Skip to content

results-scorer-check.md

Latest commit

 

History

History
37 lines (31 loc) · 2.29 KB

results-scorer-check.md

File metadata and controls

37 lines (31 loc) · 2.29 KB
Raw

Scorer-check — JS/Go profile validation

Baseline-only measurement of the bundled test-quality scorer over six cloned OSS suites. This is profile validation, not a benchmark run: there is no generated suite and no win/loss verdict — only confirmation that each auto-counted axis fires sensibly on real framework idioms.

Direction: ↓ lower-is-better, ↑ higher-is-better. n/a = axis not reliably countable for that language (excluded from any tally).

repo lang head files tests loc A1 ↓ A2 ↓ A4 ↓ A5 ↓ C1 ↓ B1 ↑ D1 ↓ D2 ↑ C2
expressjs/express js dae209a 91 1128 16439 25 2 0 0 0 28 14.57 0 879
auth0/node-jsonwebtoken js 02688b9 34 253 3799 15 0 1 0 0 21 15.02 0 9
colinhacks/zod js bbc68f9 172 1946 37466 69 497 0 0 0 915 19.25 0 0
go-chi/chi go 3b17157 24 111 5957 0 n/a 0 0 0 110 53.67 0.369 100
tidwall/gjson go 7d8b382 1 98 2760 0 n/a 0 0 0 11 28.16 0.02 0
golang-jwt/jwt go e8e5b83 19 45 3510 1 n/a 0 0 0 9 78 0.378 0

Notes (what each axis confirms / caveats)

  • express = node:assert + supertest. A.1 catches assert.throws(fn, /re/); B.1 catches assert.strictEqual(x, 'literal'); C.2 is supertest .expect.
  • jsonwebtoken = Chai BDD + chai-assert + sinon.useFakeTimers. A.1 catches .to.throw('msg') and .message…include; B.1 catches .to.equal('literal'); the sinon usage is fake-timers → C.2 (legit time control), so C.1 = 0 is correct, not a miss.
  • zod = Vitest + inline snapshots. B.1 is high (exact .toBe/snapshots). D2 = 0 is a true negative: zod parametrizes with raw for loops, not framework .each tables.
  • Go A.2 = n/a (same-package access to unexported names is idiomatic). D1_loc_per_test is inflated (counts func TestX, not t.Run subtests) and is not cross-language comparable — prefer D.2 for Go.

Profiles remain validated:false: calibrated to fire on real idioms, but not yet run through a full gen-vs-baseline benchmark. Treat numbers as a guide; read the tests for the judgement axes.