docxmlater 10.1.4 → 10.1.6

package/src/utils/acceptRevisions.ts

@@ -1,714 +1,723 @@
-import { ZipHandler } from '../zip/ZipHandler';
-import { XMLParser } from '../xml/XMLParser';
-import { RevisionWalker } from './RevisionWalker';
-
-/**
- * Accepts all tracked changes in a Word document per Microsoft's OpenXML SDK pattern
- *
- * This implementation uses DOM-based tree walking for reliability:
- * 1. Insertions (<w:ins>): Keep content, remove wrapper tags
- * 2. Deletions (<w:del>): Remove entirely (content and tags)
- * 3. Move From (<w:moveFrom>): Remove entirely (source of move)
- * 4. Move To (<w:moveTo>): Keep content, remove wrapper (destination of move)
- * 5. Property changes: Remove all *Change elements
- * 6. Range markers: Remove all boundary markers
- *
- * Also cleans up metadata in people.xml, settings.xml, and core.xml
- *
- * @see https://learn.microsoft.com/en-us/office/open-xml/how-to-accept-all-revisions
- */
-class RevisionAcceptor {
-  private zipHandler: ZipHandler;
-  /** Feature flag for DOM-based processing (default: true) */
-  private useDomBasedProcessing = true;
-
-  constructor(zipHandler: ZipHandler) {
-    this.zipHandler = zipHandler;
-  }
-
-  /**
-   * Enable or disable DOM-based processing (for testing/migration)
-   */
-  setUseDomBasedProcessing(enabled: boolean): void {
-    this.useDomBasedProcessing = enabled;
-  }
-
-  /**
-   * Main method to accept all revisions in the document
-   */
-  public async acceptAllRevisions(): Promise<void> {
-    // Process document.xml
-    await this.processDocumentPart('word/document.xml');
-    
-    // Process headers
-    const files = this.zipHandler.getFilePaths();
-    for (const file of files) {
-      if (/^word\/header\d+\.xml$/.exec(file)) {
-        await this.processDocumentPart(file);
-      }
-      if (/^word\/footer\d+\.xml$/.exec(file)) {
-        await this.processDocumentPart(file);
-      }
-    }
-
-    // Clean up metadata files
-    this.cleanupPeopleXml();
-    this.cleanupSettingsXml();
-    this.cleanupCorePropsXml();
-  }
-
-  /**
-   * Process a document part (document.xml, header, footer) to accept revisions
-   */
-  private async processDocumentPart(partPath: string): Promise<void> {
-    if (this.useDomBasedProcessing) {
-      return this.processDocumentPartDOM(partPath);
-    }
-    return this.processDocumentPartRegex(partPath);
-  }
-
-  /**
-   * DOM-based implementation of revision acceptance
-   * Uses XMLParser and RevisionWalker for reliable processing
-   */
-  private processDocumentPartDOM(partPath: string): void {
-    const xml = this.zipHandler.getFileAsString(partPath);
-    if (!xml) {
-      return;
-    }
-
-    // Step 1: Parse XML to object tree
-    // IMPORTANT: trimValues: false preserves whitespace from xml:space="preserve" attributes
-    const parsed = XMLParser.parseToObject(xml, { trimValues: false });
-
-    // Step 2: Process revisions using DOM walker
-    const processed = RevisionWalker.processTree(parsed, {
-      acceptInsertions: true,
-      acceptDeletions: true,
-      acceptMoves: true,
-      acceptPropertyChanges: true,
-    });
-
-    // Step 3: Handle image relationship ID remapping
-    this.remapImageRelationshipsInTree(processed);
-
-    // Step 4: Convert back to XML
-    const outputXml =
-      '<?xml version="1.0" encoding="UTF-8" standalone="yes"?>\n' +
-      this.objectToXml(processed);
-
-    // Step 5: Update file
-    this.zipHandler.updateFile(partPath, outputXml);
-  }
-
-  /**
-   * Legacy RegEx-based implementation (kept as fallback)
-   */
-  private processDocumentPartRegex(partPath: string): void {
-    const xml = this.zipHandler.getFileAsString(partPath);
-    if (!xml) {
-      return;
-    }
-
-    let content = xml;
-
-    // Step 1: Remove all range markers FIRST (before processing revisions)
-    // This prevents orphaned references when revision content is modified
-    content = this.removeAllRangeMarkers(content);
-
-    // Step 2: Remove all property change elements
-    // These track formatting changes and must be removed before other processing
-    content = this.removeAllPropertyChanges(content);
-
-    // Step 3: Process deletions - remove entire element INCLUDING content
-    // Must be done before insertions to handle nested scenarios
-    content = this.acceptDeletions(content);
-
-    // Step 4: Process move operations
-    // Remove moveFrom entirely (source), unwrap moveTo (destination)
-    content = this.acceptMoveFrom(content);
-    content = this.acceptMoveTo(content);
-
-    // Step 5: Process insertions - keep content, remove wrapper
-    content = this.acceptInsertions(content);
-
-    // Step 6: Final cleanup - remove any remaining orphaned tags
-    content = this.cleanupOrphanedTags(content);
-
-    // Update the file
-    this.zipHandler.updateFile(partPath, content);
-  }
-
-  /**
-   * Remove all range marker elements
-   * These are boundary markers for tracked changes and moves
-   */
-  private removeAllRangeMarkers(xml: string): string {
-    const patterns = [
-      // Move range markers
-      /<w:moveFromRangeStart[^>]*(?:\/>|>.*?<\/w:moveFromRangeStart>)/gs,
-      /<w:moveFromRangeEnd[^>]*(?:\/>|>.*?<\/w:moveFromRangeEnd>)/gs,
-      /<w:moveToRangeStart[^>]*(?:\/>|>.*?<\/w:moveToRangeStart>)/gs,
-      /<w:moveToRangeEnd[^>]*(?:\/>|>.*?<\/w:moveToRangeEnd>)/gs,
-      // Custom XML range markers
-      /<w:customXmlInsRangeStart[^>]*(?:\/>|>.*?<\/w:customXmlInsRangeStart>)/gs,
-      /<w:customXmlInsRangeEnd[^>]*(?:\/>|>.*?<\/w:customXmlInsRangeEnd>)/gs,
-      /<w:customXmlDelRangeStart[^>]*(?:\/>|>.*?<\/w:customXmlDelRangeStart>)/gs,
-      /<w:customXmlDelRangeEnd[^>]*(?:\/>|>.*?<\/w:customXmlDelRangeEnd>)/gs,
-      /<w:customXmlMoveFromRangeStart[^>]*(?:\/>|>.*?<\/w:customXmlMoveFromRangeStart>)/gs,
-      /<w:customXmlMoveFromRangeEnd[^>]*(?:\/>|>.*?<\/w:customXmlMoveFromRangeEnd>)/gs,
-      /<w:customXmlMoveToRangeStart[^>]*(?:\/>|>.*?<\/w:customXmlMoveToRangeStart>)/gs,
-      /<w:customXmlMoveToRangeEnd[^>]*(?:\/>|>.*?<\/w:customXmlMoveToRangeEnd>)/gs,
-    ];
-
-    let result = xml;
-    for (const pattern of patterns) {
-      result = result.replace(pattern, '');
-    }
-    return result;
-  }
-
-  /**
-   * Remove all property change tracking elements
-   * Per ECMA-376, these track previous state of formatting
-   */
-  private removeAllPropertyChanges(xml: string): string {
-    const patterns = [
-      // Run property changes
-      /<w:rPrChange[^>]*>[\s\S]*?<\/w:rPrChange>/g,
-      // Paragraph property changes  
-      /<w:pPrChange[^>]*>[\s\S]*?<\/w:pPrChange>/g,
-      // Table property changes
-      /<w:tblPrChange[^>]*>[\s\S]*?<\/w:tblPrChange>/g,
-      /<w:tblPrExChange[^>]*>[\s\S]*?<\/w:tblPrExChange>/g,
-      // Table cell property changes
-      /<w:tcPrChange[^>]*>[\s\S]*?<\/w:tcPrChange>/g,
-      // Table row property changes
-      /<w:trPrChange[^>]*>[\s\S]*?<\/w:trPrChange>/g,
-      // Section property changes
-      /<w:sectPrChange[^>]*>[\s\S]*?<\/w:sectPrChange>/g,
-      // Table grid changes
-      /<w:tblGridChange[^>]*>[\s\S]*?<\/w:tblGridChange>/g,
-      // Numbering changes
-      /<w:numberingChange[^>]*>[\s\S]*?<\/w:numberingChange>/g,
-    ];
-
-    let result = xml;
-    for (const pattern of patterns) {
-      result = result.replace(pattern, '');
-    }
-    return result;
-  }
-
-  /**
-   * Accept deletions - remove the entire <w:del> element including its content
-   * 
-   * Per Microsoft SDK: "DeletedRun elements should be removed along with their content"
-   */
-  private acceptDeletions(xml: string): string {
-    let result = xml;
-    let previousLength = 0;
-    
-    // Iterate until no more deletions (handles nested cases)
-    while (result.length !== previousLength) {
-      previousLength = result.length;
-      
-      // Match complete <w:del ...>...</w:del> elements and remove entirely
-      result = result.replace(/<w:del\b[^>]*>[\s\S]*?<\/w:del>/g, '');
-    }
-    
-    // Also remove self-closing deletion tags
-    result = result.replace(/<w:del\b[^>]*\/>/g, '');
-    
-    return result;
-  }
-
-  /**
-   * Accept moveFrom - remove the entire element (source of moved content)
-   * 
-   * The content exists at the moveTo destination, so we discard the source
-   */
-  private acceptMoveFrom(xml: string): string {
-    let result = xml;
-    let previousLength = 0;
-    
-    while (result.length !== previousLength) {
-      previousLength = result.length;
-      result = result.replace(/<w:moveFrom\b[^>]*>[\s\S]*?<\/w:moveFrom>/g, '');
-    }
-    
-    // Also remove self-closing tags
-    result = result.replace(/<w:moveFrom\b[^>]*\/>/g, '');
-    
-    return result;
-  }
-
-  /**
-   * Accept moveTo - keep the content, remove the wrapper tags
-   * 
-   * The moveTo location is where the content should remain
-   */
-  private acceptMoveTo(xml: string): string {
-    let result = xml;
-    
-    // Remove closing tags first (prevents issues with regex matching)
-    result = result.replace(/<\/w:moveTo>/g, '');
-    
-    // Remove opening tags (keeps content that was inside)
-    result = result.replace(/<w:moveTo\b[^>]*>/g, '');
-    
-    return result;
-  }
-
-  /**
-   * Accept insertions - keep the content, remove the wrapper tags
-   *
-   * Per Microsoft SDK: "InsertedRun elements should be unwrapped, keeping their content"
-   *
-   * IMPORTANT: This method now handles relationship ID remapping for images inside insertions.
-   * When Word tracks changes with images, it can reuse relationship IDs (like rId5) because
-   * they're in separate tracked change contexts. But when we unwrap them, duplicate IDs
-   * cause corruption. This method assigns new unique IDs to images inside insertions.
-   */
-  private acceptInsertions(xml: string): string {
-    let result = xml;
-
-    // Parse existing relationships
-    const relationships = this.parseRelationships();
-    const existingIds = new Set(relationships.keys());
-
-    // Process each w:ins element and remap images one by one
-    const insRegex = /<w:ins\b[^>]*>[\s\S]*?<\/w:ins>/g;
-
-    result = result.replace(insRegex, (insMatch) => {
-      // For each image reference inside this insertion, generate a unique new ID
-      return insMatch.replace(/r:embed="(rId\d+)"/g, (embedMatch, oldId) => {
-        // Generate new unique ID for THIS occurrence
-        const newId = this.getNextRelationshipId(existingIds);
-        existingIds.add(newId);
-
-        // Add relationship with same target as original
-        const target = relationships.get(oldId);
-        if (target) {
-          this.addRelationship(
-            newId,
-            target,
-            'http://schemas.openxmlformats.org/officeDocument/2006/relationships/image'
-          );
-        }
-
-        return `r:embed="${newId}"`;
-      });
-    });
-
-    // Now unwrap the w:ins tags (content has unique remapped IDs)
-    result = result.replace(/<\/w:ins>/g, '');
-    result = result.replace(/<w:ins\b[^>]*>/g, '');
-
-    return result;
-  }
-
-  /**
-   * Final cleanup to remove any orphaned or malformed revision-related tags
-   */
-  private cleanupOrphanedTags(xml: string): string {
-    let result = xml;
-
-    // Remove any remaining self-closing revision tags
-    result = result.replace(/<w:ins\b[^>]*\/>/g, '');
-    result = result.replace(/<w:del\b[^>]*\/>/g, '');
-    result = result.replace(/<w:moveFrom\b[^>]*\/>/g, '');
-    result = result.replace(/<w:moveTo\b[^>]*\/>/g, '');
-
-    // Remove empty w:r elements that might be left after removing deletions
-    result = result.replace(/<w:r\b[^>]*>\s*<\/w:r>/g, '');
-
-    // Remove empty w:p elements (but preserve those with properties)
-    // We keep <w:p><w:pPr>...</w:pPr></w:p> as those are intentional empty paragraphs with styling
-    result = result.replace(/<w:p>\s*<\/w:p>/g, '');
-
-    return result;
-  }
-
-  /**
-   * Parse relationship IDs from word/_rels/document.xml.rels
-   * Returns a map of relationship ID to target path
-   */
-  private parseRelationships(): Map<string, string> {
-    const relsXml = this.zipHandler.getFileAsString('word/_rels/document.xml.rels');
-    if (!relsXml) return new Map();
-
-    const map = new Map<string, string>();
-    const relationshipRegex = /<Relationship[^>]*Id="([^"]+)"[^>]*Target="([^"]+)"[^>]*\/>/g;
-    let match;
-
-    while ((match = relationshipRegex.exec(relsXml)) !== null) {
-      if (match[1] && match[2]) {
-        map.set(match[1], match[2]); // rId -> target path
-      }
-    }
-
-    return map;
-  }
-
-  /**
-   * Get the next available relationship ID
-   * Finds the highest numeric ID and increments it
-   */
-  private getNextRelationshipId(existingIds: Set<string>): string {
-    let maxId = 0;
-    for (const id of existingIds) {
-      const num = parseInt(id.replace('rId', ''));
-      if (!isNaN(num) && num > maxId) {
-        maxId = num;
-      }
-    }
-    return `rId${maxId + 1}`;
-  }
-
-  /**
-   * Add a new relationship to word/_rels/document.xml.rels
-   */
-  private addRelationship(rId: string, target: string, type: string): void {
-    const relsXml = this.zipHandler.getFileAsString('word/_rels/document.xml.rels');
-    if (!relsXml) return;
-
-    // Insert new relationship before closing tag
-    const newRel = `<Relationship Id="${rId}" Type="${type}" Target="${target}"/>`;
-    const updated = relsXml.replace('</Relationships>', `${newRel}\n</Relationships>`);
-
-    this.zipHandler.updateFile('word/_rels/document.xml.rels', updated);
-  }
-
-  /**
-   * Clean up all revision metadata files (people.xml, settings.xml, core.xml).
-   *
-   * This removes:
-   * - All revision authors from people.xml
-   * - Track changes settings from settings.xml
-   * - Resets revision count in core.xml
-   *
-   * Called as part of acceptAllRevisions() but can also be called separately
-   * when using in-memory revision acceptance.
-   */
-  public cleanupMetadata(): void {
-    this.cleanupPeopleXml();
-    this.cleanupSettingsXml();
-    this.cleanupCorePropsXml();
-  }
-
-  /**
-   * Clean up word/people.xml - remove all revision authors
-   *
-   * Handles both w: and w15: namespace variants
-   */
-  private cleanupPeopleXml(): void {
-    const peopleXml = this.zipHandler.getFileAsString('word/people.xml');
-    if (!peopleXml) {
-      return;
-    }
-
-    let content = peopleXml;
-
-    // Remove all person elements in any namespace variant
-    content = content.replace(/<w:person\b[^>]*>[\s\S]*?<\/w:person>/g, '');
-    content = content.replace(/<w15:person\b[^>]*>[\s\S]*?<\/w15:person>/g, '');
-    
-    // Handle any namespace-prefixed variants (w1:, w2:, etc.)
-    content = content.replace(/<w\d+:person\b[^>]*>[\s\S]*?<\/w\d+:person>/g, '');
-    
-    // Also remove self-closing person elements
-    content = content.replace(/<w:person\b[^>]*\/>/g, '');
-    content = content.replace(/<w15:person\b[^>]*\/>/g, '');
-    content = content.replace(/<w\d+:person\b[^>]*\/>/g, '');
-
-    this.zipHandler.updateFile('word/people.xml', content);
-  }
-
-  /**
-   * Clean up word/settings.xml - disable track changes
-   */
-  private cleanupSettingsXml(): void {
-    const settingsXml = this.zipHandler.getFileAsString('word/settings.xml');
-    if (!settingsXml) {
-      return;
-    }
-
-    let content = settingsXml;
-
-    // Remove trackRevisions element (enables tracking)
-    content = content.replace(/<w:trackRevisions\b[^>]*\/>/g, '');
-    content = content.replace(/<w:trackRevisions\b[^>]*>[\s\S]*?<\/w:trackRevisions>/g, '');
-
-    // Remove revisionView element (controls which revisions are visible)
-    content = content.replace(/<w:revisionView\b[^>]*\/>/g, '');
-    content = content.replace(/<w:revisionView\b[^>]*>[\s\S]*?<\/w:revisionView>/g, '');
-
-    // Remove doNotTrackMoves (prevents move tracking)
-    content = content.replace(/<w:doNotTrackMoves\b[^>]*\/>/g, '');
-    content = content.replace(/<w:doNotTrackMoves\b[^>]*>[\s\S]*?<\/w:doNotTrackMoves>/g, '');
-
-    // Remove doNotTrackFormatting
-    content = content.replace(/<w:doNotTrackFormatting\b[^>]*\/>/g, '');
-    content = content.replace(/<w:doNotTrackFormatting\b[^>]*>[\s\S]*?<\/w:doNotTrackFormatting>/g, '');
-
-    this.zipHandler.updateFile('word/settings.xml', content);
-  }
-
-  /**
-   * Clean up docProps/core.xml - reset revision count
-   */
-  private cleanupCorePropsXml(): void {
-    const coreXml = this.zipHandler.getFileAsString('docProps/core.xml');
-    if (!coreXml) {
-      return;
-    }
-
-    // Reset revision count to 1
-    const content = coreXml.replace(
-      /<cp:revision>\d+<\/cp:revision>/g,
-      '<cp:revision>1</cp:revision>'
-    );
-
-    this.zipHandler.updateFile('docProps/core.xml', content);
-  }
-
-  // =========================================================================
-  // DOM-based processing helper methods
-  // =========================================================================
-
-  /**
-   * Convert parsed XML object back to XML string
-   * Preserves element order using _orderedChildren metadata
-   *
-   * Based on DocumentParser.objectToXml() implementation
-   */
-  private objectToXml(obj: any): string {
-    const buildXml = (o: any, name?: string): string => {
-      // Handle simple string/number with a tag name: <tagName>value</tagName>
-      if (name && (typeof o === 'string' || typeof o === 'number')) {
-        return `<${name}>${this.escapeXml(String(o))}</${name}>`;
-      }
-      if (typeof o === 'string') return this.escapeXml(o);
-      if (typeof o !== 'object' || o === null) return String(o ?? '');
-
-      const keys = Object.keys(o);
-
-      // If a name is provided, we're building a specific element
-      // Don't return empty string for empty objects with a name - they become self-closing tags
-      if (keys.length === 0 && !name) return '';
-
-      const tagName = name || keys[0]!;
-      const element = name ? o : o[tagName];
-
-      let xml = `<${tagName}`;
-
-      // Add attributes (keys starting with @_)
-      if (element && typeof element === 'object') {
-        for (const key of Object.keys(element)) {
-          if (key.startsWith('@_')) {
-            const attrName = key.substring(2);
-            xml += ` ${attrName}="${this.escapeXml(String(element[key]))}"`;
-          }
-        }
-      }
-
-      // Check for children (non-attribute, non-text, non-metadata keys)
-      const hasChildren =
-        element &&
-        typeof element === 'object' &&
-        Object.keys(element).some(
-          (k) =>
-            !k.startsWith('@_') && k !== '#text' && k !== '_orderedChildren'
-        );
-
-      // Per ECMA-376, certain elements MUST NOT be self-closing (e.g., <w:p/> is invalid).
-      // This mirrors the CANNOT_SELF_CLOSE list in XMLBuilder.ts.
-      const CANNOT_SELF_CLOSE = [
-        'w:t', 'w:r', 'w:p', 'w:tbl', 'w:tr', 'w:tc', 'w:body',
-        'w:document', 'w:hyperlink', 'w:sdt', 'w:sdtContent', 'w:sdtPr',
-        'w:pPr', 'w:rPr', 'w:sectPr', 'w:del', 'w:ins', 'w:moveFrom', 'w:moveTo',
-      ];
-
-      if (!hasChildren && (!element?.['#text'])) {
-        if (CANNOT_SELF_CLOSE.includes(tagName)) {
-          xml += `></${tagName}>`;
-        } else {
-          xml += '/>';
-        }
-      } else {
-        xml += '>';
-
-        // Add text content
-        if (element?.['#text']) {
-          xml += this.escapeXml(String(element['#text']));
-        }
-
-        // Add child elements using _orderedChildren if available
-        if (element && typeof element === 'object') {
-          const orderedChildren = element._orderedChildren as
-            | { type: string; index: number }[]
-            | undefined;
-
-          if (orderedChildren && orderedChildren.length > 0) {
-            // Use _orderedChildren to preserve element order
-            for (const childInfo of orderedChildren) {
-              const childType = childInfo.type;
-              const childIndex = childInfo.index;
-
-              if (element[childType] !== undefined) {
-                const children = element[childType];
-
-                if (Array.isArray(children)) {
-                  if (childIndex < children.length) {
-                    const childXml = buildXml(children[childIndex], childType);
-                    xml += childXml;
-                  }
-                } else {
-                  // Single child element
-                  if (childIndex === 0) {
-                    const childXml = buildXml(children, childType);
-                    xml += childXml;
-                  }
-                }
-              }
-            }
-          } else {
-            // Fallback: iterate through keys if no _orderedChildren
-            for (const key of Object.keys(element)) {
-              if (
-                !key.startsWith('@_') &&
-                key !== '#text' &&
-                key !== '_orderedChildren'
-              ) {
-                const children = element[key];
-                if (Array.isArray(children)) {
-                  for (const child of children) {
-                    xml += buildXml(child, key);
-                  }
-                } else {
-                  xml += buildXml(children, key);
-                }
-              }
-            }
-          }
-        }
-
-        xml += `</${tagName}>`;
-      }
-
-      return xml;
-    };
-
-    return buildXml(obj);
-  }
-
-  /**
-   * Escape special XML characters
-   */
-  private escapeXml(str: string): string {
-    return str
-      .replace(/&/g, '&amp;')
-      .replace(/</g, '&lt;')
-      .replace(/>/g, '&gt;')
-      .replace(/"/g, '&quot;')
-      .replace(/'/g, '&apos;');
-  }
-
-  /**
-   * Remap image relationship IDs in the parsed tree to prevent duplicates
-   * Walks the tree looking for r:embed attributes and assigns unique IDs
-   */
-  private remapImageRelationshipsInTree(obj: any): void {
-    const relationships = this.parseRelationships();
-    const existingIds = new Set(relationships.keys());
-    const remappedIds = new Map<string, string>();
-
-    // Walk the tree and find all r:embed attributes
-    this.walkTreeForEmbeds(obj, (embedId: string, parent: any, key: string) => {
-      // Check if this ID has already been remapped
-      if (remappedIds.has(embedId)) {
-        parent[key] = remappedIds.get(embedId);
-        return;
-      }
-
-      // Check if this ID needs remapping (duplicate)
-      // For DOM-based processing, we check if we've seen this ID before in this pass
-      const target = relationships.get(embedId);
-      if (target) {
-        // Generate new unique ID
-        const newId = this.getNextRelationshipId(existingIds);
-        existingIds.add(newId);
-        remappedIds.set(embedId, newId);
-
-        // Add relationship with same target
-        this.addRelationship(
-          newId,
-          target,
-          'http://schemas.openxmlformats.org/officeDocument/2006/relationships/image'
-        );
-
-        // Update the attribute
-        parent[key] = newId;
-      }
-    });
-  }
-
-  /**
-   * Walk the tree looking for r:embed attributes
-   */
-  private walkTreeForEmbeds(
-    obj: any,
-    callback: (embedId: string, parent: any, key: string) => void
-  ): void {
-    if (!obj || typeof obj !== 'object') {
-      return;
-    }
-
-    for (const key of Object.keys(obj)) {
-      // Check for r:embed attribute
-      if (key === '@_r:embed') {
-        callback(obj[key], obj, key);
-      } else if (
-        !key.startsWith('@_') &&
-        key !== '#text' &&
-        key !== '_orderedChildren'
-      ) {
-        const value = obj[key];
-        if (Array.isArray(value)) {
-          for (const item of value) {
-            this.walkTreeForEmbeds(item, callback);
-          }
-        } else if (typeof value === 'object') {
-          this.walkTreeForEmbeds(value, callback);
-        }
-      }
-    }
-  }
-}
-
-/**
- * Convenience function to accept all revisions in a document
- */
-export async function acceptAllRevisions(zipHandler: ZipHandler): Promise<void> {
-  const acceptor = new RevisionAcceptor(zipHandler);
-  await acceptor.acceptAllRevisions();
-}
-
-/**
- * Convenience function to clean up revision metadata files.
- *
- * This removes:
- * - All revision authors from people.xml
- * - Track changes settings from settings.xml
- * - Resets revision count in core.xml
- *
- * Use this after in-memory revision acceptance to ensure metadata is also cleaned.
- * The raw XML acceptAllRevisions() function calls this automatically.
- *
- * @param zipHandler - The ZipHandler containing the DOCX package
- */
-export function cleanupRevisionMetadata(zipHandler: ZipHandler): void {
-  const acceptor = new RevisionAcceptor(zipHandler);
-  acceptor.cleanupMetadata();
-}
+import { ZipHandler } from '../zip/ZipHandler';
+import { XMLParser } from '../xml/XMLParser';
+import { RevisionWalker } from './RevisionWalker';
+
+/**
+ * Accepts all tracked changes in a Word document per Microsoft's OpenXML SDK pattern
+ *
+ * This implementation uses DOM-based tree walking for reliability:
+ * 1. Insertions (<w:ins>): Keep content, remove wrapper tags
+ * 2. Deletions (<w:del>): Remove entirely (content and tags)
+ * 3. Move From (<w:moveFrom>): Remove entirely (source of move)
+ * 4. Move To (<w:moveTo>): Keep content, remove wrapper (destination of move)
+ * 5. Property changes: Remove all *Change elements
+ * 6. Range markers: Remove all boundary markers
+ *
+ * Also cleans up metadata in people.xml, settings.xml, and core.xml
+ *
+ * @see https://learn.microsoft.com/en-us/office/open-xml/how-to-accept-all-revisions
+ */
+class RevisionAcceptor {
+  private zipHandler: ZipHandler;
+  /** Feature flag for DOM-based processing (default: true) */
+  private useDomBasedProcessing = true;
+
+  constructor(zipHandler: ZipHandler) {
+    this.zipHandler = zipHandler;
+  }
+
+  /**
+   * Enable or disable DOM-based processing (for testing/migration)
+   */
+  setUseDomBasedProcessing(enabled: boolean): void {
+    this.useDomBasedProcessing = enabled;
+  }
+
+  /**
+   * Main method to accept all revisions in the document
+   */
+  public async acceptAllRevisions(): Promise<void> {
+    // Process document.xml
+    await this.processDocumentPart('word/document.xml');
+
+    // Process headers
+    const files = this.zipHandler.getFilePaths();
+    for (const file of files) {
+      if (/^word\/header\d+\.xml$/.exec(file)) {
+        await this.processDocumentPart(file);
+      }
+      if (/^word\/footer\d+\.xml$/.exec(file)) {
+        await this.processDocumentPart(file);
+      }
+    }
+
+    // Clean up metadata files
+    this.cleanupPeopleXml();
+    this.cleanupSettingsXml();
+    this.cleanupCorePropsXml();
+  }
+
+  /**
+   * Process a document part (document.xml, header, footer) to accept revisions
+   */
+  private async processDocumentPart(partPath: string): Promise<void> {
+    if (this.useDomBasedProcessing) {
+      return this.processDocumentPartDOM(partPath);
+    }
+    return this.processDocumentPartRegex(partPath);
+  }
+
+  /**
+   * DOM-based implementation of revision acceptance
+   * Uses XMLParser and RevisionWalker for reliable processing
+   */
+  private processDocumentPartDOM(partPath: string): void {
+    const xml = this.zipHandler.getFileAsString(partPath);
+    if (!xml) {
+      return;
+    }
+
+    // Step 1: Parse XML to object tree
+    // IMPORTANT: trimValues: false preserves whitespace from xml:space="preserve" attributes
+    const parsed = XMLParser.parseToObject(xml, { trimValues: false });
+
+    // Step 2: Process revisions using DOM walker
+    const processed = RevisionWalker.processTree(parsed, {
+      acceptInsertions: true,
+      acceptDeletions: true,
+      acceptMoves: true,
+      acceptPropertyChanges: true,
+    });
+
+    // Step 3: Handle image relationship ID remapping
+    this.remapImageRelationshipsInTree(processed);
+
+    // Step 4: Convert back to XML
+    const outputXml =
+      '<?xml version="1.0" encoding="UTF-8" standalone="yes"?>\n' + this.objectToXml(processed);
+
+    // Step 5: Update file
+    this.zipHandler.updateFile(partPath, outputXml);
+  }
+
+  /**
+   * Legacy RegEx-based implementation (kept as fallback)
+   */
+  private processDocumentPartRegex(partPath: string): void {
+    const xml = this.zipHandler.getFileAsString(partPath);
+    if (!xml) {
+      return;
+    }
+
+    let content = xml;
+
+    // Step 1: Remove all range markers FIRST (before processing revisions)
+    // This prevents orphaned references when revision content is modified
+    content = this.removeAllRangeMarkers(content);
+
+    // Step 2: Remove all property change elements
+    // These track formatting changes and must be removed before other processing
+    content = this.removeAllPropertyChanges(content);
+
+    // Step 3: Process deletions - remove entire element INCLUDING content
+    // Must be done before insertions to handle nested scenarios
+    content = this.acceptDeletions(content);
+
+    // Step 4: Process move operations
+    // Remove moveFrom entirely (source), unwrap moveTo (destination)
+    content = this.acceptMoveFrom(content);
+    content = this.acceptMoveTo(content);
+
+    // Step 5: Process insertions - keep content, remove wrapper
+    content = this.acceptInsertions(content);
+
+    // Step 6: Final cleanup - remove any remaining orphaned tags
+    content = this.cleanupOrphanedTags(content);
+
+    // Update the file
+    this.zipHandler.updateFile(partPath, content);
+  }
+
+  /**
+   * Remove all range marker elements
+   * These are boundary markers for tracked changes and moves
+   */
+  private removeAllRangeMarkers(xml: string): string {
+    const patterns = [
+      // Move range markers
+      /<w:moveFromRangeStart[^>]*(?:\/>|>.*?<\/w:moveFromRangeStart>)/gs,
+      /<w:moveFromRangeEnd[^>]*(?:\/>|>.*?<\/w:moveFromRangeEnd>)/gs,
+      /<w:moveToRangeStart[^>]*(?:\/>|>.*?<\/w:moveToRangeStart>)/gs,
+      /<w:moveToRangeEnd[^>]*(?:\/>|>.*?<\/w:moveToRangeEnd>)/gs,
+      // Custom XML range markers
+      /<w:customXmlInsRangeStart[^>]*(?:\/>|>.*?<\/w:customXmlInsRangeStart>)/gs,
+      /<w:customXmlInsRangeEnd[^>]*(?:\/>|>.*?<\/w:customXmlInsRangeEnd>)/gs,
+      /<w:customXmlDelRangeStart[^>]*(?:\/>|>.*?<\/w:customXmlDelRangeStart>)/gs,
+      /<w:customXmlDelRangeEnd[^>]*(?:\/>|>.*?<\/w:customXmlDelRangeEnd>)/gs,
+      /<w:customXmlMoveFromRangeStart[^>]*(?:\/>|>.*?<\/w:customXmlMoveFromRangeStart>)/gs,
+      /<w:customXmlMoveFromRangeEnd[^>]*(?:\/>|>.*?<\/w:customXmlMoveFromRangeEnd>)/gs,
+      /<w:customXmlMoveToRangeStart[^>]*(?:\/>|>.*?<\/w:customXmlMoveToRangeStart>)/gs,
+      /<w:customXmlMoveToRangeEnd[^>]*(?:\/>|>.*?<\/w:customXmlMoveToRangeEnd>)/gs,
+    ];
+
+    let result = xml;
+    for (const pattern of patterns) {
+      result = result.replace(pattern, '');
+    }
+    return result;
+  }
+
+  /**
+   * Remove all property change tracking elements
+   * Per ECMA-376, these track previous state of formatting
+   */
+  private removeAllPropertyChanges(xml: string): string {
+    const patterns = [
+      // Run property changes
+      /<w:rPrChange[^>]*>[\s\S]*?<\/w:rPrChange>/g,
+      // Paragraph property changes
+      /<w:pPrChange[^>]*>[\s\S]*?<\/w:pPrChange>/g,
+      // Table property changes
+      /<w:tblPrChange[^>]*>[\s\S]*?<\/w:tblPrChange>/g,
+      /<w:tblPrExChange[^>]*>[\s\S]*?<\/w:tblPrExChange>/g,
+      // Table cell property changes
+      /<w:tcPrChange[^>]*>[\s\S]*?<\/w:tcPrChange>/g,
+      // Table row property changes
+      /<w:trPrChange[^>]*>[\s\S]*?<\/w:trPrChange>/g,
+      // Section property changes
+      /<w:sectPrChange[^>]*>[\s\S]*?<\/w:sectPrChange>/g,
+      // Table grid changes
+      /<w:tblGridChange[^>]*>[\s\S]*?<\/w:tblGridChange>/g,
+      // Numbering changes
+      /<w:numberingChange[^>]*>[\s\S]*?<\/w:numberingChange>/g,
+    ];
+
+    let result = xml;
+    for (const pattern of patterns) {
+      result = result.replace(pattern, '');
+    }
+    return result;
+  }
+
+  /**
+   * Accept deletions - remove the entire <w:del> element including its content
+   *
+   * Per Microsoft SDK: "DeletedRun elements should be removed along with their content"
+   */
+  private acceptDeletions(xml: string): string {
+    let result = xml;
+    let previousLength = 0;
+
+    // Iterate until no more deletions (handles nested cases)
+    while (result.length !== previousLength) {
+      previousLength = result.length;
+
+      // Match complete <w:del ...>...</w:del> elements and remove entirely
+      result = result.replace(/<w:del\b[^>]*>[\s\S]*?<\/w:del>/g, '');
+    }
+
+    // Also remove self-closing deletion tags
+    result = result.replace(/<w:del\b[^>]*\/>/g, '');
+
+    return result;
+  }
+
+  /**
+   * Accept moveFrom - remove the entire element (source of moved content)
+   *
+   * The content exists at the moveTo destination, so we discard the source
+   */
+  private acceptMoveFrom(xml: string): string {
+    let result = xml;
+    let previousLength = 0;
+
+    while (result.length !== previousLength) {
+      previousLength = result.length;
+      result = result.replace(/<w:moveFrom\b[^>]*>[\s\S]*?<\/w:moveFrom>/g, '');
+    }
+
+    // Also remove self-closing tags
+    result = result.replace(/<w:moveFrom\b[^>]*\/>/g, '');
+
+    return result;
+  }
+
+  /**
+   * Accept moveTo - keep the content, remove the wrapper tags
+   *
+   * The moveTo location is where the content should remain
+   */
+  private acceptMoveTo(xml: string): string {
+    let result = xml;
+
+    // Remove closing tags first (prevents issues with regex matching)
+    result = result.replace(/<\/w:moveTo>/g, '');
+
+    // Remove opening tags (keeps content that was inside)
+    result = result.replace(/<w:moveTo\b[^>]*>/g, '');
+
+    return result;
+  }
+
+  /**
+   * Accept insertions - keep the content, remove the wrapper tags
+   *
+   * Per Microsoft SDK: "InsertedRun elements should be unwrapped, keeping their content"
+   *
+   * IMPORTANT: This method now handles relationship ID remapping for images inside insertions.
+   * When Word tracks changes with images, it can reuse relationship IDs (like rId5) because
+   * they're in separate tracked change contexts. But when we unwrap them, duplicate IDs
+   * cause corruption. This method assigns new unique IDs to images inside insertions.
+   */
+  private acceptInsertions(xml: string): string {
+    let result = xml;
+
+    // Parse existing relationships
+    const relationships = this.parseRelationships();
+    const existingIds = new Set(relationships.keys());
+
+    // Process each w:ins element and remap images one by one
+    const insRegex = /<w:ins\b[^>]*>[\s\S]*?<\/w:ins>/g;
+
+    result = result.replace(insRegex, (insMatch) => {
+      // For each image reference inside this insertion, generate a unique new ID
+      return insMatch.replace(/r:embed="(rId\d+)"/g, (embedMatch, oldId) => {
+        // Generate new unique ID for THIS occurrence
+        const newId = this.getNextRelationshipId(existingIds);
+        existingIds.add(newId);
+
+        // Add relationship with same target as original
+        const target = relationships.get(oldId);
+        if (target) {
+          this.addRelationship(
+            newId,
+            target,
+            'http://schemas.openxmlformats.org/officeDocument/2006/relationships/image'
+          );
+        }
+
+        return `r:embed="${newId}"`;
+      });
+    });
+
+    // Now unwrap the w:ins tags (content has unique remapped IDs)
+    result = result.replace(/<\/w:ins>/g, '');
+    result = result.replace(/<w:ins\b[^>]*>/g, '');
+
+    return result;
+  }
+
+  /**
+   * Final cleanup to remove any orphaned or malformed revision-related tags
+   */
+  private cleanupOrphanedTags(xml: string): string {
+    let result = xml;
+
+    // Remove any remaining self-closing revision tags
+    result = result.replace(/<w:ins\b[^>]*\/>/g, '');
+    result = result.replace(/<w:del\b[^>]*\/>/g, '');
+    result = result.replace(/<w:moveFrom\b[^>]*\/>/g, '');
+    result = result.replace(/<w:moveTo\b[^>]*\/>/g, '');
+
+    // Remove empty w:r elements that might be left after removing deletions
+    result = result.replace(/<w:r\b[^>]*>\s*<\/w:r>/g, '');
+
+    // Remove empty w:p elements (but preserve those with properties)
+    // We keep <w:p><w:pPr>...</w:pPr></w:p> as those are intentional empty paragraphs with styling
+    result = result.replace(/<w:p>\s*<\/w:p>/g, '');
+
+    return result;
+  }
+
+  /**
+   * Parse relationship IDs from word/_rels/document.xml.rels
+   * Returns a map of relationship ID to target path
+   */
+  private parseRelationships(): Map<string, string> {
+    const relsXml = this.zipHandler.getFileAsString('word/_rels/document.xml.rels');
+    if (!relsXml) return new Map();
+
+    const map = new Map<string, string>();
+    const relationshipRegex = /<Relationship[^>]*Id="([^"]+)"[^>]*Target="([^"]+)"[^>]*\/>/g;
+    let match;
+
+    while ((match = relationshipRegex.exec(relsXml)) !== null) {
+      if (match[1] && match[2]) {
+        map.set(match[1], match[2]); // rId -> target path
+      }
+    }
+
+    return map;
+  }
+
+  /**
+   * Get the next available relationship ID
+   * Finds the highest numeric ID and increments it
+   */
+  private getNextRelationshipId(existingIds: Set<string>): string {
+    let maxId = 0;
+    for (const id of existingIds) {
+      const num = parseInt(id.replace('rId', ''));
+      if (!isNaN(num) && num > maxId) {
+        maxId = num;
+      }
+    }
+    return `rId${maxId + 1}`;
+  }
+
+  /**
+   * Add a new relationship to word/_rels/document.xml.rels
+   */
+  private addRelationship(rId: string, target: string, type: string): void {
+    const relsXml = this.zipHandler.getFileAsString('word/_rels/document.xml.rels');
+    if (!relsXml) return;
+
+    // Insert new relationship before closing tag
+    const newRel = `<Relationship Id="${rId}" Type="${type}" Target="${target}"/>`;
+    const updated = relsXml.replace('</Relationships>', `${newRel}\n</Relationships>`);
+
+    this.zipHandler.updateFile('word/_rels/document.xml.rels', updated);
+  }
+
+  /**
+   * Clean up all revision metadata files (people.xml, settings.xml, core.xml).
+   *
+   * This removes:
+   * - All revision authors from people.xml
+   * - Track changes settings from settings.xml
+   * - Resets revision count in core.xml
+   *
+   * Called as part of acceptAllRevisions() but can also be called separately
+   * when using in-memory revision acceptance.
+   */
+  public cleanupMetadata(): void {
+    this.cleanupPeopleXml();
+    this.cleanupSettingsXml();
+    this.cleanupCorePropsXml();
+  }
+
+  /**
+   * Clean up word/people.xml - remove all revision authors
+   *
+   * Handles both w: and w15: namespace variants
+   */
+  private cleanupPeopleXml(): void {
+    const peopleXml = this.zipHandler.getFileAsString('word/people.xml');
+    if (!peopleXml) {
+      return;
+    }
+
+    let content = peopleXml;
+
+    // Remove all person elements in any namespace variant
+    content = content.replace(/<w:person\b[^>]*>[\s\S]*?<\/w:person>/g, '');
+    content = content.replace(/<w15:person\b[^>]*>[\s\S]*?<\/w15:person>/g, '');
+
+    // Handle any namespace-prefixed variants (w1:, w2:, etc.)
+    content = content.replace(/<w\d+:person\b[^>]*>[\s\S]*?<\/w\d+:person>/g, '');
+
+    // Also remove self-closing person elements
+    content = content.replace(/<w:person\b[^>]*\/>/g, '');
+    content = content.replace(/<w15:person\b[^>]*\/>/g, '');
+    content = content.replace(/<w\d+:person\b[^>]*\/>/g, '');
+
+    this.zipHandler.updateFile('word/people.xml', content);
+  }
+
+  /**
+   * Clean up word/settings.xml - disable track changes
+   */
+  private cleanupSettingsXml(): void {
+    const settingsXml = this.zipHandler.getFileAsString('word/settings.xml');
+    if (!settingsXml) {
+      return;
+    }
+
+    let content = settingsXml;
+
+    // Remove trackRevisions element (enables tracking)
+    content = content.replace(/<w:trackRevisions\b[^>]*\/>/g, '');
+    content = content.replace(/<w:trackRevisions\b[^>]*>[\s\S]*?<\/w:trackRevisions>/g, '');
+
+    // Remove revisionView element (controls which revisions are visible)
+    content = content.replace(/<w:revisionView\b[^>]*\/>/g, '');
+    content = content.replace(/<w:revisionView\b[^>]*>[\s\S]*?<\/w:revisionView>/g, '');
+
+    // Remove doNotTrackMoves (prevents move tracking)
+    content = content.replace(/<w:doNotTrackMoves\b[^>]*\/>/g, '');
+    content = content.replace(/<w:doNotTrackMoves\b[^>]*>[\s\S]*?<\/w:doNotTrackMoves>/g, '');
+
+    // Remove doNotTrackFormatting
+    content = content.replace(/<w:doNotTrackFormatting\b[^>]*\/>/g, '');
+    content = content.replace(
+      /<w:doNotTrackFormatting\b[^>]*>[\s\S]*?<\/w:doNotTrackFormatting>/g,
+      ''
+    );
+
+    this.zipHandler.updateFile('word/settings.xml', content);
+  }
+
+  /**
+   * Clean up docProps/core.xml - reset revision count
+   */
+  private cleanupCorePropsXml(): void {
+    const coreXml = this.zipHandler.getFileAsString('docProps/core.xml');
+    if (!coreXml) {
+      return;
+    }
+
+    // Reset revision count to 1
+    const content = coreXml.replace(
+      /<cp:revision>\d+<\/cp:revision>/g,
+      '<cp:revision>1</cp:revision>'
+    );
+
+    this.zipHandler.updateFile('docProps/core.xml', content);
+  }
+
+  // =========================================================================
+  // DOM-based processing helper methods
+  // =========================================================================
+
+  /**
+   * Convert parsed XML object back to XML string
+   * Preserves element order using _orderedChildren metadata
+   *
+   * Based on DocumentParser.objectToXml() implementation
+   */
+  private objectToXml(obj: any): string {
+    const buildXml = (o: any, name?: string): string => {
+      // Handle simple string/number with a tag name: <tagName>value</tagName>
+      if (name && (typeof o === 'string' || typeof o === 'number')) {
+        return `<${name}>${this.escapeXml(String(o))}</${name}>`;
+      }
+      if (typeof o === 'string') return this.escapeXml(o);
+      if (typeof o !== 'object' || o === null) return String(o ?? '');
+
+      const keys = Object.keys(o);
+
+      // If a name is provided, we're building a specific element
+      // Don't return empty string for empty objects with a name - they become self-closing tags
+      if (keys.length === 0 && !name) return '';
+
+      const tagName = name || keys[0]!;
+      const element = name ? o : o[tagName];
+
+      let xml = `<${tagName}`;
+
+      // Add attributes (keys starting with @_)
+      if (element && typeof element === 'object') {
+        for (const key of Object.keys(element)) {
+          if (key.startsWith('@_')) {
+            const attrName = key.substring(2);
+            xml += ` ${attrName}="${this.escapeXml(String(element[key]))}"`;
+          }
+        }
+      }
+
+      // Check for children (non-attribute, non-text, non-metadata keys)
+      const hasChildren =
+        element &&
+        typeof element === 'object' &&
+        Object.keys(element).some(
+          (k) => !k.startsWith('@_') && k !== '#text' && k !== '_orderedChildren'
+        );
+
+      // Per ECMA-376, certain elements MUST NOT be self-closing (e.g., <w:p/> is invalid).
+      // This mirrors the CANNOT_SELF_CLOSE list in XMLBuilder.ts.
+      const CANNOT_SELF_CLOSE = [
+        'w:t',
+        'w:r',
+        'w:p',
+        'w:tbl',
+        'w:tr',
+        'w:tc',
+        'w:body',
+        'w:document',
+        'w:hyperlink',
+        'w:sdt',
+        'w:sdtContent',
+        'w:sdtPr',
+        'w:pPr',
+        'w:rPr',
+        'w:sectPr',
+        'w:del',
+        'w:ins',
+        'w:moveFrom',
+        'w:moveTo',
+      ];
+
+      if (!hasChildren && !element?.['#text']) {
+        if (CANNOT_SELF_CLOSE.includes(tagName)) {
+          xml += `></${tagName}>`;
+        } else {
+          xml += '/>';
+        }
+      } else {
+        xml += '>';
+
+        // Add text content
+        if (element?.['#text']) {
+          xml += this.escapeXml(String(element['#text']));
+        }
+
+        // Add child elements using _orderedChildren if available
+        if (element && typeof element === 'object') {
+          const orderedChildren = element._orderedChildren as
+            | { type: string; index: number }[]
+            | undefined;
+
+          if (orderedChildren && orderedChildren.length > 0) {
+            // Use _orderedChildren to preserve element order
+            for (const childInfo of orderedChildren) {
+              const childType = childInfo.type;
+              const childIndex = childInfo.index;
+
+              if (element[childType] !== undefined) {
+                const children = element[childType];
+
+                if (Array.isArray(children)) {
+                  if (childIndex < children.length) {
+                    const childXml = buildXml(children[childIndex], childType);
+                    xml += childXml;
+                  }
+                } else {
+                  // Single child element
+                  if (childIndex === 0) {
+                    const childXml = buildXml(children, childType);
+                    xml += childXml;
+                  }
+                }
+              }
+            }
+          } else {
+            // Fallback: iterate through keys if no _orderedChildren
+            for (const key of Object.keys(element)) {
+              if (!key.startsWith('@_') && key !== '#text' && key !== '_orderedChildren') {
+                const children = element[key];
+                if (Array.isArray(children)) {
+                  for (const child of children) {
+                    xml += buildXml(child, key);
+                  }
+                } else {
+                  xml += buildXml(children, key);
+                }
+              }
+            }
+          }
+        }
+
+        xml += `</${tagName}>`;
+      }
+
+      return xml;
+    };
+
+    return buildXml(obj);
+  }
+
+  /**
+   * Escape special XML characters
+   */
+  private escapeXml(str: string): string {
+    return str
+      .replace(/&/g, '&amp;')
+      .replace(/</g, '&lt;')
+      .replace(/>/g, '&gt;')
+      .replace(/"/g, '&quot;')
+      .replace(/'/g, '&apos;');
+  }
+
+  /**
+   * Remap image relationship IDs in the parsed tree to prevent duplicates
+   * Walks the tree looking for r:embed attributes and assigns unique IDs
+   */
+  private remapImageRelationshipsInTree(obj: any): void {
+    const relationships = this.parseRelationships();
+    const existingIds = new Set(relationships.keys());
+    const remappedIds = new Map<string, string>();
+
+    // Walk the tree and find all r:embed attributes
+    this.walkTreeForEmbeds(obj, (embedId: string, parent: any, key: string) => {
+      // Check if this ID has already been remapped
+      if (remappedIds.has(embedId)) {
+        parent[key] = remappedIds.get(embedId);
+        return;
+      }
+
+      // Check if this ID needs remapping (duplicate)
+      // For DOM-based processing, we check if we've seen this ID before in this pass
+      const target = relationships.get(embedId);
+      if (target) {
+        // Generate new unique ID
+        const newId = this.getNextRelationshipId(existingIds);
+        existingIds.add(newId);
+        remappedIds.set(embedId, newId);
+
+        // Add relationship with same target
+        this.addRelationship(
+          newId,
+          target,
+          'http://schemas.openxmlformats.org/officeDocument/2006/relationships/image'
+        );
+
+        // Update the attribute
+        parent[key] = newId;
+      }
+    });
+  }
+
+  /**
+   * Walk the tree looking for r:embed attributes
+   */
+  private walkTreeForEmbeds(
+    obj: any,
+    callback: (embedId: string, parent: any, key: string) => void
+  ): void {
+    if (!obj || typeof obj !== 'object') {
+      return;
+    }
+
+    for (const key of Object.keys(obj)) {
+      // Check for r:embed attribute
+      if (key === '@_r:embed') {
+        callback(obj[key], obj, key);
+      } else if (!key.startsWith('@_') && key !== '#text' && key !== '_orderedChildren') {
+        const value = obj[key];
+        if (Array.isArray(value)) {
+          for (const item of value) {
+            this.walkTreeForEmbeds(item, callback);
+          }
+        } else if (typeof value === 'object') {
+          this.walkTreeForEmbeds(value, callback);
+        }
+      }
+    }
+  }
+}
+
+/**
+ * Convenience function to accept all revisions in a document
+ */
+export async function acceptAllRevisions(zipHandler: ZipHandler): Promise<void> {
+  const acceptor = new RevisionAcceptor(zipHandler);
+  await acceptor.acceptAllRevisions();
+}
+
+/**
+ * Convenience function to clean up revision metadata files.
+ *
+ * This removes:
+ * - All revision authors from people.xml
+ * - Track changes settings from settings.xml
+ * - Resets revision count in core.xml
+ *
+ * Use this after in-memory revision acceptance to ensure metadata is also cleaned.
+ * The raw XML acceptAllRevisions() function calls this automatically.
+ *
+ * @param zipHandler - The ZipHandler containing the DOCX package
+ */
+export function cleanupRevisionMetadata(zipHandler: ZipHandler): void {
+  const acceptor = new RevisionAcceptor(zipHandler);
+  acceptor.cleanupMetadata();
+}