contracts/http-server.ts:18) requires params, query, headers, method, path; body, rawBody, remoteAddress are optional. So both sites omit three required members — headers, method, path — which is why they are TS2345 and not a missing-property nit.

The ledger's authored _note now reads, verbatim: "#13377 holds every entry that remains." Confirmed against entries — meta-public-book-grant.test.ts 1, rest-batch-size-cap.test.ts 1, rest.test.ts 2 = 4, the full class this card names.

---

Generated by Claude Code