Skip to content

ESTIMER v1.0.0

Latest
Latest

Choose a tag to compare

View all tags
@thelabcorner thelabcorner released this 11 Aug 17:03
· 11 commits to main since this release
v1.0.0
73af983

v1.0.0 — 2026-08-11

SemVer: major (first public GitHub release).

First GitHub release of ESTIMER, the quirk-solving microsecond timing facade for Adobe ExtendScript ES3 — a monotonic, wrap-corrected delta clock over $.hiresTimer + Date, with the house rejection protocol, nesting-safe stopwatches, calibration, and precise sleep.

Added

  • ESTIMER facade (functions only, ES3): prime(), now(), nowMs(), wallNow(), epoch(), measureUs(), measureMs(), samples(), median(), best(), stopwatch(), calibrate(), sleep(), stats(), medianOf(), minOf(), setSource(), setWrapPolicy(), wrapPolicy(), lane(), describe(), constants().
  • Single monotonic accumulator core (Timer) that converts $.hiresTimer's delta reads into an absolute wrap-corrected µs clock — every read advances it, so now() is safe inside measured functions and nested stopwatches (the ArcFit profiler-nesting fix, live-verified: "stopwatch outer >= inner").
  • Signed-32-bit wrap handling: 'correct' policy adds 2^32 to a single-wrap negative read (exact for true intervals in [2^31, 2^32) µs ≈ 35.8-71.6 min); 'reject' never advances on a wrapped read; multi-wrap gaps documented as unrecoverable with wallNow() as the honest long-span lane.
  • First-read discard: prime() consumes the engine-startup read (measured 675,881,881 µs on Illustrator 30.6.0) idempotently so it can never be a sample.
  • Rejection protocol: samples validated > 0 and <= maxValidUs (default 1e8 µs = 100 s), warmup runs (default 2) discarded, rejected samples carried as {index, value, reason} via collectRejected — no silent retry.
  • Calibrated sleep(): busy-wait below the 25 ms coarse threshold lands within +0.2 %; at/above it, $.sleep(ms - 1) + a final calibrated busy-wait so it never wakes before the target (measured overshoot 10-38 %, returned as the actual µs slept).
  • Three build artifacts: dist/ESTIMER.jsx (bannerless IIFE, defines var ESTIMER), dist/vendor-estimer.js (drop-in, assigns $.global.ESTIMER), dist/estimer-core.esm.mjs (Node ESM core).
  • Lane auto-detection: engine ($.hiresTimer, delta, wraps) / node (performance.now(), absolute) / date (ms wall clock, degraded); setSource() injects a fake source for deterministic wrap/negative-delta tests without 35-minute waits.

Verification

Gate: 5,165 Node assertions / 2,000,028 differential fuzz iterations vs the reference model (seed 1337) / 56/56 live-engine checks, Illustrator 30.6.0 / ExtendScript 4.5.6. All green on this commit.

  • npm run typecheck: exit 0, clean.
  • npm test: 5,165 checks, 0 failures (10 clock vectors + 9 stats vectors + 8 median vectors + 2,000 medianOf differential + 3,000 stats differential).
  • npm run fuzz: 2,000,028 differential iterations passed (seed 1337).
  • npm run build: produced dist/ESTIMER.jsx (13,765 B), dist/vendor-estimer.js (14,002 B), dist/estimer-core.esm.mjs (10,462 B).
  • npm run live-verify: 56/56 engine checks passed (ExtendScript 4.5.6, Adobe Illustrator 30.6.0) — incl. monotonicity, wrap-policy round-trip, stopwatch nesting, sleep lanes, calibrate.
  • npm run benchmark: exit 0 (Node reference lane; engine-lane tables in docs/benchmark-rounds-1.md).
  • G1-G8 engine probes: documented in docs/research-probes.md + raw JSON in evidence/raw/ (first-read magnitude, thread-locality, sleep inclusion, wrap spans, monotonicity, read overhead, ScriptUI continuity, rejection ceiling).

Release Assets

  • vendor-estimer.js
  • ESTIMER.jsx
  • estimer-core.esm.mjs
Assets 5
Loading