@file-viewer/doc 3.1.2 → 3.2.1

package/README.en.md

@@ -1,39 +1,7 @@
 # @file-viewer/doc
 
-`@file-viewer/doc` is the open-source legacy MS-DOC (`.doc`) parser and HTML renderer used by File Viewer. It is extracted from the public `msdoc-viewer@0.2.0` line and maintained as the `docjs` source package.
+Runtime package version 3.2.1.
 
-## Install
+Usage and API documentation: https://github.com/flyfish-dev/docjs
 
-```bash
-npm install @file-viewer/doc
-```
-
-Historical imports from `msdoc-viewer` remain supported through the compatibility alias package.
-
-## Usage
-
-```ts
-import { parseMsDoc, renderMsDoc } from '@file-viewer/doc'
-
-const parsed = parseMsDoc(arrayBuffer)
-const rendered = renderMsDoc(parsed)
-
-container.innerHTML = `<style>${rendered.css}</style><div class="msdoc-root">${rendered.html}</div>`
-```
-
-External links and linked images are blocked by default. Trusted integrations
-can opt in separately with `externalLinkPolicy: 'allow'` and
-`externalResourcePolicy: 'allow'`; embedded images and bookmarks keep working.
-
-The package also exports `parseMsDocToHtml`, `mountMsDoc`, `createMsDocViewer`, and `MsDocWorkerClient`.
-
-## Scope
-
-- OLE/CFB container parsing
-- WordDocument, 0Table/1Table, CLX, FKP, CHPX, PAPX, STSH, and font table parsing
-- Paragraph, character, table, image, OLE attachment, and basic field-code output
-- Embedded picture extraction via MS-DOC PICF / OfficeArt BLIP records (PNG, JPEG, DIB, TIFF), with EMF/WMF → SVG conversion for browser preview
-- Browser-native HTML/CSS rendering without server-side conversion
-- Optional module worker entry at `@file-viewer/doc/worker`
-
-This package intentionally tracks the old public `msdoc-viewer@0.2.0` code line and continues to harden that public MS-DOC path. Newer private document-engine work is not used here.
+The package's own license is MIT. Included third-party notices apply to their respective components.

package/README.md

@@ -1,269 +1,7 @@
 # @file-viewer/doc
 
-一个面向 **MS-DOC（`.doc` 二进制 Word）** 的 **TypeScript 严格类型** 解析与预览库。它把旧版 Word 文档解析为结构化 AST，再渲染为可直接嵌入任意项目的 `HTML + CSS`，同时提供 DOM Viewer 与 Web Worker 通道。
+Runtime package version 3.2.1.
 
-> 本项目处理的是 **`.doc`**，不是 `.docx`。
-> 本包基于 npm 已公开的 `msdoc-viewer@0.2.0` 旧版源码维护，不包含新版私有文档引擎实现。
+Usage and API documentation: https://github.com/flyfish-dev/docjs
 
-## 特性
-
-- **整仓 TypeScript 重构**：源码、公开 API、Worker 协议、解析 AST 全部使用 TypeScript，并开启 `strict` 模式
-- **完整声明文件产物**：构建后输出 `dist/*.d.ts`，便于社区阅读、联想、二次开发
-- **纯前端方案**：不依赖 LibreOffice、COM、ActiveX 或服务器端转换
-- **按 MS-DOC 主链路实现**：CFB/OLE → FIB → CLX/Piece Table → FKP → PAPX/CHPX/TAPX → STSH → HTML
-- **可扩展架构**：把二进制读取、属性解码、样式继承、表格布局、渲染层完全拆分
-- **支持 Web Worker**：避免大文档解析阻塞 UI 线程
-
-## 当前能力
-
-已经覆盖的主要能力：
-
-- CFB/OLE 容器读取
-- FIB（`FibBase` / `FibRgLw` / `FibRgFcLcb`）解析
-- CLX / Piece Table 文本恢复
-- FKP（PAPX / CHPX）属性读取
-- STSH 样式表解析与样式继承
-- 字体表读取
-- 常见字符样式：粗体、斜体、下划线、删除线、字号、颜色、高亮、大小写、上下标、字体切换
-- 常见段落样式：对齐、缩进、段前后距、行距、分页控制、边框
-- 表格：`sprmTDefTable`、单元格宽度、横向/纵向合并、边框、垂直对齐、nowrap、fitText
-- 图片：按 MS-DOC PICF / OfficeArt BLIP 规范提取内嵌 PNG、JPEG、DIB、TIFF，并支持 EMF/WMF 转 SVG 预览
-- OLE / ObjectPool 附件提取
-- 域代码的基础处理（例如超链接）
-- 浏览器 Viewer 与 Worker Client
-
-## 工程结构
-
-```text
-src/
-  core/
-    binary.ts        # 二进制读取器
-    cfb.ts           # CFB/OLE 容器解析
-    inflate.ts       # OfficeArt 压缩 metafile 的 zlib/deflate 解压
-    utils.ts         # 字节、编码、范围等底层工具
-  msdoc/
-    fib.ts           # FIB 读取
-    clx.ts           # CLX / Piece Table
-    fkp.ts           # FKP 页读取
-    sprm.ts          # SPRM 操作数解码
-    properties.ts    # PAPX / CHPX / TAPX 到状态对象的归并
-    styles.ts        # STSH / 样式继承
-    fonts.ts         # 字体表
-    officeart.ts     # OfficeArt 内联 shape / BLIP 辅助解析
-    objects.ts       # PICF / OfficeArt 图片与 OLE / ObjectPool
-    vector.ts        # EMF/WMF 转 SVG
-    parser.ts        # 主解析流程，输出 AST
-  render/
-    html.ts          # AST -> HTML/CSS
-  types.ts           # 完整类型声明与公共接口
-  viewer.ts          # DOM Viewer
-  worker-client.ts   # Worker 客户端
-  worker.ts          # Worker 入口
-  index.ts           # 公共导出
-
-demo/
-  index.html
-
-test/
-  smoke.mjs
-```
-
-## 安装
-
-```bash
-npm install @file-viewer/doc
-```
-
-历史 `msdoc-viewer` 包名会通过兼容别名继续导出同一套 API。
-
-## 仓库开发与构建
-
-```bash
-pnpm install
-pnpm --filter @file-viewer/doc type-check
-pnpm --filter @file-viewer/doc build
-pnpm --filter @file-viewer/doc verify:github-34
-```
-
-构建产物输出到 `dist/`：
-
-- `dist/index.js`
-- `dist/index.d.ts`
-- `dist/worker.js`
-- 其余模块的 JS / d.ts
-
-## 快速开始
-
-### 1）直接解析并渲染
-
-```ts
-import { parseMsDoc, renderMsDoc } from '@file-viewer/doc';
-
-const buffer = await file.arrayBuffer();
-const parsed = parseMsDoc(buffer);
-const rendered = renderMsDoc(parsed);
-
-document.getElementById('app')!.innerHTML = `
-  <style>${rendered.css}</style>
-  <div class="msdoc-root">${rendered.html}</div>
-`;
-```
-
-外部链接和关联图片默认阻断。确需访问可信外部资源时，应分别显式设置
-`externalLinkPolicy: 'allow'` 和 `externalResourcePolicy: 'allow'`；内嵌图片及书签不受影响。
-
-### 2）使用内置 Viewer
-
-```ts
-import { createMsDocViewer } from '@file-viewer/doc';
-
-const container = document.getElementById('app')!;
-const viewer = createMsDocViewer(container);
-
-await viewer.load(file);
-```
-
-### 3）使用 Worker 提升体验
-
-```ts
-import { createMsDocViewer, MsDocWorkerClient } from '@file-viewer/doc';
-
-const workerClient = MsDocWorkerClient.create(
-  new URL('./node_modules/@file-viewer/doc/dist/worker.js', import.meta.url)
-);
-
-const viewer = createMsDocViewer(document.getElementById('app')!, {
-  workerClient,
-});
-
-await viewer.load(file);
-```
-
-## API 概览
-
-### `parseMsDoc(input, options?)`
-
-把 `.doc` 解析成结构化结果：
-
-```ts
-import type { MsDocParseResult } from '@file-viewer/doc';
-```
-
-核心字段：
-
-- `warnings`: 解析期告警
-- `meta`: 文档元数据与统计
-- `fonts`: 字体表摘要
-- `styles`: 样式表摘要
-- `blocks`: 结构化块级 AST（段落、表格、附件）
-- `assets`: 图片 / 附件资产
-
-### `renderMsDoc(parsed, options?)`
-
-把 AST 渲染为：
-
-- `html`
-- `css`
-- `warnings`
-- `meta`
-- `assets`
-- `parsed`
-
-### `parseMsDocToHtml(input, options?)`
-
-一步完成解析与渲染。
-
-### `createMsDocViewer(container, config?)`
-
-返回一个面向浏览器的 viewer：
-
-- `load(input, options?)`
-- `mount(rendered)`
-- `clear()`
-- `destroy()`
-- `value`
-
-### `MsDocWorkerClient`
-
-为 Worker 提供强类型 RPC 包装：
-
-- `parse(input, options?)`
-- `render(parsed, options?)`
-- `parseToHtml(input, options?)`
-- `destroy()`
-
-## 关键类型
-
-所有公共类型都从根入口导出，例如：
-
-```ts
-import type {
-  MsDocParseResult,
-  MsDocRenderResult,
-  MsDocBlock,
-  ParagraphBlock,
-  TableBlock,
-  MsDocAsset,
-  MsDocParseOptions,
-  MsDocRenderOptions,
-  MsDocViewer,
-} from '@file-viewer/doc';
-```
-
-这意味着：
-
-- 社区开发者可以直接消费 `AST` 与 `asset` 数据自行渲染
-- 也可以在现有渲染器之外实现 Markdown、Canvas、PDF 等输出层
-- Worker 主线程通信协议也有完整类型约束，避免“隐式 any 消息格式”问题
-
-## 严格类型与可维护性说明
-
-这次重构的重点不只是把文件后缀改成 `.ts`，而是把整条主链路都显式类型化：
-
-- 二进制结构：CFB/FIB/CLX/FKP
-- 样式模型：字符、段落、表格状态对象
-- 中间 AST：段落、行内节点、表格、附件
-- 渲染结果：HTML/CSS/资产
-- Viewer / Worker 协议
-
-同时给关键模块补充了注释，尤其是：
-
-- `BinaryReader` 的边界读取行为
-- Worker 请求/响应通道
-- parser 主流程
-- 样式与属性归并逻辑
-
-## 本地验证
-
-```bash
-npm run smoke
-```
-
-这个命令会：
-
-1. 先运行 TypeScript 构建
-2. 再读取 `test/test.doc`
-3. 输出 `test/rendered-sample.html`
-
-如果你要手工打开 demo：
-
-```bash
-npm run build
-# 然后用任意静态服务器打开 demo/index.html
-```
-
-## 已知边界
-
-MS-DOC 是历史包袱很重的二进制格式，虽然主链路已经完整重构，但以下场景仍建议用更多真实样本持续压测：
-
-- 极端复杂的嵌套表格
-- 非常规域代码组合
-- 少见的 OLE 嵌入形式
-- 某些兼容模式下的边角 sprm 变体
-- 仅保存外部路径、未嵌入字节的链接图片（会给出占位提示，无法在离线环境加载本地盘路径）
-
-这些场景并不影响当前的整体架构，后续继续扩展时可直接在现有 TS 类型体系上增量补强。
-
-## 许可证
-
-MIT
+The package's own license is MIT. Included third-party notices apply to their respective components.

package/dist/core/binary.d.ts

@@ -2,11 +2,6 @@ import type { BinaryInput } from '../types.js';
 type ReaderMethod = ((offset: number) => number) & {
     byteWidth?: number;
 };
-/**
- * Lightweight little-endian binary reader used across the MS-DOC parser.
- * The implementation intentionally returns neutral defaults for out-of-range
- * reads so upper layers can stay resilient to slightly malformed documents.
- */
 export declare class BinaryReader {
     readonly bytes: Uint8Array;
     readonly view: DataView;

package/dist/core/binary.js

@@ -1,11 +1,6 @@
 import { textDecoder } from './utils.js';
 const utf16leDecoder = textDecoder('utf-16le');
 const latin1Decoder = typeof TextDecoder !== 'undefined' ? textDecoder('windows-1252') : textDecoder('latin1');
-/**
- * Lightweight little-endian binary reader used across the MS-DOC parser.
- * The implementation intentionally returns neutral defaults for out-of-range
- * reads so upper layers can stay resilient to slightly malformed documents.
- */
 export class BinaryReader {
     bytes;
     view;

package/dist/core/cfb.d.ts

@@ -1,7 +1,2 @@
 import type { BinaryInput, ParsedCFB } from '../types.js';
-/**
- * Parses the CFB/OLE container that wraps a legacy `.doc` file.
- * The returned object exposes stream helpers so higher layers can focus on
- * Word-specific structures instead of low-level sector navigation.
- */
 export declare function parseCFB(input: BinaryInput, _options?: Record<string, unknown>): ParsedCFB;

package/dist/core/cfb.js

@@ -105,11 +105,6 @@ function buildPathMap(entries) {
 function normalizeName(name) {
     return name.replace(/\u0000+$/, '');
 }
-/**
- * Parses the CFB/OLE container that wraps a legacy `.doc` file.
- * The returned object exposes stream helpers so higher layers can focus on
- * Word-specific structures instead of low-level sector navigation.
- */
 export function parseCFB(input, _options = {}) {
     const bytes = toUint8Array(input);
     const reader = new BinaryReader(bytes);

package/dist/core/inflate.d.ts

@@ -1,9 +1 @@
-/**
- * Small RFC1950/zlib + RFC1951/DEFLATE inflater.
- *
- * OfficeArt stores EMF/WMF BLIPs behind OfficeArtMetafileHeader.compression=0x00
- * using the zlib wrapper defined by RFC1950. Keeping this inflater dependency-free
- * preserves the browser-only build while still allowing the synchronous MS-DOC
- * parser to expose browser-displayable SVG assets.
- */
 export declare function inflateZlib(bytes: Uint8Array, expectedLength?: number): Uint8Array | null;

package/dist/core/inflate.js

@@ -1,11 +1,3 @@
-/**
- * Small RFC1950/zlib + RFC1951/DEFLATE inflater.
- *
- * OfficeArt stores EMF/WMF BLIPs behind OfficeArtMetafileHeader.compression=0x00
- * using the zlib wrapper defined by RFC1950. Keeping this inflater dependency-free
- * preserves the browser-only build while still allowing the synchronous MS-DOC
- * parser to expose browser-displayable SVG assets.
- */
 class InflateError extends Error {
     name = 'InflateError';
 }
@@ -81,8 +73,6 @@ class ByteSink {
     toUint8Array(expectedLength) {
         const out = this.buffer.slice(0, this.length);
         if (expectedLength != null && expectedLength >= 0 && out.length !== expectedLength) {
-            // The OfficeArt header is advisory in some legacy files. Do not fail the
-            // parse because a producer rounded the saved size differently.
             return out;
         }
         return out;

package/dist/msdoc/constants.js

@@ -42,7 +42,7 @@ export const JC_MAP = {
 };
 export const VERTICAL_ALIGN_MAP = {
     0: 'top',
-    1: 'center',
+    1: 'middle',
     2: 'bottom',
 };
 export const HIGHLIGHT_COLORS = {

package/dist/msdoc/fkp.js

@@ -170,7 +170,6 @@ export function readPapxProperties(wordBytes, offset, dataBytes) {
         bodyStart = offset + 2;
     }
     else {
-        // PapxInFkp.cb counts 2-byte words; the payload occupies (2 * cb) - 1 bytes.
         bodySize = cb * 2 - 1;
         bodyStart = offset + 1;
     }

package/dist/msdoc/html-word.js

@@ -216,8 +216,6 @@ function extractParagraphs(html) {
         }
         cursor = tagEnd + 1;
     }
-    // Malformed HTML saved by old Word versions is common. Preserve its visible
-    // tail as a paragraph without restarting a regex search at every opening tag.
     if (active) {
         appendParagraph(active.attributes, html.slice(active.contentStart));
     }

package/dist/msdoc/objects.d.ts

@@ -12,18 +12,7 @@ export declare function detectImageSegment(bytes: Uint8Array): {
     start: number;
     end: number;
 } | null;
-/**
- * OfficeArt records all start with the common 8-byte OfficeArtRecordHeader.
- * Parsing this explicitly lets us walk inline shape containers instead of guessing
- * image payloads by magic bytes, which was the root cause of the broken image output.
- */
 export declare function parseOfficeArtRecordHeader(bytes: Uint8Array, offset: number): OfficeArtRecordHeader | null;
-/**
- * Resolves a picture character (U+0001 + sprmCPicLocation) to an HTML-friendly asset.
- * The happy path is: PICF -> optional linked picture name -> OfficeArtInlineSpContainer
- * -> OfficeArtBlip*. When no structured BLIP can be found we still fall back to a
- * signature scan so slightly malformed files remain usable.
- */
 export declare function extractPictureAsset(dataStreamBytes: Uint8Array, pictureOffset: number | null | undefined, options?: MsDocParseOptions): ImageAsset | null;
 export declare function extractFbseImageAsset(drawingBytes: Uint8Array, offset: number, header?: OfficeArtRecordHeader | null): ImageAsset | null;
 export declare function parseOle10Native(streamBytes: Uint8Array): {

package/dist/msdoc/objects.js

@@ -182,9 +182,6 @@ function parsePicfHeader(bytes) {
     if (bytes.length < PICF_HEADER_SIZE)
         return null;
     const reader = new BinaryReader(bytes);
-    // PICMID stores the intended display goal size in twips plus mx/my scaling
-    // factors (tenths of a percent). Word uses these values for inline picture
-    // layout, including images that live in header/footer stories.
     const dxaGoalTwips = reader.i16(28);
     const dyaGoalTwips = reader.i16(30);
     const scaleXPermille = reader.u16(32);
@@ -209,11 +206,6 @@ function parsePicfHeader(bytes) {
         displayHeightTwips,
     };
 }
-/**
- * PICFAndOfficeArtData can optionally carry a length-prefixed ANSI path right after
- * the 68-byte PICF header. The spec ties this to MM_SHAPEFILE, but real-world files
- * are not always perfectly spec-compliant, so we also accept clearly path-like data.
- */
 function readOptionalPictureName(pictureChunk, picf) {
     const offset = Math.min(picf.cbHeader || PICF_HEADER_SIZE, pictureChunk.length);
     if (offset >= pictureChunk.length)
@@ -231,11 +223,6 @@ function readOptionalPictureName(pictureChunk, picf) {
     }
     return { value, nextOffset: offset + 1 + length };
 }
-/**
- * OfficeArt records all start with the common 8-byte OfficeArtRecordHeader.
- * Parsing this explicitly lets us walk inline shape containers instead of guessing
- * image payloads by magic bytes, which was the root cause of the broken image output.
- */
 export function parseOfficeArtRecordHeader(bytes, offset) {
     if (offset < 0 || offset + 8 > bytes.length)
         return null;
@@ -300,11 +287,6 @@ function dibToBmp(dibBytes) {
     out.set(dibBytes, 14);
     return out;
 }
-/**
- * OfficeArtBStoreContainerFileBlock can wrap an OfficeArtFBSE record instead of a
- * BLIP atom directly. In that case the actual image payload sits in the FBSE
- * embeddedBlip field after the fixed 36-byte FBSE header and optional nameData.
- */
 function extractEmbeddedBlipFromFbse(bytes, offset, header) {
     const payloadOffset = offset + 8;
     const payloadEnd = payloadOffset + header.recLen;
@@ -448,12 +430,6 @@ function collectOfficeArtCandidates(bytes, offset, end, out) {
         cursor = next;
     }
 }
-/**
- * The `picture` field of PICFAndOfficeArtData is an OfficeArtInlineSpContainer.
- * We recursively walk nested OfficeArt containers until we reach actual BLIP atoms
- * (PNG/JPEG/DIB/TIFF/EMF/WMF/PICT) and then extract the payload according to the
- * BLIP record-specific layout from MS-ODRAW.
- */
 function findOfficeArtCandidates(pictureChunk, startOffset) {
     if (startOffset < 0 || startOffset + 8 > pictureChunk.length)
         return [];
@@ -543,12 +519,6 @@ function createImageAsset(candidate, pictureOffset, picf, linkedPath, inlineShap
         meta,
     };
 }
-/**
- * Resolves a picture character (U+0001 + sprmCPicLocation) to an HTML-friendly asset.
- * The happy path is: PICF -> optional linked picture name -> OfficeArtInlineSpContainer
- * -> OfficeArtBlip*. When no structured BLIP can be found we still fall back to a
- * signature scan so slightly malformed files remain usable.
- */
 export function extractPictureAsset(dataStreamBytes, pictureOffset, options = {}) {
     if (!dataStreamBytes || pictureOffset == null || pictureOffset < 0 || pictureOffset + PICF_HEADER_SIZE > dataStreamBytes.length)
         return null;
@@ -712,7 +682,6 @@ export function parseOle10Native(streamBytes) {
             }
         }
         catch {
-            // try next variant
         }
     }
     return null;

package/dist/msdoc/officeart.d.ts

@@ -1,10 +1,4 @@
 export declare function detectMimeFromBlipType(blipType: number): string | null;
-/**
- * Inline picture characters often carry a miniature OfficeArt container that
- * points at the BLIP used by the picture. We only need a compact subset of the
- * shape metadata so the higher-level parser can annotate extracted assets with
- * source shape ids and BLIP indices.
- */
 export declare function parseInlineOfficeArtShape(bytes: Uint8Array, startOffset?: number): {
     shapeId: number;
     blipIndex?: number;

package/dist/msdoc/officeart.js

@@ -125,12 +125,6 @@ function scanRecords(bytes, start, end, visitor) {
     }
     return false;
 }
-/**
- * Inline picture characters often carry a miniature OfficeArt container that
- * points at the BLIP used by the picture. We only need a compact subset of the
- * shape metadata so the higher-level parser can annotate extracted assets with
- * source shape ids and BLIP indices.
- */
 export function parseInlineOfficeArtShape(bytes, startOffset = 0) {
     let result = null;
     scanRecords(bytes, startOffset, bytes.length, (offset, header) => {

package/dist/msdoc/parser.d.ts

@@ -1,10 +1,5 @@
 import type { MsDocParseOptions, MsDocParseResult, ParagraphModel, TableRowBlock } from '../types.js';
 export declare function finalizeTableGrid(rows: TableRowBlock[]): void;
 export declare function normalizeRowEndOnlyTables(paragraphs: ParagraphModel[]): void;
-/**
- * Main MS-DOC entry point.
- * It parses the OLE container, restores text through the piece table, resolves
- * paragraph/character/table properties, and finally produces a normalized AST
- * that the HTML renderer can consume.
- */
+export declare function buildBlocks(paragraphs: ParagraphModel[]): MsDocParseResult['blocks'];
 export declare function parseMsDoc(input: ArrayBuffer | Uint8Array | ArrayBufferView, options?: MsDocParseOptions): MsDocParseResult;

package/dist/msdoc/parser.js

@@ -337,7 +337,6 @@ function buildParagraphModel(range, paragraphText, styles, fonts, chpxRuns, reso
     const resolveFont = (fontId) => fonts.byIndex(fontId);
     const segments = buildCharSegments(range, paragraphText, chpxRuns, styles, baseCharProps, resolveFont, chpxCursor, range.storyKind === 'textbox' || range.storyKind === 'header-textbox');
     const inlines = buildInlineNodes(segments, resolveAsset);
-    // The terminator has its own CHPX run and is not part of paragraphText.
     const markState = range.terminator === DOC_CONTROL.paragraph
         ? buildCharSegments({ ...range, cpStart: range.cpEnd - 1 }, range.terminator, chpxRuns, styles, baseCharProps, resolveFont, { index: chpxCursor.index })[0]?.state
         : undefined;
@@ -440,10 +439,6 @@ export function finalizeTableGrid(rows) {
             }
         }
     }
-    // HTML tables need one shared column grid even when Word stores a different
-    // TDefTable boundary list per row (usually because cells are merged). Map
-    // every visible cell back onto the union of physical boundaries so colspan
-    // preserves the original geometry instead of squeezing later row cells.
     const boundaries = Array.from(new Set(rows.flatMap((row) => row.cells.flatMap((cell) => {
         const left = cell.meta?.leftBoundary;
         const right = cell.meta?.rightBoundary;
@@ -468,8 +463,6 @@ export function finalizeTableGrid(rows) {
 function resolveTableGridBorders(rows) {
     const hasOwnBorder = (borders, side) => (Object.prototype.hasOwnProperty.call(borders, side) && (() => {
         const border = borders[side];
-        // An all-zero BRC is an empty cell-level value and still inherits the
-        // table edge/inside border. NilBrc is the explicit "no border" value.
         return Boolean(border?.nil || border?.borderType);
     })());
     rows.forEach((row, rowIndex) => {
@@ -557,11 +550,6 @@ export function normalizeRowEndOnlyTables(paragraphs) {
         }
         if (remainingCells > 0)
             continue;
-        // Some Word/WPS producers store all table properties only on the row-end
-        // PAPX. Paragraph marks inside a cell then look like ordinary paragraphs.
-        // When the immediately preceding row has the same outer grid and table
-        // border identity, everything between the two row terminators belongs to
-        // the current row, including leading paragraphs in its first cell.
         let previousRowEndIndex = cursor;
         while (previousRowEndIndex >= 0
             && !paragraphs[previousRowEndIndex].paraState.tableRowEnd
@@ -581,87 +569,91 @@ export function normalizeRowEndOnlyTables(paragraphs) {
         rowParagraphs.forEach((paragraph) => inheritRowEndTableState(paragraph, rowEnd));
     }
 }
-function buildTableBlock(tableParagraphs) {
-    const rows = [];
-    let pendingRow = { cells: [] };
-    let pendingCellParagraphs = [];
-    for (const paragraph of tableParagraphs) {
-        pendingCellParagraphs.push(paragraph);
-        if (paragraph.terminator === DOC_CONTROL.cellMark) {
-            pendingRow.cells.push({
-                id: uniqueId('cell'),
-                paragraphs: pendingCellParagraphs.map(paragraphToBlock),
-                meta: null,
-            });
-            pendingCellParagraphs = [];
-            if (paragraph.paraState.tableRowEnd || paragraph.paraState.innerTableRowEnd) {
-                const cellDefs = applyTableStateToCells(paragraph.tableState);
-                while (cellDefs.length &&
-                    pendingRow.cells.length > cellDefs.length &&
-                    pendingRow.cells[pendingRow.cells.length - 1].paragraphs.every((block) => !block.text && !(block.inlines || []).length)) {
-                    pendingRow.cells.pop();
-                }
-                pendingRow.cells.forEach((cell, index) => {
-                    cell.meta = cellDefs[index] || { index };
-                });
-                const gridWidthTwips = cellDefs.length
-                    ? ((cellDefs[cellDefs.length - 1].rightBoundary || 0) - (cellDefs[0].leftBoundary || 0))
-                    : 0;
-                rows.push({
-                    id: uniqueId('row'),
-                    cells: pendingRow.cells,
-                    state: paragraph.tableState,
-                    gridWidthTwips,
-                });
-                pendingRow = { cells: [] };
-            }
-        }
-    }
-    if (pendingCellParagraphs.length) {
-        pendingRow.cells.push({ id: uniqueId('cell'), paragraphs: pendingCellParagraphs.map(paragraphToBlock), meta: null });
-    }
-    if (pendingRow.cells.length) {
-        rows.push({ id: uniqueId('row'), cells: pendingRow.cells, state: tableParagraphs[0]?.tableState || tablePropsToState([]), gridWidthTwips: 0 });
-    }
-    finalizeTableGrid(rows);
-    resolveTableGridBorders(rows);
-    const gridWidthTwips = rows.find((row) => row.gridWidthTwips)?.gridWidthTwips || 0;
-    const depth = Math.max(...tableParagraphs.map((paragraph) => getTableDepth(paragraph.paraState)), 1);
-    return {
-        type: 'table',
-        id: uniqueId('table'),
-        depth,
-        rows,
-        state: rows[0]?.state || tablePropsToState([]),
-        gridWidthTwips,
-    };
-}
-function buildBlocks(paragraphs) {
+export function buildBlocks(paragraphs) {
     const blocks = [];
-    let index = 0;
-    while (index < paragraphs.length) {
-        const paragraph = paragraphs[index];
+    const stack = [];
+    const finishCell = (pending) => {
+        pending.cells.push({
+            id: uniqueId('cell'),
+            paragraphs: pending.content.filter((block) => block.type === 'paragraph'),
+            blocks: pending.content,
+            meta: null,
+        });
+        pending.content = [];
+    };
+    const finishRow = (pending, state = pending.lastState) => {
+        if (pending.content.length)
+            finishCell(pending);
+        const definitions = applyTableStateToCells(state);
+        while (pending.cells.length < definitions.length)
+            finishCell(pending);
+        if (!pending.cells.length)
+            finishCell(pending);
+        pending.cells.forEach((cell, index) => { cell.meta = definitions[index] || { index }; });
+        pending.table.rows.push({
+            id: uniqueId('row'),
+            cells: pending.cells,
+            state,
+            gridWidthTwips: definitions.length
+                ? (definitions[definitions.length - 1].rightBoundary || 0) - (definitions[0].leftBoundary || 0)
+                : 0,
+        });
+        pending.cells = [];
+    };
+    const finishTable = () => {
+        const pending = stack.pop();
+        if (pending.content.length || pending.cells.length)
+            finishRow(pending);
+        const table = pending.table;
+        finalizeTableGrid(table.rows);
+        resolveTableGridBorders(table.rows);
+        table.state = table.rows[0]?.state || pending.lastState;
+        table.gridWidthTwips = table.rows.find(row => row.gridWidthTwips)?.gridWidthTwips || 0;
+        const parent = stack[stack.length - 1];
+        if (parent)
+            parent.content.push(table);
+        else
+            blocks.push(table);
+    };
+    for (const paragraph of paragraphs) {
         const depth = getTableDepth(paragraph.paraState);
+        if (stack.length && stack[0].storyKind !== paragraph.storyKind) {
+            while (stack.length)
+                finishTable();
+        }
+        while (stack.length && stack[stack.length - 1].table.depth > depth)
+            finishTable();
         if (depth <= 0) {
             blocks.push(paragraphToBlock(paragraph));
-            index += 1;
             continue;
         }
-        const tableParagraphs = [];
-        while (index < paragraphs.length && getTableDepth(paragraphs[index].paraState) > 0) {
-            tableParagraphs.push(paragraphs[index]);
-            index += 1;
+        let pending = stack[stack.length - 1];
+        if (!pending || pending.table.depth < depth) {
+            pending = {
+                table: { type: 'table', id: uniqueId('table'), depth, rows: [], state: paragraph.tableState, gridWidthTwips: 0 },
+                cells: [], content: [], lastState: paragraph.tableState, storyKind: paragraph.storyKind,
+            };
+            stack.push(pending);
+        }
+        pending.lastState = paragraph.tableState;
+        const rowEnd = depth > 1 ? paragraph.paraState.innerTableRowEnd : paragraph.paraState.tableRowEnd;
+        if (rowEnd) {
+            if (paragraph.text || paragraph.inlines.length)
+                pending.content.push(paragraphToBlock(paragraph));
+            finishRow(pending, paragraph.tableState);
+            continue;
         }
-        blocks.push(buildTableBlock(tableParagraphs));
+        pending.content.push(paragraphToBlock(paragraph));
+        const cellEnd = depth > 1
+            ? paragraph.paraState.innerTableCell
+            : paragraph.terminator === DOC_CONTROL.cellMark;
+        if (cellEnd)
+            finishCell(pending);
     }
+    while (stack.length)
+        finishTable();
     return blocks;
 }
-/**
- * Main MS-DOC entry point.
- * It parses the OLE container, restores text through the piece table, resolves
- * paragraph/character/table properties, and finally produces a normalized AST
- * that the HTML renderer can consume.
- */
 export function parseMsDoc(input, options = {}) {
     const warnings = [];
     const cfb = parseCFB(input, options);

package/dist/msdoc/properties.js

@@ -155,7 +155,6 @@ export function paraPropsToState(properties) {
                 break;
             case 'lineSpacing': {
                 state.lineSpacing = prop.value || 0;
-                // LSPD's second short distinguishes 1/240-line multiples from signed twips.
                 const bytes = prop.operandBytes;
                 const absolute = bytes && bytes.length >= 4
                     ? bytes[2] === 0 && bytes[3] === 0
@@ -375,9 +374,6 @@ export function applyTableStateToCells(tableState) {
     for (const op of tableState.operations || []) {
         switch (op.name) {
             case 'insertCells': {
-                // Word commonly emits TDefTable for legacy readers and TInsert for
-                // readers that process sprmPTableProps. They are alternative cell
-                // definitions, not cumulative instructions.
                 if (def?.cells?.length)
                     break;
                 const value = op.value;
@@ -429,8 +425,6 @@ export function applyTableStateToCells(tableState) {
                 rangeApply(cells, op.value.range, (cell) => {
                     const value = op.value;
                     cell.ftsWidth = value.ftsWidth;
-                    // ftsNil and ftsAuto carry no usable preferred width. Preserve the
-                    // TDefTable boundary width instead of collapsing the cell to zero.
                     if ((value.ftsWidth === 2 || value.ftsWidth === 3) && value.width != null) {
                         cell.width = value.width;
                     }

package/dist/msdoc/styles.js

@@ -155,7 +155,6 @@ function parseStyleStd(stdBytes, cbSTDBaseInFile, istd) {
         }
     }
     catch {
-        // keep best-effort partial decoding
     }
     return {
         istd,

package/dist/msdoc/vector.js

@@ -1104,7 +1104,6 @@ function parseEmf(bytes) {
             }
             case EMF_RECORD.EMR_EXTCREATEFONTINDIRECTW: {
                 const index = record.u32(8);
-                // LOGFONTW starts at offset 12: lfHeight(+0), ..., lfFaceName(+28).
                 const height = Math.abs(record.i32(12));
                 const escapement = record.i32(20);
                 const weight = record.i32(28);

package/dist/render/fonts.d.ts

@@ -1,4 +1,5 @@
-import type { CharState } from '../types.js';
+import type { CharState, FontInfo } from '../types.js';
+export declare function msDocFontFallbacks(family: string, info?: FontInfo): readonly string[];
 export declare function textFontRuns(text: string, style: CharState): Array<{
     text: string;
     fontFamily?: string;

package/dist/render/fonts.js

@@ -1,5 +1,65 @@
 const EAST_ASIAN = /[\p{Script=Han}\p{Script=Hiragana}\p{Script=Katakana}\p{Script=Hangul}\p{Script=Bopomofo}\u3000-\u303f\uff00-\uffef]/u;
 const INHERIT_FONT = /[\p{Mark}\u200c\u200d]/u;
+function cjkFontFallbacks(family) {
+    switch (family.toLowerCase().replace(/[ _-]/g, '')) {
+        case '宋体':
+        case '新宋体':
+        case 'simsun':
+        case 'nsimsun':
+            return ['SimSun', 'Songti SC', 'Noto Serif CJK SC', 'serif'];
+        case '仿宋':
+        case '仿宋gb2312':
+        case 'fangsong':
+        case 'fangsonggb2312':
+            return ['FangSong', 'STFangsong', 'Songti SC', 'Noto Serif CJK SC', 'serif'];
+        case '楷体':
+        case '楷体gb2312':
+        case 'kaiti':
+        case 'kaitigb2312':
+            return ['KaiTi', 'STKaiti', 'Kaiti SC', 'Songti SC', 'Noto Serif CJK SC', 'serif'];
+        case '黑体':
+        case 'simhei':
+            return ['SimHei', 'Heiti SC', 'Noto Sans CJK SC', 'sans-serif'];
+        case '微软雅黑':
+        case 'microsoftyahei':
+            return ['Microsoft YaHei', 'PingFang SC', 'Noto Sans CJK SC', 'sans-serif'];
+    }
+    const normalized = family.normalize('NFKC').toLowerCase();
+    if (/fangsong|仿宋/.test(normalized))
+        return ['STFangsong', 'Songti SC', 'Noto Serif CJK SC', 'serif'];
+    if (/simsun|nsimsun|宋体|宋體|songti/.test(normalized))
+        return ['Songti SC', 'Songti TC', 'Noto Serif CJK SC', 'serif'];
+    if (/kaiti|楷体|楷體/.test(normalized))
+        return ['Kaiti SC', 'STKaiti', 'Noto Serif CJK SC', 'serif'];
+    if (/mingliu|細明體|新細明體/.test(normalized))
+        return ['Songti TC', 'Noto Serif CJK TC', 'serif'];
+    if (/mincho|明朝/.test(normalized))
+        return ['Hiragino Mincho ProN', 'Yu Mincho', 'Noto Serif CJK JP', 'serif'];
+    if (/simhei|黑体|黑體|heiti|dengxian|等线|等線|yahei/.test(normalized))
+        return ['PingFang SC', 'Microsoft YaHei', 'Noto Sans CJK SC', 'sans-serif'];
+    return undefined;
+}
+export function msDocFontFallbacks(family, info) {
+    const alternate = info?.altName && info.altName !== family ? [info.altName] : [];
+    const cjk = cjkFontFallbacks(family);
+    if (cjk)
+        return [...alternate, ...cjk];
+    const genericByFamily = {
+        1: 'serif', 2: 'sans-serif', 3: 'monospace', 4: 'cursive', 5: 'fantasy',
+    };
+    const familyClass = ((info?.ffid ?? 0) >>> 4) & 7;
+    let generic = genericByFamily[familyClass];
+    if (!generic && info?.panose?.[0] === 2) {
+        const serif = info.panose[1];
+        if (info.panose[3] === 9)
+            generic = 'monospace';
+        else if (serif >= 2 && serif <= 10)
+            generic = 'serif';
+        else if (serif >= 11 && serif <= 15)
+            generic = 'sans-serif';
+    }
+    return [...alternate, generic ?? 'sans-serif'];
+}
 export function textFontRuns(text, style) {
     if (!text)
         return [];

package/dist/render/html.d.ts

@@ -1,16 +1,6 @@
 import type { MsDocParseResult, MsDocRenderOptions, MsDocRenderResult } from '../types.js';
 type ExternalLinkPolicy = NonNullable<MsDocRenderOptions['externalLinkPolicy']>;
-/**
- * Normalizes a document-provided hyperlink before it reaches generated HTML.
- * Internal bookmarks always remain available. External navigation is opt-in
- * and is still restricted to browser-safe schemes and relative paths.
- */
 export declare function sanitizeMsDocLinkHref(href: string | undefined | null, policy?: ExternalLinkPolicy): string | null;
 export declare function defaultMsDocCss(): string;
-/**
- * Converts the parsed AST into HTML and a companion CSS string.
- * Keeping rendering separate from parsing makes it easier for downstream apps
- * to customize styles or consume the AST directly.
- */
 export declare function renderMsDoc(parsed: MsDocParseResult, options?: MsDocRenderOptions): MsDocRenderResult;
 export {};

package/dist/render/html.js

@@ -1,7 +1,7 @@
 import { escapeHtml, slugify, twipsToPx } from '../core/utils.js';
 import { HIGHLIGHT_COLORS } from '../msdoc/constants.js';
 import { cssTextAlign, cssUnderline, cssVerticalAlign } from '../msdoc/properties.js';
-import { textFontRuns } from './fonts.js';
+import { msDocFontFallbacks, textFontRuns } from './fonts.js';
 const COLOR_INDEX_MAP = {
     1: '#000000',
     2: '#0000ff',
@@ -34,11 +34,10 @@ function quoteCssString(value) {
         .replace(/[\u0000-\u001f\u007f-\u009f]/g, ' ');
     return `'${escaped}'`;
 }
-/**
- * Normalizes a document-provided hyperlink before it reaches generated HTML.
- * Internal bookmarks always remain available. External navigation is opt-in
- * and is still restricted to browser-safe schemes and relative paths.
- */
+function fontFamilyToCss(family, context) {
+    const fallbacks = msDocFontFallbacks(family, context.fonts.get(family));
+    return [quoteCssString(family), ...fallbacks.map((name, index) => index === fallbacks.length - 1 ? name : quoteCssString(name))].join(',');
+}
 export function sanitizeMsDocLinkHref(href, policy = 'block') {
     const value = String(href ?? '')
         .replace(/[\u0000-\u0020\u007f-\u009f]/g, '')
@@ -160,7 +159,6 @@ function paragraphStyleToCss(paraState) {
         }
         else {
             const multiple = Math.abs(paraState.lineSpacing) / 240;
-            // A font's single line includes its metrics and leading, not just a 1em box.
             style['line-height'] = multiple === 1 ? 'normal' : String(multiple);
         }
     }
@@ -203,7 +201,7 @@ function buildUnderlineStyle(underline) {
         css['text-decoration-style'] = 'solid';
     return css;
 }
-function inlineStyleToCss(styleState) {
+function inlineStyleToCss(styleState, context) {
     const style = {};
     if (styleState.bold || styleState.boldBi)
         style['font-weight'] = '700';
@@ -215,7 +213,7 @@ function inlineStyleToCss(styleState) {
     if (styleState.fontSizeHalfPoints)
         style['font-size'] = `${styleState.fontSizeHalfPoints / 2}pt`;
     if (styleState.fontFamily)
-        style['font-family'] = `${quoteCssString(styleState.fontFamily)},sans-serif`;
+        style['font-family'] = fontFamilyToCss(styleState.fontFamily, context);
     if (styleState.colorIndex && COLOR_INDEX_MAP[styleState.colorIndex])
         style.color = COLOR_INDEX_MAP[styleState.colorIndex];
     const highlightIndex = typeof styleState.highlight === 'number' ? styleState.highlight : styleState.highlight?.index;
@@ -253,10 +251,10 @@ function renderTextNode(node, context) {
         const text = escapeHtml(run.text);
         if (!run.fontFamily || run.fontFamily === node.style.fontFamily)
             return text;
-        const font = styleObjectToCss({ 'font-family': `${quoteCssString(run.fontFamily)},sans-serif` });
+        const font = styleObjectToCss({ 'font-family': fontFamilyToCss(run.fontFamily, context) });
         return `<span style="${font}">${text}</span>`;
     }).join('');
-    const inlineStyle = inlineStyleToCss(node.style);
+    const inlineStyle = inlineStyleToCss(node.style, context);
     inlineStyle['white-space'] = 'break-spaces';
     const revision = context.reviewMode === 'all'
         ? node.style.revisionDeleted ? 'delete' : node.style.revisionInserted ? 'insert' : null
@@ -373,7 +371,7 @@ function sanitizeAssetHref(href, context) {
 function renderImageNode(node, context) {
     const allowGeneratedVector = node.asset.meta?.vectorConverted === true;
     const src = sanitizeImageSource(node.asset.sourceUrl, context.externalResourcePolicy, allowGeneratedVector) || sanitizeImageSource(node.asset.dataUrl, context.externalResourcePolicy, allowGeneratedVector);
-    const baseStyle = inlineStyleToCss(node.style);
+    const baseStyle = inlineStyleToCss(node.style, context);
     const displaySize = inlineImageDisplaySizePx(node.asset);
     if (displaySize.widthPx || displaySize.heightPx) {
         applyInlineImageDisplaySize(baseStyle, node.asset);
@@ -432,7 +430,6 @@ function reviewParagraphs(paragraphs, context) {
         const previous = result[result.length - 1];
         if (previous && previous.storyKind === paragraph.storyKind
             && hiddenRevision(previous.paragraphMark, context)) {
-            // The surviving paragraph mark owns the merged paragraph's formatting.
             result[result.length - 1] = {
                 ...paragraph,
                 inlines: [...previous.inlines, ...paragraph.inlines],
@@ -508,7 +505,15 @@ function tableStyle(block) {
     else
         style.width = '100%';
     const marginLeft = twipsToPx(block.state?.leftIndent);
-    if (marginLeft)
+    if (block.state?.alignment === 1) {
+        style['margin-inline-start'] = 'auto';
+        style['margin-inline-end'] = 'auto';
+    }
+    else if (block.state?.alignment === 2) {
+        style['margin-inline-start'] = 'auto';
+        style['margin-inline-end'] = '0';
+    }
+    else if (marginLeft)
         style['margin-inline-start'] = `${marginLeft}px`;
     if (block.state?.rtl)
         style.direction = 'rtl';
@@ -531,6 +536,48 @@ function tableStyle(block) {
     style['table-layout'] = 'fixed';
     return style;
 }
+function renderTableColumns(block) {
+    const boundaries = new Set();
+    for (const row of block.rows)
+        for (const cell of row.cells) {
+            const left = cell.meta?.leftBoundary;
+            const right = cell.meta?.rightBoundary;
+            if (Number.isFinite(left) && Number.isFinite(right) && right > left) {
+                boundaries.add(left);
+                boundaries.add(right);
+            }
+        }
+    const sorted = [...boundaries].sort((a, b) => a - b);
+    if (sorted.length < 2)
+        return '';
+    return `<colgroup>${sorted.slice(1).map((right, i) => `<col style="width:${twipsToPx(right - sorted[i])}px">`).join('')}</colgroup>`;
+}
+function renderCellBody(cell, context) {
+    const parts = [];
+    let paragraphs = [];
+    const wrapText = (body) => cell.meta?.textFlow === 5
+        ? `<div class="msdoc-cell-vertical" style="writing-mode:vertical-rl;text-orientation:upright;inline-size:max-content;block-size:max-content;margin:0 auto">${body}</div>`
+        : body;
+    const flush = () => {
+        if (!paragraphs.length)
+            return;
+        const body = reviewParagraphs(paragraphs, context)
+            .map(paragraph => renderParagraphBlock(paragraph, context, { inline: true })).join('');
+        if (body)
+            parts.push(wrapText(body));
+        paragraphs = [];
+    };
+    for (const block of cell.blocks ?? cell.paragraphs) {
+        if (block.type === 'paragraph')
+            paragraphs.push(block);
+        else {
+            flush();
+            parts.push(renderTableBlock(block, context));
+        }
+    }
+    flush();
+    return parts.join('') || wrapText('<div class="msdoc-paragraph"><br></div>');
+}
 function renderTableBlock(block, context) {
     const rows = block.rows.map((row) => {
         const rowHeight = row.state?.rowHeight ? twipsToPx(Math.abs(row.state.rowHeight)) : null;
@@ -544,14 +591,13 @@ function renderTableBlock(block, context) {
             if ((cell.rowspan ?? 1) > 1)
                 attrs.push(` rowspan="${cell.rowspan}"`);
             const style = styleObjectToCss(cellStyle(cell));
-            const body = reviewParagraphs(cell.paragraphs, context)
-                .map((paragraph) => renderParagraphBlock(paragraph, context, { inline: true })).join('');
+            const body = renderCellBody(cell, context);
             return `<td class="msdoc-cell"${attrs.join('')}${style ? ` style="${style}"` : ''}>${body || '<div class="msdoc-paragraph"><br></div>'}</td>`;
         })
             .join('');
         return `<tr class="msdoc-row"${rowStyle}>${cells}</tr>`;
     }).join('');
-    return `<table class="msdoc-table msdoc-table-depth-${block.depth}" style="${styleObjectToCss(tableStyle(block))}"><tbody>${rows}</tbody></table>`;
+    return `<table class="msdoc-table msdoc-table-depth-${block.depth}" style="${styleObjectToCss(tableStyle(block))}">${renderTableColumns(block)}<tbody>${rows}</tbody></table>`;
 }
 function renderAttachmentsBlock(block) {
     const items = block.items.map((item) => {
@@ -572,6 +618,7 @@ export function defaultMsDocCss() {
 .msdoc-line-at-least span,.msdoc-line-at-least ins,.msdoc-line-at-least del{line-height:normal}
 .msdoc-paragraph:last-child{margin-bottom:0}
 .msdoc-table{margin:12px 0;border-collapse:collapse;border-spacing:0;max-width:100%}
+.msdoc-cell>.msdoc-table{margin-top:0;margin-bottom:0}
 .msdoc-cell{padding:6px 8px;vertical-align:top;word-break:break-word;overflow-wrap:anywhere}
 .msdoc-link{color:#1a73e8;text-decoration:none}
 .msdoc-link:hover{text-decoration:underline}
@@ -583,17 +630,13 @@ export function defaultMsDocCss() {
 .msdoc-page-break{display:block;height:0;border-top:1px dashed #cbd5e1;margin:16px 0}
 `;
 }
-/**
- * Converts the parsed AST into HTML and a companion CSS string.
- * Keeping rendering separate from parsing makes it easier for downstream apps
- * to customize styles or consume the AST directly.
- */
 export function renderMsDoc(parsed, options = {}) {
     const css = options.css ?? defaultMsDocCss();
     const context = {
         reviewMode: options.reviewMode ?? 'all',
         externalLinkPolicy: options.externalLinkPolicy ?? 'block',
         externalResourcePolicy: options.externalResourcePolicy ?? 'block',
+        fonts: new Map((parsed.fonts ?? []).flatMap(font => [font.name, font.altName].filter(Boolean).map(name => [name, font]))),
     };
     const parts = [];
     let paragraphs = [];

package/dist/types.d.ts

@@ -482,7 +482,6 @@ export interface OfficeArtBlipMeta {
     blipIndex?: number;
     sourceShapeId?: number;
 }
-/** PICF / OfficeArt BLIP / linked-picture metadata used by the HTML renderer. */
 export interface ImageAssetMeta extends PicfSummary, OfficeArtBlipMeta {
     linkedPath?: string;
     sourceKind?: 'embedded' | 'linked' | 'fallback';
@@ -582,6 +581,7 @@ export interface ParagraphBlock {
 export interface TableCellBlock {
     id: string;
     paragraphs: ParagraphBlock[];
+    blocks?: Array<ParagraphBlock | TableBlock>;
     meta: TableCellMeta | null;
     colIndex?: number;
     colspan?: number;
@@ -632,11 +632,8 @@ export interface MsDocParseOptions {
 }
 export interface MsDocRenderOptions {
     css?: string;
-    /** Preserve insertion/deletion marks by default; explicitly select final or original text. */
     reviewMode?: 'all' | 'final' | 'original';
-    /** External links are blocked by default; internal document bookmarks remain active. */
     externalLinkPolicy?: 'allow' | 'block';
-    /** Linked image resources are blocked by default; embedded data/blob images remain available. */
     externalResourcePolicy?: 'allow' | 'block';
 }
 export interface MsDocMeta {

package/dist/viewer.d.ts

@@ -1,16 +1,7 @@
 import type { Config } from 'dompurify';
 import type { MsDocParseToHtmlOptions, MsDocRenderResult, MsDocViewer, MsDocViewerConfig, ViewerInput } from './types.js';
-/**
- * Shared DOC/RTF DOMPurify boundary. A per-window purifier and Trusted Types
- * policy avoid duplicate policy creation when several Word formats are opened
- * in the same host document.
- */
 export declare function sanitizeFileViewerDocumentMarkup(html: string, targetWindow: Window, config: Config): unknown | null;
 export declare function sanitizeMsDocHtml(html: string, targetWindow: Window): DocumentFragment;
 export declare function mountMsDoc(container: HTMLElement, rendered: MsDocRenderResult): HTMLElement;
 export declare function parseMsDocToHtml(input: ViewerInput, options?: MsDocParseToHtmlOptions): Promise<MsDocRenderResult>;
-/**
- * Small DOM-oriented helper that keeps browser integration trivial.
- * Apps can either use it directly or consume the lower-level parse/render APIs.
- */
 export declare function createMsDocViewer(container: HTMLElement, config?: MsDocViewerConfig): MsDocViewer;

package/dist/viewer.js

@@ -32,11 +32,6 @@ const getDocumentTrustedTypesPolicy = (targetWindow) => {
         return null;
     }
 };
-/**
- * Shared DOC/RTF DOMPurify boundary. A per-window purifier and Trusted Types
- * policy avoid duplicate policy creation when several Word formats are opened
- * in the same host document.
- */
 export function sanitizeFileViewerDocumentMarkup(html, targetWindow, config) {
     const purifier = getDocumentPurifier(targetWindow);
     if (!purifier)
@@ -49,8 +44,6 @@ export function sanitizeFileViewerDocumentMarkup(html, targetWindow, config) {
         });
     }
     catch {
-        // If a host CSP forbids this policy name, callers fall back to a text node
-        // rather than mounting unsanitized markup or failing the whole preview.
         return null;
     }
 }
@@ -112,10 +105,6 @@ export async function parseMsDocToHtml(input, options = {}) {
     const parsed = parseMsDoc(buffer, options.parseOptions || {});
     return renderMsDoc(parsed, options.renderOptions || {});
 }
-/**
- * Small DOM-oriented helper that keeps browser integration trivial.
- * Apps can either use it directly or consume the lower-level parse/render APIs.
- */
 export function createMsDocViewer(container, config = {}) {
     let current = null;
     return {

package/dist/worker-client.d.ts

@@ -1,8 +1,4 @@
 import type { MsDocParseOptions, MsDocParseResult, MsDocParseToHtmlOptions, MsDocRenderOptions, MsDocRenderResult, MsDocWorkerClientLike, WorkerLike } from './types.js';
-/**
- * Thin typed wrapper around a module worker so applications can offload
- * expensive binary parsing without having to manage message correlation.
- */
 export declare class MsDocWorkerClient implements MsDocWorkerClientLike {
     private readonly worker;
     private readonly pending;

package/dist/worker-client.js

@@ -8,10 +8,6 @@ function toArrayBuffer(input) {
     }
     throw new TypeError('Expected ArrayBuffer or TypedArray');
 }
-/**
- * Thin typed wrapper around a module worker so applications can offload
- * expensive binary parsing without having to manage message correlation.
- */
 export class MsDocWorkerClient {
     worker;
     pending = new Map();

package/package.json

@@ -1,6 +1,6 @@
 {
   "name": "@file-viewer/doc",
-  "version": "3.1.2",
+  "version": "3.2.1",
   "private": false,
   "type": "module",
   "description": "Browser-native legacy MS-DOC (.doc) parser and HTML renderer extracted from the public msdoc-viewer 0.2.0 line.",
@@ -27,8 +27,7 @@
   },
   "repository": {
     "type": "git",
-    "url": "git+https://github.com/flyfish-dev/file-viewer.git",
-    "directory": "packages/renderers/doc"
+    "url": "git+https://github.com/flyfish-dev/file-viewer.git"
   },
   "homepage": "https://github.com/flyfish-dev/docjs",
   "bugs": {
@@ -56,26 +55,66 @@
     "./package.json": "./package.json"
   },
   "files": [
-    "dist",
-    "README.md",
+    "dist/core/binary.d.ts",
+    "dist/core/binary.js",
+    "dist/core/cfb.d.ts",
+    "dist/core/cfb.js",
+    "dist/core/inflate.d.ts",
+    "dist/core/inflate.js",
+    "dist/core/utils.d.ts",
+    "dist/core/utils.js",
+    "dist/index.d.ts",
+    "dist/index.js",
+    "dist/msdoc/clx.d.ts",
+    "dist/msdoc/clx.js",
+    "dist/msdoc/constants.d.ts",
+    "dist/msdoc/constants.js",
+    "dist/msdoc/fib.d.ts",
+    "dist/msdoc/fib.js",
+    "dist/msdoc/fkp.d.ts",
+    "dist/msdoc/fkp.js",
+    "dist/msdoc/fonts.d.ts",
+    "dist/msdoc/fonts.js",
+    "dist/msdoc/html-word.d.ts",
+    "dist/msdoc/html-word.js",
+    "dist/msdoc/objects.d.ts",
+    "dist/msdoc/objects.js",
+    "dist/msdoc/officeart.d.ts",
+    "dist/msdoc/officeart.js",
+    "dist/msdoc/parser.d.ts",
+    "dist/msdoc/parser.js",
+    "dist/msdoc/properties.d.ts",
+    "dist/msdoc/properties.js",
+    "dist/msdoc/sprm.d.ts",
+    "dist/msdoc/sprm.js",
+    "dist/msdoc/styles.d.ts",
+    "dist/msdoc/styles.js",
+    "dist/msdoc/vector.d.ts",
+    "dist/msdoc/vector.js",
+    "dist/render/fonts.d.ts",
+    "dist/render/fonts.js",
+    "dist/render/html.d.ts",
+    "dist/render/html.js",
+    "dist/types.d.ts",
+    "dist/types.js",
+    "dist/viewer.d.ts",
+    "dist/viewer.js",
+    "dist/worker-client.d.ts",
+    "dist/worker-client.js",
+    "dist/worker.d.ts",
+    "dist/worker.js",
+    "LICENSE",
     "README.en.md",
-    "LICENSE"
+    "README.md"
   ],
   "dependencies": {
-    "dompurify": "3.4.15"
-  },
-  "devDependencies": {
-    "typescript": "^6.0.3"
+    "dompurify": "3.4.16"
   },
   "license": "MIT",
-  "scripts": {
-    "build": "tsc -b tsconfig.json",
-    "type-check": "tsc -b tsconfig.json",
-    "verify:table-spec": "pnpm build && node scripts/verify-table-spec.mjs",
-    "verify:github-34": "pnpm build && node scripts/verify-github-34.mjs",
-    "verify:github-87": "pnpm build && node scripts/verify-github-87.mjs",
-    "verify:github-171": "pnpm build && node scripts/verify-github-171.mjs",
-    "verify:github-192": "pnpm build && node scripts/verify-github-192.mjs",
-    "verify:github-236": "pnpm build && node scripts/verify-github-236.mjs"
+  "fileViewer": {
+    "distribution": {
+      "schemaVersion": 1,
+      "mode": "runtime-only"
+    }
   }
-}
+}