@stll/folio-core 0.60.0 → 0.62.0

package/dist/paged-layout/pageText.js

@@ -2,18 +2,23 @@
 /**
 * The plain text of a single rendered page (1-based), read by joining each of
 * the page's layout fragments' `[pmStart, pmEnd)` document text with a
-* newline. Fragments with no `pmStart`/`pmEnd` (e.g. a page-shell placeholder)
+* newline. A review projection maps those authored ranges to the displayed
+* document, excluding text the view hides. Fragments with no `pmStart`/`pmEnd`
+* (e.g. a page-shell placeholder)
 * are skipped. Returns `null` when the layout hasn't been computed yet or the
 * page number is out of range.
 */
-const getPageTextFromLayout = (layout, doc, page) => {
+const getPageTextFromLayout = ({ layout, doc, page, projection }) => {
 	if (!layout || !Number.isInteger(page) || page < 1) return null;
 	const layoutPage = layout.pages[page - 1];
 	if (!layoutPage) return null;
+	const viewDoc = projection?.type === "resolved" ? projection.doc : doc;
 	const fragmentTexts = [];
 	for (const fragment of layoutPage.fragments) {
 		if (typeof fragment.pmStart !== "number" || typeof fragment.pmEnd !== "number") continue;
-		fragmentTexts.push(doc.textBetween(fragment.pmStart, fragment.pmEnd, "\n"));
+		const from = projection?.type === "resolved" ? projection.toView(fragment.pmStart, 1) : fragment.pmStart;
+		const to = projection?.type === "resolved" ? projection.toView(fragment.pmEnd, -1) : fragment.pmEnd;
+		fragmentTexts.push(viewDoc.textBetween(from, to, "\n"));
 	}
 	return fragmentTexts.join("\n");
 };

package/dist/pdf/fonts.js

@@ -2,6 +2,7 @@ import { parseSfnt } from "../fonts/sfnt/parse.js";
 import { subsetTrueType } from "../fonts/sfnt/subset.js";
 import { toSfntBytes } from "../fonts/sfnt/woff.js";
 import { needsShaping, placeRun } from "../shaping/placeRun.js";
+import { bytesToHex, decodeText, encodeUtf16BE, encodeUtf8 } from "../utils/textEncoding.js";
 import { pdfArray, pdfAsciiString, pdfDict, pdfFlateStream, pdfName, pdfNumber, pdfNumberArray } from "./objects.js";
 import { Result, TaggedError, panic } from "better-result";
 //#region src/pdf/fonts.ts
@@ -46,7 +47,7 @@ const SUBSET_TAG_LENGTH = 6;
 const ALPHABET_LENGTH = 26;
 const UPPERCASE_A = 65;
 const WOFF_SIGNATURES = ["wOFF", "wOF2"];
-const readTag = (bytes) => String.fromCharCode(bytes[0] ?? 0, bytes[1] ?? 0, bytes[2] ?? 0, bytes[3] ?? 0);
+const readTag = (bytes) => decodeText(Uint8Array.of(bytes[0] ?? 0, bytes[1] ?? 0, bytes[2] ?? 0, bytes[3] ?? 0), { encoding: "binary" });
 /**
 * Six uppercase letters derived from the face and the exact glyph set, so the
 * tag is stable across runs and two different subsets of one family do not
@@ -148,18 +149,8 @@ const buildWinAnsiTable = () => {
 	return table;
 };
 const WIN_ANSI_BY_CODE_POINT = buildWinAnsiTable();
-const utf16BeHexOf = (text) => [...text].map((character) => utf16BeHex(character.codePointAt(0) ?? 0)).join("");
-const utf16BeHex = (codePoint) => {
-	const SUPPLEMENTARY_START = 65536;
-	const SURROGATE_HALF_BITS = 10;
-	const LOW_SURROGATE_MASK = 1023;
-	const HIGH_SURROGATE_BASE = 55296;
-	const LOW_SURROGATE_BASE = 56320;
-	const hex = (unit) => unit.toString(16).padStart(4, "0").toUpperCase();
-	if (codePoint < SUPPLEMENTARY_START) return hex(codePoint);
-	const offset = codePoint - SUPPLEMENTARY_START;
-	return hex(HIGH_SURROGATE_BASE + (offset >> SURROGATE_HALF_BITS)) + hex(LOW_SURROGATE_BASE + (offset & LOW_SURROGATE_MASK));
-};
+const utf16BeHexOf = (text) => bytesToHex(encodeUtf16BE(text)).toUpperCase();
+const utf16BeHex = (codePoint) => utf16BeHexOf(String.fromCodePoint(codePoint));
 /** Entries per `beginbfchar` block; the CMap syntax caps this at 100. */
 const BFCHAR_BLOCK_SIZE = 100;
 const toUnicodeCMap = (glyphToText) => {
@@ -290,7 +281,7 @@ const emitFont = (document, plan) => {
 		fontFileKey: font.isCff ? "FontFile3" : "FontFile2",
 		fontFileRef
 	}));
-	const toUnicodeRef = document.add(pdfFlateStream([], new TextEncoder().encode(toUnicodeCMap(plan.glyphToText))));
+	const toUnicodeRef = document.add(pdfFlateStream([], encodeUtf8(toUnicodeCMap(plan.glyphToText))));
 	const descendantRef = document.add(pdfDict([
 		["Type", pdfName("Font")],
 		["Subtype", pdfName(font.isCff ? "CIDFontType0" : "CIDFontType2")],

package/dist/pdf/images.js

@@ -1,3 +1,4 @@
+import { decodeText } from "../utils/textEncoding.js";
 import { pdfArray, pdfHexString, pdfName, pdfNumber } from "./objects.js";
 import { Result, TaggedError } from "better-result";
 import { inflateSync } from "node:zlib";
@@ -63,7 +64,7 @@ const readPngChunks = (bytes) => {
 	const CHUNK_CRC_BYTES = 4;
 	while (cursor + CHUNK_HEADER_BYTES <= bytes.length) {
 		const length = view.getUint32(cursor);
-		const type = String.fromCharCode(bytes[cursor + 4] ?? 0, bytes[cursor + 5] ?? 0, bytes[cursor + 6] ?? 0, bytes[cursor + 7] ?? 0);
+		const type = decodeText(bytes.subarray(cursor + 4, cursor + 8), { encoding: "binary" });
 		const dataStart = cursor + CHUNK_HEADER_BYTES;
 		const dataEnd = dataStart + length;
 		if (dataEnd > bytes.length) return Result.err(new PdfImageError({ message: `PNG chunk ${type} runs past the end` }));
@@ -538,7 +539,7 @@ const decodeJpeg = (bytes) => {
 		const length = view.getUint16(cursor + 2);
 		const segment = cursor + 4;
 		const frame = JPEG_FRAME_MODES[marker];
-		if (marker === APP14 && length >= 12) adobe = String.fromCharCode(...bytes.subarray(segment, segment + 5)) === "Adobe";
+		if (marker === APP14 && length >= 12) adobe = decodeText(bytes.subarray(segment, segment + 5), { encoding: "binary" }) === "Adobe";
 		if (frame !== void 0) {
 			if (!frame.dctDecodable) return Result.err(new PdfImageError({ message: `JPEG ${frame.name} frame (marker 0xff${marker.toString(16)}) has no /DCTDecode equivalent` }));
 			const height = view.getUint16(segment + 1);
@@ -565,6 +566,7 @@ const decodeImage = (source) => {
 	switch (source.format) {
 		case "png": return decodePng(source.bytes);
 		case "jpeg": return decodeJpeg(source.bytes);
+		case "svg": return Result.err(new PdfImageError({ message: "SVG images are not supported in PDF" }));
 		default: {
 			const unreachable = source.format;
 			return Result.err(new PdfImageError({ message: `unsupported image format ${String(unreachable)}` }));

package/dist/pdf/objects.d.ts

@@ -1,15 +1,4 @@
 //#region src/pdf/objects.d.ts
-/**
- * The PDF object model: values, serialization, the cross-reference table and
- * the trailer.
- *
- * Everything a backend writes into a PDF file passes through here, which is
- * what makes byte-for-byte determinism enforceable rather than aspirational:
- * numbers go through one formatter, dictionaries carry their entries as an
- * ordered list (never an object or a `Map` whose iteration order depends on
- * how it was built), and object numbers come from one allocator in call
- * order.
- */
 /** The one number formatter. Every number in the file is emitted through it. */
 declare const formatNumber: (value: number) => string;
 type PdfRef = {

package/dist/pdf/objects.js

@@ -1,3 +1,4 @@
+import { bytesToHex, encodeBinaryString, encodeUtf16BE, encodeUtf8 } from "../utils/textEncoding.js";
 import { panic } from "better-result";
 import { deflateSync } from "node:zlib";
 //#region src/pdf/objects.ts
@@ -94,7 +95,10 @@ const pdfFlateStream = (entries, data) => pdfStream(pdfDict([...entries, ["Filte
 const NAME_REGULAR = /^[A-Za-z0-9._\-+]$/u;
 const encodeName = (value) => {
 	let out = "/";
-	for (const char of value) out += NAME_REGULAR.test(char) ? char : `#${char.charCodeAt(0).toString(16).padStart(2, "0").toUpperCase()}`;
+	for (const byte of encodeUtf8(value)) {
+		const char = String.fromCharCode(byte);
+		out += NAME_REGULAR.test(char) ? char : `#${byte.toString(16).padStart(2, "0").toUpperCase()}`;
+	}
 	return out;
 };
 const encodeAsciiString = (value) => {
@@ -108,11 +112,7 @@ const encodeAsciiString = (value) => {
 	}
 	return `${out})`;
 };
-const encodeTextString = (value) => {
-	let hex = "FEFF";
-	for (let index = 0; index < value.length; index += 1) hex += value.charCodeAt(index).toString(16).padStart(4, "0").toUpperCase();
-	return `<${hex}>`;
-};
+const encodeTextString = (value) => `<FEFF${bytesToHex(encodeUtf16BE(value)).toUpperCase()}>`;
 const createByteSink = () => {
 	const chunks = [];
 	let length = 0;
@@ -122,9 +122,7 @@ const createByteSink = () => {
 	};
 	return {
 		writeAscii: (text) => {
-			const bytes = new Uint8Array(text.length);
-			for (let index = 0; index < text.length; index += 1) bytes[index] = text.charCodeAt(index) & 255;
-			writeBytes(bytes);
+			writeBytes(encodeBinaryString(text));
 		},
 		writeBytes,
 		offset: () => length,

package/dist/pdf/writePdf.js

@@ -1,4 +1,5 @@
 import { getShaper } from "../shaping/shaper.js";
+import { encodeUtf8 } from "../utils/textEncoding.js";
 import { collectResources, renderContentStream } from "./contentStream.js";
 import { prepareFonts } from "./fonts.js";
 import { decodeImage } from "./images.js";
@@ -288,7 +289,7 @@ const writePdf = async (list, options) => {
 			fonts: byFontIndex
 		});
 		const resources = collectResources(parts);
-		const contentRef = document.add(pdfFlateStream([], new TextEncoder().encode(renderContentStream(parts, resources))));
+		const contentRef = document.add(pdfFlateStream([], encodeUtf8(renderContentStream(parts, resources))));
 		const annotationRefs = [];
 		for (const link of page.links) {
 			const annotation = buildLinkAnnotation({

package/dist/prosemirror/commands/comments.js

@@ -13,11 +13,13 @@ import "../extensions/nodes/SdtExtension.js";
 import { canonicalFieldNode } from "../fieldRepresentation.js";
 import { resolveStateStory } from "../markupViewProjection.js";
 import { joinAtParagraphMark } from "../paragraphMarkJoin.js";
+import { carryParagraphProperties, paragraphLeftAfter } from "../paragraphPropertyCarry.js";
 import { getDocumentNumbering } from "../plugins/documentNumbering.js";
 import { getDocumentStyleResolver } from "../plugins/documentStyleState.js";
 import { resolvedPreservedXmlAttrs } from "../preservedXmlReview.js";
 import { setParagraphAttrsWithRebasedRunFormatting } from "../rebaseParagraphRunFormatting.js";
 import { rejoinRunsAt } from "../rejoinRunCarriers.js";
+import { commentPoint, resolvedCommentPoints } from "../replacedAnnotations.js";
 import { getFolioNodeRevisionCarriers, nodePropertyRevisionSites } from "../revisionCarriers.js";
 import { RUN_FORMATTING_MARK_NAMES } from "../runFormattingMarkNames.js";
 import { reconstructRejectedRunFormattingMarks, restoreHistoricalRunFormatting } from "../runPropertyChangeResolution.js";
@@ -100,6 +102,16 @@ function resolveChange(from, to, mode, revisionIds) {
 		const removeKeptMarksInBulk = revisionSet === null && keepType !== void 0;
 		const matchesRevision = (mark) => revisionSet === null || typeof mark.attrs["revisionId"] === "number" && revisionSet.has(mark.attrs["revisionId"]);
 		if (dispatch) {
+			const points = mode === "accept" ? resolvedCommentPoints({
+				doc: state.doc,
+				retiresStructure: (node, position) => {
+					if (node.type.name === "paragraph") return collectPPrMarkOp(node, position, from, to, mode, revisionSet)?.action === "join";
+					if (!rangeCoversNode(from, to, position, node)) return false;
+					if (node.type.name === "tableRow") return collectTableRowStructuralOp(node, position, mode, revisionSet)?.action === "remove";
+					return collectTableCellStructuralOps(node, position, mode, revisionSet).some((op) => op.type === "membership" && op.action === "remove");
+				},
+				removes: (node, position) => position >= from && position + node.nodeSize <= to && node.marks.some((mark) => mark.type === removeType && matchesRevision(mark))
+			}) : /* @__PURE__ */ new Map();
 			const tr = state.tr;
 			tr.setMeta(JOINED_RUNS_RESTYLED_META, true);
 			const deleteRanges = [];
@@ -110,6 +122,25 @@ function resolveChange(from, to, mode, revisionIds) {
 			const tableCellStructuralOps = [];
 			let removedSectionEndpointCount = 0;
 			const removedSectionReferences = [];
+			const carryRestoredProperties = (position) => {
+				if (mode !== "reject" || revisionSet === null) return;
+				const source = tr.doc.nodeAt(position);
+				const mark = source ? expectParagraphAttrs(source).pPrMark : null;
+				if (!source || mark?.kind !== "del" && mark?.kind !== "moveFrom" || revisionSet.has(mark.info.id)) return;
+				const survivor = paragraphLeftAfter({
+					doc: tr.doc,
+					paragraphPos: position
+				});
+				if (survivor === null) return;
+				carryParagraphProperties({
+					tr,
+					position: survivor,
+					source,
+					styleResolver,
+					numbering,
+					revision: mark.info
+				});
+			};
 			state.doc.nodesBetween(from, to, (node, pos) => {
 				if (node.type.name === "paragraph") {
 					const op = collectPPrMarkOp(node, pos, from, to, mode, revisionSet);
@@ -122,18 +153,21 @@ function resolveChange(from, to, mode, revisionIds) {
 						styleResolver,
 						numbering
 					});
-					if (nextAttrs) if (nextAttrs["styleId"] !== node.attrs["styleId"] && styleResolver) setParagraphAttrsWithRebasedRunFormatting({
-						nextAttrs,
-						paragraphPosition: pos,
-						styleResolver,
-						tr,
-						...mode === "reject" && { shouldRebase: (inline) => {
-							const deleted = inline.marks.some(({ type }) => type.name === "deletion");
-							const direct = inline.marks.some(({ type }) => type.name === "runFormattingOverride");
-							return !deleted || !direct;
-						} }
-					});
-					else tr.setNodeMarkup(pos, void 0, nextAttrs);
+					if (nextAttrs) {
+						if (nextAttrs["styleId"] !== node.attrs["styleId"] && styleResolver) setParagraphAttrsWithRebasedRunFormatting({
+							nextAttrs,
+							paragraphPosition: pos,
+							styleResolver,
+							tr,
+							...mode === "reject" && { shouldRebase: (inline) => {
+								const deleted = inline.marks.some(({ type }) => type.name === "deletion");
+								const direct = inline.marks.some(({ type }) => type.name === "runFormattingOverride");
+								return !deleted || !direct;
+							} }
+						});
+						else tr.setNodeMarkup(pos, void 0, nextAttrs);
+						carryRestoredProperties(pos);
+					}
 					return true;
 				}
 				if (node.type.name === "tableRow" && rangeCoversNode(from, to, pos, node)) {
@@ -388,7 +422,10 @@ function resolveChange(from, to, mode, revisionIds) {
 						removedSectionEndpointCount++;
 						removedSectionReferences.push(...sectionReferencesOf(paragraph));
 					}
-				} catch {}
+				} catch {
+					continue;
+				}
+				if (op.markWasAdded) carryRestoredProperties(mappedPos);
 			}
 			if (removedSectionEndpointCount > 0) markTrackedSectionEndpointRemoval(tr, {
 				sourceDoc: state.doc,
@@ -416,6 +453,7 @@ function resolveChange(from, to, mode, revisionIds) {
 				if (field.type === node.type) tr.setNodeMarkup(position, void 0, node.attrs, node.marks);
 				else tr.replaceWith(position, position + field.nodeSize, node);
 			}
+			for (const [id, position] of [...points].toReversed()) tr.insert(tr.mapping.map(position, -1), commentPoint(tr.doc.type.schema, id));
 			if (tr.steps.length > 0) dispatch(tr);
 		}
 		return true;

package/dist/prosemirror/commentRangeInlineAtoms.d.ts

@@ -0,0 +1,40 @@
+import { Node } from "prosemirror-model";
+//#region src/prosemirror/commentRangeInlineAtoms.d.ts
+/**
+ * Whether an inline atom extends a comment range it sits in.
+ *
+ * - `extends`: a comment mark on the atom keeps the comment's range alive.
+ * - `inert`: the atom is invisible to the reader, so a comment mark on it would
+ *   be a ghost range nobody can see or select. Comment integrity strips the mark
+ *   and never counts the atom as range content.
+ *
+ * This is separate from revision ownership: a cached rendered page break must
+ * stay inside its tracked-change wrapper, but it can never hold a comment on
+ * its own.
+ */
+type CommentRangeInlineAtomParticipation = "extends" | "inert";
+/** Keyed by the tracked-run atom set, which the fixed-point test binds to the schema. */
+declare const COMMENT_RANGE_INLINE_ATOM_PARTICIPATION: Readonly<{
+  readonly bookmarkBoundary: "extends";
+  readonly commentReference: "inert";
+  readonly field: "extends";
+  readonly hardBreak: "extends";
+  readonly image: "extends";
+  readonly math: "extends";
+  readonly moveRangeBoundary: "extends";
+  readonly noteMarker: "extends";
+  readonly pageBreakRun: "extends";
+  readonly preservedXml: "extends";
+  readonly rangeAnchor: "inert";
+  readonly renderedPageBreak: "inert";
+  readonly shape: "extends";
+  readonly structuredField: "extends";
+  readonly symbol: "extends";
+  readonly tab: "extends";
+  readonly text: "extends";
+  readonly textBoxAnchor: "extends";
+}>;
+/** Whether a node is an inline atom that must never carry or extend a comment range. */
+declare const isCommentRangeInertAtom: (node: Node) => boolean;
+//#endregion
+export { COMMENT_RANGE_INLINE_ATOM_PARTICIPATION, CommentRangeInlineAtomParticipation, isCommentRangeInertAtom };

package/dist/prosemirror/commentRangeInlineAtoms.js

@@ -0,0 +1,33 @@
+import { panic } from "better-result";
+//#region src/prosemirror/commentRangeInlineAtoms.ts
+/** Keyed by the tracked-run atom set, which the fixed-point test binds to the schema. */
+const COMMENT_RANGE_INLINE_ATOM_PARTICIPATION = Object.freeze({
+	bookmarkBoundary: "extends",
+	commentReference: "inert",
+	field: "extends",
+	hardBreak: "extends",
+	image: "extends",
+	math: "extends",
+	moveRangeBoundary: "extends",
+	noteMarker: "extends",
+	pageBreakRun: "extends",
+	preservedXml: "extends",
+	rangeAnchor: "inert",
+	renderedPageBreak: "inert",
+	shape: "extends",
+	structuredField: "extends",
+	symbol: "extends",
+	tab: "extends",
+	text: "extends",
+	textBoxAnchor: "extends"
+});
+const participationByName = new Map(Object.entries(COMMENT_RANGE_INLINE_ATOM_PARTICIPATION));
+/** Whether a node is an inline atom that must never carry or extend a comment range. */
+const isCommentRangeInertAtom = (node) => {
+	if (!node.isInline || !node.isAtom) return false;
+	const participation = participationByName.get(node.type.name);
+	if (!participation) panic(`Inline atom ${JSON.stringify(node.type.name)} has no comment-range participation`);
+	return participation === "inert";
+};
+//#endregion
+export { COMMENT_RANGE_INLINE_ATOM_PARTICIPATION, isCommentRangeInertAtom };

package/dist/prosemirror/containerFinalParagraph.d.ts

@@ -2,15 +2,15 @@ import { Node, ResolvedPos } from "prosemirror-model";
 //#region src/prosemirror/containerFinalParagraph.d.ts
 /**
  * Whether the paragraph starting at `at` is the one its container ENDS with,
- * which is the one place a paragraph mark cannot carry a revision.
+ * where a paragraph break cannot join into a following paragraph.
  *
- * A deleted mark asks a reader to join the paragraph it ends with the one
- * after it, and an inserted mark says that break was added, so rejecting it
- * closes the paragraph back over the next one. A body, a table cell, a header
- * or footer, a note and a text box each end with a paragraph that nothing
- * follows, so neither direction can be resolved there: a consumer is left with
- * a revision it can only ignore, and the redline no longer resolves to the two
- * documents it was written from.
+ * Added boundaries normally rotate onto a preceding paragraph, so rejecting
+ * them joins back into this one. If an occupied boundary or a table prevents
+ * rotation, the final inserted paragraph must retain its own revision: the
+ * resolver removes the whole paragraph after rejecting its inserted content.
+ * A terminal deletion, by contrast, clears without removing the container's
+ * required ending. Bodies, table cells, headers, footers, notes and text boxes
+ * share these rules.
  *
  * A paragraph before a TABLE is not one of those: the table is a following
  * sibling, so the paragraph does not end anything and carries its own mark

package/dist/prosemirror/containerFinalParagraph.js

@@ -14,15 +14,15 @@ const nothingFollowsIn = (parent, fromIndex, paragraphTypeName) => {
 };
 /**
 * Whether the paragraph starting at `at` is the one its container ENDS with,
-* which is the one place a paragraph mark cannot carry a revision.
+* where a paragraph break cannot join into a following paragraph.
 *
-* A deleted mark asks a reader to join the paragraph it ends with the one
-* after it, and an inserted mark says that break was added, so rejecting it
-* closes the paragraph back over the next one. A body, a table cell, a header
-* or footer, a note and a text box each end with a paragraph that nothing
-* follows, so neither direction can be resolved there: a consumer is left with
-* a revision it can only ignore, and the redline no longer resolves to the two
-* documents it was written from.
+* Added boundaries normally rotate onto a preceding paragraph, so rejecting
+* them joins back into this one. If an occupied boundary or a table prevents
+* rotation, the final inserted paragraph must retain its own revision: the
+* resolver removes the whole paragraph after rejecting its inserted content.
+* A terminal deletion, by contrast, clears without removing the container's
+* required ending. Bodies, table cells, headers, footers, notes and text boxes
+* share these rules.
 *
 * A paragraph before a TABLE is not one of those: the table is a following
 * sibling, so the paragraph does not end anything and carries its own mark

package/dist/prosemirror/conversion/fromProseDoc.js

@@ -1,5 +1,6 @@
 import { joinCommentRangesAcrossParagraphs } from "../../docx/commentRangeJoin.js";
 import { completeCommentReferences } from "../../docx/commentReferenceCompletion.js";
+import { buildSections } from "../../docx/documentParser.js";
 import { normalizeFoldedListNumbers } from "../../docx/foldedListNumberFields.js";
 import { getDocumentSourceStyles } from "../../docx/headerFooterVerbatim.js";
 import { EDITED_PREVIEW_FINGERPRINT, imageRawXmlFingerprint } from "../../docx/imageRawXml.js";
@@ -16,6 +17,7 @@ import { createStyleEngine } from "../../style-engine/styleEngine.js";
 import { normalizeShapeTextAnchor, presetDashForOutlineAttr } from "../../types/documentEnumValues.js";
 import { presetDashForCssBorderStyle } from "../../utils/borderCss.js";
 import { canonicalJson } from "../../utils/canonicalJson.js";
+import { getFindParagraphText } from "../../utils/findReplace.js";
 import { mergeTextFormatting } from "../../utils/textFormattingMerge.js";
 import { emuToPixels, emuToStrokePixels, pixelsToEmu } from "../../utils/units.js";
 import { unchangedAlternateContentXml } from "../alternateContentAttrs.js";
@@ -397,7 +399,7 @@ function fromProseDoc(pmDoc, baseDocument, { stylesheetSource, reuse = "none" })
 	const documentBody = { content: blocks };
 	if (baseDocument?.package.document.background) documentBody.background = baseDocument.package.document.background;
 	if (baseDocument?.package.document.finalSectionProperties) documentBody.finalSectionProperties = baseDocument.package.document.finalSectionProperties;
-	if (baseDocument?.package.document.sections) documentBody.sections = baseDocument.package.document.sections;
+	if (baseDocument?.package.document.sections) documentBody.sections = buildSections(blocks, documentBody.finalSectionProperties);
 	if (baseDocument?.package.document.comments) documentBody.comments = baseDocument.package.document.comments;
 	const numbering = completeNumberingForDoc(baseDocument?.package.numbering, pmDoc);
 	if (baseDocument) {
@@ -1582,12 +1584,16 @@ function extractParagraphContent(originalParagraph, _documentCounts, emptyHyperl
 	};
 	const appendDirectRun = (node, run, coalescePlainText = false) => {
 		if (currentHyperlink) flushCurrentInline();
-		const marksKey = getMarksKey(node.marks);
+		const marksKey = getMarksKey(node.marks.filter((mark) => {
+			if (mark.type.name !== "runIdentity") return true;
+			const { emptyFormatting, preservedAttributes, preserved } = expectRunIdentityMarkAttrs(mark);
+			return emptyFormatting !== true || (preservedAttributes?.length ?? 0) > 0 || (preserved?.children?.length ?? 0) > 0;
+		}));
 		const ownerId = runIdentityId(node);
 		const currentOwnerId = currentRun ? sourceRunOwners.get(currentRun) : void 0;
 		const currentRunIsPlainText = currentRun?.content.every((runContent) => runContent.type === "text");
 		const joinsOwnedSourceRun = ownerId !== void 0 && ownerId === currentOwnerId && currentRun !== null && runsShareProperties(currentRun, run);
-		const joinsOrdinaryText = coalescePlainText && ownerId === void 0 && currentOwnerId === void 0 && currentRunIsPlainText === true && currentMarksKey === marksKey;
+		const joinsOrdinaryText = coalescePlainText && currentRun !== null && runsShareProperties(currentRun, run) && currentRunIsPlainText === true && currentMarksKey === marksKey;
 		if (currentRun && (joinsOwnedSourceRun || joinsOrdinaryText)) {
 			if (joinsOrdinaryText) {
 				for (const runContent of run.content) if (runContent.type === "text") appendTextToRun(currentRun, runContent.text);
@@ -2610,7 +2616,10 @@ const reconcileEmptyFieldResultRuns = (runs, displayRun) => {
 	return reconciled;
 };
 const synchronizeFieldDisplayText = (content, displayText, fallbackRun) => {
-	let currentText = "";
+	const currentText = getFindParagraphText({
+		type: "paragraph",
+		content
+	});
 	const visitRunsIn = (items, visit) => {
 		for (const child of items) {
 			if (child.type === "run") {
@@ -2623,9 +2632,6 @@ const synchronizeFieldDisplayText = (content, displayText, fallbackRun) => {
 	const visitRuns = (visit) => {
 		visitRunsIn(content, visit);
 	};
-	visitRuns((run) => {
-		for (const runContent of run.content) if (runContent.type === "text") currentText += runContent.text;
-	});
 	if (currentText === displayText) return content;
 	let replacedText = false;
 	visitRuns((run) => {
@@ -3319,7 +3325,7 @@ function tableCellAttrsToFormatting(attrs) {
 		if (attrs.margins && !sameResolvedValue(attrs.margins, attrs._resolvedMargins)) result.margins = buildCellMarginsFromAttrs(attrs.margins);
 		if (attrs.textDirection !== (orig.textDirection ?? void 0)) if (attrs.textDirection) result.textDirection = attrs.textDirection;
 		else delete result.textDirection;
-		return result;
+		return Object.keys(result).length > 0 ? result : void 0;
 	}
 	const authoredWidth = attrs._authoredWidth;
 	if (!(attrs.colspan > 1 || attrs.rowspan > 1 || authoredWidth !== void 0 || attrs.verticalAlign || backgroundChanged || attrs.borders || attrs.margins || attrs.textDirection)) return;

package/dist/prosemirror/conversion/toProseDoc.js

@@ -31,6 +31,7 @@ import { MOVE_RANGE_BOUNDARY_NODE_NAME } from "../extensions/nodes/MoveRangeBoun
 import { RANGE_ANCHOR_NODE_NAME } from "../extensions/nodes/RangeAnchorExtension.js";
 import "../extensions/nodes/SdtExtension.js";
 import { fieldRequiresStructuredContent } from "../fieldRepresentation.js";
+import { getInlineDisplayText } from "../inlineDisplayText.js";
 import { inlineWrapperLayer } from "../inlineWrapperStack.js";
 import { CLEARED_LIST_RENDERING_ATTRS } from "../listMarker.js";
 import { listRenderingFor } from "../listRendering.js";
@@ -42,7 +43,7 @@ import { directionFromBidi } from "../paragraphDirection.js";
 import { paragraphIndentationFromFormatting } from "../paragraphIndentation.js";
 import { lineSpacingProvenanceFromSpacing } from "../paragraphSpacing.js";
 import { RUN_IDENTITY_MARK_NAME, hasRunIdentityPayload, runIdentityAttrs } from "../runIdentity.js";
-import { continueRunIdentitiesAcrossRevisions } from "../runIdentityAcrossRevisions.js";
+import { continueSourceCarrierIdentitiesAcrossRevisions } from "../runIdentityAcrossRevisions.js";
 import { getParagraphMarkSuppressionOverrides, hasDirectRunFormatting, isParagraphMarkSuppressionEligible, resolveParagraphInheritedRunFormatting, stripParagraphMarkFormattingForBodyRuns, stripParagraphMarkOnlyFormatting, suppressParagraphMarkFormatting } from "../runStyleFormatting.js";
 import { schema } from "../schema/index.js";
 import { PRESERVED_XML_LEVELS } from "../schema/nodes.js";
@@ -112,7 +113,7 @@ const runIdentityMark = (run, nextRunIdentityId) => {
 	const payload = {
 		preservedAttributes: run.preservedAttributes,
 		preserved: run.formatting?.preserved,
-		emptyFormatting: run.formatting !== void 0 && Object.keys(run.formatting).length === 0 && (run.propertyChanges?.length ?? 0) === 0 ? true : void 0
+		emptyFormatting: run.formatting !== void 0 && Object.keys(run.formatting).length === 0 ? true : void 0
 	};
 	if (!runHasPageBreakContent(run) && !runHasMixedContent(run) && !hasRunIdentityPayload(payload)) return null;
 	return schema.mark(RUN_IDENTITY_MARK_NAME, runIdentityAttrs(nextRunIdentityId(), payload));
@@ -676,7 +677,7 @@ function convertParagraph(paragraph, styleResolver, context, extraRunFormatting,
 	if (emptyHyperlinks) attrs._emptyHyperlinks = emptyHyperlinks;
 	return createProseParagraphWithPropertySource(schema.nodes["paragraph"], paragraph, {
 		attrs,
-		content: continueRunIdentitiesAcrossRevisions(inlineNodes)
+		content: continueSourceCarrierIdentitiesAcrossRevisions(inlineNodes)
 	});
 }
 /**
@@ -1256,7 +1257,7 @@ function convertTableCell({ cell, styleResolver, context, isHeader, gridWidthPer
 		attrs.margins = effectiveFormatting.margins;
 		attrs._resolvedMargins = effectiveFormatting.margins;
 	}
-	if (formatting) attrs._originalFormatting = formatting;
+	attrs._originalFormatting = formatting ?? {};
 	if (cell.propertyChanges && cell.propertyChanges.length > 0) attrs.tcPrChange = [...cell.propertyChanges];
 	if (cell.contentControls && cell.contentControls.length > 0) attrs.contentControls = cell.contentControls;
 	if (cell.carrierStack?.length) attrs.carrierStack = cell.carrierStack;
@@ -1368,27 +1369,24 @@ function standaloneTableCellToProseMirror(cell, nodeType) {
 	});
 }
 function convertField(field, { getInheritedRunFormatting, styleResolver, nextHyperlinkInstanceIndex, runScope, textBoxAnchors }) {
-	let displayText = "";
 	let fieldFormatting;
 	let fieldPropertyChanges;
 	const inlineNodes = [];
-	const hasStructuredSourceContent = fieldResultHasPageBreakContent(field) || field.type === "simpleField" && field.content.some((content) => content.type === "hyperlink" || content.type === "preservedInline" || content.type === "inlineWrapper" || content.type === "insertion" || content.type === "deletion" || content.type === "moveFrom" || content.type === "moveTo");
+	const hasPageBreakContent = fieldResultHasPageBreakContent(field);
+	const resultRuns = field.type === "complexField" ? field.fieldResult : field.content.flatMap((content) => content.type === "run" ? [content] : []);
 	const appendRun = (run, into) => {
-		for (const content of run.content) if (content.type === "text") displayText += content.text;
 		fieldFormatting ??= run.formatting;
 		fieldPropertyChanges ??= run.propertyChanges;
-		if (field.type !== "simpleField" && !hasStructuredSourceContent) return;
 		into.push(...convertRun(run, getInheritedRunFormatting(run.formatting, field.fieldType), runScope, styleResolver, textBoxAnchors));
 	};
-	const collectDisplay = (items) => {
+	const collectFormatting = (items) => {
 		for (const item of items) switch (item.type) {
 			case "run":
-				for (const child of item.content) if (child.type === "text") displayText += child.text;
 				fieldFormatting ??= item.formatting;
 				fieldPropertyChanges ??= item.propertyChanges;
 				break;
 			case "hyperlink":
-				collectDisplay(item.children);
+				collectFormatting(item.children);
 				break;
 			case "inlineWrapper":
 			case "inlineSdt":
@@ -1396,20 +1394,16 @@ function convertField(field, { getInheritedRunFormatting, styleResolver, nextHyp
 			case "deletion":
 			case "moveFrom":
 			case "moveTo":
-				collectDisplay(item.content);
+				collectFormatting(item.content);
 				break;
 			case "simpleField":
-				collectDisplay(item.content);
+				collectFormatting(item.content);
 				break;
 			case "complexField":
-				collectDisplay(item.fieldResult);
+				collectFormatting(item.fieldResult);
 				break;
 			case "preservedInline":
-				displayText += item.text;
-				break;
-			case "mathEquation":
-				displayText += item.plainText ?? "";
-				break;
+			case "mathEquation": break;
 			case "bookmarkStart":
 			case "bookmarkEnd":
 			case "commentRangeStart":
@@ -1433,7 +1427,6 @@ function convertField(field, { getInheritedRunFormatting, styleResolver, nextHyp
 				break;
 			case "hyperlink":
 				for (const { content: child } of withInlineWrapperStacks(content.children)) if (child.type === "run") {
-					for (const runContent of child.content) if (runContent.type === "text") displayText += runContent.text;
 					fieldFormatting ??= child.formatting;
 					fieldPropertyChanges ??= child.propertyChanges;
 				}
@@ -1449,7 +1442,7 @@ function convertField(field, { getInheritedRunFormatting, styleResolver, nextHyp
 			case "deletion":
 			case "moveFrom":
 			case "moveTo":
-				collectDisplay(content.content);
+				collectFormatting(content.content);
 				itemNodes.push(...convertTrackedChange(content, content.type === "insertion" || content.type === "moveTo" ? "insertion" : "deletion", nextHyperlinkInstanceIndex, runScope, {
 					current: getInheritedRunFormatting,
 					historical: getInheritedRunFormatting
@@ -1468,8 +1461,8 @@ function convertField(field, { getInheritedRunFormatting, styleResolver, nextHyp
 		styleResolver,
 		createMark: runScope.createMark
 	});
-	const createStructuredField = fieldRequiresStructuredContent(inlineNodes);
-	const resultRuns = field.type === "simpleField" ? field.content.flatMap((content) => content.type === "run" ? [content] : []) : field.fieldResult;
+	const createStructuredField = (field.type === "simpleField" || hasPageBreakContent) && fieldRequiresStructuredContent(inlineNodes);
+	const displayText = getInlineDisplayText(schema.node("paragraph", null, inlineNodes));
 	const emptyResultRuns = displayText === "" && resultRuns.length > 0 && (field.type === "complexField" || field.content.length === resultRuns.length) && resultRuns.every((run) => run.content.every((item) => item.type === "text" && item.text === "")) ? resultRuns : void 0;
 	const resultPropertyChanges = emptyResultRuns ? emptyFieldRunPropertyChanges(emptyResultRuns) : fieldPropertyChanges;
 	if (!createStructuredField && resultPropertyChanges && resultPropertyChanges.length > 0) marks.push(schema.mark("runPropertyChange", { changes: [...resultPropertyChanges] }));