@ultimat3/time 17.0.0 → 19.0.0

package/CLAUDE.md

@@ -36,7 +36,11 @@
   agreeing.** ICU 78 (Bun 1.4) RESOLVES `CET`, `EST`, `EST5EDT`, `GMT`, `MST` and their families
   where ICU 75 threw, so a runtime upgrade alone reopened the golden rule above — silently, and in
   the direction that fails dangerous: an abbreviation names no DST rule. So the judgement is never
-  delegated to `Intl`. `canonicalTimeZone` asserts the structural property itself: a zone is
+  delegated to `Intl`. Re-measured 2026-08-27 when the Bun 1.3 reversal was trialled:
+  `new Intl.DateTimeFormat('en', { timeZone: 'CET' })` still THROWS on 1.3.14 and resolves on 1.4.0,
+  while both list the same 445 zones — so the two ICUs disagree in the direction this rule already
+  refuses, and a structural judgement is the only thing that answers the same on either runtime.
+  That is the whole argument, and it is why the trial changed nothing here. `canonicalTimeZone` asserts the structural property itself: a zone is
   `Area/Location`, and `UTC` is the one legal exception. Never a denylist of the names ICU newly
   accepts — that list grows with every tzdata and ICU release, and no rule in it keeps `CET` out
   while letting `Japan` in, both being one label. The single-label `backward` links go with them
@@ -156,6 +160,15 @@
   `parseDuration` accepts a leading `-`, and `toSeconds(-3000)` is a tested `-3` — and legitimately
   fractional. A caller that needs whole non-negative milliseconds narrows on top; narrowing here
   breaks the signed-duration contract `toSeconds` is built on.
+- **`toMs(duration, subject?, option?)` names the caller's knob in the refusal, `As of 2026-08-26`.**
+  `pass a finite duration to toMs` names a function an app author reached THROUGH rather than
+  wrote: `@ultimat3/testing`'s `clock.advance('3s')` hands its argument straight down, and so does
+  any wrapper an app builds. The two names are the same two, in the same order, that
+  `@ultimat3/jobs`' `finiteDurationMs` takes — one shape, not two. They are **optional** here and
+  required there because this one is published API: making them required is `TS2554` at every
+  existing call site in every app, a major for a better sentence. The defaults reproduce today's
+  message byte for byte, so the only observable change is at a call site that supplies them.
+  `toSeconds` threads them too and defaults its subject to **`toSeconds`**, not to the delegate.
 - Tests must cover a spring-forward gap, a fall-back overlap and a non-hour offset zone.
 
 ## Commands

package/package.json

@@ -1,6 +1,6 @@
 {
   "name": "@ultimat3/time",
-  "version": "17.0.0",
+  "version": "19.0.0",
   "description": "UTC instants, DST-correct zone math, cron, durations and Intl formatting with an explicit timezone",
   "license": "MIT",
   "type": "module",
@@ -27,13 +27,13 @@
     "LICENSE"
   ],
   "engines": {
-    "bun": ">=1.3.0"
+    "bun": ">=1.4.0"
   },
   "scripts": {
     "typecheck": "tsc --noEmit -p tsconfig.json",
     "test": "bun test"
   },
   "dependencies": {
-    "@ultimat3/core": "17.0.0"
+    "@ultimat3/core": "19.0.0"
   }
 }

package/src/duration.ts

@@ -75,10 +75,22 @@ export function parseDuration(input: string): number {
  * accepts a leading `-` and `toSeconds(-3000)` is a tested `-3` — and legitimately fractional.
  * A caller that needs whole non-negative milliseconds narrows on top of this, which is what
  * `@ultimat3/notify`'s `toDurationMs` does; narrowing here would break both.
+ *
+ * `subject` and `option` name the knob the APP AUTHOR wrote, for the refusal only. They default to
+ * this function's own name, which is the right answer for the caller who typed `toMs(…)` and the
+ * wrong one for every caller reached THROUGH it: `clock.advance('3s')` in `@ultimat3/testing`
+ * passes its argument straight down, and `pass a finite duration to toMs` sends that author
+ * looking for a knob their code does not contain. Same two names, same order, as
+ * `@ultimat3/jobs`' `finiteDurationMs` — one shape, not two.
+ *
+ * OPTIONAL rather than required, unlike that one, because this function is published API: a
+ * required parameter is `TS2554` at every existing call site in every app, which is a major for a
+ * better sentence. The default reproduces today's message byte for byte, so supplying them is the
+ * only observable change.
  */
-export function toMs(duration: string | number): number {
+export function toMs(duration: string | number, subject = 'toMs', option = 'duration'): number {
   return typeof duration === 'number'
-    ? finiteOption('toMs', 'duration', duration)
+    ? finiteOption(subject, option, duration)
     : parseDuration(duration);
 }
 
@@ -87,9 +99,17 @@ export function toMs(duration: string | number): number {
  * and `'-1500ms'` was -1, so a signed duration and its mirror did not answer mirrored seconds.
  * The sign is carried out and the MAGNITUDE rounded — `packages/money/src/rounding.ts` is the
  * framework's one statement of this, and `signed()` there is why zero never comes back as `-0`.
+ *
+ * The screen is `toMs`'s, threaded — but the default subject is `toSeconds`, because that is the
+ * name a direct caller wrote. Naming the delegate would point them at a function they never typed,
+ * which is the whole defect this pair of parameters exists to fix.
  */
-export function toSeconds(duration: string | number): number {
-  const ms = toMs(duration);
+export function toSeconds(
+  duration: string | number,
+  subject = 'toSeconds',
+  option = 'duration',
+): number {
+  const ms = toMs(duration, subject, option);
   const seconds = Math.round(Math.abs(ms) / SECOND);
   if (seconds === 0) return 0;
   return ms < 0 ? -seconds : seconds;