fix(parser): stop DOCX parser dropping bare-integer text runs (#120)

src/parser/docx/document.test.ts

@@ -54,6 +54,19 @@ describe('parseDocument — text extraction', () => {
     const result = parseDocument(xml, emptyNumberingMap());
     expect(result[0]?.text).toBe('<Insert text here>');
   });
+
+  // Regression (#120): Word splits a number like "09 91 26" across runs at
+  // edit/rsid boundaries. A run whose text is a bare integer ("9") was coerced
+  // to a JS number by fast-xml-parser and dropped — corrupting "09 91 26" → "09 1 26".
+  it('preserves a bare-integer run split across a number (#120: numeric run drop)', () => {
+    const xml = makeDocXml(
+      `<w:p><w:r><w:t xml:space="preserve">09 </w:t></w:r>` +
+        `<w:r><w:t>9</w:t></w:r>` +
+        `<w:r><w:t xml:space="preserve">1 26</w:t></w:r></w:p>`
+    );
+    const result = parseDocument(xml, emptyNumberingMap());
+    expect(result[0]?.text).toBe('09 91 26');
+  });
 });
 
 describe('parseDocument — pPr field extraction', () => {

src/parser/docx/document.ts

@@ -9,12 +9,17 @@ import type { DocxParagraph, NumberingMap } from './types.js';
 // and angle brackets; setting false would corrupt those characters in paragraph text.
 // trimValues: false preserves trailing/leading spaces in w:t text nodes — trimming would
 // corrupt concatenated paragraph text across adjacent runs.
+// parseTagValue: false keeps w:t text as strings (#120): fxp's default numeric coercion
+// turns a bare-integer run (<w:t>9</w:t>) into the number 9, which extractRunText cannot
+// read and silently drops — deleting digits from numbers Word split across runs, e.g.
+// "09 91 26" stored as ["09 ", "9", "1 26"] rendered as "09 1 26".
 const xmlParser = new XMLParser({
   ignoreAttributes: false,
   attributeNamePrefix: '@_',
   textNodeName: '#text',
   processEntities: true,
   trimValues: false,
+  parseTagValue: false,
   isArray: (name) => ['w:p', 'w:r', 'w:hyperlink'].includes(name),
 });