@stll/folio-core 0.44.0 → 0.45.0

package/dist/docx/commentParser.js

@@ -1,7 +1,8 @@
+import { commentThreadParaId } from "./commentThreadKey.js";
 import { parseParagraph } from "./paragraphParser.js";
 import { cloneParagraphWithPropertySource } from "./paragraphPropertySource.js";
 import { parseRunProperties } from "./runParser.js";
-import { findChild, getAttribute, getChildElements, getLocalName, parseOnOffValue, parseXml } from "./xmlParser.js";
+import { NAMESPACES, WORDPROCESSINGML_NAMESPACE_URIS, findChild, getAttribute, getAttributeByNamespaceUri, getChildElements, getLocalName, parseOnOffValue, parseXml } from "./xmlParser.js";
 import { PARSE_WARNING_CODES } from "@stll/docx-core/model";
 //#region src/docx/commentParser.ts
 /**
@@ -25,6 +26,26 @@ import { PARSE_WARNING_CODES } from "@stll/docx-core/model";
 /** The whole lexical form of `ST_DecimalNumber`: an optional sign and digits. */
 const DECIMAL_NUMBER = /^[+-]?\d+$/u;
 const DEFAULT_ANNOTATION_REFERENCE_STYLE_ID = "CommentReference";
+/**
+* The namespaces a paraId may be written in: Word 2010 and Word 2012 wordml.
+*
+* Resolved by URI rather than by prefix, because the value is a thread key: an
+* unrelated `vendor:paraId` picked up by local name alone would key a comment
+* to the wrong thread and carry that thread's date, parent and resolved state.
+*/
+const PARA_ID_NAMESPACE_URIS = /* @__PURE__ */ new Set([NAMESPACES.w14, NAMESPACES.w15]);
+/**
+* A paraId join key, wherever an exporter writes it: on `w:comment` or `w:p`.
+*
+* One reading for both, so the wrapper and the paragraphs cannot come to
+* disagree about what counts as a key. The literal reads are the fallback for
+* a part that writes a conventional prefix without binding it, which no URI
+* lookup can resolve.
+*/
+const paraIdAttribute = (element) => {
+	const raw = getAttributeByNamespaceUri(element, PARA_ID_NAMESPACE_URIS, "paraId") ?? element.attributes?.["w14:paraId"] ?? element.attributes?.["w15:paraId"] ?? getAttributeByNamespaceUri(element, WORDPROCESSINGML_NAMESPACE_URIS, "paraId") ?? element.attributes?.["w:paraId"];
+	return raw === null || raw === void 0 ? void 0 : String(raw);
+};
 const normalizeAnnotationReferenceFormatting = (formatting) => {
 	if (formatting?.styleId === DEFAULT_ANNOTATION_REFERENCE_STYLE_ID && Object.keys(formatting).length === 1) return;
 	return formatting;
@@ -56,7 +77,8 @@ function parseCommentsExtensible(xml) {
 		if ((child.name?.replace(/^.*:/u, "") ?? "") !== "comment") continue;
 		const paraId = getAttribute(child, "w16cex", "paraId") ?? getAttribute(child, "w15", "paraId") ?? child.attributes?.["w16cex:paraId"] ?? child.attributes?.["w15:paraId"];
 		const dateUtc = getAttribute(child, "w16cex", "dateUtc") ?? getAttribute(child, "w15", "dateUtc") ?? child.attributes?.["w16cex:dateUtc"] ?? child.attributes?.["w15:dateUtc"];
-		if (paraId && dateUtc) dateUtcByParaId.set(String(paraId).toUpperCase(), String(dateUtc));
+		const key = paraId === null || paraId === void 0 ? null : String(paraId).toUpperCase();
+		if (key && dateUtc && !dateUtcByParaId.has(key)) dateUtcByParaId.set(key, String(dateUtc));
 	}
 	return dateUtcByParaId;
 }
@@ -82,12 +104,14 @@ function parseCommentsExtended(xml) {
 		if ((child.name?.replace(/^.*:/u, "") ?? "") !== "commentEx") continue;
 		const paraId = getAttribute(child, "w15", "paraId") ?? child.attributes?.["w15:paraId"];
 		if (!paraId) continue;
+		const key = String(paraId).toUpperCase();
+		if (infoByParaId.has(key)) continue;
 		const parentParaId = getAttribute(child, "w15", "paraIdParent") ?? child.attributes?.["w15:paraIdParent"];
 		const doneAttr = getAttribute(child, "w15", "done") ?? child.attributes?.["w15:done"];
 		const info = {};
 		if (parentParaId) info.parentParaId = String(parentParaId).toUpperCase();
 		if (doneAttr !== void 0) info.done = parseOnOffValue(String(doneAttr).toLowerCase()) ?? false;
-		infoByParaId.set(String(paraId).toUpperCase(), info);
+		infoByParaId.set(key, info);
 	}
 	return infoByParaId;
 }
@@ -105,9 +129,8 @@ function parseComments(commentsXml, styles, theme, rels, media, commentsExtensib
 	const dateUtcByParaId = commentsExtensibleXml ? parseCommentsExtensible(commentsExtensibleXml) : /* @__PURE__ */ new Map();
 	const extendedByParaId = commentsExtendedXml ? parseCommentsExtended(commentsExtendedXml) : /* @__PURE__ */ new Map();
 	const children = getChildElements(findChild(root, "w", "comments") ?? root);
-	const comments = [];
+	const parsed = [];
 	const commentIdByParaId = /* @__PURE__ */ new Map();
-	const paraIdByCommentIndex = /* @__PURE__ */ new Map();
 	for (const child of children) {
 		if ((child.name?.replace(/^.*:/u, "") ?? "") !== "comment") continue;
 		const rawId = getAttribute(child, "w", "id");
@@ -126,12 +149,7 @@ function parseComments(commentsXml, styles, theme, rels, media, commentsExtensib
 		const initials = rawInitials !== null ? String(rawInitials) : void 0;
 		const rawDate = getAttribute(child, "w", "date");
 		const localDate = rawDate !== null ? String(rawDate) : void 0;
-		let rawParaId = getAttribute(child, "w14", "paraId") ?? child.attributes?.["w14:paraId"] ?? getAttribute(child, "w15", "paraId") ?? child.attributes?.["w15:paraId"] ?? getAttribute(child, "w", "paraId");
-		if (!rawParaId) for (const sub of getChildElements(child)) {
-			if ((sub.name?.replace(/^.*:/u, "") ?? "") !== "p") continue;
-			const subParaId = getAttribute(sub, "w14", "paraId") ?? sub.attributes?.["w14:paraId"] ?? getAttribute(sub, "w15", "paraId") ?? sub.attributes?.["w15:paraId"] ?? getAttribute(sub, "w", "paraId");
-			if (subParaId) rawParaId = subParaId;
-		}
+		const rawParaId = paraIdAttribute(child) ?? commentThreadParaId(getChildElements(child).filter((sub) => (sub.name?.replace(/^.*:/u, "") ?? "") === "p").map(paraIdAttribute));
 		const paraId = rawParaId ? String(rawParaId).toUpperCase() : null;
 		const date = (paraId ? dateUtcByParaId.get(paraId) : void 0) ?? localDate;
 		const done = (paraId ? extendedByParaId.get(paraId) : void 0)?.done;
@@ -147,33 +165,26 @@ function parseComments(commentsXml, styles, theme, rels, media, commentsExtensib
 			annotationReferenceFormatting = normalized.annotationReferenceFormatting;
 			paragraphs.push(normalized.paragraph);
 		}
-		const commentIndex = comments.length;
-		if (paraId) {
-			commentIdByParaId.set(paraId, id);
-			paraIdByCommentIndex.set(commentIndex, paraId);
-		}
-		comments.push({
-			id,
-			author,
-			...initials !== void 0 ? { initials } : {},
-			...date !== void 0 ? { date } : {},
-			...done !== void 0 ? { done } : {},
-			...annotationReferenceFormatting !== void 0 ? { annotationReferenceFormatting } : {},
-			content: paragraphs
+		if (paraId && !commentIdByParaId.has(paraId)) commentIdByParaId.set(paraId, id);
+		parsed.push({
+			comment: {
+				id,
+				author,
+				...initials !== void 0 ? { initials } : {},
+				...date !== void 0 ? { date } : {},
+				...done !== void 0 ? { done } : {},
+				...annotationReferenceFormatting !== void 0 ? { annotationReferenceFormatting } : {},
+				content: paragraphs
+			},
+			threadParaId: paraId
 		});
 	}
-	for (let i = 0; i < comments.length; i++) {
-		const comment = comments[i];
-		if (!comment) continue;
-		const paraId = paraIdByCommentIndex.get(i);
-		if (!paraId) continue;
-		const parentParaId = extendedByParaId.get(paraId)?.parentParaId;
-		if (!parentParaId) continue;
-		const parentId = commentIdByParaId.get(parentParaId);
-		if (parentId !== void 0 && parentId !== comment.id) comments[i] = {
-			...comment,
-			parentId
-		};
+	const comments = [];
+	for (const { comment, threadParaId } of parsed) {
+		const parentParaId = threadParaId ? extendedByParaId.get(threadParaId)?.parentParaId : void 0;
+		const parentId = parentParaId ? commentIdByParaId.get(parentParaId) : void 0;
+		if (parentId !== void 0 && parentId !== comment.id) comment.parentId = parentId;
+		comments.push(comment);
 	}
 	return comments;
 }

package/dist/docx/commentThreadKey.d.ts

@@ -0,0 +1,18 @@
+//#region src/docx/commentThreadKey.d.ts
+/**
+ * The one rule for "which `w14:paraId` a comment is threaded by".
+ *
+ * `w15:commentEx/@w15:paraId` and `w15:paraIdParent` name a comment through the
+ * `w14:paraId` of its LAST paragraph, never through its `w:id`. Files in the
+ * wild leave that paragraph's id off and carry one on an earlier paragraph, so
+ * the rule folio reads by is the last paragraph that carries an id.
+ *
+ * The rule has two readers over two representations: the parser applies it to
+ * the `w:p` children of a `w:comment`, the serializer to the model's
+ * `Comment.content`. They must not drift — a save that keyed on a different
+ * paragraph than the parse would write a `commentsExtended.xml` whose entries
+ * belong to other comments — so both call this instead of restating it.
+ */
+declare const commentThreadParaId: (paragraphParaIds: Iterable<string | null | undefined>) => string | undefined;
+//#endregion
+export { commentThreadParaId };

package/dist/docx/commentThreadKey.js

@@ -0,0 +1,22 @@
+//#region src/docx/commentThreadKey.ts
+/**
+* The one rule for "which `w14:paraId` a comment is threaded by".
+*
+* `w15:commentEx/@w15:paraId` and `w15:paraIdParent` name a comment through the
+* `w14:paraId` of its LAST paragraph, never through its `w:id`. Files in the
+* wild leave that paragraph's id off and carry one on an earlier paragraph, so
+* the rule folio reads by is the last paragraph that carries an id.
+*
+* The rule has two readers over two representations: the parser applies it to
+* the `w:p` children of a `w:comment`, the serializer to the model's
+* `Comment.content`. They must not drift — a save that keyed on a different
+* paragraph than the parse would write a `commentsExtended.xml` whose entries
+* belong to other comments — so both call this instead of restating it.
+*/
+const commentThreadParaId = (paragraphParaIds) => {
+	let key;
+	for (const paraId of paragraphParaIds) if (paraId) key = paraId;
+	return key;
+};
+//#endregion
+export { commentThreadParaId };

package/dist/docx/diagramPreview.js

@@ -1,5 +1,7 @@
-import { resolveRelativePath } from "./relsParser.js";
-import { findChildByNamespaceUri, getAttribute, getLocalName, parseNumericAttribute, parseXmlDocument } from "./xmlParser.js";
+import { bytesToDataUrl } from "../utils/base64.js";
+import { PREVIEW_KINDS } from "./previewBudget.js";
+import { resolveRelationshipIdOfType, resolveRelativePath } from "./relsParser.js";
+import { OFFICE_RELATIONSHIP_NAMESPACE_URIS, findChildByNamespaceUri, getAttribute, getAttributeByNamespaceUri, getLocalName, parseNumericAttribute, parseXmlDocument } from "./xmlParser.js";
 //#region src/docx/diagramPreview.ts
 const MAX_PREVIEW_SHAPES = 128;
 const MAX_PREVIEW_PIXELS = 144e4;
@@ -7,21 +9,47 @@ const MAX_PREVIEW_PAINT_PIXELS = MAX_PREVIEW_PIXELS * 4;
 const DRAWINGML_NAMESPACE_URIS = /* @__PURE__ */ new Set(["http://schemas.openxmlformats.org/drawingml/2006/main", "http://purl.oclc.org/ooxml/drawingml/main"]);
 const DIAGRAM_NAMESPACE_URIS = /* @__PURE__ */ new Set(["http://schemas.openxmlformats.org/drawingml/2006/diagram", "http://purl.oclc.org/ooxml/drawingml/diagram"]);
 const DIAGRAM_DRAWING_NAMESPACE_URIS = /* @__PURE__ */ new Set(["http://schemas.microsoft.com/office/drawing/2008/diagram"]);
+const DIAGRAM_DATA_RELATIONSHIP_TYPE = "http://schemas.openxmlformats.org/officeDocument/2006/relationships/diagramData";
+const DIAGRAM_DRAWING_RELATIONSHIP_TYPE = "http://schemas.microsoft.com/office/2007/relationships/diagramDrawing";
+const DOCUMENT_PART_PATH = "word/document.xml";
 const WORD_DRAWING_NAMESPACE_URIS = /* @__PURE__ */ new Set(["http://schemas.openxmlformats.org/drawingml/2006/wordprocessingDrawing", "http://purl.oclc.org/ooxml/drawingml/wordprocessingDrawing"]);
+/**
+* The preview is a megapixel raster, so both checksums run over megabytes.
+* `for (const byte of bytes)` drives the array iterator protocol once per
+* byte, which profiles as the dominant cost of parsing a SmartArt document;
+* indexed loops and a table-driven CRC produce the same numbers without it.
+*/
+const CRC32_TABLE = (() => {
+	const table = /* @__PURE__ */ new Uint32Array(256);
+	for (let index = 0; index < 256; index += 1) {
+		let value = index;
+		for (let bit = 0; bit < 8; bit += 1) value = value >>> 1 ^ (value & 1 ? 3988292384 : 0);
+		table[index] = value >>> 0;
+	}
+	return table;
+})();
 const crc32 = (bytes) => {
 	let crc = 4294967295;
-	for (const byte of bytes) {
-		crc ^= byte;
-		for (let bit = 0; bit < 8; bit += 1) crc = crc >>> 1 ^ (crc & 1 ? 3988292384 : 0);
-	}
+	for (let index = 0; index < bytes.length; index += 1) crc = crc >>> 8 ^ CRC32_TABLE[(crc ^ bytes[index]) & 255];
 	return (crc ^ 4294967295) >>> 0;
 };
+/**
+* `a` and `b` stay below 2^31 for 5552 iterations from any legal state, so the
+* modulo runs per block rather than per byte.
+*/
+const ADLER32_BLOCK = 5552;
 const adler32 = (bytes) => {
 	let a = 1;
 	let b = 0;
-	for (const byte of bytes) {
-		a = (a + byte) % 65521;
-		b = (b + a) % 65521;
+	let index = 0;
+	while (index < bytes.length) {
+		const end = Math.min(index + ADLER32_BLOCK, bytes.length);
+		for (; index < end; index += 1) {
+			a += bytes[index];
+			b += a;
+		}
+		a %= 65521;
+		b %= 65521;
 	}
 	return b << 16 | a;
 };
@@ -88,9 +116,8 @@ const previewPng = (width, height, shapes) => {
 		cursor += length;
 		offset += length;
 	}
-	const output = new Uint8Array(cursor + 4);
-	output.set(compressed.subarray(0, cursor));
-	new DataView(output.buffer).setUint32(cursor, adler32(pixels));
+	new DataView(compressed.buffer).setUint32(cursor, adler32(pixels));
+	const output = compressed;
 	const signature = new Uint8Array([
 		137,
 		80,
@@ -127,13 +154,38 @@ const extent = (drawing) => {
 		height: parseNumericAttribute(value, null, "cy") ?? 0
 	};
 };
-const cachedDiagramShapes = (rels, media) => {
-	const matches = [...rels.values()].filter((relationship) => relationship.type === "http://schemas.microsoft.com/office/2007/relationships/diagramDrawing");
-	if (matches.length !== 1 || !matches[0]?.target) return [];
-	const path = resolveRelativePath("word/document.xml", matches[0].target);
-	const file = media.get(path);
-	if (!file?.data) return [];
-	const root = parseXmlDocument(new TextDecoder().decode(file.data));
+/** Parse the part a relationship of the given type names, by that relationship's id. */
+const partByRelationshipId = ({ rels, media, rId, type }) => {
+	const resolved = resolveRelationshipIdOfType(rels, rId ?? void 0, type);
+	if (resolved.status !== "resolved" || !resolved.relationship.target) return null;
+	const file = media.get(resolveRelativePath(DOCUMENT_PART_PATH, resolved.relationship.target));
+	return file?.data ? parseXmlDocument(new TextDecoder().decode(file.data)) : null;
+};
+/**
+* The drawing cache this diagram points at, reached through its own ids.
+*
+* `dgm:relIds/@r:dm` names the data part, and that part's `dsp:dataModelExt`
+* extension names the cached drawing. Scanning the relationship map for a type
+* instead of following the ids reads the wrong diagram whenever a document has
+* more than one, which is why the scan refused outright on a second match:
+* both diagrams in a two-diagram document then got no preview at all.
+*/
+const cachedDiagramShapes = (graphicData, rels, media) => {
+	const relIds = findChildByNamespaceUri(graphicData, DIAGRAM_NAMESPACE_URIS, "relIds");
+	const data = partByRelationshipId({
+		rels,
+		media,
+		rId: getAttributeByNamespaceUri(relIds, OFFICE_RELATIONSHIP_NAMESPACE_URIS, "dm"),
+		type: DIAGRAM_DATA_RELATIONSHIP_TYPE
+	});
+	if (!data) return [];
+	const dataModelExt = descendantsByNamespace(data, DIAGRAM_DRAWING_NAMESPACE_URIS, "dataModelExt").at(0);
+	const root = partByRelationshipId({
+		rels,
+		media,
+		rId: getAttribute(dataModelExt, null, "relId"),
+		type: DIAGRAM_DRAWING_RELATIONSHIP_TYPE
+	});
 	if (!root) return [];
 	const shapes = [];
 	for (const shape of descendantsByNamespace(root, DIAGRAM_DRAWING_NAMESPACE_URIS, "sp").slice(0, MAX_PREVIEW_SHAPES)) {
@@ -155,31 +207,39 @@ const cachedDiagramShapes = (rels, media) => {
 	}
 	return shapes;
 };
+const matchesNamespace = (element, namespaceUris, localName) => element.namespaceUri !== void 0 && namespaceUris.has(element.namespaceUri) && getLocalName(element.name ?? "") === localName;
 const descendantsByNamespace = (root, namespaceUris, localName) => {
 	const result = [];
 	const visit = (element) => {
-		if (element.namespaceUri && namespaceUris.has(element.namespaceUri) && getLocalName(element.name ?? "") === localName) result.push(element);
+		if (matchesNamespace(element, namespaceUris, localName)) result.push(element);
 		for (const child of element.elements ?? []) if (child.type === "element") visit(child);
 	};
 	visit(root);
 	return result;
 };
+/** The first match in document order, without walking the rest of the subtree. */
+const firstDescendantByNamespace = (root, namespaceUris, localName) => {
+	if (matchesNamespace(root, namespaceUris, localName)) return root;
+	for (const child of root.elements ?? []) {
+		if (child.type !== "element") continue;
+		const found = firstDescendantByNamespace(child, namespaceUris, localName);
+		if (found) return found;
+	}
+	return null;
+};
 /** Create a deliberately simple, bounded preview; it is never an editable diagram projection. */
 const parseDiagramPreview = (drawing, rels, media) => {
 	if (!rels || !media) return null;
-	const graphicData = descendantsByNamespace(drawing, DRAWINGML_NAMESPACE_URIS, "graphicData").at(0);
+	const graphicData = firstDescendantByNamespace(drawing, DRAWINGML_NAMESPACE_URIS, "graphicData");
 	if (!graphicData || !DIAGRAM_NAMESPACE_URIS.has(getAttribute(graphicData, null, "uri") ?? "")) return null;
 	const { width, height } = extent(drawing);
 	if (width <= 0 || height <= 0) return null;
-	const png = previewPng(width, height, cachedDiagramShapes(rels, media));
-	let binary = "";
-	for (let offset = 0; offset < png.length; offset += 32768) binary += String.fromCodePoint(...png.subarray(offset, offset + 32768));
 	return {
 		type: "image",
 		rId: "",
-		src: `data:image/png;base64,${btoa(binary)}`,
-		mimeType: "image/png",
-		filename: "smartart-preview.png",
+		src: bytesToDataUrl(previewPng(width, height, cachedDiagramShapes(graphicData, rels, media)), PREVIEW_KINDS.smartArt.mimeType),
+		mimeType: PREVIEW_KINDS.smartArt.mimeType,
+		filename: PREVIEW_KINDS.smartArt.filename,
 		size: {
 			width,
 			height

package/dist/docx/groupDrawingParser.js

@@ -1,6 +1,7 @@
 import { emuToPixels } from "../utils/units.js";
 import { parseImage, resolveImageData } from "./imageParser.js";
 import { findAllDeep, findChildByLocalName, findChildrenByLocalName, getAttribute, getChildElements, getLocalName, getTextContent, parseNumericAttribute } from "./xmlParser.js";
+import { escapeXmlAttribute, escapeXmlText } from "@stll/docx-core";
 //#region src/docx/groupDrawingParser.ts
 const HEX_COLOR = /^[0-9A-Fa-f]{6}$/u;
 const DEFAULT_TEXT_COLOR = "000000";
@@ -12,7 +13,6 @@ const CROP_SCALE = 1e5;
 const MAX_PATH_COMMANDS = 1e4;
 const MAX_TEXT_CHARACTERS = 2e4;
 const MAX_SVG_CHARACTERS = 1e6;
-const escapeXml = (value) => value.replaceAll("&", "&amp;").replaceAll("<", "&lt;").replaceAll(">", "&gt;").replaceAll("\"", "&quot;").replaceAll("'", "&apos;");
 const numericAttr = (element, name) => {
 	const direct = parseNumericAttribute(element, null, name);
 	if (direct !== void 0) return direct;
@@ -108,7 +108,7 @@ const renderTextBox = (wsp) => {
 	const color = colorFrom(findAllDeep(wsp, "w", "color").at(0) ?? null, DEFAULT_TEXT_COLOR);
 	const fontSize = halfPoints * HALF_POINT_TO_EMU;
 	const maxCharacters = Math.max(1, Math.floor(width / (fontSize * .38)));
-	const lines = paragraphs.flatMap((paragraph) => wrapLine(getTextContent(paragraph).slice(0, MAX_TEXT_CHARACTERS), maxCharacters)).map(escapeXml);
+	const lines = paragraphs.flatMap((paragraph) => wrapLine(getTextContent(paragraph).slice(0, MAX_TEXT_CHARACTERS), maxCharacters)).map(escapeXmlText);
 	if (lines.length === 0) return "";
 	const lineHeight = fontSize * 1.15;
 	const svgFontSize = 1e3;
@@ -131,7 +131,7 @@ const renderPicture = (picture, index, rels, media) => {
 	const visibleWidth = 1 - left - right;
 	const visibleHeight = 1 - top - bottom;
 	if (visibleWidth <= 0 || visibleHeight <= 0) return "";
-	const image = `<image x="${x - width * left / visibleWidth}" y="${y - height * top / visibleHeight}" width="${width / visibleWidth}" height="${height / visibleHeight}" href="${escapeXml(src)}" preserveAspectRatio="none"/>`;
+	const image = `<image x="${x - width * left / visibleWidth}" y="${y - height * top / visibleHeight}" width="${width / visibleWidth}" height="${height / visibleHeight}" href="${escapeXmlAttribute(src)}" preserveAspectRatio="none"/>`;
 	if (left === 0 && top === 0 && right === 0 && bottom === 0) return image;
 	const clipId = `group-picture-${index}`;
 	return `<defs><clipPath id="${clipId}"><rect x="${x}" y="${y}" width="${width}" height="${height}"/></clipPath></defs><g clip-path="url(#${clipId})">${image}</g>`;

package/dist/docx/hyperlinkParser.js

@@ -1,4 +1,4 @@
-import { sanitizeExternalUrl, sanitizeLinkTarget } from "../utils/urlSecurity.js";
+import { sanitizeExternalUrl } from "../utils/urlSecurity.js";
 import { RELATIONSHIP_TYPES, resolveRelationshipIdOfType } from "./relsParser.js";
 import { parseRun } from "./runParser.js";
 import { getAttribute, getChildElements, getLocalName, mergeXmlnsDeclarations, parseNumericAttribute, parseOnOffAttribute } from "./xmlParser.js";
@@ -63,7 +63,7 @@ function parseHyperlink(node, rels, styles = null, theme = null, media = null, r
 	const tooltip = getAttribute(node, "w", "tooltip");
 	if (tooltip) hyperlink.tooltip = tooltip;
 	const tgtFrame = getAttribute(node, "w", "tgtFrame");
-	if (tgtFrame) hyperlink.target = sanitizeLinkTarget(tgtFrame);
+	if (tgtFrame) hyperlink.target = tgtFrame;
 	if (parseOnOffAttribute(node, "w", "history") === true) hyperlink.history = true;
 	const docLocation = getAttribute(node, "w", "docLocation");
 	if (docLocation) hyperlink.docLocation = docLocation;

package/dist/docx/imageParser.d.ts

@@ -1,6 +1,14 @@
 import { document_d_exports } from "../types/document.js";
 import { XmlElement } from "./xmlParser.js";
 //#region src/docx/imageParser.d.ts
+/**
+ * The `a:ext` uri under which Word records that a drawing is decorative. The
+ * extension holds `<adec:decorative val="…"/>`; `CT_NonVisualDrawingProps` has
+ * no `@decorative` attribute for it to be written as.
+ */
+declare const DECORATIVE_EXTENSION_URI = "{C183D7F6-B498-43B3-948B-1728B52AA6E4}";
+/** The namespace the decorative extension's element is bound to. */
+declare const DECORATIVE_NAMESPACE = "http://schemas.microsoft.com/office/drawing/2017/decorative";
 /**
  * Resolve image data from relationships and media map
  *
@@ -89,4 +97,4 @@ declare function getWrapDistancesPx(image: document_d_exports.Image): {
  */
 declare function needsTextWrapping(image: document_d_exports.Image): boolean;
 //#endregion
-export { getImageDimensionsPx, getImageHeightPx, getImageWidthPx, getWrapDistancesPx, hasAltText, isBehindText, isDecorativeImage, isFloatingImage, isInFrontOfText, isInlineImage, needsTextWrapping, parseDrawing, parseImage, resolveImageData };
+export { DECORATIVE_EXTENSION_URI, DECORATIVE_NAMESPACE, getImageDimensionsPx, getImageHeightPx, getImageWidthPx, getWrapDistancesPx, hasAltText, isBehindText, isDecorativeImage, isFloatingImage, isInFrontOfText, isInlineImage, needsTextWrapping, parseDrawing, parseImage, resolveImageData };

package/dist/docx/imageParser.js

@@ -3,9 +3,11 @@ import { emuToPixels } from "../utils/units.js";
 import { sanitizeExternalUrl } from "../utils/urlSecurity.js";
 import { WRAP_ELEMENT_NAMES, parseAnchorBehindDoc, parsePositionH, parsePositionV, parseWrapElement } from "./drawingUtils.js";
 import { parseGraphicFrameLocks } from "./graphicFrameLocks.js";
+import { parseNonVisualDrawingNames } from "./nonVisualDrawingProps.js";
 import { RELATIONSHIP_TYPES, resolveRelationshipIdOfType } from "./relsParser.js";
 import { isTextBoxDrawing } from "./textBoxParser.js";
-import { findByFullName, findChild, getAttribute, getChildElements, parseNumericAttribute, parseOnOffValue } from "./xmlParser.js";
+import { captureVerbatimXml } from "./verbatimCapture.js";
+import { findByFullName, findChild, findChildByNamespaceUri, findChildrenByNamespaceUri, getAttribute, getChildElements, getLocalName, getNamespaceUri, parseNumericAttribute, parseOnOffAttribute, parseOnOffValue } from "./xmlParser.js";
 //#region src/docx/imageParser.ts
 /**
 * Convert rotation value (1/60000 of a degree) to degrees
@@ -64,6 +66,47 @@ function parseEffectExtent(effectExtent) {
 	};
 }
 /**
+* The `a:ext` uri under which Word records that a drawing is decorative. The
+* extension holds `<adec:decorative val="…"/>`; `CT_NonVisualDrawingProps` has
+* no `@decorative` attribute for it to be written as.
+*/
+const DECORATIVE_EXTENSION_URI = "{C183D7F6-B498-43B3-948B-1728B52AA6E4}";
+/** The namespace the decorative extension's element is bound to. */
+const DECORATIVE_NAMESPACE = "http://schemas.microsoft.com/office/drawing/2017/decorative";
+/**
+* `wp:docPr`'s extension list is DrawingML's (`a:extLst` holding `a:ext`), in
+* the Transitional or the Strict namespace. Matching on the local name alone
+* would take an `extLst` some other namespace owns and replay its children
+* inside an `a:extLst`, which is a different container than the source wrote.
+*/
+const DRAWINGML_NAMESPACE_URIS = /* @__PURE__ */ new Set(["http://schemas.openxmlformats.org/drawingml/2006/main", "http://purl.oclc.org/ooxml/drawingml/main"]);
+/**
+* Read the `wp:docPr` extension list.
+*
+* Only the decorative extension is modeled. The rest — a creation id, a
+* local-DPI hint, whatever a later Word writes — are captured as they were
+* written so the save path can put them back: an extension folio drops is a
+* fact the document had and no longer does.
+*/
+const parseDocPropsExtensions = (docPr) => {
+	const extLst = findChildByNamespaceUri(docPr, DRAWINGML_NAMESPACE_URIS, "extLst");
+	if (!extLst) return { other: [] };
+	const result = { other: [] };
+	for (const ext of findChildrenByNamespaceUri(extLst, DRAWINGML_NAMESPACE_URIS, "ext")) {
+		if (getAttribute(ext, null, "uri")?.toUpperCase() !== "{C183D7F6-B498-43B3-948B-1728B52AA6E4}") {
+			result.other.push(captureVerbatimXml(ext));
+			continue;
+		}
+		const flag = getChildElements(ext).find((child) => getLocalName(child.name) === "decorative" && getNamespaceUri(child) === "http://schemas.microsoft.com/office/drawing/2017/decorative");
+		if (!flag) {
+			result.other.push(captureVerbatimXml(ext));
+			continue;
+		}
+		result.decorative = parseOnOffValue(getAttribute(flag, null, "val")) ?? true;
+	}
+	return result;
+};
+/**
 * Parse document properties (wp:docPr)
 *
 * @param docPr - wp:docPr element
@@ -72,18 +115,17 @@ function parseEffectExtent(effectExtent) {
 function parseDocProps(docPr) {
 	if (!docPr) return {};
 	const id = getAttribute(docPr, null, "id");
-	const name = getAttribute(docPr, null, "name");
-	const descr = getAttribute(docPr, null, "descr");
-	const title = getAttribute(docPr, null, "title");
-	const decorative = parseOnOffValue(getAttribute(docPr, null, "decorative")) === true;
+	const names = parseNonVisualDrawingNames(docPr);
+	const hidden = parseOnOffAttribute(docPr, null, "hidden");
+	const { decorative, other: docPrExtensions } = parseDocPropsExtensions(docPr);
 	const hlinkClickEl = findChild(docPr, "a", "hlinkClick");
 	const hlinkRId = hlinkClickEl ? getAttribute(hlinkClickEl, "r", "id") : null;
 	return {
 		...id != null ? { id } : {},
-		...name != null ? { name } : {},
-		...descr != null ? { alt: descr } : {},
-		...title != null ? { title } : {},
-		...decorative ? { decorative } : {},
+		...names,
+		...decorative === void 0 ? {} : { decorative },
+		...hidden === void 0 ? {} : { hidden },
+		...docPrExtensions.length > 0 ? { docPrExtensions } : {},
 		...hlinkRId != null ? { hlinkRId } : {}
 	};
 }
@@ -329,7 +371,9 @@ function parseInline(inlineEl, rels, media) {
 	if (props.name !== void 0) image.docPrName = props.name;
 	if (props.alt !== void 0) image.alt = props.alt;
 	if (props.title !== void 0) image.title = props.title;
-	if (props.decorative) image.decorative = true;
+	if (props.decorative !== void 0) image.decorative = props.decorative;
+	if (props.hidden !== void 0) image.hidden = props.hidden;
+	if (props.docPrExtensions !== void 0) image.docPrExtensions = props.docPrExtensions;
 	const safeSrc = sanitizeImageSrc(imageData.src);
 	if (safeSrc) image.src = safeSrc;
 	if (imageData.mimeType) image.mimeType = imageData.mimeType;
@@ -402,7 +446,9 @@ function parseAnchor(anchorEl, rels, media) {
 	if (props.name !== void 0) image.docPrName = props.name;
 	if (props.alt !== void 0) image.alt = props.alt;
 	if (props.title !== void 0) image.title = props.title;
-	if (props.decorative) image.decorative = true;
+	if (props.decorative !== void 0) image.decorative = props.decorative;
+	if (props.hidden !== void 0) image.hidden = props.hidden;
+	if (props.docPrExtensions !== void 0) image.docPrExtensions = props.docPrExtensions;
 	const safeSrc = sanitizeImageSrc(imageData.src);
 	if (safeSrc) image.src = safeSrc;
 	if (imageData.mimeType) image.mimeType = imageData.mimeType;
@@ -537,4 +583,4 @@ function needsTextWrapping(image) {
 	].includes(image.wrap.type);
 }
 //#endregion
-export { getImageDimensionsPx, getImageHeightPx, getImageWidthPx, getWrapDistancesPx, hasAltText, isBehindText, isDecorativeImage, isFloatingImage, isInFrontOfText, isInlineImage, needsTextWrapping, parseDrawing, parseImage, resolveImageData };
+export { DECORATIVE_EXTENSION_URI, DECORATIVE_NAMESPACE, getImageDimensionsPx, getImageHeightPx, getImageWidthPx, getWrapDistancesPx, hasAltText, isBehindText, isDecorativeImage, isFloatingImage, isInFrontOfText, isInlineImage, needsTextWrapping, parseDrawing, parseImage, resolveImageData };

package/dist/docx/imageRawXml.d.ts

@@ -12,7 +12,20 @@ declare const isDrawingRawXmlMode: (value: unknown) => value is document_d_expor
 declare const EDITED_PREVIEW_FINGERPRINT = "editedPreview";
 /** Fingerprints modeled image fields which make raw DrawingML stale when edited. */
 declare const imageRawXmlFingerprint: (image: document_d_exports.Image) => string;
-/** Editable raw drawing XML can replay only while its modeled projection is unchanged. */
+/**
+ * Whether the serializer writes the captured raw XML back.
+ *
+ * For an unclassified drawing the capture is a cache of the model, so it
+ * replays only while the two agree. For a classified one the capture *is* the
+ * drawing: a preserve-only drawing has a placeholder image, and a preview-only
+ * group has a raster render of shapes the model cannot hold. Regenerating
+ * either writes something the source never contained — for the group, at best
+ * the one child picture the rasterizer saw first under that child's
+ * relationship, in place of the whole group. So a classified capture is always
+ * written back; a stale fingerprint means the edit that made it stale is lost,
+ * which {@link classifyDrawingSafety} reports as `opaque`, and never that
+ * folio may build a replacement.
+ */
 declare const canReplayEditableImageRawXml: (drawing: document_d_exports.DrawingContent) => boolean;
 /**
  * What a drawing costs the document when Folio writes it back.

package/dist/docx/imageRawXml.js

@@ -26,10 +26,29 @@ const EDITED_PREVIEW_FINGERPRINT = "editedPreview";
 const editableImageProjection = ({ id: _id, rId: _rId, src: _src, mimeType: _mimeType, filename: _filename, ...image }) => image;
 /** Fingerprints modeled image fields which make raw DrawingML stale when edited. */
 const imageRawXmlFingerprint = (image) => canonicalJson(editableImageProjection(image));
-/** Editable raw drawing XML can replay only while its modeled projection is unchanged. */
+/** Whether the modeled image still says what the captured raw drawing says. */
+const modelAgreesWithRawXml = (drawing) => drawing.rawImageFingerprint === void 0 || drawing.rawImageFingerprint === imageRawXmlFingerprint(drawing.image);
+/**
+* Whether the serializer writes the captured raw XML back.
+*
+* For an unclassified drawing the capture is a cache of the model, so it
+* replays only while the two agree. For a classified one the capture *is* the
+* drawing: a preserve-only drawing has a placeholder image, and a preview-only
+* group has a raster render of shapes the model cannot hold. Regenerating
+* either writes something the source never contained — for the group, at best
+* the one child picture the rasterizer saw first under that child's
+* relationship, in place of the whole group. So a classified capture is always
+* written back; a stale fingerprint means the edit that made it stale is lost,
+* which {@link classifyDrawingSafety} reports as `opaque`, and never that
+* folio may build a replacement.
+*/
 const canReplayEditableImageRawXml = (drawing) => {
-	if (drawing.rawXmlMode === DRAWING_RAW_XML_MODES.PRESERVE_ONLY) return true;
-	return drawing.rawImageFingerprint === void 0 || drawing.rawImageFingerprint === imageRawXmlFingerprint(drawing.image);
+	switch (drawing.rawXmlMode) {
+		case DRAWING_RAW_XML_MODES.PRESERVE_ONLY:
+		case DRAWING_RAW_XML_MODES.PREVIEW_ONLY: return true;
+		case void 0: return modelAgreesWithRawXml(drawing);
+		default: return drawing;
+	}
 };
 /**
 * What a drawing costs the document when Folio writes it back.
@@ -60,9 +79,14 @@ const canRegenerateDrawing = (drawing) => drawing.image.rId !== void 0;
 */
 const classifyDrawingSafety = (drawing) => {
 	if (drawing.rawXml === void 0) return DRAWING_SAFETY_CLASSES.NATIVE;
-	if (canReplayEditableImageRawXml(drawing)) return DRAWING_SAFETY_CLASSES.REPLAYABLE;
-	if (drawing.rawXmlMode === DRAWING_RAW_XML_MODES.PREVIEW_ONLY) return DRAWING_SAFETY_CLASSES.OPAQUE;
-	return canRegenerateDrawing(drawing) ? DRAWING_SAFETY_CLASSES.NATIVE : DRAWING_SAFETY_CLASSES.OPAQUE;
+	switch (drawing.rawXmlMode) {
+		case DRAWING_RAW_XML_MODES.PRESERVE_ONLY: return DRAWING_SAFETY_CLASSES.REPLAYABLE;
+		case DRAWING_RAW_XML_MODES.PREVIEW_ONLY: return modelAgreesWithRawXml(drawing) ? DRAWING_SAFETY_CLASSES.REPLAYABLE : DRAWING_SAFETY_CLASSES.OPAQUE;
+		case void 0:
+			if (modelAgreesWithRawXml(drawing)) return DRAWING_SAFETY_CLASSES.REPLAYABLE;
+			return canRegenerateDrawing(drawing) ? DRAWING_SAFETY_CLASSES.NATIVE : DRAWING_SAFETY_CLASSES.OPAQUE;
+		default: return drawing;
+	}
 };
 //#endregion
 export { DRAWING_SAFETY_CLASSES, EDITED_PREVIEW_FINGERPRINT, allowsDirectDrawingEdit, canReplayEditableImageRawXml, classifyDrawingSafety, imageRawXmlFingerprint, isDrawingRawXmlMode };