nl2time 0.1.0 → 0.1.1

package/README.md

@@ -73,11 +73,19 @@ Cultural/ambiguous semantics are explicit `TimeContext` knobs with CLDR-derived
 | Does "last 3 days" include today? | `partialPeriod: 'include' \| 'exclude'` | `include` |
 | "At 4" — am or pm? | — | both candidates, plausibility-ordered |
 
+## Install
+
+[![npm](https://img.shields.io/npm/v/nl2time)](https://www.npmjs.com/package/nl2time)
+
+```
+npm install nl2time
+```
+
 ## Status
 
-v0.1 — English parsing, en-* describe, core engine with a DST/edge-case battery. Recurrence (`every Tuesday`) is representable in the IR but resolves in v2.
+v0.1 — English parsing, en-* describe, core engine with a DST/edge-case battery, holidays (fixed-date, nth-weekday, computed Easter), business-day spans. Recurrence (`every Tuesday`) is representable in the IR but resolves in v2.
 
-**Corpus** ([corpus/](corpus/), runner exported as `nl2time/corpus`): a bidirectional golden set with per-case license provenance — 41 hand-authored NL→time cases (all passing), a 65-case datetime→NL golden set (all passing; to our knowledge the first published eval set for humanized time generation), and 693 cases imported from Microsoft Recognizers-Text (MIT) of which 12.6% currently pass — the honest coverage baseline the parser climbs against (`npm run eval`, `npm run baselines`). See [corpus/ATTRIBUTIONS.md](corpus/ATTRIBUTIONS.md) and [docs/porting.md](docs/porting.md).
+**Corpus** ([corpus/](corpus/), runner exported as `nl2time/corpus`): a bidirectional golden set with per-case license provenance — 41 hand-authored NL→time cases (100% passing), a 65-case datetime→NL golden set (100% passing; to our knowledge the first published eval set for humanized time generation), and 693 cases imported from Microsoft Recognizers-Text (MIT) of which **80.7% currently pass** — the coverage baseline the parser climbs against (`npm run eval`, `npm run baselines`; every previously-passing case is a CI regression gate). Most remaining failures are documented divergences from upstream idiosyncrasies rather than parser gaps. See [corpus/ATTRIBUTIONS.md](corpus/ATTRIBUTIONS.md) and [docs/porting.md](docs/porting.md).
 
 ## Development
 

package/dist/parse/en.js

@@ -125,12 +125,17 @@ function yearAt(tokens, i) {
         ? t.value
         : undefined;
 }
-/** "in the morning" / "at night" / "tonight" lookahead. Returns period + consumed. */
-function readPeriodSuffix(tokens, i) {
+/** "in the morning" / "at night" / "this morning" / "tonight" lookahead.
+ * Returns period + consumed. "this <period>" binds a preceding bare hour
+ * exactly like "in the <period>" ("8 this morning" → 8am): the period
+ * supplies the meridiem, and the reference day is the anchor either way.
+ * Deliberately NOT "last night": that also shifts the day, which a
+ * meridiem-only suffix can't express (#9). */
+function readPeriodSuffix(tokens, i, allowThis = true) {
     let at = i;
     if (word(tokens[at]) === 'in' || word(tokens[at]) === 'at')
         at += 1;
-    if (word(tokens[at]) === 'the')
+    if (word(tokens[at]) === 'the' || (allowThis && word(tokens[at]) === 'this'))
         at += 1;
     const w = word(tokens[at]);
     if (w === 'tonight')
@@ -143,6 +148,12 @@ function readPeriodSuffix(tokens, i) {
 function meridiemFor(period) {
     return period === 'morning' ? 'am' : 'pm';
 }
+/** Words that license a following bare number as a clock time
+ * ("at 8 this morning", "by 5 this afternoon"). */
+const TIME_PREPOSITIONS = new Set([
+    'at', 'around', 'about', 'by', 'from', 'until', 'till', 'to',
+    'before', 'after', 'since',
+]);
 // --- rules -----------------------------------------------------------------
 /** today / yesterday / tomorrow / tonight / day before yesterday / day after tomorrow */
 const ruleDeicticDay = (tokens, i) => {
@@ -2353,7 +2364,17 @@ const ruleClockTime = (tokens, i, ctx) => {
         if (after === "o'clock" || after === 'oclock') {
             return withPeriodSuffix({ hour: hn.value, meridiem: 'unknown' }, hn.consumed + 1);
         }
-        const ps = readPeriodSuffix(tokens, i + hn.consumed + mc);
+        // A BARE number before "this <period>" is only a clock time when the
+        // number isn't already bound by a preceding noun — "8 this morning"
+        // is 8am, but in "building 4 this afternoon" the 4 is a building
+        // number and the period keeps its range reading (Recognizers
+        // rt-dtm-0523). Utterance start or a time-preposition before the
+        // number licenses the binding; an arbitrary preceding word does not.
+        // The in/at forms ("4 in the afternoon") are explicit enough to bind
+        // regardless, as before.
+        const prevWord = word(tokens[i - 1]);
+        const allowThis = i === 0 || prevWord === undefined || TIME_PREPOSITIONS.has(prevWord);
+        const ps = readPeriodSuffix(tokens, i + hn.consumed + mc, allowThis);
         if (ps) {
             const time = { hour: hn.value, meridiem: meridiemFor(ps.period) };
             if (minute !== undefined)

package/package.json

@@ -1,6 +1,6 @@
 {
   "name": "nl2time",
-  "version": "0.1.0",
+  "version": "0.1.1",
   "description": "Bidirectional natural language ⇄ date/time: parse expressions like 'last week' into concrete intervals, and describe instants as '9pm last night' — deterministic core, optional LLM assist.",
   "license": "MIT",
   "type": "module",
@@ -19,6 +19,10 @@
     "type": "git",
     "url": "git+https://github.com/AndyFooBlah/nl2time.git"
   },
+  "homepage": "https://github.com/AndyFooBlah/nl2time#readme",
+  "bugs": {
+    "url": "https://github.com/AndyFooBlah/nl2time/issues"
+  },
   "scripts": {
     "build": "tsc -p tsconfig.build.json",
     "prepare": "npm run build",