js.documents 14.6.0 → 14.6.1

package/README.md

@@ -1011,7 +1011,7 @@ Read as **row → column**. `✓` lossless, `~` bounded, `✗` lossy, `✗✗` s
 
 **odf → PDF and embedded formulas** render faithful mathematical typesetting through STIX Two Math: real box-model layout, per-glyph metrics, font-wide constants from the `MATH` table, stretchy fences and braces via `MathVariants`. `pdfToOdf` is not attempted — recovering a semantic operator tree from glyphs is OCR-adjacent.
 
-**PDF → docx/pptx/odt/odp** is best-effort reconstruction from geometry. Reading order, font properties, page count survive; paragraph boundaries are inferred from baseline spacing. Tables recover only from a real gridline lattice. Vector primitives recover into a nested drawing document.
+**PDF → docx/pptx/odt/odp** is best-effort reconstruction from geometry. Reading order, font properties, page count survive; paragraph boundaries are inferred from baseline spacing. Text set in columns is read one column at a time: a vertical strip that no text touches over a run of lines, with running prose on both sides (a measure of at least twenty characters, lines that wrap, word gaps narrower than the strip), splits the page there before lines are clustered, so two columns whose baselines line up no longer interleave line by line. Lines crossing the strip (a full-width heading, a boxed passage) are read in page order between the column bands above and below them, and images and recovered tables are placed by the column they sit in. A table's rows are still read across, since its columns are too narrow or too ragged to pass as prose; a page with no such strip, or with text whose horizontal extent the reader did not report, reads exactly as clustering the whole page always did (`src/layout/reading-order.ts`). Tables recover only from a real gridline lattice. Vector primitives recover into a nested drawing document.
 
 **PDF → odg** is near-1:1 mapping (no clustering needed). Kind narrows upstream: rotated rects, freeform curves, multi-subpath figures become `path`.
 

package/dist/{composition-plan-CSHdRyji.cjs → composition-plan-DJQgPvBj.cjs}

@@ -23,7 +23,7 @@ const require_csv_write = require("./csv/write.cjs");
 const require_read$1 = require("./read-CahCZMwG.cjs");
 const require_svg_write = require("./svg/write.cjs");
 const require_ports_abort = require("./ports/abort.cjs");
-const require_reconstruct_presentations_prod = require("./reconstruct-presentations-prod-CA4no_aw.cjs");
+const require_reconstruct_presentations_prod = require("./reconstruct-presentations-prod-CsvGlvP2.cjs");
 const require_convert_pdf_package_tables = require("./convert/pdf-package-tables.cjs");
 const require_convert_variant_bridges = require("./convert/variant-bridges.cjs");
 const require_convert_capability = require("./convert/capability.cjs");

package/dist/{composition-plan-DQ8lPKIW.js → composition-plan-DMTkmk8n.js}

@@ -23,7 +23,7 @@ import { buildCsvText } from "./csv/write.js";
 import { i as readSvgContent } from "./read-BSH--XB4.js";
 import { buildSvgText } from "./svg/write.js";
 import { throwIfAborted } from "./ports/abort.js";
-import { _ as reconstructWordprocessing, g as reconstructSpreadsheet, i as reconstructPresentation, r as reconstructDrawing } from "./reconstruct-presentations-prod-CC8ct-5O.js";
+import { _ as reconstructWordprocessing, g as reconstructSpreadsheet, i as reconstructPresentation, r as reconstructDrawing } from "./reconstruct-presentations-prod-D1KbGPxa.js";
 import { stampPdfPackageTables } from "./convert/pdf-package-tables.js";
 import { drawingToPresentation, presentationToDrawing, presentationToWordprocessing, spreadsheetToWordprocessing, wordprocessingToPresentation } from "./convert/variant-bridges.js";
 import { UnsupportedConversionError } from "./convert/capability.js";

package/dist/convert/composition-plan.cjs

@@ -1,5 +1,5 @@
 Object.defineProperty(exports, Symbol.toStringTag, { value: "Module" });
-const require_composition_plan = require("../composition-plan-CSHdRyji.cjs");
+const require_composition_plan = require("../composition-plan-DJQgPvBj.cjs");
 exports.buildCompositionGraph = require_composition_plan.buildCompositionGraph;
 exports.isContentFormat = require_composition_plan.isContentFormat;
 exports.isSourceContentFormat = require_composition_plan.isSourceContentFormat;

package/dist/convert/composition-plan.js

@@ -1,2 +1,2 @@
-import { a as shortestPath, i as resolveCompositionPlan, n as isContentFormat, r as isSourceContentFormat, t as buildCompositionGraph } from "../composition-plan-DQ8lPKIW.js";
+import { a as shortestPath, i as resolveCompositionPlan, n as isContentFormat, r as isSourceContentFormat, t as buildCompositionGraph } from "../composition-plan-DMTkmk8n.js";
 export { buildCompositionGraph, isContentFormat, isSourceContentFormat, resolveCompositionPlan, shortestPath };

package/dist/convert/composition-to-pdf.cjs

@@ -7,7 +7,7 @@ const require_ports_abort = require("../ports/abort.cjs");
 const require_sheets = require("../sheets-BQCpNXsE.cjs");
 const require_layout_drawing = require("../layout/drawing.cjs");
 const require_convert_capability = require("./capability.cjs");
-const require_composition_plan = require("../composition-plan-CSHdRyji.cjs");
+const require_composition_plan = require("../composition-plan-DJQgPvBj.cjs");
 let document_schema_js = require("document-schema.js");
 let pdf_codec = require("pdf-codec");
 //#region src/convert/composition-to-pdf.ts

package/dist/convert/composition-to-pdf.js

@@ -6,7 +6,7 @@ import { throwIfAborted } from "../ports/abort.js";
 import { l as convertSpreadsheetToLayout } from "../sheets-DWDvGN-g.js";
 import { convertDrawingToLayout } from "../layout/drawing.js";
 import { UnsupportedConversionError } from "./capability.js";
-import { _ as executeFromPdf, b as runCompositionPlan, c as LAYOUT_CAPABLE, d as READ_ONLY_FORMAT_NODES, g as executeBridge, i as resolveCompositionPlan, s as FORMAT_NODES, v as isReadOnlyContentFormat, y as isTextFormatNode } from "../composition-plan-DQ8lPKIW.js";
+import { _ as executeFromPdf, b as runCompositionPlan, c as LAYOUT_CAPABLE, d as READ_ONLY_FORMAT_NODES, g as executeBridge, i as resolveCompositionPlan, s as FORMAT_NODES, v as isReadOnlyContentFormat, y as isTextFormatNode } from "../composition-plan-DMTkmk8n.js";
 import { assembleTree } from "document-schema.js";
 import { createFontMeasurer, createFontRegistry, loadMathFont, writePdf } from "pdf-codec";
 //#region src/convert/composition-to-pdf.ts

package/dist/convert/composition.cjs

@@ -1,5 +1,5 @@
 Object.defineProperty(exports, Symbol.toStringTag, { value: "Module" });
-const require_composition_plan = require("../composition-plan-CSHdRyji.cjs");
+const require_composition_plan = require("../composition-plan-DJQgPvBj.cjs");
 exports.CONTENT_FORMATS = require_composition_plan.CONTENT_FORMATS;
 exports.FORMAT_NODES = require_composition_plan.FORMAT_NODES;
 exports.LAYOUT_CAPABLE = require_composition_plan.LAYOUT_CAPABLE;

package/dist/convert/composition.js

@@ -1,2 +1,2 @@
-import { _ as executeFromPdf, b as runCompositionPlan, c as LAYOUT_CAPABLE, d as READ_ONLY_FORMAT_NODES, f as RECONSTRUCTORS, g as executeBridge, h as convertDocumentFromPdf, l as MAX_CONVERSION_PATH_NODES, m as TRANSFORMS, o as CONTENT_FORMATS, p as THROUGH_PDF_EDGE_COST, s as FORMAT_NODES, u as READ_ONLY_CONTENT_FORMATS, v as isReadOnlyContentFormat, y as isTextFormatNode } from "../composition-plan-DQ8lPKIW.js";
+import { _ as executeFromPdf, b as runCompositionPlan, c as LAYOUT_CAPABLE, d as READ_ONLY_FORMAT_NODES, f as RECONSTRUCTORS, g as executeBridge, h as convertDocumentFromPdf, l as MAX_CONVERSION_PATH_NODES, m as TRANSFORMS, o as CONTENT_FORMATS, p as THROUGH_PDF_EDGE_COST, s as FORMAT_NODES, u as READ_ONLY_CONTENT_FORMATS, v as isReadOnlyContentFormat, y as isTextFormatNode } from "../composition-plan-DMTkmk8n.js";
 export { CONTENT_FORMATS, FORMAT_NODES, LAYOUT_CAPABLE, MAX_CONVERSION_PATH_NODES, READ_ONLY_CONTENT_FORMATS, READ_ONLY_FORMAT_NODES, RECONSTRUCTORS, THROUGH_PDF_EDGE_COST, TRANSFORMS, convertDocumentFromPdf, executeBridge, executeFromPdf, isReadOnlyContentFormat, isTextFormatNode, runCompositionPlan };

package/dist/convert/from-pdf.cjs

@@ -1,7 +1,7 @@
 Object.defineProperty(exports, Symbol.toStringTag, { value: "Module" });
-const require_reconstruct_presentations_prod = require("../reconstruct-presentations-prod-CA4no_aw.cjs");
+const require_reconstruct_presentations_prod = require("../reconstruct-presentations-prod-CsvGlvP2.cjs");
 const require_convert_pdf_package_tables = require("./pdf-package-tables.cjs");
-const require_composition_plan = require("../composition-plan-CSHdRyji.cjs");
+const require_composition_plan = require("../composition-plan-DJQgPvBj.cjs");
 const require_codecs_read = require("../codecs/read.cjs");
 let document_schema_js = require("document-schema.js");
 let pdf_codec_read = require("pdf-codec/read");

package/dist/convert/from-pdf.js

@@ -1,6 +1,6 @@
-import { _ as reconstructWordprocessing } from "../reconstruct-presentations-prod-CC8ct-5O.js";
+import { _ as reconstructWordprocessing } from "../reconstruct-presentations-prod-D1KbGPxa.js";
 import { stampPdfPackageTables } from "./pdf-package-tables.js";
-import { h as convertDocumentFromPdf } from "../composition-plan-DQ8lPKIW.js";
+import { h as convertDocumentFromPdf } from "../composition-plan-DMTkmk8n.js";
 import { CONTENT_READERS } from "../codecs/read.js";
 import { assembleTree } from "document-schema.js";
 import { readPdf } from "pdf-codec/read";

package/dist/convert/local.cjs

@@ -1,7 +1,7 @@
 Object.defineProperty(exports, Symbol.toStringTag, { value: "Module" });
 const require_convert_capability = require("./capability.cjs");
 const require_convert_port = require("./port.cjs");
-const require_composition_plan = require("../composition-plan-CSHdRyji.cjs");
+const require_composition_plan = require("../composition-plan-DJQgPvBj.cjs");
 const require_convert_composition_to_pdf = require("./composition-to-pdf.cjs");
 const require_convert_convert = require("./convert.cjs");
 //#region src/convert/local.ts

package/dist/convert/local.js

@@ -1,6 +1,6 @@
 import { UnsupportedConversionError } from "./capability.js";
 import { DOCUMENT_FORMATS } from "./port.js";
-import { i as resolveCompositionPlan } from "../composition-plan-DQ8lPKIW.js";
+import { i as resolveCompositionPlan } from "../composition-plan-DMTkmk8n.js";
 import { convertDocument } from "./composition-to-pdf.js";
 import { odfToPdf } from "./convert.js";
 //#region src/convert/local.ts

package/dist/index.cjs

@@ -93,7 +93,7 @@ const require_ports_abort = require("./ports/abort.cjs");
 const require_sheets = require("./sheets-BQCpNXsE.cjs");
 const require_layout_drawing = require("./layout/drawing.cjs");
 const require_layout_lattice = require("./layout/lattice.cjs");
-const require_reconstruct_presentations_prod = require("./reconstruct-presentations-prod-CA4no_aw.cjs");
+const require_reconstruct_presentations_prod = require("./reconstruct-presentations-prod-CsvGlvP2.cjs");
 const require_hsqldb_script = require("./hsqldb/script.cjs");
 const require_odb_csv = require("./odb/csv.cjs");
 const require_hsqldb_rowformat = require("./hsqldb/rowformat.cjs");
@@ -107,7 +107,7 @@ const require_firebird_backup = require("./firebird/backup.cjs");
 const require_odb_read = require("./odb/read.cjs");
 const require_odb_spreadsheet = require("./odb/spreadsheet.cjs");
 const require_convert_port = require("./convert/port.cjs");
-const require_composition_plan = require("./composition-plan-CSHdRyji.cjs");
+const require_composition_plan = require("./composition-plan-DJQgPvBj.cjs");
 const require_convert_composition_to_pdf = require("./convert/composition-to-pdf.cjs");
 const require_convert_convert = require("./convert/convert.cjs");
 const require_package_codec = require("./package-codec.cjs");

package/dist/index.js

@@ -92,7 +92,7 @@ import { throwIfAborted } from "./ports/abort.js";
 import { l as convertSpreadsheetToLayout } from "./sheets-DWDvGN-g.js";
 import { convertDrawingToLayout } from "./layout/drawing.js";
 import { detectGridLattice } from "./layout/lattice.js";
-import { _ as reconstructWordprocessing, g as reconstructSpreadsheet, i as reconstructPresentation, r as reconstructDrawing } from "./reconstruct-presentations-prod-CC8ct-5O.js";
+import { _ as reconstructWordprocessing, g as reconstructSpreadsheet, i as reconstructPresentation, r as reconstructDrawing } from "./reconstruct-presentations-prod-D1KbGPxa.js";
 import { HsqldbScriptParseError, displayTextFor, parseHsqldbScript } from "./hsqldb/script.js";
 import { OdbTableNotFoundError, OdbTableNotSpecifiedError } from "./odb/csv.js";
 import { HsqldbRowFormatError } from "./hsqldb/rowformat.js";
@@ -106,7 +106,7 @@ import { FirebirdBackupFormatError, SUPPORTED_BACKUP_FORMAT_VERSION, readFirebir
 import { OdbNoEmbeddedDataSourceError, OdbUnsupportedFormatError, readOdbTables } from "./odb/read.js";
 import { odbTablesToSpreadsheetDocument } from "./odb/spreadsheet.js";
 import { DOCUMENT_FORMATS, DocumentFormatSchema } from "./convert/port.js";
-import { i as resolveCompositionPlan } from "./composition-plan-DQ8lPKIW.js";
+import { i as resolveCompositionPlan } from "./composition-plan-DMTkmk8n.js";
 import { convertDocument } from "./convert/composition-to-pdf.js";
 import { OdmUnresolvedSectionError, csvToMarkdown, csvToOds, csvToPdf, csvToXlsx, docToMarkdown, docToPdf, docxToMarkdown, docxToOdt, docxToPdf, docxToPptx, epubToPdf, markdownToCsv, markdownToDocx, markdownToOdt, markdownToPdf, markdownToXlsx, odbReportToDocx, odbReportToOdt, odbReportToPdf, odbToCsv, odbToXlsx, odfToPdf, odgToPdf, odgToSvg, odmToPdf, odpToOdt, odpToPdf, odpToPptx, odsToCsv, odsToPdf, odsToXlsx, odtToDocx, odtToMarkdown, odtToOdp, odtToPdf, pptToMarkdown, pptToPdf, pptxToDocx, pptxToOdp, pptxToPdf, rtfToPdf, svgToOdg, svgToPdf, xlsToMarkdown, xlsToPdf, xlsxToCsv, xlsxToMarkdown, xlsxToOds, xlsxToPdf } from "./convert/convert.js";
 import { UnsupportedPackageFormatError, decodeDocumentPackage, decodeOdbPackage, encodeDocumentPackage } from "./package-codec.js";

package/dist/layout/reading-order.cjs

@@ -0,0 +1,202 @@
+Object.defineProperty(exports, Symbol.toStringTag, { value: "Module" });
+const require_layout_reconstruct_lines = require("./reconstruct-lines.cjs");
+let pdf_codec_fonts = require("pdf-codec/fonts");
+let pdf_codec_afm_widths = require("pdf-codec/afm-widths");
+//#region src/layout/reading-order.ts
+const MIN_COLUMN_LINES = 3;
+const MIN_PROSE_MEASURE_CHARS = 20;
+const SPACE_CHAR_CODE = 32;
+const AFM_UNITS_PER_EM = 1e3;
+/**
+* Splits a page's text items into reading-order flows at the column gutters the page's geometry states.
+* @param items - The page's text items in content-stream order, after any table recovery has removed the text it claimed.
+* @returns The flows in reading order, and a locator placing anything else among them. A page with no gutter yields one flow whose lines are `clusterIntoLines(items)`.
+*/
+function readingOrder(items) {
+	const bands = segment(items);
+	const regions = [];
+	collectRegions(bands, [], regions);
+	return {
+		regions,
+		locate: (xPt, yPt) => locate(bands, xPt, yPt)
+	};
+}
+/**
+* Orders two placed blocks for reading: by reading-order path first, then top to bottom by y.
+* @returns Negative when `a` is read first, positive when `b` is, zero when neither precedes the other.
+*/
+function compareReadingPosition(a, b) {
+	const length = Math.min(a.path.length, b.path.length);
+	for (let i = 0; i < length; i++) {
+		const difference = a.path[i] - b.path[i];
+		if (difference !== 0) return difference;
+	}
+	return a.path.length - b.path.length || b.yPt - a.yPt;
+}
+function segment(items) {
+	const lines = require_layout_reconstruct_lines.clusterIntoLines(items);
+	const split = findColumnSplit(lines);
+	if (split === void 0) return [flowBand(lines)];
+	const before = lines.slice(0, split.startLine);
+	const band = lines.slice(split.startLine, split.endLine);
+	const after = lines.slice(split.endLine);
+	return [
+		...before.length > 0 ? segment(itemsOf(before)) : [],
+		{
+			kind: "columns",
+			gutterXPt: (split.loPt + split.hiPt) / 2,
+			left: segment(split.left),
+			right: segment(split.right),
+			topPt: band[0].baselineY,
+			bottomPt: band.at(-1).baselineY
+		},
+		...after.length > 0 ? segment(itemsOf(after)) : []
+	];
+}
+function flowBand(lines) {
+	return {
+		kind: "flow",
+		lines,
+		topPt: lines[0]?.baselineY ?? Number.POSITIVE_INFINITY,
+		bottomPt: lines.at(-1)?.baselineY ?? Number.NEGATIVE_INFINITY
+	};
+}
+function itemsOf(lines) {
+	return lines.flatMap((line) => line.items);
+}
+function collectRegions(bands, prefix, into) {
+	bands.forEach((band, index) => {
+		if (band.kind === "flow") into.push({
+			lines: band.lines,
+			path: [...prefix, index]
+		});
+		else {
+			collectRegions(band.left, [
+				...prefix,
+				index,
+				0
+			], into);
+			collectRegions(band.right, [
+				...prefix,
+				index,
+				1
+			], into);
+		}
+	});
+}
+function locate(bands, xPt, yPt) {
+	let index = bands.length - 1;
+	for (let i = 0; i + 1 < bands.length; i++) if (yPt >= (bands[i].bottomPt + bands[i + 1].topPt) / 2) {
+		index = i;
+		break;
+	}
+	const band = bands[index];
+	if (band.kind === "flow") return [index];
+	const side = xPt < band.gutterXPt ? 0 : 1;
+	return [
+		index,
+		side,
+		...locate(side === 0 ? band.left : band.right, xPt, yPt)
+	];
+}
+function findColumnSplit(lines) {
+	const items = itemsOf(lines);
+	if (lines.length < MIN_COLUMN_LINES || !items.every(isMeasurable)) return;
+	const runs = gutterRuns(lines, Math.min(...items.map(spaceAdvancePt))).sort((a, b) => b.endLine - b.startLine - (a.endLine - a.startLine) || b.hiPt - b.loPt - (a.hiPt - a.loPt) || a.startLine - b.startLine);
+	for (const run of runs) {
+		const runItems = itemsOf(lines.slice(run.startLine, run.endLine));
+		if (run.hiPt - run.loPt <= Math.max(...runItems.map(spaceAdvancePt))) continue;
+		const left = runItems.filter((item) => rightEdge(item) <= run.loPt);
+		const right = runItems.filter((item) => item.xPt >= run.hiPt);
+		if (isRunningProse(left, run.hiPt - run.loPt) && isRunningProse(right, run.hiPt - run.loPt)) return {
+			...run,
+			left,
+			right
+		};
+	}
+}
+function isMeasurable(item) {
+	return item.widthPt !== void 0 && item.writingMode === void 0 && (item.rotationDeg ?? 0) === 0;
+}
+function rightEdge(item) {
+	return item.xPt + (item.widthPt ?? 0);
+}
+function spaceAdvancePt(item) {
+	const { standardName } = (0, pdf_codec_fonts.resolveStandardFont)(item.font.family, item.font.weight === "bold", item.font.style === "italic");
+	const advance = (0, pdf_codec_afm_widths.afmWidthOfCode)(standardName, SPACE_CHAR_CODE);
+	if (advance === void 0) throw new Error(`${standardName} has no AFM width for the space glyph`);
+	return advance / AFM_UNITS_PER_EM * item.sizePt;
+}
+function freeSpans(line) {
+	const sorted = [...line.items].sort((a, b) => a.xPt - b.xPt);
+	const spans = [];
+	let reachedPt = Number.NEGATIVE_INFINITY;
+	for (const item of sorted) {
+		if (item.xPt > reachedPt) spans.push({
+			loPt: reachedPt,
+			hiPt: item.xPt
+		});
+		reachedPt = Math.max(reachedPt, rightEdge(item));
+	}
+	spans.push({
+		loPt: reachedPt,
+		hiPt: Number.POSITIVE_INFINITY
+	});
+	return spans;
+}
+function gutterRuns(lines, minWidthPt) {
+	const runs = [];
+	let open = /* @__PURE__ */ new Map();
+	const record = (span, endLine) => {
+		if (Number.isFinite(span.loPt) && Number.isFinite(span.hiPt) && endLine - span.startLine >= MIN_COLUMN_LINES) runs.push({
+			...span,
+			endLine
+		});
+	};
+	lines.forEach((line, lineIndex) => {
+		const next = /* @__PURE__ */ new Map();
+		const keep = (loPt, hiPt, startLine) => {
+			if (hiPt - loPt <= minWidthPt) return;
+			const key = `${String(loPt)}|${String(hiPt)}`;
+			const existing = next.get(key);
+			if (existing === void 0 || startLine < existing.startLine) next.set(key, {
+				loPt,
+				hiPt,
+				startLine
+			});
+		};
+		const spans = freeSpans(line);
+		for (const strip of open.values()) for (const span of spans) keep(Math.max(strip.loPt, span.loPt), Math.min(strip.hiPt, span.hiPt), strip.startLine);
+		for (const span of spans) keep(span.loPt, span.hiPt, lineIndex);
+		for (const [key, strip] of open) if (next.get(key)?.startLine !== strip.startLine) record(strip, lineIndex);
+		open = next;
+	});
+	for (const strip of open.values()) record(strip, lines.length);
+	return runs;
+}
+function isRunningProse(items, gutterWidthPt) {
+	const lines = require_layout_reconstruct_lines.clusterIntoLines(items);
+	if (lines.length < MIN_COLUMN_LINES) return false;
+	const characters = items.reduce((sum, item) => sum + item.text.length, 0);
+	const advancePt = items.reduce((sum, item) => sum + (item.widthPt ?? 0), 0);
+	if (characters === 0 || advancePt === 0) return false;
+	const leftPt = Math.min(...items.map((item) => item.xPt));
+	const rightPt = Math.max(...items.map(rightEdge));
+	if ((rightPt - leftPt) / (advancePt / characters) < MIN_PROSE_MEASURE_CHARS) return false;
+	for (const line of lines) for (const span of freeSpans(line)) if (Number.isFinite(span.loPt) && Number.isFinite(span.hiPt) && span.hiPt - span.loPt >= gutterWidthPt) return false;
+	let wrapped = 0;
+	for (let i = 0; i + 1 < lines.length; i++) {
+		const lineRightPt = Math.max(...lines[i].items.map(rightEdge));
+		const nextFirst = lines[i + 1].items[0];
+		if (rightPt - lineRightPt < firstWordWidthPt(nextFirst) + spaceAdvancePt(nextFirst)) wrapped++;
+	}
+	return wrapped * 2 > lines.length - 1;
+}
+function firstWordWidthPt(item) {
+	const text = item.text.trimStart();
+	const firstWord = text.split(/\s/u)[0] ?? text;
+	return text.length === 0 ? 0 : (item.widthPt ?? 0) * firstWord.length / item.text.length;
+}
+//#endregion
+exports.compareReadingPosition = compareReadingPosition;
+exports.readingOrder = readingOrder;

package/dist/layout/reading-order.d.cts

@@ -0,0 +1,40 @@
+import { o as TextLine } from "../reconstruct-lines-B7t66ICm.cjs";
+import { LayoutText } from "pdf-codec";
+//#region src/layout/reading-order.d.ts
+/** One flow of text in reading order: the lines clustered from it, and its position among the page's flows. */
+interface ReadingRegion {
+  /** The flow's own lines, top to bottom, exactly as clusterIntoLines returns them for the flow's items. */
+  readonly lines: readonly TextLine[];
+  /** The flow's place in reading order: compared element by element, a smaller path is read first. */
+  readonly path: readonly number[];
+}
+/** A page's text split into reading-order flows. */
+interface ReadingOrder {
+  /** Every flow on the page, in reading order. A page with no column gutter has exactly one, holding every item. */
+  readonly regions: readonly ReadingRegion[];
+  /**
+   * The reading-order path of a point on the page, for placing something that is not text (an image, a recovered table, a drawing) among the flows: the band its y falls in and, inside a column band, the column its x falls in.
+   * @param xPt - PDF-space x of the thing being placed; `Number.NEGATIVE_INFINITY` for something with no single column, which places it in the leftmost one.
+   * @param yPt - PDF-space y the caller orders that thing by among the flows' own baselines.
+   */
+  locate: (xPt: number, yPt: number) => readonly number[];
+}
+/**
+ * Splits a page's text items into reading-order flows at the column gutters the page's geometry states.
+ * @param items - The page's text items in content-stream order, after any table recovery has removed the text it claimed.
+ * @returns The flows in reading order, and a locator placing anything else among them. A page with no gutter yields one flow whose lines are `clusterIntoLines(items)`.
+ */
+declare function readingOrder(items: readonly LayoutText[]): ReadingOrder;
+/**
+ * Orders two placed blocks for reading: by reading-order path first, then top to bottom by y.
+ * @returns Negative when `a` is read first, positive when `b` is, zero when neither precedes the other.
+ */
+declare function compareReadingPosition(a: {
+  readonly path: readonly number[];
+  readonly yPt: number;
+}, b: {
+  readonly path: readonly number[];
+  readonly yPt: number;
+}): number;
+//#endregion
+export { ReadingOrder, ReadingRegion, compareReadingPosition, readingOrder };

package/dist/layout/reading-order.d.ts

@@ -0,0 +1,40 @@
+import { o as TextLine } from "../reconstruct-lines-B7t66ICm.js";
+import { LayoutText } from "pdf-codec";
+//#region src/layout/reading-order.d.ts
+/** One flow of text in reading order: the lines clustered from it, and its position among the page's flows. */
+interface ReadingRegion {
+  /** The flow's own lines, top to bottom, exactly as clusterIntoLines returns them for the flow's items. */
+  readonly lines: readonly TextLine[];
+  /** The flow's place in reading order: compared element by element, a smaller path is read first. */
+  readonly path: readonly number[];
+}
+/** A page's text split into reading-order flows. */
+interface ReadingOrder {
+  /** Every flow on the page, in reading order. A page with no column gutter has exactly one, holding every item. */
+  readonly regions: readonly ReadingRegion[];
+  /**
+   * The reading-order path of a point on the page, for placing something that is not text (an image, a recovered table, a drawing) among the flows: the band its y falls in and, inside a column band, the column its x falls in.
+   * @param xPt - PDF-space x of the thing being placed; `Number.NEGATIVE_INFINITY` for something with no single column, which places it in the leftmost one.
+   * @param yPt - PDF-space y the caller orders that thing by among the flows' own baselines.
+   */
+  locate: (xPt: number, yPt: number) => readonly number[];
+}
+/**
+ * Splits a page's text items into reading-order flows at the column gutters the page's geometry states.
+ * @param items - The page's text items in content-stream order, after any table recovery has removed the text it claimed.
+ * @returns The flows in reading order, and a locator placing anything else among them. A page with no gutter yields one flow whose lines are `clusterIntoLines(items)`.
+ */
+declare function readingOrder(items: readonly LayoutText[]): ReadingOrder;
+/**
+ * Orders two placed blocks for reading: by reading-order path first, then top to bottom by y.
+ * @returns Negative when `a` is read first, positive when `b` is, zero when neither precedes the other.
+ */
+declare function compareReadingPosition(a: {
+  readonly path: readonly number[];
+  readonly yPt: number;
+}, b: {
+  readonly path: readonly number[];
+  readonly yPt: number;
+}): number;
+//#endregion
+export { ReadingOrder, ReadingRegion, compareReadingPosition, readingOrder };

package/dist/layout/reading-order.js

@@ -0,0 +1,200 @@
+import { clusterIntoLines } from "./reconstruct-lines.js";
+import { resolveStandardFont } from "pdf-codec/fonts";
+import { afmWidthOfCode } from "pdf-codec/afm-widths";
+//#region src/layout/reading-order.ts
+const MIN_COLUMN_LINES = 3;
+const MIN_PROSE_MEASURE_CHARS = 20;
+const SPACE_CHAR_CODE = 32;
+const AFM_UNITS_PER_EM = 1e3;
+/**
+* Splits a page's text items into reading-order flows at the column gutters the page's geometry states.
+* @param items - The page's text items in content-stream order, after any table recovery has removed the text it claimed.
+* @returns The flows in reading order, and a locator placing anything else among them. A page with no gutter yields one flow whose lines are `clusterIntoLines(items)`.
+*/
+function readingOrder(items) {
+	const bands = segment(items);
+	const regions = [];
+	collectRegions(bands, [], regions);
+	return {
+		regions,
+		locate: (xPt, yPt) => locate(bands, xPt, yPt)
+	};
+}
+/**
+* Orders two placed blocks for reading: by reading-order path first, then top to bottom by y.
+* @returns Negative when `a` is read first, positive when `b` is, zero when neither precedes the other.
+*/
+function compareReadingPosition(a, b) {
+	const length = Math.min(a.path.length, b.path.length);
+	for (let i = 0; i < length; i++) {
+		const difference = a.path[i] - b.path[i];
+		if (difference !== 0) return difference;
+	}
+	return a.path.length - b.path.length || b.yPt - a.yPt;
+}
+function segment(items) {
+	const lines = clusterIntoLines(items);
+	const split = findColumnSplit(lines);
+	if (split === void 0) return [flowBand(lines)];
+	const before = lines.slice(0, split.startLine);
+	const band = lines.slice(split.startLine, split.endLine);
+	const after = lines.slice(split.endLine);
+	return [
+		...before.length > 0 ? segment(itemsOf(before)) : [],
+		{
+			kind: "columns",
+			gutterXPt: (split.loPt + split.hiPt) / 2,
+			left: segment(split.left),
+			right: segment(split.right),
+			topPt: band[0].baselineY,
+			bottomPt: band.at(-1).baselineY
+		},
+		...after.length > 0 ? segment(itemsOf(after)) : []
+	];
+}
+function flowBand(lines) {
+	return {
+		kind: "flow",
+		lines,
+		topPt: lines[0]?.baselineY ?? Number.POSITIVE_INFINITY,
+		bottomPt: lines.at(-1)?.baselineY ?? Number.NEGATIVE_INFINITY
+	};
+}
+function itemsOf(lines) {
+	return lines.flatMap((line) => line.items);
+}
+function collectRegions(bands, prefix, into) {
+	bands.forEach((band, index) => {
+		if (band.kind === "flow") into.push({
+			lines: band.lines,
+			path: [...prefix, index]
+		});
+		else {
+			collectRegions(band.left, [
+				...prefix,
+				index,
+				0
+			], into);
+			collectRegions(band.right, [
+				...prefix,
+				index,
+				1
+			], into);
+		}
+	});
+}
+function locate(bands, xPt, yPt) {
+	let index = bands.length - 1;
+	for (let i = 0; i + 1 < bands.length; i++) if (yPt >= (bands[i].bottomPt + bands[i + 1].topPt) / 2) {
+		index = i;
+		break;
+	}
+	const band = bands[index];
+	if (band.kind === "flow") return [index];
+	const side = xPt < band.gutterXPt ? 0 : 1;
+	return [
+		index,
+		side,
+		...locate(side === 0 ? band.left : band.right, xPt, yPt)
+	];
+}
+function findColumnSplit(lines) {
+	const items = itemsOf(lines);
+	if (lines.length < MIN_COLUMN_LINES || !items.every(isMeasurable)) return;
+	const runs = gutterRuns(lines, Math.min(...items.map(spaceAdvancePt))).sort((a, b) => b.endLine - b.startLine - (a.endLine - a.startLine) || b.hiPt - b.loPt - (a.hiPt - a.loPt) || a.startLine - b.startLine);
+	for (const run of runs) {
+		const runItems = itemsOf(lines.slice(run.startLine, run.endLine));
+		if (run.hiPt - run.loPt <= Math.max(...runItems.map(spaceAdvancePt))) continue;
+		const left = runItems.filter((item) => rightEdge(item) <= run.loPt);
+		const right = runItems.filter((item) => item.xPt >= run.hiPt);
+		if (isRunningProse(left, run.hiPt - run.loPt) && isRunningProse(right, run.hiPt - run.loPt)) return {
+			...run,
+			left,
+			right
+		};
+	}
+}
+function isMeasurable(item) {
+	return item.widthPt !== void 0 && item.writingMode === void 0 && (item.rotationDeg ?? 0) === 0;
+}
+function rightEdge(item) {
+	return item.xPt + (item.widthPt ?? 0);
+}
+function spaceAdvancePt(item) {
+	const { standardName } = resolveStandardFont(item.font.family, item.font.weight === "bold", item.font.style === "italic");
+	const advance = afmWidthOfCode(standardName, SPACE_CHAR_CODE);
+	if (advance === void 0) throw new Error(`${standardName} has no AFM width for the space glyph`);
+	return advance / AFM_UNITS_PER_EM * item.sizePt;
+}
+function freeSpans(line) {
+	const sorted = [...line.items].sort((a, b) => a.xPt - b.xPt);
+	const spans = [];
+	let reachedPt = Number.NEGATIVE_INFINITY;
+	for (const item of sorted) {
+		if (item.xPt > reachedPt) spans.push({
+			loPt: reachedPt,
+			hiPt: item.xPt
+		});
+		reachedPt = Math.max(reachedPt, rightEdge(item));
+	}
+	spans.push({
+		loPt: reachedPt,
+		hiPt: Number.POSITIVE_INFINITY
+	});
+	return spans;
+}
+function gutterRuns(lines, minWidthPt) {
+	const runs = [];
+	let open = /* @__PURE__ */ new Map();
+	const record = (span, endLine) => {
+		if (Number.isFinite(span.loPt) && Number.isFinite(span.hiPt) && endLine - span.startLine >= MIN_COLUMN_LINES) runs.push({
+			...span,
+			endLine
+		});
+	};
+	lines.forEach((line, lineIndex) => {
+		const next = /* @__PURE__ */ new Map();
+		const keep = (loPt, hiPt, startLine) => {
+			if (hiPt - loPt <= minWidthPt) return;
+			const key = `${String(loPt)}|${String(hiPt)}`;
+			const existing = next.get(key);
+			if (existing === void 0 || startLine < existing.startLine) next.set(key, {
+				loPt,
+				hiPt,
+				startLine
+			});
+		};
+		const spans = freeSpans(line);
+		for (const strip of open.values()) for (const span of spans) keep(Math.max(strip.loPt, span.loPt), Math.min(strip.hiPt, span.hiPt), strip.startLine);
+		for (const span of spans) keep(span.loPt, span.hiPt, lineIndex);
+		for (const [key, strip] of open) if (next.get(key)?.startLine !== strip.startLine) record(strip, lineIndex);
+		open = next;
+	});
+	for (const strip of open.values()) record(strip, lines.length);
+	return runs;
+}
+function isRunningProse(items, gutterWidthPt) {
+	const lines = clusterIntoLines(items);
+	if (lines.length < MIN_COLUMN_LINES) return false;
+	const characters = items.reduce((sum, item) => sum + item.text.length, 0);
+	const advancePt = items.reduce((sum, item) => sum + (item.widthPt ?? 0), 0);
+	if (characters === 0 || advancePt === 0) return false;
+	const leftPt = Math.min(...items.map((item) => item.xPt));
+	const rightPt = Math.max(...items.map(rightEdge));
+	if ((rightPt - leftPt) / (advancePt / characters) < MIN_PROSE_MEASURE_CHARS) return false;
+	for (const line of lines) for (const span of freeSpans(line)) if (Number.isFinite(span.loPt) && Number.isFinite(span.hiPt) && span.hiPt - span.loPt >= gutterWidthPt) return false;
+	let wrapped = 0;
+	for (let i = 0; i + 1 < lines.length; i++) {
+		const lineRightPt = Math.max(...lines[i].items.map(rightEdge));
+		const nextFirst = lines[i + 1].items[0];
+		if (rightPt - lineRightPt < firstWordWidthPt(nextFirst) + spaceAdvancePt(nextFirst)) wrapped++;
+	}
+	return wrapped * 2 > lines.length - 1;
+}
+function firstWordWidthPt(item) {
+	const text = item.text.trimStart();
+	const firstWord = text.split(/\s/u)[0] ?? text;
+	return text.length === 0 ? 0 : (item.widthPt ?? 0) * firstWord.length / item.text.length;
+}
+//#endregion
+export { compareReadingPosition, readingOrder };

package/dist/layout/reconstruct-presentations-prod.cjs

@@ -1,5 +1,5 @@
 Object.defineProperty(exports, Symbol.toStringTag, { value: "Module" });
-const require_reconstruct_presentations_prod = require("../reconstruct-presentations-prod-CA4no_aw.cjs");
+const require_reconstruct_presentations_prod = require("../reconstruct-presentations-prod-CsvGlvP2.cjs");
 exports.layoutItemToVector = require_reconstruct_presentations_prod.layoutItemToVector;
 exports.lineToParagraph = require_reconstruct_presentations_prod.lineToParagraph;
 exports.reconstructDrawing = require_reconstruct_presentations_prod.reconstructDrawing;

package/dist/layout/reconstruct-presentations-prod.js

@@ -1,2 +1,2 @@
-import { a as reconstructSheet, i as reconstructPresentation, n as lineToParagraph, o as recoverPageVectors, r as reconstructDrawing, s as splitLineByLargeGaps, t as layoutItemToVector } from "../reconstruct-presentations-prod-CC8ct-5O.js";
+import { a as reconstructSheet, i as reconstructPresentation, n as lineToParagraph, o as recoverPageVectors, r as reconstructDrawing, s as splitLineByLargeGaps, t as layoutItemToVector } from "../reconstruct-presentations-prod-D1KbGPxa.js";
 export { layoutItemToVector, lineToParagraph, reconstructDrawing, reconstructPresentation, reconstructSheet, recoverPageVectors, splitLineByLargeGaps };

package/dist/layout/reconstruct-tables.cjs

@@ -1,5 +1,5 @@
 Object.defineProperty(exports, Symbol.toStringTag, { value: "Module" });
-const require_reconstruct_presentations_prod = require("../reconstruct-presentations-prod-CA4no_aw.cjs");
+const require_reconstruct_presentations_prod = require("../reconstruct-presentations-prod-CsvGlvP2.cjs");
 exports.buildGridFromLattice = require_reconstruct_presentations_prod.buildGridFromLattice;
 exports.buildGridFromTextClustering = require_reconstruct_presentations_prod.buildGridFromTextClustering;
 exports.recoverTables = require_reconstruct_presentations_prod.recoverTables;

package/dist/layout/reconstruct-tables.js

@@ -1,2 +1,2 @@
-import { b as recoverTables, v as buildGridFromLattice, y as buildGridFromTextClustering } from "../reconstruct-presentations-prod-CC8ct-5O.js";
+import { b as recoverTables, v as buildGridFromLattice, y as buildGridFromTextClustering } from "../reconstruct-presentations-prod-D1KbGPxa.js";
 export { buildGridFromLattice, buildGridFromTextClustering, recoverTables };

package/dist/layout/reconstruct.cjs

@@ -1,5 +1,5 @@
 Object.defineProperty(exports, Symbol.toStringTag, { value: "Module" });
-const require_reconstruct_presentations_prod = require("../reconstruct-presentations-prod-CA4no_aw.cjs");
+const require_reconstruct_presentations_prod = require("../reconstruct-presentations-prod-CsvGlvP2.cjs");
 exports.CELL_TYPES = require_reconstruct_presentations_prod.CELL_TYPES;
 exports.DEFAULT_COLUMN_WIDTH_FALLBACK_PT = require_reconstruct_presentations_prod.DEFAULT_COLUMN_WIDTH_FALLBACK_PT;
 exports.NO_ITEMS = require_reconstruct_presentations_prod.NO_ITEMS;

package/dist/layout/reconstruct.js

@@ -1,2 +1,2 @@
-import { _ as reconstructWordprocessing, c as CELL_TYPES, d as ROW_TYPES, f as TABLE_TYPES, g as reconstructSpreadsheet, h as nearestOfType, l as DEFAULT_COLUMN_WIDTH_FALLBACK_PT, m as indexStructure, p as ZERO_MARGINS, u as NO_ITEMS } from "../reconstruct-presentations-prod-CC8ct-5O.js";
+import { _ as reconstructWordprocessing, c as CELL_TYPES, d as ROW_TYPES, f as TABLE_TYPES, g as reconstructSpreadsheet, h as nearestOfType, l as DEFAULT_COLUMN_WIDTH_FALLBACK_PT, m as indexStructure, p as ZERO_MARGINS, u as NO_ITEMS } from "../reconstruct-presentations-prod-D1KbGPxa.js";
 export { CELL_TYPES, DEFAULT_COLUMN_WIDTH_FALLBACK_PT, NO_ITEMS, ROW_TYPES, TABLE_TYPES, ZERO_MARGINS, indexStructure, nearestOfType, reconstructSpreadsheet, reconstructWordprocessing };

package/dist/{reconstruct-presentations-prod-CA4no_aw.cjs → reconstruct-presentations-prod-CsvGlvP2.cjs}

@@ -5,6 +5,7 @@ const require_layout_cell_typing = require("./layout/cell-typing.cjs");
 const require_ports_abort = require("./ports/abort.cjs");
 const require_layout_reconstruct_lines = require("./layout/reconstruct-lines.cjs");
 const require_layout_lattice = require("./layout/lattice.cjs");
+const require_layout_reading_order = require("./layout/reading-order.cjs");
 let document_schema_js = require("document-schema.js");
 let pdf_codec_layout = require("pdf-codec/layout");
 //#region src/layout/reconstruct-tables.ts
@@ -519,19 +520,19 @@ function reconstructPageBlocks(page, pageIndex, images, headingLevels, form, str
 	}
 	const textItems = page.items.filter((i) => i.kind === "text" && !consumedText.has(i));
 	const imageItems = page.items.filter((i) => i.kind === "image");
-	const lines = require_layout_reconstruct_lines.clusterIntoLines(textItems);
-	const paragraphs = require_layout_reconstruct_lines.clusterIntoParagraphs(lines);
+	const order = require_layout_reading_order.readingOrder(textItems);
 	const recordDivisions = (block, items) => {
 		if (structure !== void 0) divisionByBlock.set(block, divisionChainOf(items, structure));
 	};
 	const positioned = [];
-	for (const paragraph of paragraphs) {
+	for (const { lines, path } of order.regions) for (const paragraph of require_layout_reconstruct_lines.clusterIntoParagraphs(lines)) {
 		const paragraphItems = paragraph.lines.flatMap((line) => line.items);
 		const structural = structure === void 0 ? void 0 : structureHeadingLevel(paragraphItems, structure);
 		const headingLevel = structural !== void 0 ? structural.level : headingLevelOf(paragraph, headingLevels);
 		const block = require_layout_reconstruct_lines.paragraphToContentParagraph(paragraph, pageIndex, headingLevel);
 		recordDivisions(block, paragraphItems);
 		positioned.push({
+			path,
 			yPt: paragraph.lines[0].baselineY,
 			block
 		});
@@ -554,6 +555,7 @@ function reconstructPageBlocks(page, pageIndex, images, headingLevels, form, str
 			});
 			recordDivisions(block, [img]);
 			positioned.push({
+				path: order.locate(img.xPt + img.widthPt / 2, img.yPt),
 				yPt: img.yPt,
 				block
 			});
@@ -562,6 +564,7 @@ function reconstructPageBlocks(page, pageIndex, images, headingLevels, form, str
 	for (const recovered of recoveredTables) {
 		recordDivisions(recovered.table, [...recovered.consumedText]);
 		positioned.push({
+			path: order.locate(recovered.frame.xPt + recovered.frame.widthPt / 2, recovered.topYPt),
 			yPt: recovered.topYPt,
 			block: recovered.table
 		});
@@ -571,11 +574,12 @@ function reconstructPageBlocks(page, pageIndex, images, headingLevels, form, str
 		const vectorItems = page.items.filter((item) => !claimedItems.has(item) && layoutItemToVector(item, page.heightPt) !== void 0);
 		recordDivisions(recoveredVectors.block, vectorItems);
 		positioned.push({
+			path: order.locate(Number.NEGATIVE_INFINITY, recoveredVectors.topYPt),
 			yPt: recoveredVectors.topYPt,
 			block: recoveredVectors.block
 		});
 	}
-	positioned.sort((a, b) => b.yPt - a.yPt);
+	positioned.sort(require_layout_reading_order.compareReadingPosition);
 	const blocks = positioned.map((p) => p.block);
 	reconcileLinks(blocks, page, pageIndex);
 	appendAnnotationConstructs(blocks, page, pageIndex);
@@ -772,7 +776,7 @@ function reconstructSlide(page, pageIndex, images, structure) {
 	}
 	const textItems = page.items.filter((i) => i.kind === "text" && !consumedText.has(i));
 	const imageItems = page.items.filter((i) => i.kind === "image");
-	const textShapes = clusterIntoBlocks(require_layout_reconstruct_lines.clusterIntoLines(textItems)).map((block) => blockToShape(block, page.heightPt, pageIndex));
+	const textShapes = require_layout_reading_order.readingOrder(textItems).regions.flatMap((region) => clusterIntoBlocks(region.lines)).map((block) => blockToShape(block, page.heightPt, pageIndex));
 	const imageShapes = [];
 	for (const img of imageItems) {
 		const shape = imageToShape(img, page.heightPt, pageIndex, images);

package/dist/{reconstruct-presentations-prod-CC8ct-5O.js → reconstruct-presentations-prod-D1KbGPxa.js}

@@ -5,6 +5,7 @@ import { inferCellValue } from "./layout/cell-typing.js";
 import { throwIfAborted } from "./ports/abort.js";
 import { HALF_POINT_BUCKET_PT, bucketCounts, clusterIntoLines, clusterIntoParagraphs, estimateModalLineSpacing, fontSizesClose, gapExceeds, lineBox, modeOf, paragraphToContentParagraph, pushRunsForLine, textBoxOfItem, textItemToContentRun, textItemVerticalExtent } from "./layout/reconstruct-lines.js";
 import { detectGridLattice, findColumnIndex, findRowIndex } from "./layout/lattice.js";
+import { compareReadingPosition, readingOrder } from "./layout/reading-order.js";
 import { denseTableRows } from "document-schema.js";
 import { sourcePageIndex } from "pdf-codec/layout";
 //#region src/layout/reconstruct-tables.ts
@@ -519,19 +520,19 @@ function reconstructPageBlocks(page, pageIndex, images, headingLevels, form, str
 	}
 	const textItems = page.items.filter((i) => i.kind === "text" && !consumedText.has(i));
 	const imageItems = page.items.filter((i) => i.kind === "image");
-	const lines = clusterIntoLines(textItems);
-	const paragraphs = clusterIntoParagraphs(lines);
+	const order = readingOrder(textItems);
 	const recordDivisions = (block, items) => {
 		if (structure !== void 0) divisionByBlock.set(block, divisionChainOf(items, structure));
 	};
 	const positioned = [];
-	for (const paragraph of paragraphs) {
+	for (const { lines, path } of order.regions) for (const paragraph of clusterIntoParagraphs(lines)) {
 		const paragraphItems = paragraph.lines.flatMap((line) => line.items);
 		const structural = structure === void 0 ? void 0 : structureHeadingLevel(paragraphItems, structure);
 		const headingLevel = structural !== void 0 ? structural.level : headingLevelOf(paragraph, headingLevels);
 		const block = paragraphToContentParagraph(paragraph, pageIndex, headingLevel);
 		recordDivisions(block, paragraphItems);
 		positioned.push({
+			path,
 			yPt: paragraph.lines[0].baselineY,
 			block
 		});
@@ -554,6 +555,7 @@ function reconstructPageBlocks(page, pageIndex, images, headingLevels, form, str
 			});
 			recordDivisions(block, [img]);
 			positioned.push({
+				path: order.locate(img.xPt + img.widthPt / 2, img.yPt),
 				yPt: img.yPt,
 				block
 			});
@@ -562,6 +564,7 @@ function reconstructPageBlocks(page, pageIndex, images, headingLevels, form, str
 	for (const recovered of recoveredTables) {
 		recordDivisions(recovered.table, [...recovered.consumedText]);
 		positioned.push({
+			path: order.locate(recovered.frame.xPt + recovered.frame.widthPt / 2, recovered.topYPt),
 			yPt: recovered.topYPt,
 			block: recovered.table
 		});
@@ -571,11 +574,12 @@ function reconstructPageBlocks(page, pageIndex, images, headingLevels, form, str
 		const vectorItems = page.items.filter((item) => !claimedItems.has(item) && layoutItemToVector(item, page.heightPt) !== void 0);
 		recordDivisions(recoveredVectors.block, vectorItems);
 		positioned.push({
+			path: order.locate(Number.NEGATIVE_INFINITY, recoveredVectors.topYPt),
 			yPt: recoveredVectors.topYPt,
 			block: recoveredVectors.block
 		});
 	}
-	positioned.sort((a, b) => b.yPt - a.yPt);
+	positioned.sort(compareReadingPosition);
 	const blocks = positioned.map((p) => p.block);
 	reconcileLinks(blocks, page, pageIndex);
 	appendAnnotationConstructs(blocks, page, pageIndex);
@@ -772,7 +776,7 @@ function reconstructSlide(page, pageIndex, images, structure) {
 	}
 	const textItems = page.items.filter((i) => i.kind === "text" && !consumedText.has(i));
 	const imageItems = page.items.filter((i) => i.kind === "image");
-	const textShapes = clusterIntoBlocks(clusterIntoLines(textItems)).map((block) => blockToShape(block, page.heightPt, pageIndex));
+	const textShapes = readingOrder(textItems).regions.flatMap((region) => clusterIntoBlocks(region.lines)).map((block) => blockToShape(block, page.heightPt, pageIndex));
 	const imageShapes = [];
 	for (const img of imageItems) {
 		const shape = imageToShape(img, page.heightPt, pageIndex, images);

package/package.json

@@ -1,6 +1,6 @@
 {
   "name": "js.documents",
-  "version": "14.6.0",
+  "version": "14.6.1",
   "description": "Converts between docx, pptx, odt, odp, ods, odg, xlsx, csv, svg, markdown, rtf, doc, xls, ppt and PDF through one shared, Zod-typed content schema (wpd read-only), plus read+write editors for OOXML and OpenDocument. Worker-isomorphic.",
   "type": "module",
   "repository": {