@stll/folio-core 0.43.0 → 0.45.0

package/dist/docx/tableParser.js

@@ -1,11 +1,16 @@
+import { attachPendingRangeMarkers, attachTrailingRangeMarkers, isBlockRangeMarker } from "./blockRangeMarkers.js";
 import { parseBookmarkEnd, parseBookmarkStart } from "./bookmarkParser.js";
 import { appendBookmarkMarkerToLastParagraphInBlocks, appendBookmarkMarkerToLastParagraphInCells, prependBookmarkMarkersToFirstParagraphInBlocks, prependBookmarkMarkersToFirstParagraphInCell } from "./bookmarkPlacement.js";
+import { parseBorderSpec } from "./borderParser.js";
 import { parseParagraph } from "./paragraphParser.js";
 import { enrichParagraphTextBoxes } from "./paragraphTextBoxEnrichment.js";
-import { BorderStyleSchema, FloatingTableXSpecSchema, FloatingTableYSpecSchema, ShadingPatternSchema, TableCellTextDirectionSchema, ThemeColorSlotSchema, narrowEnum } from "./parserEnums.js";
+import { FloatingTableXSpecSchema, FloatingTableYSpecSchema, TableCellTextDirectionSchema, narrowEnum } from "./parserEnums.js";
+import { parseShading } from "./shadingParser.js";
+import { TABLE_LOOK_FLAGS } from "./tableLook.js";
 import { parsePropertyChangeInfo, parseTrackedChangeInfo } from "./trackedChangeInfo.js";
+import { percentageSpelling, transitionalSlotEncoding } from "./transitionalSpelling.js";
 import { captureVerbatimXml } from "./verbatimCapture.js";
-import { cloneElement, findChild, findChildByLocalName, findChildren, getAttribute, getChildElements, getLocalName, mergeXmlnsDeclarations, parseBooleanElement, parseNumericAttribute, parseTableMeasurementValue, selectAlternateContentBranch } from "./xmlParser.js";
+import { cloneElement, findChild, findChildByLocalName, findChildren, getAttribute, getChildElements, getLocalName, mergeXmlnsDeclarations, parseBooleanElement, parseNumericAttribute, parseOnOffAttribute, parseTableMeasurementValue, selectAlternateContentBranch } from "./xmlParser.js";
 //#region src/docx/tableParser.ts
 /**
 * Sanity cap on `w:gridSpan` (and the derived table column count). Word's
@@ -22,15 +27,32 @@ const MAX_TABLE_COLUMNS = 63;
 */
 function parseTableMeasurement(element) {
 	if (!element) return;
-	const typeStr = getAttribute(element, "w", "type") ?? "dxa";
-	let type = "dxa";
-	if (typeStr === "auto" || typeStr === "dxa" || typeStr === "nil" || typeStr === "pct") type = typeStr;
+	const declared = getAttribute(element, "w", "type");
+	const type = tableWidthType(element, declared === "auto" || declared === "dxa" || declared === "nil" || declared === "pct" ? declared : void 0);
 	return {
 		value: parseTableMeasurementValue(element, type) ?? 0,
 		type
 	};
 }
 /**
+* What unit `w:w` counts in.
+*
+* `w:type` is optional on `CT_TblWidth` and the schema gives it no default, so
+* reading an absent one as `dxa` turned `w:w="50%"` into 50 twips: a
+* full-width table became a hairline. `w:w` is `ST_MeasurementOrPercent`, so a
+* value spelled with a `%` is a percentage whatever `w:type` says — the
+* spelling carries its own unit, and no number of twips is ever written that
+* way. `auto` and `nil` are left alone: neither reads `w:w` as a width at all.
+*/
+const tableWidthType = (element, declared) => {
+	if (declared === "auto" || declared === "nil") return declared;
+	const raw = getAttribute(element, "w", "w");
+	const spelledAsPercent = raw !== null && percentageSpelling(raw) !== void 0;
+	const slotTakesPercent = transitionalSlotEncoding(element.namespaceUri, getLocalName(element.name), "w")?.percent !== void 0;
+	if (spelledAsPercent && slotTakesPercent) return "pct";
+	return declared ?? "dxa";
+};
+/**
 * Parse width from an element (shorthand for common case)
 */
 function parseWidth(element) {
@@ -42,34 +64,6 @@ function parseWidth(element) {
 * @param element - Border element (w:top, w:bottom, etc.)
 * @returns Parsed border or undefined
 */
-function parseBorderSpec(element) {
-	if (!element) return;
-	const rawStyle = getAttribute(element, "w", "val") ?? "none";
-	const border = { style: narrowEnum(rawStyle, BorderStyleSchema) ?? rawStyle };
-	const sz = parseNumericAttribute(element, "w", "sz");
-	if (sz !== void 0) border.size = sz;
-	const space = parseNumericAttribute(element, "w", "space");
-	if (space !== void 0) border.space = space;
-	const color = getAttribute(element, "w", "color");
-	const themeColor = getAttribute(element, "w", "themeColor");
-	const themeTint = getAttribute(element, "w", "themeTint");
-	const themeShade = getAttribute(element, "w", "themeShade");
-	if (color || themeColor || themeTint || themeShade) {
-		const colorVal = {};
-		if (color === "auto") colorVal.auto = true;
-		else if (color !== null) colorVal.rgb = color;
-		const validatedThemeColor = narrowEnum(themeColor, ThemeColorSlotSchema);
-		if (validatedThemeColor) colorVal.themeColor = validatedThemeColor;
-		if (themeTint !== null) colorVal.themeTint = themeTint;
-		if (themeShade !== null) colorVal.themeShade = themeShade;
-		border.color = colorVal;
-	}
-	const shadow = getAttribute(element, "w", "shadow");
-	if (shadow === "1" || shadow === "true") border.shadow = true;
-	const frame = getAttribute(element, "w", "frame");
-	if (frame === "1" || frame === "true") border.frame = true;
-	return border;
-}
 /**
 * Parse table borders (w:tblBorders or w:tcBorders)
 *
@@ -125,33 +119,13 @@ function parseCellMargins(marginsElement) {
 	return margins;
 }
 /**
-* Parse shading properties (w:shd)
+* Read a `w:tblLook` (the only reader; `styleParser` calls this one).
 *
-* @param shdElement - The w:shd element
-* @returns Parsed shading or undefined
-*/
-function parseShading(shdElement) {
-	if (!shdElement) return;
-	const shading = {};
-	const fillStr = getAttribute(shdElement, "w", "fill");
-	if (fillStr && fillStr !== "auto") shading.fill = { rgb: fillStr };
-	const themeFill = narrowEnum(getAttribute(shdElement, "w", "themeFill"), ThemeColorSlotSchema);
-	if (themeFill) {
-		shading.fill = { themeColor: themeFill };
-		const themeFillTint = getAttribute(shdElement, "w", "themeFillTint");
-		if (themeFillTint) shading.fill.themeTint = themeFillTint;
-		const themeFillShade = getAttribute(shdElement, "w", "themeFillShade");
-		if (themeFillShade) shading.fill.themeShade = themeFillShade;
-	}
-	const colorStr = getAttribute(shdElement, "w", "color");
-	if (colorStr && colorStr !== "auto") shading.color = { rgb: colorStr };
-	const pattern = narrowEnum(getAttribute(shdElement, "w", "val"), ShadingPatternSchema);
-	if (pattern) shading.pattern = pattern;
-	if (Object.keys(shading).length === 0) return;
-	return shading;
-}
-/**
-* Parse table look flags (w:tblLook)
+* What the author wrote, and nothing more: `w:val` verbatim and each flag as
+* stated, absent, `false` or `true`. Folding `w:val`'s bits into the flags here
+* would forget which of the two the document said, and writing the result back
+* would invent attributes the author never had. `resolveTableLook` owns the
+* other direction.
 *
 * @param lookElement - The w:tblLook element
 * @returns Parsed table look or undefined
@@ -159,29 +133,11 @@ function parseShading(shdElement) {
 function parseTableLook(lookElement) {
 	if (!lookElement) return;
 	const look = {};
-	const firstRow = getAttribute(lookElement, "w", "firstRow");
-	if (firstRow === "1" || firstRow === "true") look.firstRow = true;
-	const lastRow = getAttribute(lookElement, "w", "lastRow");
-	if (lastRow === "1" || lastRow === "true") look.lastRow = true;
-	const firstColumn = getAttribute(lookElement, "w", "firstColumn");
-	if (firstColumn === "1" || firstColumn === "true") look.firstColumn = true;
-	const lastColumn = getAttribute(lookElement, "w", "lastColumn");
-	if (lastColumn === "1" || lastColumn === "true") look.lastColumn = true;
-	const noHBand = getAttribute(lookElement, "w", "noHBand");
-	if (noHBand === "1" || noHBand === "true") look.noHBand = true;
-	const noVBand = getAttribute(lookElement, "w", "noVBand");
-	if (noVBand === "1" || noVBand === "true") look.noVBand = true;
 	const val = getAttribute(lookElement, "w", "val");
-	if (val) {
-		const flags = Number.parseInt(val, 16);
-		if (!Number.isNaN(flags)) {
-			if (flags & 32) look.firstRow = true;
-			if (flags & 64) look.lastRow = true;
-			if (flags & 128) look.firstColumn = true;
-			if (flags & 256) look.lastColumn = true;
-			if (flags & 512) look.noHBand = true;
-			if (flags & 1024) look.noVBand = true;
-		}
+	if (val !== null && val !== "") look.val = val;
+	for (const flag of TABLE_LOOK_FLAGS) {
+		const stated = parseOnOffAttribute(lookElement, "w", flag);
+		if (stated !== void 0) look[flag] = stated;
 	}
 	if (Object.keys(look).length === 0) return;
 	return look;
@@ -430,30 +386,18 @@ function parseTableRowProperties(trPrElement) {
 function parseConditionalFormatStyle(cnfElement) {
 	if (!cnfElement) return;
 	const style = {};
-	const firstRow = getAttribute(cnfElement, "w", "firstRow");
-	if (firstRow === "1" || firstRow === "true") style.firstRow = true;
-	const lastRow = getAttribute(cnfElement, "w", "lastRow");
-	if (lastRow === "1" || lastRow === "true") style.lastRow = true;
-	const firstColumn = getAttribute(cnfElement, "w", "firstColumn");
-	if (firstColumn === "1" || firstColumn === "true") style.firstColumn = true;
-	const lastColumn = getAttribute(cnfElement, "w", "lastColumn");
-	if (lastColumn === "1" || lastColumn === "true") style.lastColumn = true;
-	const oddHBand = getAttribute(cnfElement, "w", "oddHBand");
-	if (oddHBand === "1" || oddHBand === "true") style.oddHBand = true;
-	const evenHBand = getAttribute(cnfElement, "w", "evenHBand");
-	if (evenHBand === "1" || evenHBand === "true") style.evenHBand = true;
-	const oddVBand = getAttribute(cnfElement, "w", "oddVBand");
-	if (oddVBand === "1" || oddVBand === "true") style.oddVBand = true;
-	const evenVBand = getAttribute(cnfElement, "w", "evenVBand");
-	if (evenVBand === "1" || evenVBand === "true") style.evenVBand = true;
-	const nwCell = getAttribute(cnfElement, "w", "firstRowFirstColumn");
-	if (nwCell === "1" || nwCell === "true") style.nwCell = true;
-	const neCell = getAttribute(cnfElement, "w", "firstRowLastColumn");
-	if (neCell === "1" || neCell === "true") style.neCell = true;
-	const swCell = getAttribute(cnfElement, "w", "lastRowFirstColumn");
-	if (swCell === "1" || swCell === "true") style.swCell = true;
-	const seCell = getAttribute(cnfElement, "w", "lastRowLastColumn");
-	if (seCell === "1" || seCell === "true") style.seCell = true;
+	if (parseOnOffAttribute(cnfElement, "w", "firstRow") === true) style.firstRow = true;
+	if (parseOnOffAttribute(cnfElement, "w", "lastRow") === true) style.lastRow = true;
+	if (parseOnOffAttribute(cnfElement, "w", "firstColumn") === true) style.firstColumn = true;
+	if (parseOnOffAttribute(cnfElement, "w", "lastColumn") === true) style.lastColumn = true;
+	if (parseOnOffAttribute(cnfElement, "w", "oddHBand") === true) style.oddHBand = true;
+	if (parseOnOffAttribute(cnfElement, "w", "evenHBand") === true) style.evenHBand = true;
+	if (parseOnOffAttribute(cnfElement, "w", "oddVBand") === true) style.oddVBand = true;
+	if (parseOnOffAttribute(cnfElement, "w", "evenVBand") === true) style.evenVBand = true;
+	if (parseOnOffAttribute(cnfElement, "w", "firstRowFirstColumn") === true) style.nwCell = true;
+	if (parseOnOffAttribute(cnfElement, "w", "firstRowLastColumn") === true) style.neCell = true;
+	if (parseOnOffAttribute(cnfElement, "w", "lastRowFirstColumn") === true) style.swCell = true;
+	if (parseOnOffAttribute(cnfElement, "w", "lastRowLastColumn") === true) style.seCell = true;
 	const val = getAttribute(cnfElement, "w", "val");
 	if (val && val.length === 12) {
 		if (val[0] === "1") style.firstRow = true;
@@ -530,6 +474,7 @@ function withContainerXmlns(options, element) {
 function parseCellContent(tcElement, styles, theme, numbering, rels, media, options) {
 	const content = [];
 	const pendingBookmarkMarkers = [];
+	const pendingRangeMarkers = [];
 	const elements = getChildElements(tcElement);
 	const parseCellChild = (child, childOptions = options) => {
 		if (!child.name) return;
@@ -538,6 +483,7 @@ function parseCellContent(tcElement, styles, theme, numbering, rels, media, opti
 			const para = parseParagraph(child, styles, theme, numbering, rels, media, childOptions);
 			enrichParagraphTextBoxes(para, child, styles, theme, numbering, rels, media, parseTable);
 			prependPendingBookmarkMarkers(para, pendingBookmarkMarkers);
+			attachPendingRangeMarkers(para, pendingRangeMarkers);
 			content.push(para);
 			return;
 		}
@@ -545,6 +491,7 @@ function parseCellContent(tcElement, styles, theme, numbering, rels, media, opti
 			const table = parseTable(child, styles, theme, numbering, rels, media, childOptions);
 			if (!table) return;
 			if (prependBookmarkMarkersToFirstParagraphInBlocks([table], pendingBookmarkMarkers)) pendingBookmarkMarkers.length = 0;
+			attachPendingRangeMarkers(table, pendingRangeMarkers);
 			content.push(table);
 			return;
 		}
@@ -565,7 +512,9 @@ function parseCellContent(tcElement, styles, theme, numbering, rels, media, opti
 		if (localName === "bookmarkStart" || localName === "bookmarkEnd") {
 			const marker = parseBookmarkMarker(child, localName);
 			if (!appendBookmarkMarkerToLastParagraphInBlocks(content, marker)) pendingBookmarkMarkers.push(marker);
+			return;
 		}
+		if (isBlockRangeMarker(localName)) pendingRangeMarkers.push(captureVerbatimXml(child));
 	};
 	for (const child of elements) parseCellChild(child);
 	if (content.length === 0) content.push({
@@ -573,6 +522,7 @@ function parseCellContent(tcElement, styles, theme, numbering, rels, media, opti
 		content: [...pendingBookmarkMarkers]
 	});
 	else if (pendingBookmarkMarkers.length > 0) appendBookmarkMarkersToLastParagraphInBlocks(content, pendingBookmarkMarkers);
+	attachTrailingRangeMarkers(content, pendingRangeMarkers);
 	return content;
 }
 /**
@@ -747,10 +697,14 @@ function parseTable(tblElement, styles, theme, numbering, rels, media, options)
 	const gridElement = findChild(tblElement, "w", "tblGrid");
 	const columnWidths = parseTableGrid(gridElement);
 	if (columnWidths) table.columnWidths = columnWidths;
-	if (gridElement) table.formatting = {
-		...table.formatting,
-		gridSourceXml: captureVerbatimXml(gridElement)
-	};
+	if (gridElement) {
+		const gridChange = findChild(gridElement, "w", "tblGridChange");
+		table.formatting = {
+			...table.formatting,
+			gridSourceXml: captureVerbatimXml(gridElement),
+			...gridChange ? { gridChangeXml: captureVerbatimXml(gridChange) } : {}
+		};
+	}
 	const tableOptions = withContainerXmlns(options, tblElement);
 	const rowsWithGridOffsets = /* @__PURE__ */ new Set();
 	const parseTableChild = (child, childOptions = tableOptions) => {
@@ -878,4 +832,4 @@ function isFloatingTable(table) {
 	return table.formatting?.floating !== void 0;
 }
 //#endregion
-export { MAX_TABLE_COLUMNS, getHeaderRows, getTableColumnCount, getTableRowCount, getTableText, hasHeaderRow, isCellHorizontallyMerged, isCellMergeContinuation, isCellMergeStart, isFloatingTable, parseBorderSpec, parseCellMargins, parseConditionalFormatStyle, parseFloatingTableProperties, parseShading, parseTable, parseTableBorders, parseTableCell, parseTableCellProperties, parseTableGrid, parseTableLook, parseTableMeasurement, parseTableProperties, parseTableRow, parseTableRowProperties };
+export { MAX_TABLE_COLUMNS, getHeaderRows, getTableColumnCount, getTableRowCount, getTableText, hasHeaderRow, isCellHorizontallyMerged, isCellMergeContinuation, isCellMergeStart, isFloatingTable, parseCellMargins, parseConditionalFormatStyle, parseFloatingTableProperties, parseTable, parseTableBorders, parseTableCell, parseTableCellProperties, parseTableGrid, parseTableLook, parseTableMeasurement, parseTableProperties, parseTableRow, parseTableRowProperties };

package/dist/docx/textBoxParser.js

@@ -1,6 +1,7 @@
 import { emuToPixels } from "../utils/units.js";
 import { parseAnchorPosition, parseAnchorWrap, parseFill, parseOutline, resolveColorValueToHex } from "./drawingUtils.js";
-import { findByFullName, findChildByLocalName, findChildByNamespaceUri, findChildrenByLocalName, findChildrenByNamespaceUri, getAttribute, getChildElements, parseNumericAttribute } from "./xmlParser.js";
+import { parseNonVisualDrawingNames } from "./nonVisualDrawingProps.js";
+import { findByFullName, findChildByLocalName, findChildByNamespaceUri, findChildrenByLocalName, findChildrenByNamespaceUri, getAttribute, getChildElements, parseNumericAttribute, parseOnOffAttribute } from "./xmlParser.js";
 //#region src/docx/textBoxParser.ts
 const DRAWINGML_MAIN_NAMESPACE_URIS = /* @__PURE__ */ new Set(["http://schemas.openxmlformats.org/drawingml/2006/main", "http://purl.oclc.org/ooxml/drawingml/main"]);
 const WORDPROCESSING_SHAPE_NAMESPACE_URIS = /* @__PURE__ */ new Set(["http://schemas.microsoft.com/office/word/2010/wordprocessingShape"]);
@@ -30,9 +31,9 @@ function parseBodyProperties(bodyPr) {
 	else if (findChildByLocalName(bodyPr, "noAutofit")) result.autoFit = "none";
 	const verticalAlign = parseTextBoxVerticalAlign(getAttribute(bodyPr, null, "anchor"));
 	if (verticalAlign) result.verticalAlign = verticalAlign;
-	const fromWordArt = getAttribute(bodyPr, null, "fromWordArt");
+	const fromWordArt = parseOnOffAttribute(bodyPr, null, "fromWordArt");
 	const warp = findChildByNamespaceUri(bodyPr, DRAWINGML_MAIN_NAMESPACE_URIS, "prstTxWarp");
-	if (fromWordArt !== null || warp) {
+	if (fromWordArt !== void 0 || warp) {
 		const adjustments = warp ? findChildrenByNamespaceUri(findChildByNamespaceUri(warp, DRAWINGML_MAIN_NAMESPACE_URIS, "avLst"), DRAWINGML_MAIN_NAMESPACE_URIS, "gd").map((adjustment) => {
 			const name = getAttribute(adjustment, null, "name");
 			const formula = getAttribute(adjustment, null, "fmla");
@@ -43,7 +44,7 @@ function parseBodyProperties(bodyPr) {
 		}).filter((adjustment) => adjustment !== void 0) : [];
 		const preset = warp ? getAttribute(warp, null, "prst") : null;
 		result.wordArt = {
-			...fromWordArt !== null ? { fromWordArt: fromWordArt === "1" || fromWordArt === "true" || fromWordArt === "on" } : {},
+			...fromWordArt !== void 0 ? { fromWordArt } : {},
 			...preset !== null ? { preset } : {},
 			...adjustments.length > 0 ? { adjustments } : {}
 		};
@@ -148,13 +149,15 @@ function parseTextBox(drawingEl) {
 	};
 	const docPr = findByFullName(container, "wp:docPr");
 	const id = docPr ? getAttribute(docPr, null, "id") ?? void 0 : void 0;
+	const names = parseNonVisualDrawingNames(docPr);
 	const fill = parseFill(spPr ?? null);
 	const outline = parseOutline(spPr ?? null);
 	const bodyProps = parseBodyProperties(bodyPr ?? null);
 	const textBox = {
 		type: "textBox",
 		size,
-		content: []
+		content: [],
+		...names
 	};
 	if (id) textBox.id = id;
 	if (fill) textBox.fill = fill;
@@ -193,13 +196,15 @@ function parseTextBoxFromShape(wsp, size, position, wrap) {
 	const bodyPr = findChildByNamespaceUri(wsp, WORDPROCESSING_SHAPE_NAMESPACE_URIS, "bodyPr");
 	const cNvPr = wspChildren.find((el) => el.name === "wps:cNvPr");
 	const id = cNvPr ? getAttribute(cNvPr, null, "id") ?? void 0 : void 0;
+	const names = parseNonVisualDrawingNames(cNvPr);
 	const fill = parseFill(spPr ?? null);
 	const outline = parseOutline(spPr ?? null);
 	const bodyProps = parseBodyProperties(bodyPr ?? null);
 	const textBox = {
 		type: "textBox",
 		size,
-		content: []
+		content: [],
+		...names
 	};
 	if (id) textBox.id = id;
 	if (fill) textBox.fill = fill;

package/dist/docx/trackedMoveRangeNormalization.d.ts

@@ -1,5 +1,7 @@
 import { document_d_exports } from "../types/document.js";
 //#region src/docx/trackedMoveRangeNormalization.d.ts
+/** The code this normalisation is reported under, owned here, not at the caller. */
+declare const UNBALANCED_MOVE_RANGE_WARNING: "unbalanced-move-range";
 type NormalizeTrackedMoveRangesInput = {
   documentBody: document_d_exports.DocumentBody;
   headers?: Map<string, document_d_exports.HeaderFooter>;
@@ -12,4 +14,4 @@ type NormalizeTrackedMoveRangesResult = {
 };
 declare const normalizeTrackedMoveRanges: ({ documentBody, headers, footers, footnotes, endnotes }: NormalizeTrackedMoveRangesInput) => NormalizeTrackedMoveRangesResult;
 //#endregion
-export { normalizeTrackedMoveRanges };
+export { UNBALANCED_MOVE_RANGE_WARNING, normalizeTrackedMoveRanges };

package/dist/docx/trackedMoveRangeNormalization.js

@@ -1,5 +1,8 @@
-import { visitDocxParagraphs } from "./paragraphTraversal.js";
+import { InlineContentRemovals, visitDocxParagraphs, visitInlineContentSlots } from "./paragraphTraversal.js";
+import { PARSE_WARNING_CODES } from "@stll/docx-core/model";
 //#region src/docx/trackedMoveRangeNormalization.ts
+/** The code this normalisation is reported under, owned here, not at the caller. */
+const UNBALANCED_MOVE_RANGE_WARNING = PARSE_WARNING_CODES.unbalancedMoveRange;
 const normalizeTrackedMoveRanges = ({ documentBody, headers, footers, footnotes, endnotes }) => {
 	const rangeMarkers = [];
 	visitDocxParagraphs({
@@ -9,10 +12,10 @@ const normalizeTrackedMoveRanges = ({ documentBody, headers, footers, footnotes,
 		footnotes,
 		endnotes
 	}, (paragraph) => {
-		for (const [index, content] of paragraph.content.entries()) {
-			const marker = toMoveRangeMarkerRef(paragraph.content, index, content);
+		visitInlineContentSlots(paragraph, ({ content, index, item }) => {
+			const marker = toMoveRangeMarkerRef(content, index, item);
 			if (marker) rangeMarkers.push(marker);
-		}
+		});
 	});
 	return { removedUnbalancedMoveRangeMarkers: removeUnbalancedMoveRangeMarkers(rangeMarkers) };
 };
@@ -27,14 +30,9 @@ const removeUnbalancedMoveRangeMarkers = (rangeMarkers) => {
 		}
 		byRange.set(key, [marker]);
 	}
-	const removals = /* @__PURE__ */ new Map();
+	const removals = new InlineContentRemovals();
 	const markForRemoval = (marker) => {
-		const indexes = removals.get(marker.content);
-		if (indexes) {
-			indexes.add(marker.index);
-			return;
-		}
-		removals.set(marker.content, /* @__PURE__ */ new Set([marker.index]));
+		removals.mark(marker);
 	};
 	for (const markers of byRange.values()) {
 		let openStart = null;
@@ -55,15 +53,7 @@ const removeUnbalancedMoveRangeMarkers = (rangeMarkers) => {
 		}
 		if (openStart) markForRemoval(openStart);
 	}
-	let removedCount = 0;
-	for (const [content, indexes] of removals.entries()) {
-		if (indexes.size === 0) continue;
-		const nextContent = content.filter((_, index) => !indexes.has(index));
-		removedCount += content.length - nextContent.length;
-		content.length = 0;
-		content.push(...nextContent);
-	}
-	return removedCount;
+	return removals.apply();
 };
 const toMoveRangeMarkerRef = (content, index, marker) => {
 	if (marker.type === "moveFromRangeStart") return {
@@ -97,4 +87,4 @@ const toMoveRangeMarkerRef = (content, index, marker) => {
 	return null;
 };
 //#endregion
-export { normalizeTrackedMoveRanges };
+export { UNBALANCED_MOVE_RANGE_WARNING, normalizeTrackedMoveRanges };

package/dist/docx/transitionalSpelling.d.ts

@@ -1,4 +1,4 @@
-import { SlotEncoding } from "./strictValueEncodings.gen.js";
+import { PercentUnit, SlotEncoding } from "./strictValueEncodings.gen.js";
 //#region src/docx/transitionalSpelling.d.ts
 /**
  * The Transitional URI a Strict one stands for, or the URI unchanged.
@@ -12,6 +12,17 @@ declare const toTransitionalNamespaceUri: (uri: string) => string;
 declare const isStrictNamespaceUri: (uri: string) => boolean;
 /** Namespaces ECMA-376 Part 4 republished; nothing else starts with this. */
 declare const STRICT_URI_PREFIX = "http://purl.oclc.org/ooxml/";
+/** How many of a unit's numbers make one whole percent. */
+declare const NUMBERS_PER_PERCENT: Readonly<Record<PercentUnit, number>>;
+/**
+ * The percent a value spells out, or undefined when it is not spelled as one.
+ *
+ * A value written `50%` says what it is; a value written `2500` needs its
+ * slot's unit to be read. Callers that must tell the two apart — the
+ * verbatim-capture conversion, and `CT_TblWidth`, whose `w:type` is optional —
+ * ask here rather than testing for a `%` themselves.
+ */
+declare const percentageSpelling: (value: string) => number | undefined;
 /**
  * How one slot's Transitional type spells its value as a number.
  *
@@ -20,4 +31,4 @@ declare const STRICT_URI_PREFIX = "http://purl.oclc.org/ooxml/";
  */
 declare const transitionalSlotEncoding: (namespaceUri: string | undefined, elementLocalName: string, attributeLocalName?: string) => SlotEncoding | undefined;
 //#endregion
-export { STRICT_URI_PREFIX, isStrictNamespaceUri, toTransitionalNamespaceUri, transitionalSlotEncoding };
+export { NUMBERS_PER_PERCENT, STRICT_URI_PREFIX, isStrictNamespaceUri, percentageSpelling, toTransitionalNamespaceUri, transitionalSlotEncoding };

package/dist/docx/transitionalSpelling.js

@@ -21,6 +21,28 @@ const toTransitionalNamespaceUri = (uri) => TRANSITIONAL_NAMESPACE_BY_STRICT_URI
 const isStrictNamespaceUri = (uri) => TRANSITIONAL_NAMESPACE_BY_STRICT_URI.has(uri);
 /** Namespaces ECMA-376 Part 4 republished; nothing else starts with this. */
 const STRICT_URI_PREFIX = "http://purl.oclc.org/ooxml/";
+/** `-?12.5%`: the one shape ECMA-376 gives a percentage that carries its unit. */
+const PERCENTAGE = /^(-?[0-9]+(?:\.[0-9]+)?)%$/u;
+/** How many of a unit's numbers make one whole percent. */
+const NUMBERS_PER_PERCENT = {
+	fiftiethPercent: 50,
+	thousandthPercent: 1e3,
+	wholePercent: 1
+};
+/**
+* The percent a value spells out, or undefined when it is not spelled as one.
+*
+* A value written `50%` says what it is; a value written `2500` needs its
+* slot's unit to be read. Callers that must tell the two apart — the
+* verbatim-capture conversion, and `CT_TblWidth`, whose `w:type` is optional —
+* ask here rather than testing for a `%` themselves.
+*/
+const percentageSpelling = (value) => {
+	const match = PERCENTAGE.exec(value.trim());
+	if (match === null) return;
+	const percent = Number(match[1]);
+	return Number.isNaN(percent) ? void 0 : percent;
+};
 /**
 * How one slot's Transitional type spells its value as a number.
 *
@@ -33,4 +55,4 @@ const transitionalSlotEncoding = (namespaceUri, elementLocalName, attributeLocal
 	return TRANSITIONAL_SLOT_ENCODINGS.get(attributeLocalName === void 0 ? slot : `${slot} @${attributeLocalName}`);
 };
 //#endregion
-export { STRICT_URI_PREFIX, isStrictNamespaceUri, toTransitionalNamespaceUri, transitionalSlotEncoding };
+export { NUMBERS_PER_PERCENT, STRICT_URI_PREFIX, isStrictNamespaceUri, percentageSpelling, toTransitionalNamespaceUri, transitionalSlotEncoding };

package/dist/docx/unzip.d.ts

@@ -6,10 +6,33 @@ declare class DocxSecurityError extends Error {
 type DocxUnzipLimits = {
   maxInputBytes: number;
   maxFiles: number;
+  /**
+   * Inflated bytes allowed in one XML part.
+   *
+   * A byte ceiling bounds the markup, not what parsing it allocates: a parsed
+   * tree costs 3x to 15x the part's bytes, and the cheapest element to write
+   * is the most expensive per byte. Use `maxXmlElementsPerPart` and the
+   * package bounds to cap memory; this one caps text-dominated parts, where
+   * bytes and cost do track each other.
+   */
   maxXmlBytes: number;
   maxMediaBytes: number;
   maxFontBytes: number;
+  /**
+   * Inflated bytes allowed across every entry in the package.
+   *
+   * Same caveat as `maxXmlBytes`: this is a ceiling on what passes through
+   * memory as bytes, not on what the parsed structure retains.
+   */
   maxTotalUncompressedBytes: number;
+  /** Elements allowed in one XML part, counted before any tree is built. */
+  maxXmlElementsPerPart: number;
+  /** Attributes allowed in one XML part, counted before any tree is built. */
+  maxXmlAttributesPerPart: number;
+  /** Elements allowed across every XML part in the package. */
+  maxXmlElementsPerPackage: number;
+  /** Attributes allowed across every XML part in the package. */
+  maxXmlAttributesPerPackage: number;
   allowedMediaMimeTypes: ReadonlySet<string>;
 };
 type DocxUnzipOptions = Partial<Omit<DocxUnzipLimits, "allowedMediaMimeTypes">> & {

package/dist/docx/unzip.js

@@ -1,6 +1,7 @@
+import { bytesToDataUrl } from "../utils/base64.js";
 import { DOCX_CONTAINER_TYPES, detectDocxContainerType } from "./encryption/containerFormat.js";
 import { openDocxBuffer } from "./encryption/openEncryptedDocx.js";
-import { FOLIO_XML_RESOURCE_LIMITS, assertXmlResourceLimits } from "./xmlResourceLimits.js";
+import { FOLIO_XML_RESOURCE_LIMITS, assertXmlResourceLimits, createXmlPackageBudget } from "./xmlResourceLimits.js";
 import JSZip from "jszip";
 //#region src/docx/unzip.ts
 /**
@@ -58,8 +59,21 @@ const DEFAULT_UNZIP_LIMITS = {
 	maxMediaBytes: 25 * MEBIBYTE,
 	maxFontBytes: 10 * MEBIBYTE,
 	maxTotalUncompressedBytes: 250 * MEBIBYTE,
+	maxXmlElementsPerPart: FOLIO_XML_RESOURCE_LIMITS.maxElementsPerPart,
+	maxXmlAttributesPerPart: FOLIO_XML_RESOURCE_LIMITS.maxAttributesPerPart,
+	maxXmlElementsPerPackage: FOLIO_XML_RESOURCE_LIMITS.maxElementsPerPackage,
+	maxXmlAttributesPerPackage: FOLIO_XML_RESOURCE_LIMITS.maxAttributesPerPackage,
 	allowedMediaMimeTypes: DEFAULT_ALLOWED_MEDIA_MIME_TYPES
 };
+/** The XML bounds an unzip enforces, in the shape the preflight takes. */
+const xmlResourceLimitsFor = (limits) => ({
+	maxBytes: limits.maxXmlBytes,
+	maxDepth: FOLIO_XML_RESOURCE_LIMITS.maxDepth,
+	maxElementsPerPart: limits.maxXmlElementsPerPart,
+	maxAttributesPerPart: limits.maxXmlAttributesPerPart,
+	maxElementsPerPackage: limits.maxXmlElementsPerPackage,
+	maxAttributesPerPackage: limits.maxXmlAttributesPerPackage
+});
 const PARSED_XML_PARTS = /* @__PURE__ */ new Set([
 	"[content_types].xml",
 	"_rels/.rels",
@@ -188,10 +202,11 @@ async function unzipDocx(buffer, options = {}) {
 			}));
 		}
 	}
+	const xmlBudget = createXmlPackageBudget();
 	for (const extracted of await Promise.all(extractionTasks.map((extract) => extract()))) {
 		if (!extracted) continue;
 		if (extracted.type === "xml") {
-			assignXmlContent(content, extracted, limits);
+			assignXmlContent(content, extracted, limits, xmlBudget);
 			continue;
 		}
 		if (extracted.type === "media") {
@@ -203,19 +218,21 @@ async function unzipDocx(buffer, options = {}) {
 	return content;
 }
 /**
-* Parts that every consumer expands into an object tree. Preflighting them once
-* here puts the bound on the unzip, so `parseDocx`, the selective save and the
-* repack path all share it instead of each entry point carrying its own.
+* Preflight every XML part the unzip retains, not a named few.
+*
+* Naming the parts to bound is the bug: `word/document.xml`, `word/styles.xml`
+* and `word/numbering.xml` were counted, and headers, footers, footnotes,
+* endnotes, comments and every other `word/*.xml` part were parsed into trees
+* unbounded. Bounding the reader instead of the part list means a part added
+* later is bounded by construction, and the package budget makes the ceiling
+* the package's rather than each part's.
 */
-const PREFLIGHT_XML_PARTS = /* @__PURE__ */ new Set([
-	"word/document.xml",
-	"word/styles.xml",
-	"word/numbering.xml"
-]);
-function assignXmlContent(content, { path, lowerPath, content: xmlContent }, limits) {
-	if (PREFLIGHT_XML_PARTS.has(lowerPath)) assertXmlResourceLimits(xmlContent, {
-		...FOLIO_XML_RESOURCE_LIMITS,
-		maxBytes: limits.maxXmlBytes
+function assignXmlContent(content, { path, lowerPath, content: xmlContent }, limits, budget) {
+	assertXmlResourceLimits({
+		xml: xmlContent,
+		limits: xmlResourceLimitsFor(limits),
+		partPath: path,
+		budget
 	});
 	content.allXml.set(path, xmlContent);
 	if (lowerPath === "word/document.xml") content.documentXml = xmlContent;
@@ -426,14 +443,7 @@ function getMediaMimeType(path) {
 * @returns Data URL string
 */
 function mediaToDataUrl(data, mimeType) {
-	const bytes = new Uint8Array(data);
-	const chunks = [];
-	const chunkSize = 32768;
-	for (let offset = 0; offset < bytes.length; offset += chunkSize) {
-		const chunk = bytes.subarray(offset, offset + chunkSize);
-		chunks.push(String.fromCodePoint(...chunk));
-	}
-	return `data:${mimeType};base64,${btoa(chunks.join(""))}`;
+	return bytesToDataUrl(new Uint8Array(data), mimeType);
 }
 /**
 * Extract a specific file from the original ZIP

package/dist/docx/verbatimCapture.js

@@ -1,5 +1,5 @@
 import { TRANSITIONAL_NAMESPACE_BY_STRICT_URI } from "./strictValueEncodings.gen.js";
-import { isStrictNamespaceUri, toTransitionalNamespaceUri, transitionalSlotEncoding } from "./transitionalSpelling.js";
+import { NUMBERS_PER_PERCENT, isStrictNamespaceUri, percentageSpelling, toTransitionalNamespaceUri, transitionalSlotEncoding } from "./transitionalSpelling.js";
 import { roundHalfAwayFromZero, universalMeasureAs } from "./universalMeasure.js";
 import { NAMESPACES, OOXML_NAMESPACE_SCOPE, cloneElement, elementToXml, getChildElements, getLocalName, getNamespacePrefix, parseXml } from "./xmlParser.js";
 import { assertXmlResourceLimits } from "./xmlResourceLimits.js";
@@ -52,13 +52,6 @@ const DRAWING_EXTENSION_MARKERS = [
 	"/wordprocessingShape"
 ];
 const isUnschemadDrawingNamespace = (uri) => !TRANSITIONAL_URIS.has(uri) && DRAWING_EXTENSION_MARKERS.some((marker) => uri.includes(marker));
-/** `-?12.5%`: the one shape ECMA-376 gives a percentage that carries its unit. */
-const PERCENTAGE = /^(-?[0-9]+(?:\.[0-9]+)?)%$/u;
-const NUMBERS_PER_PERCENT = {
-	fiftiethPercent: 50,
-	thousandthPercent: 1e3,
-	wholePercent: 1
-};
 /** The Transitional spelling of one Strict-produced value, or the value unchanged. */
 const transitionalValue = (value, namespaceUri, encoding) => {
 	const measureUnit = encoding?.measure;
@@ -66,11 +59,11 @@ const transitionalValue = (value, namespaceUri, encoding) => {
 		const measure = universalMeasureAs(value, measureUnit);
 		if (measure !== void 0) return String(measure);
 	}
-	const percentage = PERCENTAGE.exec(value);
-	if (percentage === null) return value;
+	const percentage = percentageSpelling(value);
+	if (percentage === void 0) return value;
 	const percentUnit = encoding?.percent ?? (isUnschemadDrawingNamespace(namespaceUri) ? "thousandthPercent" : void 0);
 	if (percentUnit === void 0) return value;
-	return String(roundHalfAwayFromZero(Number(percentage[1]) * NUMBERS_PER_PERCENT[percentUnit]));
+	return String(roundHalfAwayFromZero(percentage * NUMBERS_PER_PERCENT[percentUnit]));
 };
 const originOf = (element, inherited) => {
 	const uri = element.namespaceUri;
@@ -277,7 +270,7 @@ const parsedReplayEnvelope = (xml, inheritedNamespaceScope) => {
 };
 const withinXmlResourceLimits = (xml) => {
 	try {
-		assertXmlResourceLimits(xml);
+		assertXmlResourceLimits({ xml });
 		return true;
 	} catch {
 		return false;