Recognize `.f.mjs` in proof discovery and coverage tooling by sergey-shandar · Pull Request #1422 · functionalscript/functionalscript

sergey-shandar · 2026-08-04T10:27:02Z

Implements the tooling half of
fjs/emergent_testing/todo/f-mjs-test-and-coverage.md
(P1).

Authored .f.mjs is the extension a FunctionalScript module moves to once the
current parser and compiler accept it. Before the first rename, the test and
coverage tooling has to treat it like .f.ts — otherwise migrating a module
silently drops its proofs and its coverage rows while CI stays green.

Changes

Proof discovery — shouldLoad (fjs/dev/module.f.ts) matches .f.mjs,
so a migrated module.f.mjs and its internal proof export stay in
bulk-load discovery. Ordinary .mjs files remain opt-in through the existing
proof.mjs convention. Covered by an extended shouldLoad proof and a new
virtual-filesystem allFiles proof.
Node coverage — npm run cov now passes --test-coverage-include for
both **/module.f.ts and **/module.f.mjs.
Deno coverage — deno task cov and the generated CI step use
.*module\.f\.(ts|mjs).
CI generator — fjs/ci/deno/module.f.ts exports coverageInclude as the
single source of that filter. New fjs/ci/deno/proof.f.ts fails if .f.mjs
support is removed from it, and fjs/ci/proof.f.ts now asserts against the
same constant instead of a duplicated literal. .github/workflows/ci.yml is
regenerated from the generator.
Policy — AGENTS.md §2/§3.1/§3.2 and the matching summary in
CONTRIBUTING.md apply mandatory 100% proof coverage to both authored
extensions and document the mixed module.f.mjs / proof.f.ts layout used
during incremental migration, plus the dependency-closure rule a
proof.f.mjs must satisfy.

What is not in this PR, and why

The issue also asked for two runtime fixtures. Both turned out to depend on
fjs/ci/todo/f-mjs-package-support.md:

A proof.f.ts importing a module.f.mjs fails npx tsc with TS7016
while allowJs is off. Turning allowJs/checkJs on makes npx tsc pass,
but npm run prepack (tsc --NoEmit false) then fails with TS5055
because authored .mjs becomes both an input and a JavaScript emit target —
exactly the repeatable-emission work owned by the package-support issue.
An internal proof inside a .f.mjs module has no assert helper it may
import: AGENTS.md §3.3 requires assert/assertEq from
fjs/asserts/module.f.ts, but the dependency-closure rule forbids authored
.f.mjs from importing a relative .f.ts module.

Rather than duplicate the package-support work here, the issue file is rewritten
to cover just the remaining fixtures, marked blocked on package support, with
both obstacles and a proposed resolution for the assert-helper question recorded.
f-mjs-package-support.md and fjs/fsc/README.md are updated so the ordering
between the two prerequisites is stated in one direction everywhere.

Verification

npx tsc — clean.
npm test — 2290 pass, 0 fail.
npm run ci-update — regenerated workflow matches the committed one.
node --test --experimental-test-coverage with both include flags — accepted,
report unchanged for .f.ts modules.

Deno is not installed in this environment, so deno task cov was not executed;
the filter change is covered by the generator proof instead.

Generated by Claude Code

Authored `.f.mjs` is the FunctionalScript extension a module moves to once the current parser and compiler accept it. Before the first rename, the test and coverage tooling has to treat it like `.f.ts`, otherwise migrating a module silently drops its proofs and its coverage rows. - `shouldLoad` matches `.f.mjs`, so a migrated module (and its internal `proof` export) stays in bulk-load discovery. Ordinary `.mjs` files remain opt-in through the `proof.mjs` convention. - `npm run cov` and `deno task cov` include `module.f.mjs` alongside `module.f.ts`. - `fjs/ci/deno/module.f.ts` exports `coverageInclude` as the single source of the generated Deno coverage filter, with a proof that fails if `.f.mjs` support is removed; `fjs/ci/proof.f.ts` asserts against the same constant instead of a duplicated literal. The checked-in workflow is regenerated. - `AGENTS.md` §3.2 and the `CONTRIBUTING.md` summary apply mandatory proof coverage to both authored extensions and document the mixed `module.f.mjs` / `proof.f.ts` layout used during migration. The issue's two runtime fixtures do not land here: a `proof.f.ts` cannot import a `module.f.mjs` until `allowJs`/`checkJs` are on, and enabling those alone breaks `npm run prepack` with TS5055 — the repeatable-emission work owned by `f-mjs-package-support.md`. The issue is updated to record that dependency, the assert-helper question for an internal `.f.mjs` proof, and the resulting ordering between the two prerequisites. Co-Authored-By: Claude Opus 5 <noreply@anthropic.com> Claude-Session: https://claude.ai/code/session_01JimQDf1MdP9oPskJj7nfrn

Co-Authored-By: Claude Opus 5 <noreply@anthropic.com> Claude-Session: https://claude.ai/code/session_01JimQDf1MdP9oPskJj7nfrn

sergey-shandar · 2026-08-04T13:14:58Z

 */
 export const shouldLoad = (s: string): boolean =>
-    s.endsWith('.f.ts')    || s.endsWith('.f.js')    ||
+    s.endsWith('.f.ts')    || s.endsWith('.f.js')    || s.endsWith('.f.mjs') ||


Make it symmetrical, accept:

All FunctionalScript files: ends with:.f.ts, .f.mts, .f.js and .f.mjs,

All unpure JavaScript/TypeScript proofs: ends with: proof.ts, proof.mts, proof.js and proof.mjs.

Done in 60a534f — shouldLoad is now the two symmetrical families you described, and the doc comment states them as such:

FunctionalScript modules, loaded whatever the file name: .f.ts, .f.mts, .f.js, .f.mjs

impure JavaScript/TypeScript, loaded only under the proof name: proof.ts, proof.mts, proof.js, proof.mjs

The proof now covers both families exhaustively, including the negative side for every impure extension (module.ts / module.mts / module.js / module.mjs are all rejected), and the allFiles discovery proof gained a .f.mts entry.

I left the coverage filters (npm run cov, deno task cov, the generated Deno CI step) on module.f.ts + module.f.mjs, since those are the two authored implementation extensions in fjs/fsc/README.md. Say the word if you'd like .f.mts there too.

Generated by Claude Code

Review feedback: state the rule as two symmetrical families instead of an ad-hoc list. FunctionalScript modules are loaded whatever the file name (`.f.ts`, `.f.mts`, `.f.js`, `.f.mjs`); impure JavaScript/TypeScript is loaded only under the `proof` name (`proof.ts`, `proof.mts`, `proof.js`, `proof.mjs`). This adds `.f.mts` to discovery and covers both families exhaustively in the proofs, including the negative side for every impure extension. Co-Authored-By: Claude Opus 5 <noreply@anthropic.com> Claude-Session: https://claude.ai/code/session_01JimQDf1MdP9oPskJj7nfrn

sergey-shandar · 2026-08-04T13:27:04Z

@codex review

chatgpt-codex-connector · 2026-08-04T13:31:32Z

Codex Review: Didn't find any major issues. Keep them coming!

Reviewed commit: 9225b09cfd

ℹ️ About Codex in GitHub

Your team has set up Codex to review pull requests in this repo. Reviews are triggered when you

Open a pull request for review
Mark a draft as ready
Comment "@codex review".

If Codex has suggestions, it will comment; otherwise it will react with 👍.

Codex can also answer questions or update the PR. Try commenting "@codex address that feedback".

My original item was wrong twice. It proposed adding `--test-coverage-include=**/module.f.mjs`, which #1422 already added, and it blamed the include globs for the vacuous report -- but those only filter which files appear in the report, so they cannot make a run that executed nothing report something. The cause is discovery: node --test's default patterns don't match proof.f.mjs / module.f.mjs, and the only file that does match is fjs/emergent_testing/all.test.ts, where behaviour is Node-version dependent -- 0 tests on v23 per review, while on v22.22.2 it discovers that file and runs the suite through it. Rewritten around that, with the version split recorded so the next person pins one. Also qualifies the "continue upward" checkbox: the scenarios are authored TypeScript proof source, so the bare [x] read as contradicting the new section that says so. Reported by o2alexanderfedin, who re-measured every other number in the section and found them accurate.

claude added 2 commits August 4, 2026 10:26

Add CHANGELOG entry for #1422

58cbd2f

Co-Authored-By: Claude Opus 5 <noreply@anthropic.com> Claude-Session: https://claude.ai/code/session_01JimQDf1MdP9oPskJj7nfrn

sergey-shandar commented Aug 4, 2026

View reviewed changes

claude and others added 2 commits August 4, 2026 13:16

Merge branch 'main' into claude/epic-fermi-lnsw8v

bb9a24e

Merge branch 'main' into claude/epic-fermi-lnsw8v

9225b09

sergey-shandar merged commit 28d8b1b into main Aug 4, 2026
19 checks passed

sergey-shandar deleted the claude/epic-fermi-lnsw8v branch August 4, 2026 13:32

o2alexanderfedin mentioned this pull request Aug 12, 2026

Record what remains after stage 1 in the migration TODO #1506

Merged

3 tasks

Navigation Menu

Uh oh!

Uh oh!

Recognize `.f.mjs` in proof discovery and coverage tooling - #1422

Recognize `.f.mjs` in proof discovery and coverage tooling#1422
sergey-shandar merged 5 commits into
mainfrom
claude/epic-fermi-lnsw8v

sergey-shandar commented Aug 4, 2026

Uh oh!

sergey-shandar Aug 4, 2026

Uh oh!

sergey-shandar Aug 4, 2026

Uh oh!

sergey-shandar commented Aug 4, 2026

Uh oh!

chatgpt-codex-connector Bot commented Aug 4, 2026

Uh oh!

Uh oh!

Reviewers

Assignees

Labels

Projects

Milestone

Development

Uh oh!

2 participants

Uh oh!

Uh oh!

Conversation

sergey-shandar commented Aug 4, 2026

Changes

What is not in this PR, and why

Verification

Uh oh!

sergey-shandar Aug 4, 2026

Choose a reason for hiding this comment

Uh oh!

sergey-shandar Aug 4, 2026

Choose a reason for hiding this comment

Uh oh!

sergey-shandar commented Aug 4, 2026

Uh oh!

chatgpt-codex-connector Bot commented Aug 4, 2026

Uh oh!

Uh oh!

Reviewers

Assignees

Labels

Projects

Milestone

Development

Uh oh!

2 participants