Fixed ligatures in Sinhala

Replace zero-width joiners with temporary strip markers to prevent ICU from stripping them.

build/generate-unicode-data.ts

@@ -38,7 +38,7 @@ fs.writeFileSync('src/data/unicode_properties.ts',
 export function canCombineGraphemes(former: string, latter: string): boolean {
     // Zero-width joiner
     // Indic_Syllabic_Category=Invisible_Stacker as of Unicode ${indicSyllabicCategory.version}, published ${indicSyllabicCategory.date}.
-    const terminalJoinersRegExp = /[\u200D${indicSyllabicCategory.characterClass}]$/u;
+    const terminalJoinersRegExp = /[\\u200D${indicSyllabicCategory.characterClass}]$/u;
     return terminalJoinersRegExp.test(former) || /^\\p{gc=Mc}/u.test(latter);
 }
 `);

src/data/unicode_properties.ts

@@ -6,6 +6,6 @@
 export function canCombineGraphemes(former: string, latter: string): boolean {
     // Zero-width joiner
     // Indic_Syllabic_Category=Invisible_Stacker as of Unicode 16.0.0, published 2024-04-30.
-    const terminalJoinersRegExp = /[‍\u1039\u17D2\u1A60\u1BAB\uAAF6\u{10A3F}\u{11133}\u{113D0}\u{1193E}\u{11A47}\u{11A99}\u{11D45}\u{11D97}\u{11F42}]$/u;
+    const terminalJoinersRegExp = /[\u200D\u1039\u17D2\u1A60\u1BAB\uAAF6\u{10A3F}\u{11133}\u{113D0}\u{1193E}\u{11A47}\u{11A99}\u{11D45}\u{11D97}\u{11F42}]$/u;
     return terminalJoinersRegExp.test(former) || /^\p{gc=Mc}/u.test(latter);
 }

src/symbol/shaping.ts

@@ -272,11 +272,14 @@ function shapeText(
         lines = [];
         // ICU operates on code units.
         lineBreaks = lineBreaks.map(index => logicalInput.toCodeUnitIndex(index));
+        // Replace zero-width joiners with temporary strip markers (from the Private Use Area) to prevent ICU from stripping them out.
+        const markedInput = logicalInput.toString().replace(/\u200D/g, '\uF8FF');
         const untaggedLines =
-            processBidirectionalText(logicalInput.toString(), lineBreaks);
+            processBidirectionalText(markedInput, lineBreaks);
         for (const line of untaggedLines) {
             const taggedLine = new TaggedString();
-            taggedLine.text = line;
+            // Restore zero-width joiners from temporary strip markers.
+            taggedLine.text = line.replace(/\uF8FF/g, '\u200D');
             taggedLine.sections = logicalInput.sections;
             // eslint-disable-next-line @typescript-eslint/no-unused-vars
             for (const char of splitByGraphemeCluster(line)) {
@@ -290,20 +293,23 @@ function shapeText(
         lines = [];
         // ICU operates on code units.
         lineBreaks = lineBreaks.map(index => logicalInput.toCodeUnitIndex(index));
+        // Replace zero-width joiners with temporary strip markers (from the Private Use Area) to prevent ICU from stripping them out.
+        const markedInput = logicalInput.toString().replace(/\u200D/g, '\uF8FF');
 
         // Convert grapheme cluster–based section index to be based on code units.
         let i = 0;
         const sectionIndex = [];
-        for (const grapheme of splitByGraphemeCluster(logicalInput.text)) {
+        for (const grapheme of splitByGraphemeCluster(markedInput)) {
             sectionIndex.push(...Array(grapheme.length).fill(logicalInput.sectionIndex[i]));
             i++;
         }
 
         const processedLines =
-            processStyledBidirectionalText(logicalInput.text, sectionIndex, lineBreaks);
+            processStyledBidirectionalText(markedInput, sectionIndex, lineBreaks);
         for (const line of processedLines) {
             const taggedLine = new TaggedString();
-            taggedLine.text = line[0];
+            // Restore zero-width joiners from temporary strip markers.
+            taggedLine.text = line[0].replace(/\uF8FF/g, '\u200D');
             taggedLine.sectionIndex = line[1];
             taggedLine.sections = logicalInput.sections;
             lines.push(taggedLine);