Build your personal knowledge base with Trilium Notes
TypeScript electronelectron-appknowledge-baseknowledge-graphknowledge-managementknowledge-management-graphlocal-firstnote-managmentnote-takernote-takingnotebooknotesnotes-apppersonal-knowledge-basepersonal-wikiscriptableself-hostedself-hostingwiki
fix(ocr): port the office processor to officeparser 8
Elian Doran4cc3884e9379
diff --git a/apps/server/src/services/ocr/processors/office_processor.spec.ts b/apps/server/src/services/ocr/processors/office_processor.spec.tsindex 88d6edc92e..b1f1234537 100644--- a/apps/server/src/services/ocr/processors/office_processor.spec.ts+++ b/apps/server/src/services/ocr/processors/office_processor.spec.ts@@ -53,7 +53,7 @@ describe('OfficeProcessor', () => {it('extracts text and reports high confidence when text is present', async () => {const processor = new OfficeProcessor();- mockParseOffice.mockResolvedValue({ toText: () => ' document body ' });+ mockParseOffice.mockResolvedValue({ to: async () => ({ value: ' document body ' }) });const result = await processor.extractText(buffer, { mimeType: DOCX, language: 'deu' });@@ -62,7 +62,6 @@ describe('OfficeProcessor', () => {expect(result.pageCount).toBe(1);expect(result.language).toBe('deu');expect(mockParseOffice).toHaveBeenCalledWith(buffer, {- outputErrorToConsole: false,newlineDelimiter: '\n',ignoreNotes: false});@@ -70,13 +69,12 @@ describe('OfficeProcessor', () => {it('passes an explicit fileType hint for RTF, whose buffer auto-detection is unreliable', async () => {const processor = new OfficeProcessor();- mockParseOffice.mockResolvedValue({ toText: () => 'rtf body' });+ mockParseOffice.mockResolvedValue({ to: async () => ({ value: 'rtf body' }) });const result = await processor.extractText(buffer, { mimeType: RTF });expect(result.text).toBe('rtf body');expect(mockParseOffice).toHaveBeenCalledWith(buffer, {- outputErrorToConsole: false,newlineDelimiter: '\n',ignoreNotes: false,fileType: 'rtf'@@ -85,13 +83,12 @@ describe('OfficeProcessor', () => {it('passes an explicit fileType hint for EPUB, which shares the ZIP container with DOCX/ODT', async () => {const processor = new OfficeProcessor();- mockParseOffice.mockResolvedValue({ toText: () => 'epub body' });+ mockParseOffice.mockResolvedValue({ to: async () => ({ value: 'epub body' }) });const result = await processor.extractText(buffer, { mimeType: EPUB });expect(result.text).toBe('epub body');expect(mockParseOffice).toHaveBeenCalledWith(buffer, {- outputErrorToConsole: false,newlineDelimiter: '\n',ignoreNotes: false,fileType: 'epub'@@ -100,7 +97,7 @@ describe('OfficeProcessor', () => {it('reports zero confidence and defaults language when no text is extracted', async () => {const processor = new OfficeProcessor();- mockParseOffice.mockResolvedValue({ toText: () => ' ' });+ mockParseOffice.mockResolvedValue({ to: async () => ({ value: ' ' }) });const result = await processor.extractText(buffer, { mimeType: ODT });diff --git a/apps/server/src/services/ocr/processors/office_processor.ts b/apps/server/src/services/ocr/processors/office_processor.tsindex 4680dd8039..ff93341083 100644--- a/apps/server/src/services/ocr/processors/office_processor.ts+++ b/apps/server/src/services/ocr/processors/office_processor.ts@@ -6,7 +6,6 @@ import { OCRProcessingOptions, OCRResult } from '../ocr_service.js';import { FileProcessor } from './file_processor.js';const PARSER_CONFIG: OfficeParserConfig = {- outputErrorToConsole: false,newlineDelimiter: '\n',ignoreNotes: false};@@ -39,7 +38,11 @@ export class OfficeProcessor extends FileProcessor {// Dynamically imported so officeparser only loads when an Office file is actually processed.const { OfficeParser } = await import('officeparser');const ast = await OfficeParser.parseOffice(buffer, config);- const trimmed = ast.toText().trim();+ // `preserveLayout` pads cells into aligned columns and prefixes list markers, which the+ // search index has no use for and which makes a document's text depend on the format it+ // was authored in. A flat stream of text nodes indexes the same words either way.+ const { value } = await ast.to('text', { textConfig: { preserveLayout: false } });+ const trimmed = value.trim();return {text: trimmed,diff --git a/packages/trilium-core/src/services/office_preview.ts b/packages/trilium-core/src/services/office_preview.tsindex 8120d5a101..7b19873a4d 100644--- a/packages/trilium-core/src/services/office_preview.ts+++ b/packages/trilium-core/src/services/office_preview.ts@@ -56,13 +56,15 @@ export async function convertOfficeToHtml(content: string | Uint8Array, mime: stconst { OfficeConverter } = await import("officeparser");const { value } = await OfficeConverter.convert(buffer, "html", {// Pass the explicit fileType only when auto-detection is unreliable (RTF/EPUB).- // Embedded images need no extra config: the parser always emits them as data: URIs,- // which the client-side sanitizer permits on <img>.parseConfig: fileType ? { fileType } : undefined,generatorConfig: {// Charts would be emitted as a Chart.js <script> (loaded from a CDN) that the// client-side sanitizer strips anyway, so don't bother generating them.includeCharts: false,+ // Inline every embedded image as a data: URI, which the client-side sanitizer permits+ // on <img>. Above the default cap a fragment emits `<img src="<attachment name>">`,+ // which resolves against the note's URL and 404s.+ maxInlineImageBytes: Infinity,// Emit an embeddable fragment instead of a full standalone <html> document.htmlConfig: { standalone: false }}