#106 - Basic, English-only OCR

src/components/ResultItemVault.svelte

@@ -18,7 +18,6 @@
   $: {
     imagePath = null
     if (isFileImage(note.path)) {
-      console.log(note.path)
       // @ts-ignore
       const file = app.vault.getFiles().find(f => f.path === note.path)
       if (file) {

src/file-loader.ts

@@ -3,6 +3,7 @@ import {
   extractHeadingsFromCache,
   getAliasesFromMetadata,
   getTagsFromMetadata,
+  isFileImage,
   isFilePlaintext,
   removeDiacritics,
 } from './tools/utils'
@@ -10,7 +11,7 @@ import * as NotesIndex from './notes-index'
 import type { TFile } from 'obsidian'
 import type { IndexedDocument } from './globals'
 import { getNonExistingNotes } from './tools/notes'
-import { getPdfText } from 'obsidian-text-extract'
+import { getPdfText, getImageText } from 'obsidian-text-extract'
 
 /**
  * Return all plaintext files as IndexedDocuments
@@ -27,20 +28,33 @@ export async function getPlainTextFiles(): Promise<IndexedDocument[]> {
 }
 
 /**
- * Return all PDF files as IndexedDocuments.
- * If a PDF isn't cached, it will be read from the disk and added to the IndexedDB
+ * Return all PDFs as IndexedDocuments.
  */
-export async function getPDFFiles(): Promise<IndexedDocument[]> {
+export async function getPDFAsDocuments(): Promise<IndexedDocument[]> {
   const files = app.vault.getFiles().filter(f => f.path.endsWith('.pdf'))
+  return await getBinaryFiles(files)
+}
+
+/**
+ * Return all imageas as IndexedDocuments.
+ */
+export async function getImagesAsDocuments(): Promise<IndexedDocument[]> {
+  const files = app.vault.getFiles().filter(f => isFileImage(f.path))
+  return await getBinaryFiles(files)
+}
+
+async function getBinaryFiles(files: TFile[]): Promise<IndexedDocument[]> {
   const data: IndexedDocument[] = []
   const input = []
   for (const file of files) {
-    input.push(new Promise(async (resolve, reject) => {
-      const doc = await fileToIndexedDocument(file)
-      await cacheManager.updateLiveDocument(file.path, doc)
-      data.push(doc)
-      return resolve(null)
-    }))
+    input.push(
+      new Promise(async (resolve, reject) => {
+        const doc = await fileToIndexedDocument(file)
+        await cacheManager.updateLiveDocument(file.path, doc)
+        data.push(doc)
+        return resolve(null)
+      })
+    )
   }
   await Promise.all(input)
   return data
@@ -51,13 +65,15 @@ export async function getPDFFiles(): Promise<IndexedDocument[]> {
  * Will use the cache if possible.
  */
 export async function fileToIndexedDocument(
-  file: TFile,
+  file: TFile
 ): Promise<IndexedDocument> {
   let content: string
   if (isFilePlaintext(file.path)) {
     content = await app.vault.cachedRead(file)
   } else if (file.path.endsWith('.pdf')) {
     content = await getPdfText(file)
+  } else if (isFileImage(file.path)) {
+    content = await getImageText(file)
   } else {
     throw new Error('Invalid file: ' + file.path)
   }

src/main.ts

@@ -84,9 +84,7 @@ export default class OmnisearchPlugin extends Plugin {
     showWelcomeNotice(this)
   }
 
-  onunload(): void {
-
-  }
+  onunload(): void {}
 
   addRibbonButton(): void {
     this.ribbonButton = this.addRibbonIcon('search', 'Omnisearch', _evt => {
@@ -110,13 +108,13 @@ async function populateIndex(): Promise<void> {
   // Initialize minisearch
   let engine = SearchEngine.getEngine()
 
-  // No cache for iOS
+  // if not iOS, load data from cache
   if (!Platform.isIosApp) {
     engine = await SearchEngine.initFromCache()
   }
 
   // Load plaintext files
-  console.log('Omnisearch - Fetching notes')
+  console.log('Omnisearch - Reading notes')
   const plainTextFiles = await FileLoader.getPlainTextFiles()
   let allFiles = [...plainTextFiles]
   // iOS: since there's no cache, directly index the documents
@@ -127,15 +125,28 @@ async function populateIndex(): Promise<void> {
 
   // Load PDFs
   if (settings.PDFIndexing) {
-    console.log('Omnisearch - Fetching PDFs')
-    const pdfs = await FileLoader.getPDFFiles()
+    console.log('Omnisearch - Reading PDFs')
+    const pdfDocuments = await FileLoader.getPDFAsDocuments()
     // iOS: since there's no cache, just index the documents
     if (Platform.isIosApp) {
       await wait(1000)
-      await engine.addAllToMinisearch(pdfs)
+      await engine.addAllToMinisearch(pdfDocuments)
     }
     // Add PDFs to the files list
-    allFiles = [...allFiles, ...pdfs]
+    allFiles = [...allFiles, ...pdfDocuments]
+  }
+
+  // Load Images
+  if (settings.imagesIndexing) {
+    console.log('Omnisearch - Reading Images')
+    const imagesDocuments = await FileLoader.getImagesAsDocuments()
+    // iOS: since there's no cache, just index the documents
+    if (Platform.isIosApp) {
+      await wait(1000)
+      await engine.addAllToMinisearch(imagesDocuments)
+    }
+    // Add Images to the files list
+    allFiles = [...allFiles, ...imagesDocuments]
   }
 
   console.log('Omnisearch - Total number of files: ' + allFiles.length)
@@ -146,6 +157,7 @@ async function populateIndex(): Promise<void> {
     console.log('Omnisearch - Checking index cache diff...')
     // Check which documents need to be removed/added/updated
     const diffDocs = await cacheManager.getDiffDocuments(allFiles)
+    console.log(`Omnisearch - Files to add/remove/update: ${diffDocs.toAdd.length}/${diffDocs.toDelete.length}/${diffDocs.toUpdate.length}`)
     needToUpdateCache = !!(
       diffDocs.toAdd.length ||
       diffDocs.toDelete.length ||
@@ -154,23 +166,19 @@ async function populateIndex(): Promise<void> {
 
     // Add
     await engine.addAllToMinisearch(diffDocs.toAdd)
-    console.log(`Omnisearch - ${diffDocs.toAdd.length} files to add`)
     diffDocs.toAdd.forEach(doc =>
       cacheManager.updateLiveDocument(doc.path, doc)
     )
 
     // Delete
-    console.log(`Omnisearch - ${diffDocs.toDelete.length} files to remove`)
     diffDocs.toDelete.forEach(d => engine.removeFromMinisearch(d))
     diffDocs.toDelete.forEach(doc => cacheManager.deleteLiveDocument(doc.path))
 
     // Update (delete + add)
-    console.log(`Omnisearch - ${diffDocs.toUpdate.length} files to update`)
-    diffDocs.toUpdate
-      .forEach(({ oldDoc, newDoc }) => {
-        engine.removeFromMinisearch(oldDoc)
-        cacheManager.updateLiveDocument(oldDoc.path, newDoc)
-      })
+    diffDocs.toUpdate.forEach(({ oldDoc, newDoc }) => {
+      engine.removeFromMinisearch(oldDoc)
+      cacheManager.updateLiveDocument(oldDoc.path, newDoc)
+    })
     await engine.addAllToMinisearch(diffDocs.toUpdate.map(d => d.newDoc))
   }
 
@@ -205,13 +213,13 @@ async function cleanOldCacheFiles() {
 }
 
 function showWelcomeNotice(plugin: Plugin) {
-  const code = '1.7.6'
+  const code = '1.8.0-beta.3'
   if (settings.welcomeMessage !== code) {
     const welcome = new DocumentFragment()
     welcome.createSpan({}, span => {
-      span.innerHTML = `<strong>Omnisearch has been updated</strong>
-New beta feature: PDF search 🔎📄
-<small>Toggle "<i>BETA - Index PDFs</i>" in Omnisearch settings page.</small>`
+      span.innerHTML = `<strong>Omnisearch BETA has been updated</strong>
+You can now enable "Images Indexing" to use Optical Character Recognition on your scanned documents
+🔎🖼`
     })
     new Notice(welcome, 30000)
   }

src/settings.ts

@@ -25,6 +25,8 @@ export interface OmnisearchSettings extends WeightingSettings {
   indexedFileTypes: string[]
   /** Enable PDF indexing */
   PDFIndexing: boolean
+  /** Enable PDF indexing */
+  imagesIndexing: boolean
   /** Display Omnisearch popup notices over Obsidian */
   showIndexingNotices: boolean
   /** Activate the small 🔍 button on Obsidian's ribbon */
@@ -147,7 +149,7 @@ export class SettingsTab extends PluginSettingTab {
     indexPDFsDesc.createSpan({}, span => {
       span.innerHTML = `Omnisearch will include PDFs in search results.
         <ul>
-          <li>⚠️ Depending on their size, PDFs can take anywhere from a few seconds to 2 minutes to be processed.</li>
+          <li>⚠️ Each PDF can take anywhere from a few seconds to 2 minutes to be processed.</li>
           <li>⚠️ Texts extracted from PDFs may contain errors such as missing spaces, or spaces in the middle of words.</li>
           <li>⚠️ Some PDFs can't be processed correctly and will return an empty text.</li>
           <li>This feature is currently a work-in-progress, please report issues that you might experience.</li>
@@ -164,6 +166,26 @@ export class SettingsTab extends PluginSettingTab {
         })
       )
 
+    // PDF Indexing
+    const indexImagesDesc = new DocumentFragment()
+    indexImagesDesc.createSpan({}, span => {
+      span.innerHTML = `Omnisearch will use <a href="https://en.wikipedia.org/wiki/Tesseract_(software)">Tesseract</a> to index images from their text.
+        <ul>
+          <li>Only English is supported at the moment.</li>
+          <li>Not all images can be correctly read by the OCR, this feature works best with scanned documents.</li>
+        </ul>      
+        <strong style="color: var(--text-accent)">Needs a restart to fully take effect.</strong>`
+    })
+    new Setting(containerEl)
+      .setName('BETA - Images Indexing')
+      .setDesc(indexImagesDesc)
+      .addToggle(toggle =>
+        toggle.setValue(settings.imagesIndexing).onChange(async v => {
+          settings.imagesIndexing = v
+          await saveSettings(this.plugin)
+        })
+      )
+
     // #endregion Behavior
 
     // #region User Interface
@@ -315,6 +337,7 @@ export const DEFAULT_SETTINGS: OmnisearchSettings = {
   ignoreDiacritics: true,
   indexedFileTypes: [] as string[],
   PDFIndexing: false,
+  imagesIndexing: false,
 
   showIndexingNotices: false,
   showShortName: false,

src/tools/utils.ts

@@ -184,7 +184,9 @@ export function getCtrlKeyLabel(): 'ctrl' | '⌘' {
 
 export function isFileIndexable(path: string): boolean {
   return (
-    (settings.PDFIndexing && path.endsWith('.pdf')) || isFilePlaintext(path)
+    (settings.PDFIndexing && path.endsWith('.pdf')) ||
+    isFilePlaintext(path) ||
+    (settings.imagesIndexing && isFileImage(path))
   )
 }