@stll/folio-core 0.16.0 → 0.17.1

package/dist/controller/fontReadiness.js

@@ -120,6 +120,8 @@ function addLayoutFontFamilyFace(faces, value, descriptor) {
 	const fontFamily = value;
 	addLayoutFontFamilyFace(faces, fontFamily.ascii, descriptor);
 	addLayoutFontFamilyFace(faces, fontFamily.hAnsi, descriptor);
+	addLayoutFontFamilyFace(faces, fontFamily.cs, descriptor);
+	addLayoutFontFamilyFace(faces, fontFamily.eastAsia, descriptor);
 }
 function addLayoutFontFamilyNameFace(faces, family, descriptor) {
 	const normalized = family.trim();

package/dist/docx/server/extractDocxText.d.ts

@@ -11,11 +11,30 @@ type DocxParagraphSource = "header" | "body" | "footer";
  * - `delimiter` — the `| --- |` line GFM requires under the header.
  */
 type DocxTableRowKind = "cells" | "syntheticHeader" | "delimiter";
+/** Source paragraph inside a structured table cell. */
+type ExtractedDocxTableCellParagraph = {
+  text: string;
+  style?: string;
+  bold?: boolean;
+  fontSize?: number;
+  alignment?: "left" | "center" | "right" | "both";
+};
+/** Source paragraphs contained by one physical table cell. */
+type ExtractedDocxTableCell = {
+  /** Source `w:p` records in document order; empty padding cells have no entries. */
+  paragraphs: readonly ExtractedDocxTableCellParagraph[];
+};
 /** Table membership of a paragraph whose `text` is a markdown table row. */
 type DocxTableRowPosition = {
   /** 0-based index of the source `w:tbl`, in extraction order across all parts. */
   table: number;
-  kind: DocxTableRowKind;
+  kind: "cells";
+  /** Source cells aligned to the rendered GFM columns. */
+  cells: readonly ExtractedDocxTableCell[];
+} | {
+  /** 0-based index of the source `w:tbl`, in extraction order across all parts. */
+  table: number;
+  kind: "syntheticHeader" | "delimiter";
 };
 /** Paragraph text and lightweight formatting metadata from a DOCX archive. */
 type ExtractedDocxParagraph = {
@@ -43,4 +62,4 @@ type ExtractedDocxText = {
 /** Extract paragraph text and formatting metadata from a DOCX archive. */
 declare const extractDocxText: (bytes: ArrayBuffer | Uint8Array) => Promise<ExtractedDocxText>;
 //#endregion
-export { DocxParagraphSource, DocxTableRowKind, DocxTableRowPosition, ExtractedDocxParagraph, ExtractedDocxText, extractDocxText };
+export { DocxParagraphSource, DocxTableRowKind, DocxTableRowPosition, ExtractedDocxParagraph, ExtractedDocxTableCell, ExtractedDocxTableCellParagraph, ExtractedDocxText, extractDocxText };

package/dist/docx/server/extractDocxText.js

@@ -1,14 +1,21 @@
 import { escapeTableCell } from "../../markdown/escape.js";
 import { RELATIONSHIP_TYPES, parseRelationships } from "../relsParser.js";
-import { findAllDeep, findChild, findDeep, getAttribute, getAttributeAnyPrefix, getLocalName, getTextContent, parseXml } from "../xmlParser.js";
+import { findAllDeep, findDeep, getAttribute, getAttributeByNamespaceUri, getLocalName, getNamespaceUri, getTextContent, parseXml } from "../xmlParser.js";
 import { loadDocxArchive } from "./boundedArchive.js";
 //#region src/docx/server/extractDocxText.ts
 const DOCUMENT_RELS_PATH = "word/_rels/document.xml.rels";
+const WORDPROCESSINGML_NAMESPACES = /* @__PURE__ */ new Set(["http://schemas.openxmlformats.org/wordprocessingml/2006/main", "http://purl.oclc.org/ooxml/wordprocessingml/main"]);
 const childElements = (element) => element.elements?.filter((child) => child.type === "element") ?? [];
+const wordElementName = (element) => WORDPROCESSINGML_NAMESPACES.has(getNamespaceUri(element) ?? "") ? getLocalName(element.name) : null;
+const findWordChild = (parent, localName) => {
+	if (!parent) return null;
+	return childElements(parent).find((child) => wordElementName(child) === localName) ?? null;
+};
+const getWordAttribute = (element, localName) => getAttributeByNamespaceUri(element, WORDPROCESSINGML_NAMESPACES, localName);
 const collectText = (element) => {
 	let text = "";
 	const walk = (node) => {
-		const localName = getLocalName(node.name ?? "");
+		const localName = wordElementName(node);
 		if (localName === "t") {
 			text += getTextContent(node);
 			return;
@@ -27,29 +34,45 @@ const collectText = (element) => {
 	walk(element);
 	return text;
 };
+const countAcceptedTextChars = (element) => {
+	let chars = 0;
+	const walk = (node) => {
+		const localName = wordElementName(node);
+		if (localName === "t") {
+			chars += getTextContent(node).length;
+			return;
+		}
+		if (localName === "del" || localName === "delText" || localName === "moveFrom") return;
+		for (const child of childElements(node)) walk(child);
+	};
+	walk(element);
+	return chars;
+};
 const readParagraphProperties = (paragraph) => {
-	const properties = findChild(paragraph, "w", "pPr");
+	const properties = findWordChild(paragraph, "pPr");
 	if (!properties) return {};
 	const result = {};
-	const styleValue = getAttributeAnyPrefix(findChild(properties, "w", "pStyle"), "val");
+	const style = findWordChild(properties, "pStyle");
+	const styleValue = getWordAttribute(style, "val");
 	if (styleValue !== null) result.style = styleValue;
-	const alignment = getAttributeAnyPrefix(findChild(properties, "w", "jc"), "val");
+	const justification = findWordChild(properties, "jc");
+	const alignment = getWordAttribute(justification, "val");
 	if (alignment === "left" || alignment === "center" || alignment === "right" || alignment === "both") result.alignment = alignment;
 	return result;
 };
 const readRunMetrics = (paragraph) => {
 	const metrics = [];
 	for (const run of childElements(paragraph)) {
-		if (getLocalName(run.name ?? "") !== "r") continue;
-		const properties = findChild(run, "w", "rPr");
-		const boldProperty = findChild(properties, "w", "b");
-		const boldValue = getAttributeAnyPrefix(boldProperty, "val");
+		if (wordElementName(run) !== "r") continue;
+		const properties = findWordChild(run, "rPr");
+		const boldProperty = findWordChild(properties, "b");
+		const boldValue = getWordAttribute(boldProperty, "val");
 		const bold = boldProperty !== null && boldValue !== "0" && boldValue !== "false";
-		const sizeValue = getAttributeAnyPrefix(findChild(properties, "w", "sz"), "val");
+		const sizeProperty = findWordChild(properties, "sz");
+		const sizeValue = getWordAttribute(sizeProperty, "val");
 		const parsedSize = sizeValue === null ? NaN : Number.parseInt(sizeValue, 10);
 		const fontSize = Number.isFinite(parsedSize) && parsedSize > 0 ? parsedSize : void 0;
-		let chars = 0;
-		for (const textNode of findAllDeep(run, "w", "t")) chars += getTextContent(textNode).length;
+		const chars = countAcceptedTextChars(run);
 		if (chars === 0) continue;
 		const entry = {
 			bold,
@@ -60,6 +83,19 @@ const readRunMetrics = (paragraph) => {
 	}
 	return metrics;
 };
+const readParagraph = (paragraph) => {
+	const entry = { text: collectText(paragraph) };
+	const { style, alignment } = readParagraphProperties(paragraph);
+	if (style !== void 0) entry.style = style;
+	if (alignment !== void 0) entry.alignment = alignment;
+	const runs = readRunMetrics(paragraph);
+	if (runs.length === 0) return entry;
+	const totalChars = runs.reduce((sum, run) => sum + run.chars, 0);
+	if (runs.reduce((sum, run) => sum + (run.bold ? run.chars : 0), 0) > totalChars / 2) entry.bold = true;
+	const firstFontSize = runs.find((run) => run.fontSize !== void 0)?.fontSize;
+	if (firstFontSize !== void 0) entry.fontSize = firstFontSize;
+	return entry;
+};
 const TABLE_DELIMITER_CELL = "---";
 /** GFM cannot nest tables; an inner table joins its cells inside the outer cell. */
 const NESTED_TABLE_CELL_SEPARATOR = " / ";
@@ -77,7 +113,7 @@ const collectTableParts = (parent, localName) => {
 	const parts = [];
 	const walk = (node) => {
 		for (const child of childElements(node)) {
-			const childName = getLocalName(child.name ?? "");
+			const childName = wordElementName(child);
 			if (childName === localName) {
 				parts.push(child);
 				continue;
@@ -94,47 +130,75 @@ const collectTableParts = (parent, localName) => {
 * Blank paragraphs are dropped so a cell padded with empty paragraphs does not
 * render as a run of `<br>`. A nested table contributes one line per inner row.
 */
-const readCellText = (cell, depth) => {
+const readCellSourceParagraphs = (cell, depth) => {
+	const paragraphs = [];
+	const walk = (node) => {
+		for (const child of childElements(node)) {
+			const childName = wordElementName(child);
+			if (childName === "p") {
+				const paragraph = readParagraph(child);
+				if (paragraph.text.length > 0) paragraphs.push(paragraph);
+				continue;
+			}
+			if (childName === "tbl") {
+				if (depth >= MAX_NESTED_TABLE_DEPTH) continue;
+				for (const row of collectTableParts(child, "tr")) for (const nestedCell of collectTableParts(row, "tc")) for (const paragraph of readCellSourceParagraphs(nestedCell, depth + 1)) paragraphs.push(paragraph);
+				continue;
+			}
+			walk(child);
+		}
+	};
+	walk(cell);
+	return paragraphs;
+};
+const readCellRenderedLines = (cell, depth) => {
 	const lines = [];
 	const walk = (node) => {
 		for (const child of childElements(node)) {
-			const childName = getLocalName(child.name ?? "");
+			const childName = wordElementName(child);
 			if (childName === "p") {
 				const text = collectText(child);
 				if (text.length > 0) lines.push(text);
 				continue;
 			}
 			if (childName === "tbl") {
-				if (depth >= MAX_NESTED_TABLE_DEPTH) continue;
-				for (const line of flattenNestedTable(child, depth + 1)) lines.push(line);
+				if (depth < MAX_NESTED_TABLE_DEPTH) for (const line of flattenNestedTable(child, depth + 1)) lines.push(line);
 				continue;
 			}
 			walk(child);
 		}
 	};
 	walk(cell);
-	return lines.join("\n");
+	return lines;
 };
 const flattenNestedTable = (table, depth) => {
 	const lines = [];
 	for (const row of collectTableParts(table, "tr")) {
-		const cells = collectTableParts(row, "tc").map((cell) => readCellText(cell, depth));
+		const cells = collectTableParts(row, "tc").map((cell) => readCellRenderedLines(cell, depth).join("\n"));
 		if (cells.some((text) => text.length > 0)) lines.push(cells.join(NESTED_TABLE_CELL_SEPARATOR));
 	}
 	return lines;
 };
+const emptyTableCell = () => ({
+	text: "",
+	paragraphs: [],
+	gridSpan: 1
+});
 const readTableCell = (cell, depth) => {
-	const properties = findChild(cell, "w", "tcPr");
-	const gridSpanValue = getAttributeAnyPrefix(findChild(properties, "w", "gridSpan"), "val");
+	const properties = findWordChild(cell, "tcPr");
+	const gridSpanValue = getWordAttribute(findWordChild(properties, "gridSpan"), "val");
 	const parsedGridSpan = gridSpanValue === null ? 1 : Number.parseInt(gridSpanValue, 10);
 	const gridSpan = Number.isFinite(parsedGridSpan) && parsedGridSpan > 1 ? parsedGridSpan : 1;
-	const vMerge = findChild(properties, "w", "vMerge");
-	if (vMerge !== null && getAttributeAnyPrefix(vMerge, "val") !== "restart") return {
+	const vMerge = findWordChild(properties, "vMerge");
+	if (vMerge !== null && getWordAttribute(vMerge, "val") !== "restart") return {
 		text: "",
+		paragraphs: [],
 		gridSpan
 	};
+	const paragraphs = readCellSourceParagraphs(cell, depth);
 	return {
-		text: readCellText(cell, depth),
+		text: readCellRenderedLines(cell, depth).join("\n"),
+		paragraphs,
 		gridSpan
 	};
 };
@@ -150,11 +214,17 @@ const readTableCell = (cell, depth) => {
 * names invented here would be read back as facts about the document.
 */
 const declaresHeaderRow = (row) => {
-	const header = findChild(findChild(row, "w", "trPr"), "w", "tblHeader");
+	const header = findWordChild(findWordChild(row, "trPr"), "tblHeader");
 	if (header === null) return false;
-	const value = getAttributeAnyPrefix(header, "val");
+	const value = getWordAttribute(header, "val");
 	return value !== "0" && value !== "false";
 };
+const readRowGridOffset = (row, localName) => {
+	const properties = findWordChild(row, "trPr");
+	const value = getWordAttribute(findWordChild(properties, localName), "val");
+	const parsed = value === null ? 0 : Number.parseInt(value, 10);
+	return Number.isFinite(parsed) && parsed > 0 ? Math.min(parsed, MAX_TABLE_COLUMNS) : 0;
+};
 const readTableGrid = (table) => {
 	const rows = [];
 	let columnCount = 0;
@@ -162,13 +232,17 @@ const readTableGrid = (table) => {
 	for (const [rowIndex, row] of collectTableParts(table, "tr").entries()) {
 		if (rowIndex === 0) firstRowIsHeader = declaresHeaderRow(row);
 		const columns = [];
+		const gridBefore = readRowGridOffset(row, "gridBefore");
+		for (let index = 0; index < gridBefore; index += 1) columns.push(emptyTableCell());
 		for (const cell of collectTableParts(row, "tc")) {
 			if (columns.length >= MAX_TABLE_COLUMNS) break;
-			const { text, gridSpan } = readTableCell(cell, 0);
-			columns.push(text);
-			const padding = Math.min(gridSpan - 1, MAX_TABLE_COLUMNS - columns.length);
-			for (let index = 0; index < padding; index++) columns.push("");
+			const extractedCell = readTableCell(cell, 0);
+			columns.push(extractedCell);
+			const padding = Math.min(extractedCell.gridSpan - 1, MAX_TABLE_COLUMNS - columns.length);
+			for (let index = 0; index < padding; index++) columns.push(emptyTableCell());
 		}
+		const gridAfter = Math.min(readRowGridOffset(row, "gridAfter"), MAX_TABLE_COLUMNS - columns.length);
+		for (let index = 0; index < gridAfter; index += 1) columns.push(emptyTableCell());
 		if (columns.length > columnCount) columnCount = columns.length;
 		rows.push(columns);
 	}
@@ -181,7 +255,7 @@ const readTableGrid = (table) => {
 /** Pad a row to the table's column count and escape each cell into a pipe row. */
 const toRowLine = (columns, columnCount) => {
 	const cells = [];
-	for (let column = 0; column < columnCount; column++) cells.push(escapeTableCell(columns[column] ?? ""));
+	for (let column = 0; column < columnCount; column++) cells.push(escapeTableCell(columns[column]?.text ?? ""));
 	return `| ${cells.join(" | ")} |`;
 };
 /** Render a `w:tbl` as GFM rows. A table with no cell at all renders nothing. */
@@ -190,7 +264,7 @@ const renderTableRows = (table, tableIndex) => {
 	const [firstRow, ...remainingRows] = rows;
 	if (columnCount === 0 || firstRow === void 0) return [];
 	const rendered = [];
-	const push = (text, kind) => {
+	const pushScaffolding = (text, kind) => {
 		rendered.push({
 			text,
 			position: {
@@ -199,10 +273,24 @@ const renderTableRows = (table, tableIndex) => {
 			}
 		});
 	};
-	if (firstRowIsHeader) push(toRowLine(firstRow, columnCount), "cells");
-	else push(toRowLine([], columnCount), "syntheticHeader");
-	push(toRowLine(Array.from({ length: columnCount }, () => TABLE_DELIMITER_CELL), columnCount), "delimiter");
-	for (const row of firstRowIsHeader ? remainingRows : rows) push(toRowLine(row, columnCount), "cells");
+	const pushCells = (cells) => {
+		rendered.push({
+			text: toRowLine(cells, columnCount),
+			position: {
+				table: tableIndex,
+				kind: "cells",
+				cells: Array.from({ length: columnCount }, (_, column) => ({ paragraphs: cells.at(column)?.paragraphs ?? [] }))
+			}
+		});
+	};
+	if (firstRowIsHeader) pushCells(firstRow);
+	else pushScaffolding(toRowLine([], columnCount), "syntheticHeader");
+	pushScaffolding(toRowLine(Array.from({ length: columnCount }, () => ({
+		text: TABLE_DELIMITER_CELL,
+		paragraphs: [],
+		gridSpan: 1
+	})), columnCount), "delimiter");
+	for (const row of firstRowIsHeader ? remainingRows : rows) pushCells(row);
 	return rendered;
 };
 const extractContainer = ({ container, source, startIndex, startTableIndex }) => {
@@ -210,22 +298,13 @@ const extractContainer = ({ container, source, startIndex, startTableIndex }) =>
 	let charCount = 0;
 	let tableCount = 0;
 	const pushProse = (paragraph) => {
-		const text = collectText(paragraph);
+		const extracted = readParagraph(paragraph);
+		const { text } = extracted;
 		const entry = {
 			index: startIndex + paragraphs.length,
-			text,
-			source
+			source,
+			...extracted
 		};
-		const { style, alignment } = readParagraphProperties(paragraph);
-		if (style !== void 0) entry.style = style;
-		if (alignment !== void 0) entry.alignment = alignment;
-		const runs = readRunMetrics(paragraph);
-		if (runs.length > 0) {
-			const totalChars = runs.reduce((sum, run) => sum + run.chars, 0);
-			if (runs.reduce((sum, run) => sum + (run.bold ? run.chars : 0), 0) > totalChars / 2) entry.bold = true;
-			const firstFontSize = runs.find((run) => run.fontSize !== void 0)?.fontSize;
-			if (firstFontSize !== void 0) entry.fontSize = firstFontSize;
-		}
 		paragraphs.push(entry);
 		charCount += text.length;
 	};
@@ -246,7 +325,7 @@ const extractContainer = ({ container, source, startIndex, startTableIndex }) =>
 	*/
 	const walkBlocks = (node) => {
 		for (const child of childElements(node)) {
-			const childName = getLocalName(child.name ?? "");
+			const childName = wordElementName(child);
 			if (childName === "tbl") {
 				for (const row of renderTableRows(child, startTableIndex + tableCount)) pushTableRow(row);
 				tableCount += 1;
@@ -279,7 +358,7 @@ const extractParts = async ({ archive, source, rootName, startIndex, startTableI
 			startIndex: nextIndex,
 			startTableIndex: startTableIndex + tableCount
 		});
-		paragraphs.push(...result.paragraphs);
+		for (const paragraph of result.paragraphs) paragraphs.push(paragraph);
 		charCount += result.charCount;
 		tableCount += result.tableCount;
 		nextIndex += result.paragraphs.length;

package/dist/docx/xmlParser.d.ts

@@ -36,8 +36,17 @@ type XmlElement = {
   text?: string | number | boolean;
   type?: string;
   name?: string;
+  /** Resolved namespace URI; non-enumerable so legacy structural comparisons stay stable. */
+  namespaceUri?: string;
+  /** In-scope namespace declarations; non-enumerable so legacy comparisons stay stable. */
+  namespaceScope?: XmlNamespaceScope;
   elements?: XmlElement[];
 };
+/** Parent-linked XML namespace declarations; each scope stores only local overrides. */
+type XmlNamespaceScope = {
+  bindings: ReadonlyMap<string, string>;
+  parent?: XmlNamespaceScope;
+};
 /**
  * Common OOXML namespace URIs — re-exported from @stll/docx-utils.
  */
@@ -85,6 +94,10 @@ declare function getLocalName(name: string | undefined): string;
  * e.g., "w:p" -> "w", "a:graphic" -> "a"
  */
 declare function getNamespacePrefix(name: string): string | null;
+/** Namespace URI resolved from the element's in-scope XML declarations. */
+declare const getNamespaceUri: (element: XmlElement) => string | undefined;
+/** Get an attribute whose prefix resolves to one of the accepted namespace URIs. */
+declare function getAttributeByNamespaceUri(element: XmlElement | null | undefined, namespaceUris: ReadonlySet<string>, localName: string): string | null;
 /**
  * Check if an element matches a given namespaced name
  *
@@ -313,4 +326,4 @@ declare function mergeXmlnsDeclarations(inherited: Record<string, string>, eleme
  */
 declare function cloneWithXmlnsDeclarations(element: XmlElement, xmlnsDecls: Record<string, string>): XmlElement;
 //#endregion
-export { NAMESPACES, XmlElement, cloneWithXmlnsDeclarations, collectXmlnsDeclarations, elementToXml, findAllDeep, findByFullName, findChild, findChildByLocalName, findChildren, findChildrenByLocalName, findDeep, getAttribute, getAttributeAny, getAttributeAnyPrefix, getAttributes, getChildElements, getLocalName, getNamespacePrefix, getTextContent, hasChild, hasFlag, matchesName, mergeXmlnsDeclarations, parseBooleanElement, parseColorElement, parseNumberingLevelAttribute, parseNumericAttribute, parseTableMeasurementValue, parseXml, parseXmlDocument };
+export { NAMESPACES, XmlElement, XmlNamespaceScope, cloneWithXmlnsDeclarations, collectXmlnsDeclarations, elementToXml, findAllDeep, findByFullName, findChild, findChildByLocalName, findChildren, findChildrenByLocalName, findDeep, getAttribute, getAttributeAny, getAttributeAnyPrefix, getAttributeByNamespaceUri, getAttributes, getChildElements, getLocalName, getNamespacePrefix, getNamespaceUri, getTextContent, hasChild, hasFlag, matchesName, mergeXmlnsDeclarations, parseBooleanElement, parseColorElement, parseNumberingLevelAttribute, parseNumericAttribute, parseTableMeasurementValue, parseXml, parseXmlDocument };

package/dist/docx/xmlParser.js

@@ -55,6 +55,15 @@ const TEXT_KEY = "#text";
 const ATTR_KEY = ":@";
 /** Character reference required to keep carriage returns through XML end-of-line normalization. */
 const XML_CARRIAGE_RETURN_REFERENCE = "
";
+const EMPTY_NAMESPACE_SCOPE = { bindings: /* @__PURE__ */ new Map() };
+const resolveNamespaceUri = (scope, prefix) => {
+	let current = scope;
+	while (current) {
+		const value = current.bindings.get(prefix);
+		if (value !== void 0) return value;
+		current = current.parent;
+	}
+};
 /**
 * Convert a fast-xml-parser preserveOrder node into an XmlElement.
 *
@@ -62,7 +71,7 @@ const XML_CARRIAGE_RETURN_REFERENCE = "
";
 * (the tag name or `#text`) whose value is the children array, plus an
 * optional `:@` key holding the attributes object.
 */
-function fxpNodeToElement(node) {
+function fxpNodeToElement(node, inheritedNamespaceScope = EMPTY_NAMESPACE_SCOPE) {
 	if (TEXT_KEY in node) return {
 		type: "text",
 		text: node[TEXT_KEY]
@@ -76,8 +85,34 @@ function fxpNodeToElement(node) {
 			name: key
 		};
 		if (attrs) element.attributes = attrs;
+		let localBindings = null;
+		if (attrs) for (const [attribute, value] of Object.entries(attrs)) {
+			if (attribute !== "xmlns" && !attribute.startsWith("xmlns:")) continue;
+			localBindings ??= /* @__PURE__ */ new Map();
+			const prefix = attribute === "xmlns" ? "" : attribute.slice(6);
+			localBindings.set(prefix, value);
+		}
+		const namespaceScope = localBindings === null ? inheritedNamespaceScope : {
+			bindings: localBindings,
+			parent: inheritedNamespaceScope
+		};
+		Object.defineProperty(element, "namespaceScope", {
+			configurable: false,
+			enumerable: false,
+			value: namespaceScope,
+			writable: false
+		});
+		const colonIndex = key.indexOf(":");
+		const prefix = colonIndex === -1 ? "" : key.slice(0, colonIndex);
+		const namespaceUri = resolveNamespaceUri(namespaceScope, prefix);
+		if (namespaceUri !== void 0) Object.defineProperty(element, "namespaceUri", {
+			configurable: false,
+			enumerable: false,
+			value: namespaceUri,
+			writable: false
+		});
 		if (children.length > 0) {
-			for (let index = 0; index < children.length; index += 1) children[index] = fxpNodeToElement(children[index]);
+			for (let index = 0; index < children.length; index += 1) children[index] = fxpNodeToElement(children[index], namespaceScope);
 			element.elements = children;
 		}
 		return element;
@@ -159,6 +194,18 @@ function getNamespacePrefix(name) {
 	const colonIndex = name.indexOf(":");
 	return colonIndex !== -1 ? name.slice(0, colonIndex) : null;
 }
+/** Namespace URI resolved from the element's in-scope XML declarations. */
+const getNamespaceUri = (element) => element.namespaceUri;
+/** Get an attribute whose prefix resolves to one of the accepted namespace URIs. */
+function getAttributeByNamespaceUri(element, namespaceUris, localName) {
+	if (!element?.attributes) return null;
+	for (const [name, value] of Object.entries(element.attributes)) {
+		if (value === void 0 || getLocalName(name) !== localName) continue;
+		const prefix = getNamespacePrefix(name);
+		if (prefix !== null && namespaceUris.has(resolveNamespaceUri(element.namespaceScope, prefix) ?? "")) return String(value);
+	}
+	return null;
+}
 function hasLocalName(name, localName) {
 	if (!name) return false;
 	if (name === localName) return true;
@@ -640,4 +687,4 @@ function cloneWithXmlnsDeclarations(element, xmlnsDecls) {
 	return element;
 }
 //#endregion
-export { NAMESPACES, cloneWithXmlnsDeclarations, collectXmlnsDeclarations, elementToXml, findAllDeep, findByFullName, findChild, findChildByLocalName, findChildren, findChildrenByLocalName, findDeep, getAttribute, getAttributeAny, getAttributeAnyPrefix, getAttributes, getChildElements, getLocalName, getNamespacePrefix, getTextContent, hasChild, hasFlag, matchesName, mergeXmlnsDeclarations, parseBooleanElement, parseColorElement, parseNumberingLevelAttribute, parseNumericAttribute, parseTableMeasurementValue, parseXml, parseXmlDocument };
+export { NAMESPACES, cloneWithXmlnsDeclarations, collectXmlnsDeclarations, elementToXml, findAllDeep, findByFullName, findChild, findChildByLocalName, findChildren, findChildrenByLocalName, findDeep, getAttribute, getAttributeAny, getAttributeAnyPrefix, getAttributeByNamespaceUri, getAttributes, getChildElements, getLocalName, getNamespacePrefix, getNamespaceUri, getTextContent, hasChild, hasFlag, matchesName, mergeXmlnsDeclarations, parseBooleanElement, parseColorElement, parseNumberingLevelAttribute, parseNumericAttribute, parseTableMeasurementValue, parseXml, parseXmlDocument };

package/dist/layout-bridge/convert/toFlowBlocks.js

@@ -246,6 +246,8 @@ function extractRunFormatting(marks, theme) {
 			if (font) formatting.fontFamily = font;
 			const eastAsiaFont = resolveEastAsiaThemeFont(attrs, theme);
 			if (eastAsiaFont) formatting.eastAsiaFontFamily = eastAsiaFont;
+			const complexScriptFont = resolveComplexScriptThemeFont(attrs, theme);
+			if (complexScriptFont) formatting.complexScriptFontFamily = complexScriptFont;
 			break;
 		}
 		case "language": {
@@ -361,6 +363,9 @@ const resolveWesternThemeFont = (fontFamily, theme) => {
 	const themeRef = fontFamily.asciiTheme ?? fontFamily.hAnsiTheme;
 	return (themeRef ? resolveThemeFont(themeRef, theme?.fontScheme) : null) ?? fontFamily.ascii ?? fontFamily.hAnsi ?? void 0;
 };
+const resolveComplexScriptThemeFont = (fontFamily, theme) => {
+	return (fontFamily.cstheme ? resolveThemeFont(fontFamily.cstheme, theme?.fontScheme) : null) ?? fontFamily.cs ?? void 0;
+};
 const resolveEastAsiaThemeFont = (fontFamily, theme) => {
 	return (fontFamily.eastAsiaTheme ? resolveThemeFont(fontFamily.eastAsiaTheme, theme?.fontScheme) : null) ?? fontFamily.eastAsia ?? void 0;
 };
@@ -404,6 +409,8 @@ function paragraphRunDefaults(pmAttrs, theme) {
 	if (fontFamily) result.fontFamily = fontFamily;
 	const eastAsiaFontFamily = defaultTextFormatting.fontFamily ? resolveEastAsiaThemeFont(defaultTextFormatting.fontFamily, theme) : void 0;
 	if (eastAsiaFontFamily) result.eastAsiaFontFamily = eastAsiaFontFamily;
+	const complexScriptFontFamily = defaultTextFormatting.fontFamily ? resolveComplexScriptThemeFont(defaultTextFormatting.fontFamily, theme) : void 0;
+	if (complexScriptFontFamily) result.complexScriptFontFamily = complexScriptFontFamily;
 	if (defaultTextFormatting.language) result.language = { ...defaultTextFormatting.language };
 	if (defaultTextFormatting.fontSize !== void 0) result.fontSize = defaultTextFormatting.fontSize / 2;
 	if (defaultTextFormatting.bold !== void 0) result.bold = defaultTextFormatting.bold;

package/dist/layout-engine/measure/cache.js

@@ -170,7 +170,7 @@ const paragraphMeasureCache = /* @__PURE__ */ new Map();
 */
 function hashParagraphBlock(block) {
 	const parts = [`lbp:${getLineBreakProviderGeneration()}`];
-	for (const run of block.runs) if (run.kind === "text") parts.push(`t:${run.text}|${run.fontFamily}|${run.eastAsiaFontFamily}|${run.fontSize}|${run.bold}|${run.italic}|${run.allCaps}|${run.smallCaps}|${run.horizontalScale}|${run.letterSpacing}|${run.language?.val}|${run.language?.eastAsia}|${run.language?.bidi}`);
+	for (const run of block.runs) if (run.kind === "text") parts.push(`t:${run.text}|${run.fontFamily}|${run.eastAsiaFontFamily}|${run.complexScriptFontFamily}|${run.fontSize}|${run.bold}|${run.italic}|${run.allCaps}|${run.smallCaps}|${run.horizontalScale}|${run.letterSpacing}|${run.language?.val}|${run.language?.eastAsia}|${run.language?.bidi}`);
 	else if (run.kind === "tab") parts.push(`tab:${run.width}`);
 	else if (run.kind === "image") parts.push(`img:${run.width}x${run.height}:${run.exactLineHeight === true ? "exact" : "text"}`);
 	else if (run.kind === "lineBreak") parts.push("br");

package/dist/layout-engine/measure/measureContainer.js

@@ -1,4 +1,4 @@
-import { hasCjk, isCjkCodePoint, segmentByScript } from "../../utils/scriptSegments.js";
+import { SCRIPT_CLASS, hasCjk, hasComplexScript, scriptClassOf, segmentByScript } from "../../utils/scriptSegments.js";
 import { getCachedFontMetrics, getCachedTextWidth, getTextWidthCacheGeneration, setCachedFontMetrics, setCachedTextWidth } from "./cache.js";
 import { buildFontString, getResolvedData, ptToPx } from "./measureHelpers.js";
 import { setMeasureProvider } from "./measureProvider.js";
@@ -118,7 +118,7 @@ function canvasGetFontMetrics(style) {
 function canvasMeasureTextWidth(text, style) {
 	if (!text) return 0;
 	const measuredText = applyTextTransform(text, style);
-	if (style.eastAsiaFontFamily && !style.letterSpacing && hasCjk(measuredText)) return measureMixedScriptWidth(measuredText, style);
+	if (!style.letterSpacing && needsPerScriptFonts(style, measuredText)) return measureMixedScriptWidth(measuredText, style);
 	const ctx = getCanvasContext();
 	const font = buildFontString(style);
 	const letterSpacing = style.letterSpacing ?? 0;
@@ -173,14 +173,25 @@ function glyphAdvanceStyle(style, fontFamily) {
 * horizontal scale are applied once over the whole string so the total matches
 * the painter, which renders the same segments as sibling spans.
 */
+/**
+* Whether this run needs more than one font, i.e. it carries a per-script slot
+* AND text that would select it. Keeps the all-Latin path on a single font.
+*/
+function needsPerScriptFonts(style, measuredText) {
+	if (style.eastAsiaFontFamily && hasCjk(measuredText)) return true;
+	return Boolean(style.complexScriptFontFamily) && hasComplexScript(measuredText);
+}
+/** The font slot a script class selects, falling back to the western one. */
+function scriptFontFamily(style, script) {
+	if (script === SCRIPT_CLASS.eastAsia) return style.eastAsiaFontFamily ?? style.fontFamily;
+	if (script === SCRIPT_CLASS.complex) return style.complexScriptFontFamily ?? style.fontFamily;
+	return style.fontFamily;
+}
 function measureMixedScriptWidth(measuredText, style) {
 	const letterSpacing = style.letterSpacing ?? 0;
 	const horizontalScale = getHorizontalScaleFactor(style);
 	let glyphWidth = 0;
-	for (const segment of segmentByScript(measuredText)) {
-		const fontFamily = segment.isCjk ? style.eastAsiaFontFamily : style.fontFamily;
-		glyphWidth += canvasMeasureTextWidth(segment.text, glyphAdvanceStyle(style, fontFamily));
-	}
+	for (const segment of segmentByScript(measuredText)) glyphWidth += canvasMeasureTextWidth(segment.text, glyphAdvanceStyle(style, scriptFontFamily(style, segment.script)));
 	let width = glyphWidth;
 	if (letterSpacing) {
 		const codePoints = countCodePoints(measuredText);
@@ -251,6 +262,10 @@ function canvasMeasureRun(text, style) {
 		...style,
 		fontFamily: style.eastAsiaFontFamily
 	}) : void 0;
+	const complexScriptFont = style.complexScriptFontFamily !== void 0 && !style.letterSpacing ? buildFontString({
+		...style,
+		fontFamily: style.complexScriptFontFamily
+	}) : void 0;
 	const letterSpacing = style.letterSpacing ?? 0;
 	const scale = getHorizontalScaleFactor(style);
 	const charWidths = [];
@@ -259,7 +274,10 @@ function canvasMeasureRun(text, style) {
 	for (const char of text) {
 		const cp = char.codePointAt(0);
 		const measured = applyTextTransform(char, style);
-		if (eastAsiaFont !== void 0) ctx.font = isCjkCodePoint(cp) ? eastAsiaFont : baseFont;
+		if (eastAsiaFont !== void 0 || complexScriptFont !== void 0) {
+			const script = scriptClassOf(cp);
+			ctx.font = (script === SCRIPT_CLASS.eastAsia ? eastAsiaFont : void 0) ?? (script === SCRIPT_CLASS.complex ? complexScriptFont : void 0) ?? baseFont;
+		}
 		let charWidth = ctx.measureText(measured).width;
 		if (letterSpacing && offset + char.length < text.length) charWidth += letterSpacing;
 		charWidth *= scale;

package/dist/layout-engine/measure/measureHelpers.js

@@ -28,6 +28,7 @@ function buildRunFontStyle(run, fallbackFontFamily, fallbackFontSize) {
 	return {
 		fontFamily: run.fontFamily ?? fallbackFontFamily,
 		...run.eastAsiaFontFamily !== void 0 ? { eastAsiaFontFamily: run.eastAsiaFontFamily } : {},
+		...run.complexScriptFontFamily !== void 0 ? { complexScriptFontFamily: run.complexScriptFontFamily } : {},
 		fontSize,
 		...run.bold !== void 0 ? { bold: run.bold } : {},
 		...run.italic !== void 0 ? { italic: run.italic } : {},

package/dist/layout-engine/measure/measureTypes.d.ts

@@ -21,6 +21,11 @@ type FontStyle = {
    * and click positioning stay in sync.
    */
   eastAsiaFontFamily?: string;
+  /**
+   * Complex-script font for Arabic, Hebrew, Indic and South-East Asian code
+   * points. Same contract as `eastAsiaFontFamily`, over a different slot.
+   */
+  complexScriptFontFamily?: string;
   fontSize?: number;
   bold?: boolean;
   italic?: boolean;

package/dist/layout-engine/types.d.ts

@@ -37,6 +37,15 @@ type RunFormatting = {
    * to `fontFamily`.
    */
   eastAsiaFontFamily?: string;
+  /**
+   * Resolved complex-script font (`w:cs` / `cstheme`). Arabic, Hebrew, Indic
+   * and South-East Asian code points in this run measure and paint with this
+   * font; the rest keeps `fontFamily` (ascii/hAnsi). Mirrors
+   * `eastAsiaFontFamily` exactly, including the shared segmentation. Absent
+   * means complex-script text falls back to `fontFamily`, which is what Word
+   * does NOT do and is why this slot has to reach layout at all.
+   */
+  complexScriptFontFamily?: string;
   /** Run language metadata resolved from `w:lang`. */
   language?: {
     val?: string;