@stll/folio-core 0.16.0 → 0.17.0

package/dist/utils/scriptSegments.d.ts

@@ -1,19 +1,34 @@
 //#region src/utils/scriptSegments.d.ts
 /**
- * Script segmentation for per-character East-Asian font selection.
+ * Script segmentation for per-character font selection.
  *
- * Word resolves a run's font per character: East-Asian (CJK) code points use
- * the run's `w:eastAsia` font slot, everything else uses `w:ascii`/`w:hAnsi`.
+ * Word resolves a run's font per character across three slots: East-Asian code
+ * points use `w:eastAsia`, complex-script code points (Arabic, Hebrew, Indic,
+ * South-East Asian) use `w:cs`, and everything else uses `w:ascii`/`w:hAnsi`.
  * folio mirrors this by splitting a run's text into maximal same-script
  * segments that the measurer and the painter both consume, so line wrapping
  * stays in sync with rendering.
  *
+ * A three-value discriminator rather than the `isCjk` boolean this started as:
+ * the question is "which font slot does this character take?", which already
+ * has three answers and may grow again.
+ *
  * Ranges are authored with `\u` escapes only — never pasted glyphs (a pasted
  * glyph once silently corrupted an RTL character class here).
  */
+/** Which of Word's font slots a code point selects. */
+declare const SCRIPT_CLASS: {
+  /** `w:eastAsia` — CJK ideographs, kana, Hangul. */
+  readonly eastAsia: "eastAsia";
+  /** `w:cs` — Arabic, Hebrew, Indic, South-East Asian. */
+  readonly complex: "complex";
+  /** `w:ascii` / `w:hAnsi` — everything else. */
+  readonly western: "western";
+};
+type ScriptClass = (typeof SCRIPT_CLASS)[keyof typeof SCRIPT_CLASS];
 type ScriptSegment = {
   text: string;
-  isCjk: boolean;
+  script: ScriptClass;
 };
 /**
  * True when a code point belongs to an East-Asian script that Word renders with
@@ -26,6 +41,22 @@ declare function isCjkCodePoint(cp: number): boolean;
  * this to skip segmentation entirely on the common all-Latin path.
  */
 declare function hasCjk(text: string): boolean;
+/**
+ * True when a code point belongs to a script Word renders with the `w:cs`
+ * (complex script) font slot: the bidirectional scripts plus the Indic and
+ * South-East Asian scripts that need contextual shaping.
+ *
+ * Checked after CJK, which owns the ranges above U+2E80 that would otherwise
+ * overlap the presentation forms below.
+ */
+declare function isComplexScriptCodePoint(cp: number): boolean;
+/** Which font slot a single code point selects. CJK wins where ranges meet. */
+declare function scriptClassOf(cp: number): ScriptClass;
+/**
+ * True when the text contains at least one complex-script code point. Callers
+ * use this to skip segmentation entirely on the common all-Latin path.
+ */
+declare function hasComplexScript(text: string): boolean;
 /**
  * Split text into maximal runs of one script class (CJK vs. non-CJK). Iterates
  * by code point so astral ideographs (surrogate pairs) are never split between
@@ -33,4 +64,4 @@ declare function hasCjk(text: string): boolean;
  */
 declare function segmentByScript(text: string): ScriptSegment[];
 //#endregion
-export { ScriptSegment, hasCjk, isCjkCodePoint, segmentByScript };
+export { SCRIPT_CLASS, ScriptClass, ScriptSegment, hasCjk, hasComplexScript, isCjkCodePoint, isComplexScriptCodePoint, scriptClassOf, segmentByScript };

package/dist/utils/scriptSegments.js

@@ -1,5 +1,31 @@
 //#region src/utils/scriptSegments.ts
 /**
+* Script segmentation for per-character font selection.
+*
+* Word resolves a run's font per character across three slots: East-Asian code
+* points use `w:eastAsia`, complex-script code points (Arabic, Hebrew, Indic,
+* South-East Asian) use `w:cs`, and everything else uses `w:ascii`/`w:hAnsi`.
+* folio mirrors this by splitting a run's text into maximal same-script
+* segments that the measurer and the painter both consume, so line wrapping
+* stays in sync with rendering.
+*
+* A three-value discriminator rather than the `isCjk` boolean this started as:
+* the question is "which font slot does this character take?", which already
+* has three answers and may grow again.
+*
+* Ranges are authored with `\u` escapes only — never pasted glyphs (a pasted
+* glyph once silently corrupted an RTL character class here).
+*/
+/** Which of Word's font slots a code point selects. */
+const SCRIPT_CLASS = {
+	/** `w:eastAsia` — CJK ideographs, kana, Hangul. */
+	eastAsia: "eastAsia",
+	/** `w:cs` — Arabic, Hebrew, Indic, South-East Asian. */
+	complex: "complex",
+	/** `w:ascii` / `w:hAnsi` — everything else. */
+	western: "western"
+};
+/**
 * True when a code point belongs to an East-Asian script that Word renders with
 * the `w:eastAsia` font slot: CJK ideographs, kana, Hangul, CJK symbols and
 * punctuation, and fullwidth/halfwidth forms.
@@ -16,6 +42,31 @@ function hasCjk(text) {
 	return false;
 }
 /**
+* True when a code point belongs to a script Word renders with the `w:cs`
+* (complex script) font slot: the bidirectional scripts plus the Indic and
+* South-East Asian scripts that need contextual shaping.
+*
+* Checked after CJK, which owns the ranges above U+2E80 that would otherwise
+* overlap the presentation forms below.
+*/
+function isComplexScriptCodePoint(cp) {
+	return cp >= 1424 && cp <= 1535 || cp >= 1536 && cp <= 1791 || cp >= 1792 && cp <= 1871 || cp >= 1872 && cp <= 1919 || cp >= 1920 && cp <= 1983 || cp >= 1984 && cp <= 2047 || cp >= 2048 && cp <= 2111 || cp >= 2112 && cp <= 2143 || cp >= 2144 && cp <= 2159 || cp >= 2160 && cp <= 2207 || cp >= 2208 && cp <= 2303 || cp >= 2304 && cp <= 3455 || cp >= 3456 && cp <= 3583 || cp >= 3584 && cp <= 3711 || cp >= 3712 && cp <= 3839 || cp >= 3840 && cp <= 4095 || cp >= 4096 && cp <= 4255 || cp >= 6016 && cp <= 6143 || cp >= 64285 && cp <= 64335 || cp >= 64336 && cp <= 65023 || cp >= 65136 && cp <= 65279 || cp >= 68864 && cp <= 68927 || cp >= 69312 && cp <= 69375 || cp >= 69424 && cp <= 69487 || cp >= 69488 && cp <= 69551 || cp >= 125184 && cp <= 125279 || cp >= 126464 && cp <= 126719;
+}
+/** Which font slot a single code point selects. CJK wins where ranges meet. */
+function scriptClassOf(cp) {
+	if (isCjkCodePoint(cp)) return SCRIPT_CLASS.eastAsia;
+	if (isComplexScriptCodePoint(cp)) return SCRIPT_CLASS.complex;
+	return SCRIPT_CLASS.western;
+}
+/**
+* True when the text contains at least one complex-script code point. Callers
+* use this to skip segmentation entirely on the common all-Latin path.
+*/
+function hasComplexScript(text) {
+	for (const ch of text) if (isComplexScriptCodePoint(ch.codePointAt(0))) return true;
+	return false;
+}
+/**
 * Split text into maximal runs of one script class (CJK vs. non-CJK). Iterates
 * by code point so astral ideographs (surrogate pairs) are never split between
 * fonts. Empty input yields no segments; single-class input yields one.
@@ -23,30 +74,30 @@ function hasCjk(text) {
 function segmentByScript(text) {
 	const segments = [];
 	let current = "";
-	let currentIsCjk = false;
+	let currentScript = SCRIPT_CLASS.western;
 	for (const ch of text) {
-		const cjk = isCjkCodePoint(ch.codePointAt(0));
+		const script = scriptClassOf(ch.codePointAt(0));
 		if (current.length === 0) {
 			current = ch;
-			currentIsCjk = cjk;
+			currentScript = script;
 			continue;
 		}
-		if (cjk === currentIsCjk) {
+		if (script === currentScript) {
 			current += ch;
 			continue;
 		}
 		segments.push({
 			text: current,
-			isCjk: currentIsCjk
+			script: currentScript
 		});
 		current = ch;
-		currentIsCjk = cjk;
+		currentScript = script;
 	}
 	if (current.length > 0) segments.push({
 		text: current,
-		isCjk: currentIsCjk
+		script: currentScript
 	});
 	return segments;
 }
 //#endregion
-export { hasCjk, isCjkCodePoint, segmentByScript };
+export { SCRIPT_CLASS, hasCjk, hasComplexScript, isCjkCodePoint, isComplexScriptCodePoint, scriptClassOf, segmentByScript };

package/package.json

@@ -1,6 +1,6 @@
 {
   "name": "@stll/folio-core",
-  "version": "0.16.0",
+  "version": "0.17.0",
   "description": "Headless, framework-neutral core of folio: the OOXML (.docx) parser, document model, ProseMirror integration, and page-layout engine. No React.",
   "keywords": [
     "document-model",