@stll/folio-core 0.44.0 → 0.46.0

package/dist/docx/paragraphPropertySource.js

@@ -262,10 +262,11 @@ const invalidTableCellParagraphSourcePayload = (classification, path) => {
 const tableCellBlockTraversalByType = {
 	blockSdt: "blockSdt",
 	paragraph: "paragraph",
+	preservedBlock: "leaf",
 	table: "table"
 };
 const tableCellParagraphContentTraversalByType = {
-	bidiWrapper: "content",
+	inlineWrapper: "content",
 	bookmarkEnd: "leaf",
 	bookmarkStart: "leaf",
 	commentRangeEnd: "leaf",
@@ -283,6 +284,7 @@ const tableCellParagraphContentTraversalByType = {
 	moveTo: "content",
 	moveToRangeEnd: "leaf",
 	moveToRangeStart: "leaf",
+	preservedInline: "leaf",
 	run: "run",
 	simpleField: "content"
 };
@@ -294,6 +296,7 @@ const tableCellRunContentTraversalByType = {
 	footnoteRef: "leaf",
 	instrText: "leaf",
 	noBreakHyphen: "leaf",
+	preservedXml: "leaf",
 	renderedPageBreak: "leaf",
 	shape: "shape",
 	softHyphen: "leaf",
@@ -453,6 +456,7 @@ function visitDecodedTableCellBlock(value, path, context) {
 			for (const [index, child] of content.entries()) visitDecodedTableCellBlock(child, `${path}.content[${index}]`, context);
 			return;
 		}
+		case "leaf": return;
 		default: return traversal;
 	}
 }

package/dist/docx/paragraphTextBoxEnrichment.d.ts

@@ -1,7 +1,7 @@
 import { document_d_exports } from "../types/document.js";
 import { NumberingMap } from "./numberingParser.js";
-import { StyleMap } from "./styleParser.js";
 import { XmlElement } from "./xmlParser.js";
+import { StyleMap } from "./styleParser.js";
 import { TableParserFn } from "./textBoxParser.js";
 //#region src/docx/paragraphTextBoxEnrichment.d.ts
 declare const enrichParagraphTextBoxes: (paragraph: document_d_exports.Paragraph, paraXml: XmlElement, styles: StyleMap | null, theme: document_d_exports.Theme | null, numbering: NumberingMap | null, rels: document_d_exports.RelationshipMap | null, media: Map<string, document_d_exports.MediaFile> | null, parseTable: TableParserFn) => void;

package/dist/docx/paragraphTextBoxEnrichment.js

@@ -1,6 +1,7 @@
 import { pixelsToEmu } from "../utils/units.js";
 import { parseParagraph } from "./paragraphParser.js";
-import { getTextBoxContentElement, isTextBoxDrawing, parseTextBox, parseTextBoxContent } from "./textBoxParser.js";
+import { getTextBoxContentElement, parseTextBox, parseTextBoxContent, scanRunForTextBoxDrawings } from "./textBoxParser.js";
+import { isVmlPictParsedByRunParser } from "./vmlImageParser.js";
 import { findDeep, getAttribute, getChildElements, getLocalName } from "./xmlParser.js";
 //#region src/docx/paragraphTextBoxEnrichment.ts
 const VML_HORIZONTAL_RELATIVES = /* @__PURE__ */ new Set([
@@ -182,7 +183,10 @@ const enrichTextBoxRuns = ({ content, xmlChildren, styles, theme, numbering, rel
 			if (parsedIndex < content.length && parsedContent?.type !== "run") parsedIndex += 1;
 			continue;
 		}
-		const { textBoxDrawings, vmlTextBoxes, hasNonTextBoxContent } = scanRunForTextBoxDrawings(xmlChild);
+		const { textBoxDrawings, vmlTextBoxes, hasNonTextBoxContent } = scanRunForTextBoxDrawings({
+			xmlRun: xmlChild,
+			claimedByRunParser: (pictElement) => isVmlPictParsedByRunParser(pictElement, rels, media)
+		});
 		const parsedRun = parsedContent?.type === "run" ? parsedContent : void 0;
 		const targetRun = parsedRun ?? (hasNonTextBoxContent ? lastConsumedRun : void 0);
 		const targetRunMatchesXml = targetRun !== void 0 && (hasNonTextBoxContent || parsedRun?.content.length === 0);
@@ -198,6 +202,9 @@ const enrichTextBoxRuns = ({ content, xmlChildren, styles, theme, numbering, rel
 				type: "shape",
 				shapeType: "textBox",
 				size: textBox.size,
+				...textBox.name !== void 0 ? { name: textBox.name } : {},
+				...textBox.alt !== void 0 ? { alt: textBox.alt } : {},
+				...textBox.title !== void 0 ? { title: textBox.title } : {},
 				...textBox.position !== void 0 ? { position: textBox.position } : {},
 				...textBox.wrap !== void 0 ? { wrap: textBox.wrap } : {},
 				...textBox.fill !== void 0 ? { fill: textBox.fill } : {},
@@ -251,56 +258,6 @@ const enrichTextBoxRuns = ({ content, xmlChildren, styles, theme, numbering, rel
 		}
 	}
 };
-const scanRunForTextBoxDrawings = (xmlRun) => {
-	const textBoxDrawings = [];
-	const vmlTextBoxes = [];
-	let hasNonTextBoxContent = false;
-	const visitDrawing = (drawingEl) => {
-		if (isTextBoxDrawing(drawingEl)) {
-			textBoxDrawings.push(drawingEl);
-			return;
-		}
-		hasNonTextBoxContent = true;
-	};
-	for (const el of getChildElements(xmlRun)) {
-		const name = getLocalName(el.name ?? "");
-		if (name === "rPr") continue;
-		if (name === "drawing") {
-			visitDrawing(el);
-			continue;
-		}
-		if (name === "pict") {
-			if (findDeep(el, "v", "textbox")) if (findDeep(el, "v", "imagedata")) hasNonTextBoxContent = true;
-			else vmlTextBoxes.push(el);
-			else hasNonTextBoxContent = true;
-			continue;
-		}
-		if (name === "AlternateContent") {
-			const branches = getChildElements(el);
-			const choice = branches.find((branch) => getLocalName(branch.name ?? "") === "Choice");
-			const fallback = branches.find((branch) => getLocalName(branch.name ?? "") === "Fallback");
-			const tryBranch = (branch) => {
-				if (!branch) return false;
-				let found = false;
-				for (const innerEl of getChildElements(branch)) if (getLocalName(innerEl.name ?? "") === "drawing") {
-					visitDrawing(innerEl);
-					found = true;
-				}
-				return found;
-			};
-			let foundInBranch = tryBranch(choice);
-			if (!foundInBranch) foundInBranch = tryBranch(fallback);
-			if (!foundInBranch) hasNonTextBoxContent = true;
-			continue;
-		}
-		hasNonTextBoxContent = true;
-	}
-	return {
-		textBoxDrawings,
-		vmlTextBoxes,
-		hasNonTextBoxContent
-	};
-};
 const parseVmlTextBoxShape = (pictEl, styles, theme, numbering, rels, media, parseTable) => {
 	const shapeEl = findDeep(pictEl, "v", "shape");
 	const textBoxEl = shapeEl ? findDeep(shapeEl, "v", "textbox") : null;

package/dist/docx/paragraphTraversal.js

@@ -16,7 +16,7 @@ const visitParagraphRuns = (paragraph, visit) => {
 			case "moveFrom":
 			case "moveTo":
 			case "inlineSdt":
-			case "bidiWrapper":
+			case "inlineWrapper":
 				for (const child of content.content) visitParagraphContent(child);
 				return;
 			case "complexField":
@@ -32,7 +32,8 @@ const visitParagraphRuns = (paragraph, visit) => {
 			case "moveFromRangeEnd":
 			case "moveToRangeStart":
 			case "moveToRangeEnd":
-			case "mathEquation": return;
+			case "mathEquation":
+			case "preservedInline": return;
 			default: return content;
 		}
 	};
@@ -69,7 +70,7 @@ const visitInlineContentSlots = (paragraph, visit) => {
 				case "deletion":
 				case "moveFrom":
 				case "moveTo":
-				case "bidiWrapper":
+				case "inlineWrapper":
 					visitContent(item.content);
 					break;
 				case "run":
@@ -83,7 +84,8 @@ const visitInlineContentSlots = (paragraph, visit) => {
 				case "moveFromRangeEnd":
 				case "moveToRangeStart":
 				case "moveToRangeEnd":
-				case "mathEquation": break;
+				case "mathEquation":
+				case "preservedInline": break;
 				default: panic(`Unsupported paragraph content: ${JSON.stringify(item)}`);
 			}
 		}
@@ -145,6 +147,7 @@ const visitDocxParagraphs = ({ documentBody, headers, footers, footnotes, endnot
 			visitTable(block);
 			return;
 		}
+		if (block.type === "preservedBlock") return;
 		visitBlocks(block.content);
 	};
 	const visitBlocks = (blocks) => {

package/dist/docx/parseWarningMessage.js

@@ -36,7 +36,8 @@ const PARSE_WARNING_MESSAGES = {
 	[PARSE_WARNING_CODES.unrecognisedOnOffValue]: (warning) => `Ignored on/off value${quoted(warning.value)}${where(warning)}; ST_OnOff is 1, 0, true, false, on or off.`,
 	[PARSE_WARNING_CODES.borderWithoutValue]: (warning) => `Read a border with no w:val${where(warning)} as having no border style.`,
 	[PARSE_WARNING_CODES.styleSetDuplicateStyleId]: (warning) => `Dropped a style repeating the id${quoted(warning.value)} another style in the set already defines.`,
-	[PARSE_WARNING_CODES.styleSetInitialStyleMissing]: (warning) => `The style set names initial paragraph style${quoted(warning.value)}, which it does not contain; used the set's default instead${where(warning)}.`
+	[PARSE_WARNING_CODES.styleSetInitialStyleMissing]: (warning) => `The style set names initial paragraph style${quoted(warning.value)}, which it does not contain; used the set's default instead${where(warning)}.`,
+	[PARSE_WARNING_CODES.pageBreakProjectionApproximated]: (warning) => `${warning.detail ?? "An explicit page break is laid out approximately"}${where(warning)}.`
 };
 const formatParseWarning = (warning) => PARSE_WARNING_MESSAGES[warning.code](warning);
 const formatParseWarnings = (warnings) => warnings.map(formatParseWarning);

package/dist/docx/parser.js

@@ -23,6 +23,7 @@ import { UNNUMBERED_PARAGRAPH_WARNING, UNNUMBERED_STYLE_WARNING, normalizeNumber
 import { assignDocumentParagraphPropertySourceContract } from "./paragraphPropertySource.js";
 import { createParseWarningCollector } from "./parseContext.js";
 import { formatParseWarnings } from "./parseWarningMessage.js";
+import { enforcePackagePreviewBudget } from "./previewBudget.js";
 import { RELATIONSHIP_TYPES, parseRelationships, resolveRelativePath } from "./relsParser.js";
 import { normalizeRenderedPageBreakHints } from "./renderedPageBreakNormalization.js";
 import { parseSettings } from "./settingsParser.js";
@@ -30,7 +31,6 @@ import { parseStylesPackage } from "./styleParser.js";
 import { applyThemeFontLang, parseTheme } from "./themeParser.js";
 import { UNBALANCED_MOVE_RANGE_WARNING, normalizeTrackedMoveRanges } from "./trackedMoveRangeNormalization.js";
 import { getMediaMimeType, mediaToDataUrl, unzipDocx } from "./unzip.js";
-import { enforcePackageVmlPreviewBudget } from "./vmlPreview.js";
 import { FOLIO_XML_RESOURCE_LIMITS } from "./xmlResourceLimits.js";
 import { TaggedError } from "better-result";
 import { PARSE_WARNING_CODES } from "@stll/docx-core/model";
@@ -281,7 +281,7 @@ async function parseDocx(input, options = {}) {
 			...requiredFonts.length > 0 ? { requiredFonts } : {}
 		};
 		assignDocumentParagraphPropertySourceContract(document, await paragraphPropertySourceDigest);
-		enforcePackageVmlPreviewBudget(document.package);
+		enforcePackagePreviewBudget(document.package);
 		const validation = validateFolioDocumentModel(document);
 		const parsedCompleteModel = parseHeadersFooters && parseNotes;
 		if (!validation.valid && parsedCompleteModel) throw new DocxModelValidationError("Parsed DOCX produced an invalid document model", validation.issues);

package/dist/docx/preservedRunContent.d.ts

@@ -0,0 +1,32 @@
+import { document_d_exports } from "../types/document.js";
+import { XmlElement } from "./xmlParser.js";
+//#region src/docx/preservedRunContent.d.ts
+/**
+ * The visible text a preserved run child contributes, or `""` when it shows
+ * nothing. Unknown markup contributes nothing on purpose: guessing at the text
+ * of an element folio has never seen would put invented words in a document.
+ */
+declare const preservedRunChildText: (element: XmlElement) => string;
+/**
+ * Capture one unmodelled run child.
+ *
+ * `captureVerbatimXml` materialises exactly the namespace bindings the
+ * fragment's own prefixes need and leaves the canonical ones to the rebuilt
+ * root, so ordinary `w:` markup comes back byte for byte while a foreign
+ * prefix still arrives bound. Copying the whole root scope onto the fragment
+ * instead would rewrite every capture with declarations it does not use.
+ */
+declare const preserveRunChild: (element: XmlElement) => document_d_exports.PreservedXmlContent;
+/**
+ * One capture from the shared dispatcher's sink, as a member of the
+ * container's own inline union.
+ *
+ * The sink holds markup the schema does not declare for the container at all,
+ * so nothing is known about it beyond its bytes: `text` is empty because a
+ * name folio has never seen shows nothing folio can read.
+ */
+declare const preservedInlineCapture: (xml: string) => document_d_exports.PreservedInline;
+/** Capture one unmodelled inline child of a paragraph or an inline wrapper. */
+declare const preserveInlineChild: (element: XmlElement) => document_d_exports.PreservedInline;
+//#endregion
+export { preserveInlineChild, preserveRunChild, preservedInlineCapture, preservedRunChildText };

package/dist/docx/preservedRunContent.js

@@ -0,0 +1,86 @@
+import { captureVerbatimXml } from "./verbatimCapture.js";
+import { findChildrenByLocalName, getChildElements, getLocalName, getTextContent } from "./xmlParser.js";
+//#region src/docx/preservedRunContent.ts
+/**
+* Where a preserved run child hides text a reader sees, by local name: the
+* child elements whose `w:t` descendants are on the line rather than above,
+* beside, or nowhere.
+*
+* `w:ruby` is the whole list today. Its `w:rt` is the annotation printed above
+* the base and is not the sentence's text; its `w:rubyBase` is.
+*/
+const VISIBLE_TEXT_CHILDREN = /* @__PURE__ */ new Map([["ruby", ["rubyBase"]]]);
+/** Cap on the text one preserved child contributes, mirroring `w:t` handling. */
+const MAX_PRESERVED_TEXT_LENGTH = 1e5;
+const collectTextContent = (element, into) => {
+	for (const child of getChildElements(element)) {
+		if (getLocalName(child.name) === "t") {
+			into.push(getTextContent(child));
+			continue;
+		}
+		collectTextContent(child, into);
+	}
+};
+/**
+* The visible text a preserved run child contributes, or `""` when it shows
+* nothing. Unknown markup contributes nothing on purpose: guessing at the text
+* of an element folio has never seen would put invented words in a document.
+*/
+const preservedRunChildText = (element) => {
+	const sources = VISIBLE_TEXT_CHILDREN.get(getLocalName(element.name));
+	if (sources === void 0) return "";
+	const parts = [];
+	for (const localName of sources) for (const source of findChildrenByLocalName(element, localName)) collectTextContent(source, parts);
+	return parts.join("").slice(0, MAX_PRESERVED_TEXT_LENGTH);
+};
+/**
+* Capture one unmodelled run child.
+*
+* `captureVerbatimXml` materialises exactly the namespace bindings the
+* fragment's own prefixes need and leaves the canonical ones to the rebuilt
+* root, so ordinary `w:` markup comes back byte for byte while a foreign
+* prefix still arrives bound. Copying the whole root scope onto the fragment
+* instead would rewrite every capture with declarations it does not use.
+*/
+const preserveRunChild = (element) => ({
+	type: "preservedXml",
+	xml: captureVerbatimXml(element),
+	text: preservedRunChildText(element)
+});
+/**
+* Where an unmodelled *inline* child hides text a reader sees, by local name.
+*
+* `w:customXml` and `w:smartTag` are transparent wrappers (ECMA-376 §17.5.1,
+* §17.5.1.9): their content is ordinary inline content, so their `w:t`
+* descendants are on the line. Every other inline child folio captures —
+* `w:permStart`, `w:proofErr` and the custom-XML revision ranges — is an
+* empty marker, and an element folio has never seen contributes nothing on
+* purpose, because guessing at its text would put invented words in a
+* document.
+*/
+const VISIBLE_TEXT_INLINE_CHILDREN = /* @__PURE__ */ new Set(["customXml", "smartTag"]);
+/**
+* One capture from the shared dispatcher's sink, as a member of the
+* container's own inline union.
+*
+* The sink holds markup the schema does not declare for the container at all,
+* so nothing is known about it beyond its bytes: `text` is empty because a
+* name folio has never seen shows nothing folio can read.
+*/
+const preservedInlineCapture = (xml) => ({
+	type: "preservedInline",
+	xml,
+	text: ""
+});
+/** Capture one unmodelled inline child of a paragraph or an inline wrapper. */
+const preserveInlineChild = (element) => {
+	const parts = [];
+	if (VISIBLE_TEXT_INLINE_CHILDREN.has(getLocalName(element.name))) collectTextContent(element, parts);
+	return {
+		type: "preservedInline",
+		xml: captureVerbatimXml(element),
+		text: parts.join("").slice(0, MAX_PRESERVED_TEXT_LENGTH)
+	};
+};
+//#endregion
+export { preserveInlineChild, preserveRunChild, preservedInlineCapture, preservedRunChildText };

package/dist/docx/previewBudget.d.ts

@@ -0,0 +1,64 @@
+//#region src/docx/previewBudget.d.ts
+/**
+ * Every synthetic preview a parse attaches to the model, and the budget each
+ * one answers to.
+ *
+ * A preview is not document content. It is a drawing folio makes up so a shape
+ * it cannot project still occupies the page, and the package round-trips from
+ * its preserved XML whether the preview exists or not. So a preview is the one
+ * thing in the model that may be dropped, and a package that would retain more
+ * of it than it is worth has it dropped rather than being refused.
+ *
+ * Producer and budget have to agree on what a preview looks like. They used to
+ * agree by coincidence: the VML producer wrote its filename as a literal and
+ * the budget recognized it as a constant in another file, so renaming one
+ * would have stopped the other charging for it without anything failing. The
+ * table below is that agreement written once, and a producer builds its image
+ * out of the entry the budget matches against.
+ */
+declare const VML_PREVIEW_DATA_URL_PREFIX = "data:image/svg+xml;charset=utf-8,";
+declare const PREVIEW_KINDS: {
+  /** A VML shape folio renders rather than projects (`v:shape`, `v:rect`, ...). */
+  readonly vmlShape: {
+    readonly mimeType: "image/svg+xml";
+    readonly filename: "vml-shape-preview.svg";
+    readonly srcPrefix: "data:image/svg+xml;charset=utf-8,";
+    readonly maxPackageCharacters: number;
+  };
+  /**
+   * A SmartArt diagram: its extent filled with one flat rectangle per shape.
+   *
+   * A raster rather than a vector because the display list decodes only base64
+   * PNG and JPEG, so a vector preview would be missing from every display-list
+   * backend (PDF among them) while still showing in the DOM. That is what
+   * makes this preview expensive: one of them is 7.3 MB of data URL whatever
+   * the package weighs, because its cost follows the extent the author chose
+   * rather than anything the drawing contains.
+   *
+   * Across the public corpus, the fifty packages that produce one retain a
+   * median of 7.3 MB and a maximum of 51.3 MB (ten previews, from a package
+   * under a megabyte). The cap is set above that maximum: it refuses no
+   * legitimate file in the corpus while bounding what had no bound at all, and
+   * it is a ceiling rather than a fix. The fix is for the preview to be a
+   * descriptor the renderer rasterizes, which needs the display-list contract
+   * to carry one.
+   */
+  readonly smartArt: {
+    readonly mimeType: "image/png";
+    readonly filename: "smartart-preview.png";
+    readonly srcPrefix: "data:image/png;base64,";
+    readonly maxPackageCharacters: number;
+  };
+};
+type PreviewKindName = keyof typeof PREVIEW_KINDS;
+/** Per-kind allowances for one package, defaulting to the table's caps. */
+type PreviewBudgetOverrides = Partial<Record<PreviewKindName, number>>;
+/**
+ * Charge every generated preview in the model against its kind's allowance and
+ * drop the `src` of those past it. Dropping leaves the image in place with its
+ * size and wrap, so the page still reserves the space the drawing occupies,
+ * and never touches the preserved XML the package saves from.
+ */
+declare const enforcePackagePreviewBudget: (root: unknown, overrides?: PreviewBudgetOverrides) => void;
+//#endregion
+export { PREVIEW_KINDS, PreviewBudgetOverrides, VML_PREVIEW_DATA_URL_PREFIX, enforcePackagePreviewBudget };

package/dist/docx/previewBudget.js

@@ -0,0 +1,88 @@
+//#region src/docx/previewBudget.ts
+const VML_PREVIEW_DATA_URL_PREFIX = "data:image/svg+xml;charset=utf-8,";
+const MEBIBYTE = 1024 * 1024;
+const PREVIEW_KINDS = {
+	/** A VML shape folio renders rather than projects (`v:shape`, `v:rect`, ...). */
+	vmlShape: {
+		mimeType: "image/svg+xml",
+		filename: "vml-shape-preview.svg",
+		srcPrefix: VML_PREVIEW_DATA_URL_PREFIX,
+		maxPackageCharacters: 8 * MEBIBYTE
+	},
+	/**
+	* A SmartArt diagram: its extent filled with one flat rectangle per shape.
+	*
+	* A raster rather than a vector because the display list decodes only base64
+	* PNG and JPEG, so a vector preview would be missing from every display-list
+	* backend (PDF among them) while still showing in the DOM. That is what
+	* makes this preview expensive: one of them is 7.3 MB of data URL whatever
+	* the package weighs, because its cost follows the extent the author chose
+	* rather than anything the drawing contains.
+	*
+	* Across the public corpus, the fifty packages that produce one retain a
+	* median of 7.3 MB and a maximum of 51.3 MB (ten previews, from a package
+	* under a megabyte). The cap is set above that maximum: it refuses no
+	* legitimate file in the corpus while bounding what had no bound at all, and
+	* it is a ceiling rather than a fix. The fix is for the preview to be a
+	* descriptor the renderer rasterizes, which needs the display-list contract
+	* to carry one.
+	*/
+	smartArt: {
+		mimeType: "image/png",
+		filename: "smartart-preview.png",
+		srcPrefix: "data:image/png;base64,",
+		maxPackageCharacters: 64 * MEBIBYTE
+	}
+};
+const KIND_NAMES = Object.keys(PREVIEW_KINDS);
+/**
+* The kind a model image was generated as, or `undefined` for one the package
+* actually carries. A generated preview has no relationship behind it, so
+* `rId` is empty; the filename and data-URL prefix name which producer made it.
+*/
+const previewKindOf = (value) => {
+	if (!("type" in value) || value.type !== "image" || !("rId" in value) || value.rId !== "" || !("src" in value) || typeof value.src !== "string" || !("mimeType" in value) || !("filename" in value)) return;
+	const { src, mimeType, filename } = value;
+	return KIND_NAMES.find((name) => {
+		const kind = PREVIEW_KINDS[name];
+		return mimeType === kind.mimeType && filename === kind.filename && src.startsWith(kind.srcPrefix);
+	});
+};
+/**
+* Charge every generated preview in the model against its kind's allowance and
+* drop the `src` of those past it. Dropping leaves the image in place with its
+* size and wrap, so the page still reserves the space the drawing occupies,
+* and never touches the preserved XML the package saves from.
+*/
+const enforcePackagePreviewBudget = (root, overrides = {}) => {
+	const remaining = new Map(KIND_NAMES.map((name) => [name, Math.max(0, overrides[name] ?? PREVIEW_KINDS[name].maxPackageCharacters)]));
+	const visited = /* @__PURE__ */ new WeakSet();
+	const visit = (value) => {
+		if (value === null || typeof value !== "object" || visited.has(value)) return;
+		visited.add(value);
+		if (value instanceof ArrayBuffer || ArrayBuffer.isView(value)) return;
+		if (value instanceof Map) {
+			for (const child of value.values()) visit(child);
+			return;
+		}
+		if (Array.isArray(value)) {
+			for (const child of value) visit(child);
+			return;
+		}
+		const kind = previewKindOf(value);
+		if (kind !== void 0) {
+			const image = value;
+			const length = image.src.length;
+			const left = remaining.get(kind);
+			if (length <= left) remaining.set(kind, left - length);
+			else {
+				remaining.set(kind, 0);
+				delete image.src;
+			}
+		}
+		for (const child of Object.values(value)) visit(child);
+	};
+	visit(root);
+};
+//#endregion
+export { PREVIEW_KINDS, VML_PREVIEW_DATA_URL_PREFIX, enforcePackagePreviewBudget };

package/dist/docx/renderedPageBreakNormalization.js

@@ -37,9 +37,7 @@ const scanHyperlink = (hyperlink, scan) => {
 };
 const scanInlineContent = (content, scan) => {
 	for (const child of content) {
-		let result;
-		if (child.type === "run") result = scanRun(child, scan);
-		else if (child.type === "hyperlink") result = scanHyperlink(child, scan);
+		const result = scanParagraphContent(child, scan);
 		if (result !== void 0) return result;
 	}
 };
@@ -48,7 +46,7 @@ const scanParagraphContent = (content, scan) => {
 	if (content.type === "hyperlink") return scanHyperlink(content, scan);
 	if (content.type === "simpleField") return scanInlineContent(content.content, scan);
 	if (content.type === "insertion" || content.type === "deletion" || content.type === "moveFrom" || content.type === "moveTo") return scanInlineContent(content.content, scan);
-	if (content.type === "inlineSdt") {
+	if (content.type === "inlineSdt" || content.type === "inlineWrapper") {
 		for (const child of content.content) {
 			const result = scanParagraphContent(child, scan);
 			if (result !== void 0) return result;

package/dist/docx/revisionIdNormalization.js

@@ -117,7 +117,10 @@ const normalizeRevisionIdsInXmlParts = (parts) => {
 	const occurrencesByPath = /* @__PURE__ */ new Map();
 	const reserved = /* @__PURE__ */ new Set();
 	for (const [path, xml] of candidates) {
-		assertXmlResourceLimits(xml);
+		assertXmlResourceLimits({
+			xml,
+			partPath: path
+		});
 		const ids = [];
 		if (rewriteStreamingXmlDecimalAttributes(xml, (element) => {
 			const identified = identifiedElement(element);
@@ -127,7 +130,9 @@ const normalizeRevisionIdsInXmlParts = (parts) => {
 			return null;
 		}).status === "unsupported") throw new XmlResourceLimitError({
 			message: `Revision-id normalization could not safely scan ${path}`,
-			limit: "syntax"
+			limit: "syntax",
+			observed: 0,
+			allowed: 0
 		});
 		occurrencesByPath.set(path, ids);
 	}
@@ -137,14 +142,19 @@ const normalizeRevisionIdsInXmlParts = (parts) => {
 	else firstSeen.add(id);
 	if (repeatedPaths.size > 0) for (const [path, xml] of parts) {
 		if (occurrencesByPath.has(path) || !ANNOTATION_ELEMENT_CANDIDATE.test(xml)) continue;
-		assertXmlResourceLimits(xml);
+		assertXmlResourceLimits({
+			xml,
+			partPath: path
+		});
 		if (rewriteStreamingXmlDecimalAttributes(xml, (element) => {
 			const identified = identifiedElement(element);
 			if (identified !== null) reserved.add(identified.id);
 			return null;
 		}).status === "unsupported") throw new XmlResourceLimitError({
 			message: `Revision-id normalization could not safely scan ${path}`,
-			limit: "syntax"
+			limit: "syntax",
+			observed: 0,
+			allowed: 0
 		});
 	}
 	let nextId = 0;
@@ -185,7 +195,9 @@ const normalizeRevisionIdsInXmlParts = (parts) => {
 		});
 		if (rewritten.status === "unsupported") throw new XmlResourceLimitError({
 			message: `Revision-id normalization could not safely rewrite ${path}`,
-			limit: "syntax"
+			limit: "syntax",
+			observed: 0,
+			allowed: 0
 		});
 		normalized.set(path, rewritten.value);
 	}