reamkit 1.32.0 → 1.32.1

package/dist/esm/excel/formula/functions.js

@@ -537,7 +537,7 @@ function parseCriteria(s) {
 	return {
 		kind: "text",
 		negate: op === "<>",
-		re: wildcardToRegExp(text)
+		wildcard: wildcard(text)
 	};
 }
 function matchCriteria(cell, crit) {
@@ -555,22 +555,45 @@ function matchCriteria(cell, crit) {
 	}
 	const text = cell.t === "str" ? cell.v : cell.t === "blank" ? "" : void 0;
 	if (text === void 0) return false;
-	const hit = crit.re.test(text);
+	const hit = crit.wildcard.test(text);
 	return crit.negate ? !hit : hit;
 }
-function wildcardToRegExp(pattern) {
-	let out = "^";
+function wildcard(pattern) {
+	const pieces = [[]];
 	for (let i = 0; i < pattern.length; i++) {
 		const ch = pattern[i];
+		const atoms = pieces[pieces.length - 1];
 		if (ch === "~" && (pattern[i + 1] === "*" || pattern[i + 1] === "?" || pattern[i + 1] === "~")) {
-			out += escapeRe(pattern[i + 1]);
+			atoms.push(escapeRe(pattern[i + 1]));
 			i++;
-		} else if (ch === "*") out += ".*";
-		else if (ch === "?") out += ".";
-		else out += escapeRe(ch);
+		} else if (ch === "*") pieces.push([]);
+		else atoms.push(ch === "?" ? "." : escapeRe(ch));
 	}
-	out += "$";
-	return new RegExp(out, "is");
+	const all = pieces.map((atoms) => ({
+		at: new RegExp(atoms.join(""), "isy"),
+		search: new RegExp(atoms.join(""), "gis"),
+		width: atoms.length
+	}));
+	const fits = (piece, text, at) => {
+		piece.at.lastIndex = at;
+		return piece.at.test(text);
+	};
+	const first = all[0];
+	if (all.length === 1) return { test: (text) => text.length === first.width && fits(first, text, 0) };
+	const last = all[all.length - 1];
+	const between = all.slice(1, -1).filter((piece) => piece.width > 0);
+	return { test(text) {
+		const end = text.length - last.width;
+		if (end < first.width || !fits(first, text, 0) || !fits(last, text, end)) return false;
+		let at = first.width;
+		for (const piece of between) {
+			piece.search.lastIndex = at;
+			const found = piece.search.exec(text);
+			if (found === null || found.index + piece.width > end) return false;
+			at = found.index + piece.width;
+		}
+		return true;
+	} };
 }
 function escapeRe(ch) {
 	return /[.*+?^${}()|[\]\\]/.test(ch) ? `\\${ch}` : ch;
@@ -1090,10 +1113,10 @@ function matchFn(args, ev, ctx) {
 	if (len > MAX_SCAN) return err("#N/A");
 	const at = (i) => horizontal ? refGet(range.sheet, ctx, rect.r0, rect.c0 + i) : refGet(range.sheet, ctx, rect.r0 + i, rect.c0);
 	if (type === 0) {
-		const re = target.t === "str" ? wildcardToRegExp(target.v) : void 0;
+		const wild = target.t === "str" ? wildcard(target.v) : void 0;
 		for (let i = 0; i < len; i++) {
 			const cell = at(i);
-			if (re ? cell.t === "str" && re.test(cell.v) : scalarEquals(cell, target)) return num(i + 1);
+			if (wild ? cell.t === "str" && wild.test(cell.v) : scalarEquals(cell, target)) return num(i + 1);
 		}
 		return err("#N/A");
 	}
@@ -1156,10 +1179,10 @@ function lookupFn(args, ev, ctx, vh) {
 	const keyAt = (i) => vertical ? refGet(sh, ctx, rect.r0 + i, rect.c0) : refGet(sh, ctx, rect.r0, rect.c0 + i);
 	const resultAt = (i) => vertical ? refGet(sh, ctx, rect.r0 + i, rect.c0 + line - 1) : refGet(sh, ctx, rect.r0 + line - 1, rect.c0 + i);
 	if (!approx) {
-		const re = target.t === "str" ? wildcardToRegExp(target.v) : void 0;
+		const wild = target.t === "str" ? wildcard(target.v) : void 0;
 		for (let i = 0; i < vecLen; i++) {
 			const k = keyAt(i);
-			if (re ? k.t === "str" && re.test(k.v) : scalarEquals(k, target)) return resultAt(i);
+			if (wild ? k.t === "str" && wild.test(k.v) : scalarEquals(k, target)) return resultAt(i);
 		}
 		return err("#N/A");
 	}

package/dist/esm/excel/print-model.js

@@ -813,12 +813,23 @@ function gridBody(worksheet, sharedStrings, styles, date1904, print) {
 			}
 			const ws = cellMatrix[r]?.[c];
 			let text = ws ? resolveCellText(ws, sharedStrings, styles, date1904) : "";
+			let generalOverflow = false;
 			if (ws?.type === "n" && ws.rawValue.length > 0 && (styles.cellXfs[ws.styleIndex ?? 0]?.numFmtId ?? 0) === 0) {
 				const unit = charTwips(styles.cellXfs[ws.styleIndex ?? 0], styles, textTwipsUnit);
 				const digit = charWidthUnits("0");
 				if (unit > 0 && digit > 0) {
-					const room = columnWidths[c] / unit;
-					text = generalToWidth(ws.rawValue, (candidate) => estimateChars(candidate) / digit <= room);
+					let widthTwips = columnWidths[c];
+					if (merge) {
+						widthTwips = 0;
+						for (let a = merge.startColumn; a <= Math.min(merge.endColumn, colWindowEnd); a++) {
+							const i = a - colStart;
+							if (i >= 0 && i < colCount && !hiddenCols.has(i)) widthTwips += columnWidths[i];
+						}
+					}
+					const room = (widthTwips - 75) / unit;
+					const fits = (candidate) => estimateChars(candidate) / digit <= room;
+					text = generalToWidth(ws.rawValue, fits);
+					generalOverflow = !fits(text);
 				}
 			}
 			const cfText = text.length > 0 ? text : void 0;
@@ -964,7 +975,7 @@ function gridBody(worksheet, sharedStrings, styles, date1904, print) {
 				...dropdown ? { dropdown: true } : {},
 				...noteFlag ? { noteFlag } : {},
 				...!wrapText && !rotated ? { noWrap: true } : {},
-				...!wrapText && !rotated && !shrinkToFit && !merge && ws?.type === "n" && (xf?.numFmtId ?? 0) !== 0 && text.length > 0 && tooWideToShow(text, charTwips(xf, styles, textTwipsUnit), columnWidths[c]) ? { hashOnOverflow: true } : {},
+				...!wrapText && !rotated && !shrinkToFit && !merge && ws?.type === "n" && text.length > 0 && (generalOverflow || (xf?.numFmtId ?? 0) !== 0 && tooWideToShow(text, charTwips(xf, styles, textTwipsUnit), columnWidths[c])) ? { hashOnOverflow: true } : {},
 				verticalAlign: verticalAlignOf(xf)
 			};
 			const indentLevels = xf?.alignment?.indent ?? 0;

package/dist/esm/excel/vml-drawing.js

@@ -153,7 +153,7 @@ var CSS_UNITS = new Map([
 ]);
 function cssLengthPt(value) {
 	if (value === void 0) return void 0;
-	const m = /^\s*(-?[\d.]+)\s*([a-z]*)\s*$/i.exec(value);
+	const m = /^(-?[\d.]+)\s*([a-z]*)$/i.exec(value.trim());
 	if (!m) return void 0;
 	const n = Number(m[1]);
 	if (!Number.isFinite(n)) return void 0;

package/dist/esm/excel/xlsx-reader.js

@@ -4,7 +4,7 @@ import { detectImageFormat } from "../core/images.js";
 import { INDEXED_COLORS } from "../core/indexed-colors.js";
 import { bytesInclude, packageHasPart } from "../core/bytes.js";
 import { OFFICE_2023_THEME_PALETTE, makeColorResolver } from "../core/drawingml/colors.js";
-import { parseChart, withChartColorStyle } from "../core/drawingml/chart-parser.js";
+import { parseChart, pointsPerSeries, withChartColorStyle } from "../core/drawingml/chart-parser.js";
 import { parseTheme, parseThemeEffectStyles, parseThemeFillStyles, parseThemeLineWidths } from "../core/drawingml/theme-parser.js";
 import { isOoxmlRel } from "../core/opc/relationship-types.js";
 import { OpcPackage } from "../core/opc/package.js";
@@ -409,6 +409,11 @@ function readXlsxToSheetDoc(xlsx) {
 * them, and a reference naming a sheet this workbook does not have resolves to
 * nothing rather than to zeros.
 *
+* A reference reads no more cells than the series' share of the chart's points
+* ({@link pointsPerSeries}), and finds each in an index of its sheet. The range
+* is the file's to name and unbounded — `B2:B99999999999` — and every cell of
+* it was searched for through the whole sheet.
+*
 * @param chart         The parsed chart.
 * @param sheets        Every sheet in the workbook, by tab order.
 * @param styles        The style table (for a referenced cell's number format).
@@ -417,13 +422,27 @@ function readXlsxToSheetDoc(xlsx) {
 * @returns The chart, unchanged when nothing needed resolving.
 */
 function withWorkbookData(chart, sheets, styles, sharedStrings, date1904) {
+	const most = pointsPerSeries(chart.series.length);
+	const indexes = /* @__PURE__ */ new Map();
+	const cellAt = (grid, row, column) => {
+		let index = indexes.get(grid);
+		if (!index) {
+			index = /* @__PURE__ */ new Map();
+			for (const cell of grid.cells) {
+				const key = `${String(cell.row)}:${String(cell.column)}`;
+				if (!index.has(key)) index.set(key, cell);
+			}
+			indexes.set(grid, index);
+		}
+		return index.get(`${String(row)}:${String(column)}`);
+	};
 	const cellsOf = (ref) => {
 		if (ref === void 0) return void 0;
 		const area = resolveChartRef(ref, sheets);
 		if (!area) return void 0;
 		const out = [];
-		for (let row = area.startRow; row <= area.endRow; row++) for (let col = area.startColumn; col <= area.endColumn; col++) {
-			const cell = area.grid.cells.find((c) => c.row === row && c.column === col);
+		for (let row = area.startRow; row <= area.endRow && out.length < most; row++) for (let col = area.startColumn; col <= area.endColumn && out.length < most; col++) {
+			const cell = cellAt(area.grid, row, col);
 			out.push(cell ? resolveCellText(cell, sharedStrings, styles, date1904) : "");
 		}
 		return out;
@@ -810,9 +829,24 @@ function resolveTableSlicerItems(cache, tableIndex, styles, sharedStrings, date1
 	}
 	return out;
 }
+function slicerStyleNumber(name) {
+	const prefix = /SlicerStyle/gi;
+	const isLetter = (i) => /[A-Za-z]/.test(name[i] ?? "");
+	const isDigit = (i) => /\d/.test(name[i] ?? "");
+	for (let found = prefix.exec(name); found; found = prefix.exec(name)) {
+		let end = prefix.lastIndex;
+		while (isLetter(end)) end++;
+		if (isDigit(end)) {
+			let digits = end;
+			while (isDigit(digits)) digits++;
+			return name.slice(end, digits);
+		}
+		prefix.lastIndex = end;
+	}
+}
 function resolveSlicerStyle(styleName, palette) {
-	const m = styleName ? /SlicerStyle[A-Za-z]*?(\d+)/i.exec(styleName) : null;
-	const column = m ? (Number(m[1]) - 1) % 7 : 1;
+	const number = styleName ? slicerStyleNumber(styleName) : void 0;
+	const column = number !== void 0 ? (Number(number) - 1) % 7 : 1;
 	const base = column === 0 ? "7F7F7F" : palette.get(`accent${column}`) ?? "4472C4";
 	return {
 		headerHex: base,

package/dist/esm/layout/styled-layout.js

@@ -3174,7 +3174,7 @@ function collectFontResources(body, headersFooters, options, imageResources) {
 				if (chart.catAxisTitle) add(chart.catAxisTitle);
 				if (chart.valAxisTitle) add(chart.valAxisTitle);
 				for (const sr of chart.series) for (const label of sr.pointLabels ?? []) add(label.text);
-				add("0123456789.,-%() ");
+				add("0123456789.,-+E%() ");
 				if (chart.numberFormat) add(chart.numberFormat);
 			}
 		}

package/dist/esm/markdown/markdown-writer.js

@@ -82,9 +82,29 @@ function lose(ctx, severity, feature, detail) {
 */
 function joinBlocks(blocks) {
 	const body = blocks.map(trimHardBreaks).filter((b) => b.length > 0).join("\n\n");
-	return body.length > 0 ? `${body.replace(/[ \t]+$/gm, "")}\n` : "";
+	return body.length > 0 ? `${trimLineEnds(body)}\n` : "";
 }
 /**
+* The text without the spaces and tabs that end its lines — what
+* `/[ \t]+$/gm` leaves, a line ending wherever `$` does under `m`: at LF, CR,
+* U+2028, U+2029 and the end of the text. The expression retries every blank
+* of a long run that something else follows, and is quadratic on it; this is
+* one pass.
+*/
+function trimLineEnds(text) {
+	const out = [];
+	let start = 0;
+	for (let i = 0; i <= text.length; i++) {
+		if (i < text.length && !isLineEnd(text.charCodeAt(i))) continue;
+		let end = i;
+		while (end > start && (text[end - 1] === " " || text[end - 1] === "	")) end--;
+		out.push(text.slice(start, end), text.slice(i, i + 1));
+		start = i + 1;
+	}
+	return out.join("");
+}
+var isLineEnd = (c) => c === 10 || c === 13 || c === 8232 || c === 8233;
+/**
 * Drop a hard break sitting at either end of a block: at the end there is no
 * next line to break to, at the start no previous one, and either way the
 * backslash is left standing in the text as itself. Only a break is taken —
@@ -218,7 +238,7 @@ function emitListItem(out, resolved, marker, inline, ctx) {
 	top.markerWidth = bullet.length + 1;
 	const pad = " ".repeat(top.indent);
 	const cont = " ".repeat(top.indent + top.markerWidth);
-	const line = `${pad}${bullet} ${inline.replaceAll("\\\n", `\\\n${cont}`)}`.replace(/\s+$/, "");
+	const line = `${pad}${bullet} ${inline.replaceAll("\\\n", `\\\n${cont}`)}`.trimEnd();
 	if (wasOpen && out.length > 0) out[out.length - 1] += `\n${line}`;
 	else out.push(line);
 }
@@ -232,9 +252,24 @@ function listBullet(marker, numId, ilvl, counter, ctx) {
 	const level = numberingLevel(numId, ilvl, ctx.numbering);
 	if (!(level ? level.format !== "bullet" && level.format !== "none" : /\d/.test(marker))) return "-";
 	if (level && level.format !== "decimal" && level.format !== "decimalZero") lose(ctx, "degraded", FEATURES.lists, `${level.format} list markers render as decimal`);
-	const digits = /(\d+)\D*$/.exec(marker);
-	if (marker.replace(/\D+/g, "").length > (digits?.[1]?.length ?? 0)) lose(ctx, "degraded", FEATURES.lists, "multi-level list marker flattened to one number");
-	return `${digits ? Number(digits[1]) : counter}.`;
+	const digits = lastNumber(marker);
+	if (marker.replace(/\D+/g, "").length > (digits?.length ?? 0)) lose(ctx, "degraded", FEATURES.lists, "multi-level list marker flattened to one number");
+	return `${digits !== void 0 ? Number(digits) : counter}.`;
+}
+/**
+* The last run of digits in a marker — what `/(\d+)\D*$/` matched, read back
+* from the end: that expression tried every digit of a long run as its start.
+* The template the marker comes from is the file's, so the run can be any
+* length, and every item of the list reads it.
+*/
+function lastNumber(marker) {
+	const isDigit = (i) => /\d/.test(marker[i] ?? "");
+	let end = marker.length;
+	while (end > 0 && !isDigit(end - 1)) end--;
+	if (end === 0) return void 0;
+	let start = end - 1;
+	while (start > 0 && isDigit(start - 1)) start--;
+	return marker.slice(start, end);
 }
 function numberingLevel(numId, ilvl, numbering) {
 	if (!numbering) return void 0;
@@ -366,7 +401,7 @@ function destination(url) {
 * paragraph it belongs to — the same shape GFM's own heading anchors take.
 */
 function bookmarkSlug(name) {
-	const slug = name.toLowerCase().replace(/[^\p{L}\p{N}]+/gu, "-").replace(/^-+|-+$/g, "");
+	const slug = name.toLowerCase().split(/[^\p{L}\p{N}]+/u).filter((word) => word.length > 0).join("-");
 	return slug.length > 0 ? slug : "bookmark";
 }
 /** The bookmark names some run actually links to — the only ones worth an anchor. */
@@ -684,7 +719,7 @@ var SPAN_EDGE = String.raw`(?:\\\n|\s)`;
 function applyMarks(text, marks, ctx, before, after) {
 	const escaped = escapeInline(text);
 	const lead = new RegExp(`^${SPAN_EDGE}*`).exec(escaped)?.[0] ?? "";
-	const trail = new RegExp(`${SPAN_EDGE}*$`).exec(escaped)?.[0] ?? "";
+	const trail = trailingEdge(escaped);
 	const core = escaped.slice(lead.length, escaped.length - trail.length);
 	if (core.length === 0) return escaped;
 	const outerBefore = lead.length > 0 ? lead[lead.length - 1] : before;
@@ -718,6 +753,22 @@ function applyMarks(text, marks, ctx, before, after) {
 	}
 	return `${lead}${linkify(out, marks, ctx)}${trail}`;
 }
+/**
+* The whitespace and hard breaks the escaped text ends in — what
+* `${SPAN_EDGE}*$` matches, read back from the end instead. Tried from the
+* front, that expression rescans a long run of whitespace from each of its
+* characters whenever something other than the end follows the run, and is
+* quadratic on it.
+*/
+function trailingEdge(escaped) {
+	let from = escaped.length;
+	while (from > 0) {
+		const c = escaped[from - 1];
+		if (c === "\\" ? escaped[from] !== "\n" : !/\s/u.test(c)) break;
+		from--;
+	}
+	return escaped.slice(from);
+}
 var isSpace = (c) => c === "" || /\s/u.test(c);
 var isPunct = (c) => /[\p{P}\p{S}]/u.test(c);
 /** §6.2 — the delimiter run is left-flanking, so it can open emphasis. */

package/dist/esm/pdf-reader/annot-draw.js

@@ -260,7 +260,7 @@ function freeTextContents(file, annot) {
 	const found = file.get(annot, "Contents");
 	const raw = found instanceof PdfHexString ? String.fromCharCode(...found.bytes) : typeof found === "string" ? found : void 0;
 	if (raw === void 0) return void 0;
-	const text = textString(raw).replace(/\s+$/u, "");
+	const text = textString(raw).trimEnd();
 	return text.length > 0 ? text : void 0;
 }
 /**
@@ -331,7 +331,7 @@ function defaultAppearance(file, annot) {
 	const form = acro instanceof Map ? file.get(acro, "DA") : void 0;
 	const da = own ?? (typeof form === "string" ? form : "");
 	const tf = /\/([^\s/]+)\s+([\d.]+)\s+Tf/u.exec(da);
-	const rgb = /([\d.]+)\s+([\d.]+)\s+([\d.]+)\s+rg/u.exec(da);
+	const rgb = /(?<![\d.])([\d.]+)\s+([\d.]+)\s+([\d.]+)\s+rg/u.exec(da);
 	const gray = /(^|\s)([\d.]+)\s+g(\s|$)/u.exec(da);
 	const color = rgb ? `${num(Number(rgb[1]))} ${num(Number(rgb[2]))} ${num(Number(rgb[3]))}` : gray ? `${num(Number(gray[2]))} ${num(Number(gray[2]))} ${num(Number(gray[2]))}` : "0 0 0";
 	return {

package/dist/esm/pdf-reader/annots.js

@@ -137,7 +137,9 @@ function typedField(file, annot) {
 * field whose value is "Hello World".
 */
 function withoutVariableText(file, stream) {
-	const stripped = new TextDecoder("latin1").decode(file.streamData(stream)).replace(/\/Tx\s+BMC[\s\S]*?EMC/gu, "");
+	const text = new TextDecoder("latin1").decode(file.streamData(stream));
+	const end = text.lastIndexOf("EMC") + 3;
+	const stripped = text.slice(0, end).replace(/\/Tx\s+BMC[\s\S]*?EMC/gu, "") + text.slice(end);
 	const dict = new Map(stream.dict);
 	dict.delete("Filter");
 	dict.delete("DecodeParms");

package/dist/esm/pdf-reader/font.js

@@ -1047,7 +1047,7 @@ function familyOfFace(baseFont) {
 	const name = plainFace(baseFont);
 	const cut = /^(.+?)[-,]([^-,]+)$/u.exec(name);
 	const whole = cut && STYLE_WORDS.test(cut[2]) ? cut[1] : cut ? name.replace(/-/gu, " ") : GLUED_STYLE.exec(name)?.[1] ?? name;
-	return (whole.replace(/(?:PSMT|PS|MT)$/u, "") || whole).replace(/([a-z])([A-Z])/gu, "$1 $2").replace(/([A-Z]+)([A-Z][a-z])/gu, "$1 $2").replace(/\bDeja Vu\b/u, "DejaVu").replace(/\s+/gu, " ").trim();
+	return (whole.replace(/(?:PSMT|PS|MT)$/u, "") || whole).replace(/([a-z])([A-Z])/gu, "$1 $2").replace(/(?<![A-Z])([A-Z]+)([A-Z][a-z])/gu, "$1 $2").replace(/\bDeja Vu\b/u, "DejaVu").replace(/\s+/gu, " ").trim();
 }
 /**
 * A face name with what producers write around it taken off: the subset tag

package/dist/esm/pdf-reader/layout.js

@@ -1585,7 +1585,7 @@ function inkSpan(runs) {
 	const last = marked[marked.length - 1];
 	const space = (r) => r.spaceWidthPt !== void 0 && r.spaceWidthPt > 0 ? r.spaceWidthPt : (r.fontSizePt || 10) * .25;
 	const lead = (/^\s*/u.exec(first.text)?.[0].length ?? 0) * space(first);
-	const trail = (/\s*$/u.exec(last.text)?.[0].length ?? 0) * space(last);
+	const trail = (last.text.length - last.text.trimEnd().length) * space(last);
 	const x = first.x + lead;
 	return {
 		x,

package/dist/esm/word/document-parser.js

@@ -619,7 +619,7 @@ function collectPositionTabs(p) {
 * @returns The display text, or `undefined` when this is not a MACROBUTTON.
 */
 function macroButtonText(instr) {
-	const shown = /^\s*MACROBUTTON\s+\S+\s(.*)$/su.exec(instr)?.[1]?.replace(/\s+$/u, "");
+	const shown = /^\s*MACROBUTTON\s+\S+\s(.*)$/su.exec(instr)?.[1]?.trimEnd();
 	return shown !== void 0 && shown !== "" ? shown : void 0;
 }
 function parseFieldInstr(instr) {

package/dist/esm/word/docx-reader.js

@@ -15,6 +15,7 @@ import { OpcPackage } from "../core/opc/package.js";
 import { parseCoreProperties } from "../core/opc/core-properties.js";
 import "../core/opc/index.js";
 import { parseXml } from "../pptx/pptx-reader.js";
+import { firstTagMatch } from "../core/opc/tag-scan.js";
 import { EMPTY_NUMBERING, parseNumbering } from "./numbering-parser.js";
 import { EMPTY_SECTION, applyAuthorIds, bodyIndexForBlock, newBlockCounter, parseBackgroundColor, parseBackgroundFill, parseCommentThreads, parseDocument, parseHeaderFooter, parseNotes, parsePeople, parseSections } from "./document-parser.js";
 import { parseStyles } from "./styles-parser.js";
@@ -214,21 +215,26 @@ function makeDiagramResolver(pkg, partName, store) {
 		return resolved;
 	};
 }
+function partIndex(path) {
+	const dot = path.lastIndexOf(".");
+	if (dot < 0 || dot === path.length - 1) return "";
+	let from = dot;
+	while (from > 0 && path.charCodeAt(from - 1) >= 48 && path.charCodeAt(from - 1) <= 57) from--;
+	return path.slice(from, dot);
+}
 function siblingPart(pkg, ownerPart, dataPart, type) {
 	const own = pkg.getPartRelationships(dataPart).find((r) => r.type.endsWith(type));
 	if (own) return pkg.resolveRelatedPart(dataPart, own)?.data;
-	const index = (path) => /(\d*)\.[^.]+$/u.exec(path)?.[1] ?? "";
-	const want = index(dataPart);
+	const want = partIndex(dataPart);
 	const rels = pkg.getPartRelationships(ownerPart).filter((r) => r.type.endsWith(type));
-	const rel = rels.find((r) => index(r.target) === want) ?? (rels.length === 1 ? rels[0] : void 0);
+	const rel = rels.find((r) => partIndex(r.target) === want) ?? (rels.length === 1 ? rels[0] : void 0);
 	return rel ? pkg.resolveRelatedPart(ownerPart, rel)?.data : void 0;
 }
 function drawingFromOwner(pkg, ownerPart, dataPart) {
 	const drawings = pkg.getPartRelationships(ownerPart).filter((r) => r.type.endsWith("/diagramDrawing"));
 	if (drawings.length === 0) return void 0;
-	const index = (path) => /(\d*)\.[^.]+$/u.exec(path)?.[1] ?? "";
-	const want = index(dataPart);
-	const rel = drawings.find((r) => index(r.target) === want) ?? (drawings.length === 1 ? drawings[0] : void 0);
+	const want = partIndex(dataPart);
+	const rel = drawings.find((r) => partIndex(r.target) === want) ?? (drawings.length === 1 ? drawings[0] : void 0);
 	return rel ? pkg.resolveRelatedPart(ownerPart, rel) : void 0;
 }
 function makeImageResolver(pkg, store, partName = MAIN_DOCUMENT_PART) {
@@ -377,11 +383,10 @@ function chartOwningParts(pkg, sections, mainPart) {
 	return parts;
 }
 function detectDocxLanguage(stylesData, documentData) {
-	const re = /<w:lang\b[^>]*\bw:val="([^"]+)"/;
 	const decoder = new TextDecoder();
 	for (const data of [stylesData, documentData]) {
 		if (!data) continue;
-		const m = re.exec(decoder.decode(data));
+		const m = firstTagMatch(decoder.decode(data), "<w:lang", true, /<w:lang\b[^>]*\bw:val="([^"]+)"/y);
 		if (m?.[1]) return m[1];
 	}
 }

package/dist/esm/word/docx-to-pdf.js

@@ -6,6 +6,7 @@ import "../pdf/index.js";
 import { parseThemeFonts } from "../core/drawingml/theme-parser.js";
 import { OpcPackage } from "../core/opc/package.js";
 import "../core/opc/index.js";
+import { tagMatches } from "../core/opc/tag-scan.js";
 import { resolveWordThemeFont } from "./theme-fonts.js";
 import { readDocx } from "./docx-reader.js";
 import { flowRenderOptions } from "../core/converter/project.js";
@@ -13,6 +14,7 @@ import { flowRenderOptions } from "../core/converter/project.js";
 var STYLES_PART = "word/styles.xml";
 var THEME_PART = "word/theme/theme1.xml";
 var MAIN_DOCUMENT_PART = "word/document.xml";
+var ASCII_FONT = /<w:rFonts[^>]*?\bw:(ascii|asciiTheme)="([^"]+)"/y;
 /**
 * Synchronous one-shot conversion (requires `fonts`/`fontBytes`, no network).
 *
@@ -72,13 +74,11 @@ function detectDocxFamilyKeys(docx) {
 	}
 	const decoder = new TextDecoder("utf-8");
 	const themeFonts = docxThemeFonts(pkg);
-	const re = /<w:rFonts[^>]*?\bw:(ascii|asciiTheme)="([^"]+)"/g;
 	for (const part of [STYLES_PART, MAIN_DOCUMENT_PART]) {
 		const data = pkg.getPart(part);
 		if (!data) continue;
 		const xml = decoder.decode(data);
-		let m;
-		while ((m = re.exec(xml)) !== null) {
+		for (const m of tagMatches(xml, "<w:rFonts", false, ASCII_FONT)) {
 			const name = m[1] === "ascii" ? m[2] : resolveWordThemeFont(m[2], themeFonts);
 			if (name !== void 0) keys.add(resolveFamilyKey(name));
 		}

package/dist/esm/word/font-table.js

@@ -1,4 +1,5 @@
 import { FontRegistry } from "../core/font/font-registry.js";
+import { isWordChar, tagMatches } from "../core/opc/tag-scan.js";
 //#region src/word/font-table.ts
 var FONT_TABLE_PART = "word/fontTable.xml";
 var REL_FONT = "http://schemas.openxmlformats.org/officeDocument/2006/relationships/font";
@@ -48,19 +49,15 @@ function obfuscateEmbeddedFont(data, fontKey) {
 function parseFontTable(data) {
 	const xml = new TextDecoder("utf-8").decode(data);
 	const out = [];
-	const fontRe = /<w:font\b[^>]*\bw:name="([^"]+)"[^>]*>([\s\S]*?)<\/w:font>/g;
-	let fm;
-	while ((fm = fontRe.exec(xml)) !== null) {
+	const closed = xml.slice(0, xml.lastIndexOf("</w:font>") + 9);
+	for (const fm of tagMatches(closed, "<w:font", true, FONT)) {
 		const name = fm[1];
-		const inner = fm[2];
 		const embeds = {};
-		const embedRe = /<w:embed(Regular|Bold|Italic|BoldItalic)\b[^>]*\br:id="([^"]+)"[^>]*\bw:fontKey="([^"]+)"/g;
-		let em;
-		while ((em = embedRe.exec(inner)) !== null) {
-			const variant = VARIANT_BY_TAG[em[1]];
+		for (const em of embedRefs(fm[2])) {
+			const variant = VARIANT_BY_TAG[em.tag];
 			if (variant) embeds[variant] = {
-				rId: em[2],
-				fontKey: em[3]
+				rId: em.rId,
+				fontKey: em.fontKey
 			};
 		}
 		if (Object.keys(embeds).length > 0) out.push({
@@ -70,6 +67,88 @@ function parseFontTable(data) {
 	}
 	return out;
 }
+/** A named `w:font` and everything up to its close. Sticky, for tagMatches. */
+var FONT = /<w:font\b[^>]*\bw:name="([^"]+)"[^>]*>([\s\S]*?)<\/w:font>/y;
+var EMBED_HEAD = /<w:embed(Regular|Bold|Italic|BoldItalic)\b/y;
+/**
+* The faces a font's element embeds — what
+* `/<w:embed(Regular|Bold|Italic|BoldItalic)\b[^>]*\br:id="([^"]+)"[^>]*\bw:fontKey="([^"]+)"/g`
+* finds in it, in one pass. The expression gave the tag back one `r:id` at a
+* time and read the rest of it again for a `w:fontKey` after each, quadratic
+* in a tag with many; and searching, it read each tag again from every
+* `<w:embed` inside it. Where a start fails, every later one before the tag's
+* `>` does too, so the search moves past it.
+*/
+function embedRefs(inner) {
+	const out = [];
+	let at = inner.indexOf("<w:embed");
+	while (at >= 0) {
+		EMBED_HEAD.lastIndex = at;
+		const head = EMBED_HEAD.exec(inner);
+		if (head === null) {
+			at = inner.indexOf("<w:embed", at + 1);
+			continue;
+		}
+		const from = EMBED_HEAD.lastIndex;
+		const tagEnd = endOfStretch(inner, from);
+		const found = embedRef(inner, from, tagEnd);
+		if (found) {
+			out.push({
+				tag: head[1],
+				rId: found.rId,
+				fontKey: found.fontKey
+			});
+			at = inner.indexOf("<w:embed", found.end);
+		} else if (tagEnd < inner.length) at = inner.indexOf("<w:embed", tagEnd + 1);
+		else break;
+	}
+	return out;
+}
+/** The first `>` at or after `from`, or the end of the text. */
+function endOfStretch(s, from) {
+	const gt = s.indexOf(">", from);
+	return gt < 0 ? s.length : gt;
+}
+/**
+* The reference one `<w:embed…` makes, its attributes read from `from` up to
+* the tag's `>` at `tagEnd`. The first `[^>]*` gives back the last `r:id`
+* first; the second takes the last `w:fontKey` before the `>` that follows the
+* `r:id`'s value — and that one is the same for every `r:id` whose value ends
+* in one stretch, so it is looked for once per stretch.
+*/
+function embedRef(s, from, tagEnd) {
+	const keys = /* @__PURE__ */ new Map();
+	const lastKey = (stretchEnd) => {
+		if (keys.has(stretchEnd)) return keys.get(stretchEnd);
+		const start = s.lastIndexOf(">", stretchEnd - 1) + 1;
+		let key;
+		for (let f = s.lastIndexOf("w:fontKey=\"", stretchEnd - 11); f >= start; f = s.lastIndexOf("w:fontKey=\"", f - 1)) {
+			const close = s.indexOf("\"", f + 11);
+			if (!isWordChar(s.charCodeAt(f - 1)) && close > f + 11) {
+				key = {
+					at: f,
+					value: s.slice(f + 11, close),
+					end: close + 1
+				};
+				break;
+			}
+			if (f === 0) break;
+		}
+		keys.set(stretchEnd, key);
+		return key;
+	};
+	for (let p = s.lastIndexOf("r:id=\"", tagEnd - 6); p >= from; p = s.lastIndexOf("r:id=\"", p - 1)) {
+		if (isWordChar(s.charCodeAt(p - 1))) continue;
+		const close = s.indexOf("\"", p + 6);
+		if (close <= p + 6) continue;
+		const key = lastKey(close < tagEnd ? tagEnd : endOfStretch(s, close + 1));
+		if (key !== void 0 && key.at > close) return {
+			rId: s.slice(p + 6, close),
+			fontKey: key.value,
+			end: key.end
+		};
+	}
+}
 /**
 * De-obfuscate and load every embedded font in the package, building one
 * {@link FontRegistry} per font keyed by its normalized (trimmed, lower-cased)

package/dist/esm/word/numbering-parser.js

@@ -171,7 +171,7 @@ var CSS_UNITS = new Map([
 	["px", .75]
 ]);
 function cssLengthPt(raw) {
-	const m = raw !== void 0 ? /^\s*(-?[\d.]+)\s*([a-z%]*)\s*$/u.exec(raw) : null;
+	const m = raw !== void 0 ? /^(-?[\d.]+)\s*([a-z%]*)$/u.exec(raw.trim()) : null;
 	if (!m) return void 0;
 	const n = Number(m[1]);
 	if (!Number.isFinite(n)) return void 0;

package/package.json

@@ -1,6 +1,6 @@
 {
   "name": "reamkit",
-  "version": "1.32.0",
+  "version": "1.32.1",
   "description": "Ream — convert DOCX, XLSX, PPTX, PDF and legacy DOC, XLS and PPT to PDF, SVG, HTML, Markdown, DOCX and XLSX, built from scratch on the ECMA-376 and ISO 32000 specifications. Parse once, convert anywhere.",
   "license": "MIT",
   "author": "Alex Krassavin <info@reamkit.dev>",
@@ -91,7 +91,7 @@
     "typecheck:dist": "tsc -p tsconfig.json --noEmit"
   },
   "dependencies": {
-    "fast-xml-parser": "5.7.0",
+    "fast-xml-parser": "5.7.3",
     "fflate": "^0.8.2"
   },
   "devDependencies": {
@@ -103,7 +103,7 @@
     "tsx": "^4.19.0",
     "typescript": "^5.7.0",
     "vite": "8.0.5",
-    "vitest": "^2.1.0"
+    "vitest": "^5.0.3"
   },
   "engines": {
     "node": ">=20"