@stll/folio-core 0.32.1 → 0.33.0

package/dist/shaping/shaper.js

@@ -0,0 +1,102 @@
+import __wbg_init, { shapeRun } from "../generated/text_shaper.js";
+import { TaggedError } from "better-result";
+//#region src/shaping/shaper.ts
+/**
+* The boundary to the text shaper.
+*
+* Shaping is the one part of text layout that cannot be derived from code
+* points: Arabic letters take their form from their neighbours, lam-alef
+* ligates, Devanagari reorders and forms conjuncts, Hebrew points hang off the
+* letter they belong to. One implementation answers that question, in Rust, and
+* every caller reaches it through this module: a measurement in CSS pixels and
+* a PDF text matrix scale the same glyph ids and the same advances, so the two
+* cannot describe different text.
+*
+* The artifact is loaded on demand. It carries the OpenType layout engine and
+* the Unicode data that engine needs, and a document in Latin, Cyrillic or
+* Greek needs none of it: `getShaper` is called only once a run in a shaping
+* script is about to be measured or painted, and never before.
+*/
+/** Which way a run advances. Resolved by the producer, never guessed here. */
+const SHAPING_DIRECTION = {
+	leftToRight: "ltr",
+	rightToLeft: "rtl"
+};
+var ShaperError = class extends TaggedError("ShaperError") {};
+/**
+* The flat buffer the WebAssembly entry returns:
+* `[unitsPerEm, glyphCount, (glyphId, cluster, xAdvance, yAdvance, xOffset,
+* yOffset) * glyphCount]`. A page of Arabic is tens of thousands of glyphs, and
+* an object each would cost more in boundary crossings than the shaping itself.
+*/
+const HEADER_FIELDS = 2;
+const FIELDS_PER_GLYPH = 6;
+const decode = (buffer) => {
+	const unitsPerEm = buffer[0] ?? 0;
+	const glyphCount = buffer[1] ?? 0;
+	const glyphs = [];
+	for (let index = 0; index < glyphCount; index += 1) {
+		const at = HEADER_FIELDS + index * FIELDS_PER_GLYPH;
+		glyphs.push({
+			glyphId: buffer[at] ?? 0,
+			cluster: buffer[at + 1] ?? 0,
+			xAdvance: buffer[at + 2] ?? 0,
+			yAdvance: buffer[at + 3] ?? 0,
+			xOffset: buffer[at + 4] ?? 0,
+			yOffset: buffer[at + 5] ?? 0
+		});
+	}
+	return {
+		glyphs,
+		unitsPerEm
+	};
+};
+const shaper = { shapeRun: ({ font, faceIndex = 0, text, direction, script = "", language = "", featuresOn = [], featuresOff = [] }) => {
+	try {
+		return decode(shapeRun(font, faceIndex, text, direction === SHAPING_DIRECTION.rightToLeft, script, language, [...featuresOn], [...featuresOff]));
+	} catch (cause) {
+		throw new ShaperError({
+			message: "Could not shape a run of text",
+			cause
+		});
+	}
+} };
+let loading;
+/**
+* Loads the shaper, once, and hands back the single implementation both the
+* measurer and the PDF backend use.
+*
+* Calling this is what fetches the artifact, so a caller asks for it only after
+* establishing that some run actually needs shaping.
+*/
+const getShaper = () => {
+	loading ??= __wbg_init().then(() => {
+		resolved = shaper;
+		return shaper;
+	}).catch((cause) => {
+		loading = void 0;
+		throw new ShaperError({
+			message: "Could not initialize the text shaper",
+			cause
+		});
+	});
+	return loading;
+};
+/**
+* Whether the artifact has been loaded in this process.
+*
+* Exists for the test that a document in a non-shaping script never pays for
+* it, which is the whole reason this is a separate artifact.
+*/
+const shaperLoaded = () => loading !== void 0;
+let resolved = null;
+/**
+* The shaper if it is already loaded, and never a load.
+*
+* The measure seam is synchronous: a measurement cannot await an artifact. So
+* whoever is about to measure text that shapes resolves {@link getShaper}
+* first, and the measurer reads the result here.
+*/
+const resolvedShaper = () => resolved;
+//#endregion
+export { SHAPING_DIRECTION, ShaperError, getShaper, resolvedShaper, shaperLoaded };

package/dist/types/block-id.d.ts

@@ -11,12 +11,21 @@
  * branded type makes the divergence a compile-time error rather
  * than a silent "scrollToBlock retries 20 times and gives up".
  *
- * Two id shapes are allowed:
+ * Three id shapes are allowed:
  * - Word's `w14:paraId` (or any non-empty allocator-generated id),
  *   surfaced verbatim when the source paragraph carries one.
  * - A zero-padded `seq-NNNN` fallback derived from document order,
  *   used when the paragraph has no paraId or the paraId collides
  *   with one already taken in the same derivation pass.
+ * - A zero-padded `blank-NNNN` fallback for a paragraph that holds no
+ *   text, numbered in its own sequence.
+ *
+ * The two sequences are separate on purpose. `seq-NNNN` counts the
+ * paragraphs that carry text, and that count is the published
+ * contract: the server extractor derives the same numbers, and a
+ * citation stored against one has to keep naming the same paragraph.
+ * Numbering blank paragraphs into the same sequence would renumber
+ * every stored citation after the first blank line in a document.
  */
 type FolioBlockId = string & {
   readonly __brand: "folio.blockId";
@@ -37,7 +46,14 @@ type DeriveBlockIdInput = {
   taken: ReadonlySet<string>;
 };
 declare const deriveBlockId: ({ paraId, index, taken }: DeriveBlockIdInput) => FolioBlockId;
+/**
+ * The id for a paragraph that holds no text, when it carries no
+ * paraId to be named by. Numbered in its own sequence so the
+ * published `seq-NNNN` positions stay where they are.
+ */
+declare const deriveBlankBlockId: ({ paraId, index, taken }: DeriveBlockIdInput) => FolioBlockId;
 declare const isSequentialFolioBlockId: (id: string) => boolean;
+declare const isBlankFolioBlockId: (id: string) => boolean;
 /**
  * Extract the paraId an id was derived from, if any. Returns `null`
  * for sequential fallbacks. Accepts plain `string` so consumers that
@@ -60,4 +76,4 @@ declare const getSequentialFolioBlockIdIndex: (id: string) => number | null;
  */
 declare const isFolioBlockId: (value: unknown) => value is FolioBlockId;
 //#endregion
-export { DeriveBlockIdInput, FolioBlockId, deriveBlockId, getFolioParaIdFromBlockId, getSequentialFolioBlockIdIndex, isFolioBlockId, isSequentialFolioBlockId };
+export { DeriveBlockIdInput, FolioBlockId, deriveBlankBlockId, deriveBlockId, getFolioParaIdFromBlockId, getSequentialFolioBlockIdIndex, isBlankFolioBlockId, isFolioBlockId, isSequentialFolioBlockId };

package/dist/types/block-id.js

@@ -11,17 +11,29 @@
 * branded type makes the divergence a compile-time error rather
 * than a silent "scrollToBlock retries 20 times and gives up".
 *
-* Two id shapes are allowed:
+* Three id shapes are allowed:
 * - Word's `w14:paraId` (or any non-empty allocator-generated id),
 *   surfaced verbatim when the source paragraph carries one.
 * - A zero-padded `seq-NNNN` fallback derived from document order,
 *   used when the paragraph has no paraId or the paraId collides
 *   with one already taken in the same derivation pass.
+* - A zero-padded `blank-NNNN` fallback for a paragraph that holds no
+*   text, numbered in its own sequence.
+*
+* The two sequences are separate on purpose. `seq-NNNN` counts the
+* paragraphs that carry text, and that count is the published
+* contract: the server extractor derives the same numbers, and a
+* citation stored against one has to keep naming the same paragraph.
+* Numbering blank paragraphs into the same sequence would renumber
+* every stored citation after the first blank line in a document.
 */
 const SEQUENTIAL_BLOCK_ID_PREFIX = "seq-";
+const BLANK_BLOCK_ID_PREFIX = "blank-";
 const SEQUENTIAL_BLOCK_ID_PADDING = 4;
 const SEQUENTIAL_BLOCK_ID_PATTERN = /^seq-\d{4,}$/u;
-const formatSequentialBlockId = (index) => `${SEQUENTIAL_BLOCK_ID_PREFIX}${String(index).padStart(SEQUENTIAL_BLOCK_ID_PADDING, "0")}`;
+const BLANK_BLOCK_ID_PATTERN = /^blank-\d{4,}$/u;
+const formatWithPrefix = (prefix, index) => `${prefix}${String(index).padStart(SEQUENTIAL_BLOCK_ID_PADDING, "0")}`;
+const formatSequentialBlockId = (index) => formatWithPrefix(SEQUENTIAL_BLOCK_ID_PREFIX, index);
 /**
 * The opaque `FolioBlockId` brand has no constructor; an unchecked
 * cast is the runtime no-op that mints one. Centralising the cast
@@ -29,8 +41,22 @@ const formatSequentialBlockId = (index) => `${SEQUENTIAL_BLOCK_ID_PREFIX}${Strin
 * on.
 */
 const brand = (value) => value;
+/**
+* Whether a source id would be read back as one this module generated.
+*
+* `seq-` and `blank-` are reserved: `getFolioParaIdFromBlockId` reports an id
+* in either shape as having no source paragraph, and `isFolioBlockId` holds
+* both to their exact pattern. A `w14:paraId` that happened to look like one
+* would therefore be told apart from itself, so it is not used verbatim and
+* the paragraph takes a generated id instead. (A real `w14:paraId` is eight
+* hex digits, so this costs nothing on any package Word writes; it is here so
+* a hand-written or synthesized source cannot break the round trip.)
+*/
+const isReservedShape = (paraId) => paraId.startsWith(SEQUENTIAL_BLOCK_ID_PREFIX) || paraId.startsWith(BLANK_BLOCK_ID_PREFIX);
+const usableParaId = (paraId, taken) => paraId !== null && paraId.length > 0 && !taken.has(paraId) && !isReservedShape(paraId) ? paraId : null;
 const deriveBlockId = ({ paraId, index, taken }) => {
-	if (paraId !== null && paraId.length > 0 && !taken.has(paraId)) return brand(paraId);
+	const usable = usableParaId(paraId, taken);
+	if (usable !== null) return brand(usable);
 	let candidate = index;
 	let formatted = formatSequentialBlockId(candidate);
 	while (taken.has(formatted)) {
@@ -39,14 +65,31 @@ const deriveBlockId = ({ paraId, index, taken }) => {
 	}
 	return brand(formatted);
 };
+/**
+* The id for a paragraph that holds no text, when it carries no
+* paraId to be named by. Numbered in its own sequence so the
+* published `seq-NNNN` positions stay where they are.
+*/
+const deriveBlankBlockId = ({ paraId, index, taken }) => {
+	const usable = usableParaId(paraId, taken);
+	if (usable !== null) return brand(usable);
+	let candidate = index;
+	let formatted = formatWithPrefix(BLANK_BLOCK_ID_PREFIX, candidate);
+	while (taken.has(formatted)) {
+		candidate += 1;
+		formatted = formatWithPrefix(BLANK_BLOCK_ID_PREFIX, candidate);
+	}
+	return brand(formatted);
+};
 const isSequentialFolioBlockId = (id) => SEQUENTIAL_BLOCK_ID_PATTERN.test(id);
+const isBlankFolioBlockId = (id) => BLANK_BLOCK_ID_PATTERN.test(id);
 /**
 * Extract the paraId an id was derived from, if any. Returns `null`
 * for sequential fallbacks. Accepts plain `string` so consumers that
 * have a raw id (snapshot anchors, operation blockIds) can call it
 * without an upfront brand check.
 */
-const getFolioParaIdFromBlockId = (id) => isSequentialFolioBlockId(id) ? null : id;
+const getFolioParaIdFromBlockId = (id) => isSequentialFolioBlockId(id) || isBlankFolioBlockId(id) ? null : id;
 /**
 * The 1-based document position encoded in a sequential fallback id
 * (`seq-0011` -> 11), or `null` for paraId-backed ids. The number is
@@ -67,7 +110,8 @@ const getSequentialFolioBlockIdIndex = (id) => {
 const isFolioBlockId = (value) => {
 	if (typeof value !== "string" || value.length === 0) return false;
 	if (value.startsWith(SEQUENTIAL_BLOCK_ID_PREFIX)) return SEQUENTIAL_BLOCK_ID_PATTERN.test(value);
+	if (value.startsWith(BLANK_BLOCK_ID_PREFIX)) return BLANK_BLOCK_ID_PATTERN.test(value);
 	return true;
 };
 //#endregion
-export { deriveBlockId, getFolioParaIdFromBlockId, getSequentialFolioBlockIdIndex, isFolioBlockId, isSequentialFolioBlockId };
+export { deriveBlankBlockId, deriveBlockId, getFolioParaIdFromBlockId, getSequentialFolioBlockIdIndex, isBlankFolioBlockId, isFolioBlockId, isSequentialFolioBlockId };

package/dist/types/content.d.ts

@@ -1,2 +1,2 @@
-import { BlockContent, BlockSdt, BookmarkEnd, BookmarkStart, BreakContent, Column, Comment, CommentRangeEnd, CommentRangeStart, CommentReference, ComplexField, Deletion, DocumentBody, DrawingContent, DrawingRawXmlMode, Endnote, EndnotePosition, EndnoteProperties, Field, FieldCharContent, FieldType, FooterReference, Footnote, FootnotePosition, FootnoteProperties, HeaderFooter, HeaderFooterType, HeaderReference, Hyperlink, Image, ImageCrop, ImagePadding, ImagePosition, ImageSize, ImageTransform, ImageWrap, InlineSdt, Insertion, InstrTextContent, LineNumberRestart, MathEquation, MoveFrom, MoveFromRangeEnd, MoveFromRangeStart, MoveTo, MoveToRangeEnd, MoveToRangeStart, NoBreakHyphenContent, NoteNumberRestart, NoteReferenceContent, PageOrientation, Paragraph, ParagraphContent, ParagraphPropertyChange, PositionalTab, PropertyChangeInfo, Run, RunContent, RunPropertyChange, SdtProperties, SdtType, Section, SectionProperties, SectionStart, Shape, ShapeContent, ShapeFill, ShapeGeometryAdjustment, ShapeOutline, ShapeTextBody, ShapeType, SimpleField, SoftHyphenContent, SymbolContent, TabContent, Table, TableCell, TableCellPropertyChange, TableRow, TableRowPropertyChange, TableStructuralChangeInfo, TextBox, TextContent, TrackedChangeInfo, TrackedRunChange, VerticalAlign } from "@stll/docx-core/model";
-export type { BlockContent, BlockSdt, BookmarkEnd, BookmarkStart, BreakContent, Column, Comment, CommentRangeEnd, CommentRangeStart, CommentReference, ComplexField, Deletion, DocumentBody, DrawingContent, DrawingRawXmlMode, Endnote, EndnotePosition, EndnoteProperties, Field, FieldCharContent, FieldType, FooterReference, Footnote, FootnotePosition, FootnoteProperties, HeaderFooter, HeaderFooterType, HeaderReference, Hyperlink, Image, ImageCrop, ImagePadding, ImagePosition, ImageSize, ImageTransform, ImageWrap, InlineSdt, Insertion, InstrTextContent, LineNumberRestart, MathEquation, MoveFrom, MoveFromRangeEnd, MoveFromRangeStart, MoveTo, MoveToRangeEnd, MoveToRangeStart, NoBreakHyphenContent, NoteNumberRestart, NoteReferenceContent, PageOrientation, Paragraph, ParagraphContent, ParagraphPropertyChange, PositionalTab, PropertyChangeInfo, Run, RunContent, RunPropertyChange, SdtProperties, SdtType, Section, SectionProperties, SectionStart, Shape, ShapeContent, ShapeFill, ShapeGeometryAdjustment, ShapeOutline, ShapeTextBody, ShapeType, SimpleField, SoftHyphenContent, SymbolContent, TabContent, Table, TableCell, TableCellPropertyChange, TableRow, TableRowPropertyChange, TableStructuralChangeInfo, TextBox, TextContent, TrackedChangeInfo, TrackedRunChange, VerticalAlign };
+import { BlockContent, BlockSdt, BookmarkEnd, BookmarkStart, BreakContent, Column, Comment, CommentRangeEnd, CommentRangeStart, CommentReference, ComplexField, Deletion, DocumentBody, DrawingContent, DrawingRawXmlMode, Endnote, EndnotePosition, EndnoteProperties, Field, FieldCharContent, FieldType, FooterReference, Footnote, FootnotePosition, FootnoteProperties, HeaderFooter, HeaderFooterType, HeaderReference, Hyperlink, Image, ImageCrop, ImagePadding, ImagePosition, ImageSize, ImageTransform, ImageWrap, InlineSdt, Insertion, InstrTextContent, LineNumberRestart, MathEquation, MoveFrom, MoveFromRangeEnd, MoveFromRangeStart, MoveTo, MoveToRangeEnd, MoveToRangeStart, NoBreakHyphenContent, NoteNumberRestart, NoteReferenceContent, PageOrientation, Paragraph, ParagraphContent, ParagraphPropertyChange, PositionalTab, PropertyChangeInfo, Run, RunContent, RunPropertyChange, SdtProperties, SdtType, Section, SectionProperties, SectionStart, Shape, ShapeContent, ShapeFill, ShapeGeometryAdjustment, ShapeOutline, ShapeTextBody, ShapeType, SimpleField, SoftHyphenContent, SymbolContent, TabContent, Table, TableCell, TableCellPropertyChange, TableRow, TableRowPropertyChange, TableStructuralChangeInfo, TextBox, TextContent, TrackedChangeInfo, TrackedRunChange, TrackedRunContent, VerticalAlign } from "@stll/docx-core/model";
+export type { BlockContent, BlockSdt, BookmarkEnd, BookmarkStart, BreakContent, Column, Comment, CommentRangeEnd, CommentRangeStart, CommentReference, ComplexField, Deletion, DocumentBody, DrawingContent, DrawingRawXmlMode, Endnote, EndnotePosition, EndnoteProperties, Field, FieldCharContent, FieldType, FooterReference, Footnote, FootnotePosition, FootnoteProperties, HeaderFooter, HeaderFooterType, HeaderReference, Hyperlink, Image, ImageCrop, ImagePadding, ImagePosition, ImageSize, ImageTransform, ImageWrap, InlineSdt, Insertion, InstrTextContent, LineNumberRestart, MathEquation, MoveFrom, MoveFromRangeEnd, MoveFromRangeStart, MoveTo, MoveToRangeEnd, MoveToRangeStart, NoBreakHyphenContent, NoteNumberRestart, NoteReferenceContent, PageOrientation, Paragraph, ParagraphContent, ParagraphPropertyChange, PositionalTab, PropertyChangeInfo, Run, RunContent, RunPropertyChange, SdtProperties, SdtType, Section, SectionProperties, SectionStart, Shape, ShapeContent, ShapeFill, ShapeGeometryAdjustment, ShapeOutline, ShapeTextBody, ShapeType, SimpleField, SoftHyphenContent, SymbolContent, TabContent, Table, TableCell, TableCellPropertyChange, TableRow, TableRowPropertyChange, TableStructuralChangeInfo, TextBox, TextContent, TrackedChangeInfo, TrackedRunChange, TrackedRunContent, VerticalAlign };

package/dist/utils/clipboard.js

@@ -1,3 +1,4 @@
+import { htmlCommentEnd } from "./htmlComments.js";
 import { stripXmlDeclarations } from "./stripXmlDeclarations.js";
 import createDOMPurify from "dompurify";
 //#region src/utils/clipboard.ts
@@ -259,9 +260,9 @@ function stripHtmlComments(html) {
 			break;
 		}
 		result += html.slice(cursor, start);
-		const end = html.indexOf("-->", start + 4);
-		if (end === -1) break;
-		cursor = end + 3;
+		const end = htmlCommentEnd(html, start);
+		if (end === null) break;
+		cursor = end;
 	}
 	return result;
 }

package/dist/utils/fontResolver.js

@@ -445,6 +445,189 @@ const FONT_MAPPINGS = {
 			unitsPerEm: 2048
 		})
 	},
+	"arial narrow": {
+		googleFont: "Arimo",
+		category: "sans-serif",
+		fallbackStack: [
+			"Arial Narrow",
+			"Arimo",
+			"Arial",
+			"Helvetica",
+			"sans-serif"
+		],
+		singleLineRatio: singleLineRatioOf({
+			source: "hhea",
+			hheaAscent: 1916,
+			hheaDescent: -434,
+			hheaLineGap: 0,
+			unitsPerEm: 2048
+		})
+	},
+	helvetica: {
+		googleFont: "Arimo",
+		category: "sans-serif",
+		fallbackStack: [
+			"Helvetica",
+			"Arimo",
+			"Arial",
+			"sans-serif"
+		],
+		singleLineRatio: singleLineRatioOf({
+			source: "measured",
+			ratio: 1.2,
+			note: "Measured at 11pt against a 13.2pt reference-layout line pitch. Helvetica's hhea ascent and descent sum to exactly one em, and that 1.0 ratio is not the rendered pitch, so the formula does not apply here."
+		})
+	},
+	"bookman old style": {
+		googleFont: "EB Garamond",
+		category: "serif",
+		fallbackStack: [
+			"Bookman Old Style",
+			"EB Garamond",
+			"Georgia",
+			"serif"
+		],
+		singleLineRatio: singleLineRatioOf({
+			source: "hhea",
+			hheaAscent: 1929,
+			hheaDescent: -475,
+			hheaLineGap: 0,
+			unitsPerEm: 2048
+		})
+	},
+	"franklin gothic book": {
+		googleFont: "Libre Franklin",
+		category: "sans-serif",
+		fallbackStack: [
+			"Franklin Gothic Book",
+			"Libre Franklin",
+			"Arial",
+			"sans-serif"
+		],
+		singleLineRatio: singleLineRatioOf({
+			source: "hhea",
+			hheaAscent: 1877,
+			hheaDescent: -445,
+			hheaLineGap: 0,
+			unitsPerEm: 2048
+		})
+	},
+	"calibri light": {
+		googleFont: "Carlito",
+		category: "sans-serif",
+		fallbackStack: [
+			"Calibri Light",
+			"Calibri",
+			"Carlito",
+			"Arial",
+			"sans-serif"
+		],
+		singleLineRatio: singleLineRatioOf({
+			source: "hhea",
+			hheaAscent: 1950,
+			hheaDescent: -550,
+			hheaLineGap: 0,
+			unitsPerEm: 2048
+		})
+	},
+	candara: {
+		googleFont: "Carlito",
+		category: "sans-serif",
+		fallbackStack: [
+			"Candara",
+			"Carlito",
+			"Arial",
+			"sans-serif"
+		],
+		singleLineRatio: singleLineRatioOf({
+			source: "hhea",
+			hheaAscent: 1484,
+			hheaDescent: -564,
+			hheaLineGap: 452,
+			unitsPerEm: 2048
+		})
+	},
+	constantia: {
+		googleFont: "Caladea",
+		category: "serif",
+		fallbackStack: [
+			"Constantia",
+			"Caladea",
+			"Georgia",
+			"serif"
+		],
+		singleLineRatio: singleLineRatioOf({
+			source: "hhea",
+			hheaAscent: 1538,
+			hheaDescent: -510,
+			hheaLineGap: 452,
+			unitsPerEm: 2048
+		})
+	},
+	corbel: {
+		googleFont: "Carlito",
+		category: "sans-serif",
+		fallbackStack: [
+			"Corbel",
+			"Carlito",
+			"Arial",
+			"sans-serif"
+		],
+		singleLineRatio: singleLineRatioOf({
+			source: "hhea",
+			hheaAscent: 1523,
+			hheaDescent: -525,
+			hheaLineGap: 425,
+			unitsPerEm: 2048
+		})
+	},
+	dubai: {
+		googleFont: "Noto Sans Arabic",
+		category: "sans-serif",
+		fallbackStack: [
+			"Dubai",
+			"Noto Sans Arabic",
+			"Arial",
+			"sans-serif"
+		],
+		singleLineRatio: singleLineRatioOf({
+			source: "hhea",
+			hheaAscent: 1129,
+			hheaDescent: -559,
+			hheaLineGap: 0,
+			unitsPerEm: 1e3
+		})
+	},
+	"simplified arabic": {
+		googleFont: "Noto Naskh Arabic",
+		category: "serif",
+		fallbackStack: [
+			"Simplified Arabic",
+			"Noto Naskh Arabic",
+			"Times New Roman",
+			"serif"
+		],
+		singleLineRatio: singleLineRatioOf({
+			source: "measured",
+			ratio: 1.6582,
+			note: "Measured at 11pt against an 18.24pt reference-layout line pitch; the face ships with the office suite, not as a readable font file."
+		})
+	},
+	"sakkal majalla": {
+		googleFont: "Noto Naskh Arabic",
+		category: "serif",
+		fallbackStack: [
+			"Sakkal Majalla",
+			"Noto Naskh Arabic",
+			"Times New Roman",
+			"serif"
+		],
+		singleLineRatio: singleLineRatioOf({
+			source: "measured",
+			ratio: 1.3964,
+			note: "Measured at 11pt against a 15.36pt reference-layout line pitch; the face ships with the office suite, not as a readable font file."
+		})
+	},
 	"ms mincho": {
 		googleFont: "Noto Serif JP",
 		category: "serif",

package/dist/utils/htmlComments.d.ts

@@ -0,0 +1,17 @@
+//#region src/utils/htmlComments.d.ts
+/**
+ * Index just past the HTML comment that opens at `start`, or `null` when the
+ * comment is never closed.
+ *
+ * `<!-->` and `<!--->` are ABRUPT-CLOSING comments: the tokenizer's comment
+ * start and comment-start-dash states end the comment at that `>` (HTML
+ * §13.2.5.42-43). Scanning for `-->` from `start + 4` runs straight past them,
+ * finds nothing, and every caller then treats the rest of the document as one
+ * unterminated comment — a pasted payload that opens with `<!-->` is dropped
+ * whole and silently.
+ *
+ * Both comment strippers resolve the terminator here so the two cannot drift.
+ */
+declare const htmlCommentEnd: (html: string, start: number) => number | null;
+//#endregion
+export { htmlCommentEnd };

package/dist/utils/htmlComments.js

@@ -0,0 +1,22 @@
+//#region src/utils/htmlComments.ts
+/**
+* Index just past the HTML comment that opens at `start`, or `null` when the
+* comment is never closed.
+*
+* `<!-->` and `<!--->` are ABRUPT-CLOSING comments: the tokenizer's comment
+* start and comment-start-dash states end the comment at that `>` (HTML
+* §13.2.5.42-43). Scanning for `-->` from `start + 4` runs straight past them,
+* finds nothing, and every caller then treats the rest of the document as one
+* unterminated comment — a pasted payload that opens with `<!-->` is dropped
+* whole and silently.
+*
+* Both comment strippers resolve the terminator here so the two cannot drift.
+*/
+const htmlCommentEnd = (html, start) => {
+	if (html[start + 4] === ">") return start + 5;
+	if (html[start + 4] === "-" && html[start + 5] === ">") return start + 6;
+	const end = html.indexOf("-->", start + 4);
+	return end === -1 ? null : end + 3;
+};
+//#endregion
+export { htmlCommentEnd };

package/dist/utils/scriptSegments.d.ts

@@ -56,6 +56,17 @@ declare function scriptClassOf(cp: number): ScriptClass;
  * True when the text contains at least one complex-script code point. Callers
  * use this to skip segmentation entirely on the common all-Latin path.
  */
+/**
+ * True when a code point belongs to a right-to-left script.
+ *
+ * A subset of the complex-script ranges: Thai, Lao, Khmer, Myanmar, Tibetan and
+ * the Indic scripts shape but read left to right. Shaping needs the answer
+ * because the direction decides the order glyphs come back in, and a run shaped
+ * the wrong way round reads backwards.
+ */
+declare function isRightToLeftCodePoint(cp: number): boolean;
+/** True when the text contains at least one right-to-left code point. */
+declare function hasRightToLeft(text: string): boolean;
 declare function hasComplexScript(text: string): boolean;
 /**
  * Split text into maximal runs of one script class (CJK vs. non-CJK). Iterates
@@ -64,4 +75,4 @@ declare function hasComplexScript(text: string): boolean;
  */
 declare function segmentByScript(text: string): ScriptSegment[];
 //#endregion
-export { SCRIPT_CLASS, ScriptClass, ScriptSegment, hasCjk, hasComplexScript, isCjkCodePoint, isComplexScriptCodePoint, scriptClassOf, segmentByScript };
+export { SCRIPT_CLASS, ScriptClass, ScriptSegment, hasCjk, hasComplexScript, hasRightToLeft, isCjkCodePoint, isComplexScriptCodePoint, isRightToLeftCodePoint, scriptClassOf, segmentByScript };

package/dist/utils/scriptSegments.js

@@ -62,6 +62,22 @@ function scriptClassOf(cp) {
 * True when the text contains at least one complex-script code point. Callers
 * use this to skip segmentation entirely on the common all-Latin path.
 */
+/**
+* True when a code point belongs to a right-to-left script.
+*
+* A subset of the complex-script ranges: Thai, Lao, Khmer, Myanmar, Tibetan and
+* the Indic scripts shape but read left to right. Shaping needs the answer
+* because the direction decides the order glyphs come back in, and a run shaped
+* the wrong way round reads backwards.
+*/
+function isRightToLeftCodePoint(cp) {
+	return cp >= 1424 && cp <= 1535 || cp >= 1536 && cp <= 1791 || cp >= 1792 && cp <= 1871 || cp >= 1872 && cp <= 1919 || cp >= 1920 && cp <= 1983 || cp >= 1984 && cp <= 2047 || cp >= 2048 && cp <= 2111 || cp >= 2112 && cp <= 2143 || cp >= 2144 && cp <= 2159 || cp >= 2160 && cp <= 2207 || cp >= 2208 && cp <= 2303 || cp >= 64285 && cp <= 64335 || cp >= 64336 && cp <= 65023 || cp >= 65136 && cp <= 65279 || cp >= 68864 && cp <= 68927 || cp >= 69312 && cp <= 69375 || cp >= 69424 && cp <= 69487 || cp >= 69488 && cp <= 69551 || cp >= 125184 && cp <= 125279 || cp >= 126464 && cp <= 126719;
+}
+/** True when the text contains at least one right-to-left code point. */
+function hasRightToLeft(text) {
+	for (const ch of text) if (isRightToLeftCodePoint(ch.codePointAt(0))) return true;
+	return false;
+}
 function hasComplexScript(text) {
 	for (const ch of text) if (isComplexScriptCodePoint(ch.codePointAt(0))) return true;
 	return false;
@@ -100,4 +116,4 @@ function segmentByScript(text) {
 	return segments;
 }
 //#endregion
-export { SCRIPT_CLASS, hasCjk, hasComplexScript, isCjkCodePoint, isComplexScriptCodePoint, scriptClassOf, segmentByScript };
+export { SCRIPT_CLASS, hasCjk, hasComplexScript, hasRightToLeft, isCjkCodePoint, isComplexScriptCodePoint, isRightToLeftCodePoint, scriptClassOf, segmentByScript };

package/dist/version-comparison.d.ts

@@ -1,6 +1,6 @@
 import { FolioAIBlock } from "./ai-edits/types.js";
-import { FolioDocumentStoryHandle } from "./ai-edits/headless.js";
 import { WordDiffSegment } from "./ai-edits/word-diff.js";
+import { FolioDocumentStoryHandle } from "./ai-edits/headless.js";
 import { FOLIO_DOCUMENT_METADATA_PROPERTIES, FolioDocumentMetadataProperty, FolioDocumentPrivacyOptions, FolioDocumentPrivacyReport, FolioDocumentPrivacyTransform } from "./docx/metadataPrivacy.js";
 //#region src/version-comparison.d.ts
 /** One word-level diff segment within a `modified` block. Mirrors {@link WordDiffSegment}. */

package/package.json

@@ -1,6 +1,6 @@
 {
   "name": "@stll/folio-core",
-  "version": "0.32.1",
+  "version": "0.33.0",
   "description": "Headless, framework-neutral core of folio: the OOXML (.docx) parser, document model, ProseMirror integration, and page-layout engine. No React.",
   "keywords": [
     "document-model",
@@ -108,12 +108,14 @@
     "prepack": "bun run build",
     "pack:dry-run": "bun pm pack --dry-run",
     "typecheck": "bun ../../scripts/tsc-native.ts --noEmit -p tsconfig.build.json",
-    "test": "bun test src",
+    "test": "bun test src scripts",
+    "wasm:check": "bun ../../scripts/build-shaper-wasm.ts --check",
+    "wasm:generate": "bun ../../scripts/build-shaper-wasm.ts --write",
     "test:property": "bun test $(grep -rlF 'fc.assert' src --include='*.test.ts')",
     "perf": "bun scripts/profile-editor.ts"
   },
   "dependencies": {
-    "@stll/docx-core": "^0.18.0",
+    "@stll/docx-core": "^0.19.0",
     "@stll/docx-utils": "^0.1.0",
     "@stll/template-conditions": "^0.1.0",
     "better-result": "3.0.1",
@@ -138,6 +140,8 @@
     "yjs": "^13.6.31"
   },
   "devDependencies": {
+    "@fontsource/noto-sans-devanagari": "^5.3.0",
+    "@fontsource/noto-sans-hebrew": "^5.3.0",
     "@types/hyphen": "1.14.0"
   },
   "main": "./dist/index.js",