Embedding PDF.js in Omnisearch to avoid crashes

package.json

@@ -37,7 +37,8 @@
   },
   "dependencies": {
     "@vanakat/plugin-api": "^0.1.0",
-    "minisearch": "^5.0.0"
+    "minisearch": "^5.0.0",
+    "pdfjs-dist": "^2.16.105"
   },
   "pnpm": {
     "overrides": {

pnpm-lock.yaml

@@ -19,6 +19,7 @@ specifiers:
   jest: ^27.5.1
   minisearch: ^5.0.0
   obsidian: latest
+  pdfjs-dist: ^2.16.105
   prettier: ^2.7.1
   prettier-plugin-svelte: ^2.7.0
   svelte: ^3.50.1
@@ -30,6 +31,7 @@ specifiers:
 dependencies:
   '@vanakat/plugin-api': 0.1.0
   minisearch: 5.0.0
+  pdfjs-dist: 2.16.105
 
 devDependencies:
   '@babel/preset-env': 7.19.0
@@ -2436,6 +2438,10 @@ packages:
       webidl-conversions: 5.0.0
     dev: true
 
+  /dommatrix/1.0.3:
+    resolution: {integrity: sha512-l32Xp/TLgWb8ReqbVJAFIvXmY7go4nTxxlWiAFyhoQw9RKEOHBZNnyGvJWqDVSPmq3Y9HlM4npqF/T6VMOXhww==}
+    dev: false
+
   /electron-to-chromium/1.4.247:
     resolution: {integrity: sha512-FLs6R4FQE+1JHM0hh3sfdxnYjKvJpHZyhQDjc2qFq/xFvmmRt/TATNToZhrcGUFzpF2XjeiuozrA8lI0PZmYYw==}
     dev: true
@@ -3920,6 +3926,18 @@ packages:
     resolution: {integrity: sha512-gDKb8aZMDeD/tZWs9P6+q0J9Mwkdl6xMV8TjnGP3qJVJ06bdMgkbBlLU8IdfOsIsFz2BW1rNVT3XuNEl8zPAvw==}
     engines: {node: '>=8'}
 
+  /pdfjs-dist/2.16.105:
+    resolution: {integrity: sha512-J4dn41spsAwUxCpEoVf6GVoz908IAA3mYiLmNxg8J9kfRXc2jxpbUepcP0ocp0alVNLFthTAM8DZ1RaHh8sU0A==}
+    peerDependencies:
+      worker-loader: ^3.0.8
+    peerDependenciesMeta:
+      worker-loader:
+        optional: true
+    dependencies:
+      dommatrix: 1.0.3
+      web-streams-polyfill: 3.2.1
+    dev: false
+
   /picocolors/1.0.0:
     resolution: {integrity: sha512-1fygroTLlHu66zi26VoTDv8yRgm0Fccecssto+MhsZ0D/DGW2sm8E8AjW7NU5VVTRt5GxbeZ5qBuJr+HyLYkjQ==}
     dev: true
@@ -4559,6 +4577,11 @@ packages:
       makeerror: 1.0.12
     dev: true
 
+  /web-streams-polyfill/3.2.1:
+    resolution: {integrity: sha512-e0MO3wdXWKrLbL0DgGnUV7WHVuw9OUvL4hjgnPkIeEvESk74gAITi5G606JtZPp39cd8HA9VQzCIvA49LpPN5Q==}
+    engines: {node: '>= 8'}
+    dev: false
+
   /webidl-conversions/5.0.0:
     resolution: {integrity: sha512-VlZwKPCkYKxQgeSbH5EyngOmRp7Ww7I9rQLERETtf5ofd9pGeswWiOtogpEO850jziPRarreGxn5QIiTqpb2wA==}
     engines: {node: '>=8'}

src/notes-index.ts

@@ -1,8 +1,10 @@
 import {Notice, TAbstractFile, TFile} from 'obsidian'
 import {
+  canIndexPDFs,
   extractHeadingsFromCache,
   getAliasesFromMetadata,
-  getTagsFromMetadata, isFileIndexable,
+  getTagsFromMetadata,
+  isFileIndexable,
   isFilePlaintext,
   removeDiacritics,
   wait,
@@ -178,3 +180,31 @@ export async function saveIndexToFile(): Promise<void> {
     isIndexChanged = false
   }
 }
+
+export async function indexPDFs() {
+  if (canIndexPDFs()) {
+    const start = new Date().getTime()
+    const files = app.vault.getFiles().filter(f => f.path.endsWith('.pdf'))
+    if (files.length > 50) {
+      new Notice(`⚠️ Omnisearch is indexing ${files.length} PDFs. You can experience slowdowns while this work is in progress.`)
+    }
+
+    const promises: Promise<void>[] = []
+    for (const file of files) {
+      if (getNoteFromCache(file.path)) {
+        removeFromIndex(file.path)
+      }
+      promises.push(addToIndex(file))
+    }
+    await Promise.all(promises)
+
+    // Notice & log
+    const message = `Omnisearch - Indexed ${files.length} PDFs in ${
+      new Date().getTime() - start
+    }ms`
+    if (settings.showIndexingNotices) {
+      new Notice(message)
+    }
+    console.log(message)
+  }
+}

src/pdf-parser.ts

@@ -1,11 +1,11 @@
 import type { TFile } from 'obsidian'
-import { loadPdfJs } from 'obsidian'
+import PDFJs from 'pdfjs-dist'
+import pdfjsWorker from 'pdfjs-dist/build/pdf.worker.entry'
 
-let PDFJs: any = null
+PDFJs.GlobalWorkerOptions.workerSrc = pdfjsWorker
 
 // https://stackoverflow.com/a/59929946
 export async function getPdfText(file: TFile): Promise<string> {
-  PDFJs = PDFJs ?? (await loadPdfJs())
   const data = await app.vault.readBinary(file)
   const doc = await PDFJs.getDocument(data).promise
   const pageTexts = Array.from({ length: doc.numPages }, async (v, i) => {

src/search.ts

@@ -24,7 +24,7 @@ import {
   loadNotesCache,
   resetNotesCache,
 } from './notes'
-import { addToIndex, removeFromIndex, saveIndexToFile } from './notes-index'
+import {addToIndex, indexPDFs, removeFromIndex, saveIndexToFile} from './notes-index'
 
 export let minisearchInstance: MiniSearch<IndexedNote>
 
@@ -134,38 +134,6 @@ export async function initGlobalSearchIndex(): Promise<void> {
   }
 }
 
-async function indexPDFs() {
-  if (canIndexPDFs()) {
-    const start = new Date().getTime()
-    console.warn(
-      "Omnisearch - Warnings on 'pdf.worker.min' are due to some issues while reading PDFs file and can usually be ignored."
-    )
-    const files = app.vault.getFiles().filter(f => f.path.endsWith('.pdf'))
-    let promises: Promise<void>[] = []
-    for (const [i, file] of files.entries()) {
-      if (getNoteFromCache(file.path)) {
-        removeFromIndex(file.path)
-      }
-      promises.push(addToIndex(file))
-      if (i % 10 === 0) {
-        await wait(1)
-        await Promise.all(promises)
-        promises = []
-      }
-    }
-    await Promise.all(promises)
-
-    // Notice & log
-    const message = `Omnisearch - Indexed ${files.length} PDFs in ${
-      new Date().getTime() - start
-    }ms`
-    if (settings.showIndexingNotices) {
-      new Notice(message)
-    }
-    console.log(message)
-  }
-}
-
 /**
  * Searches the index for the given query,
  * and returns an array of raw results

src/settings.ts

@@ -106,23 +106,23 @@ export class SettingsTab extends PluginSettingTab {
           })
       })
 
-    // // Index PDFs
-    // const indexPDFsDesc = new DocumentFragment()
-    // indexPDFsDesc.createSpan({}, span => {
-    //   span.innerHTML = `Omnisearch will index your PDFs, and return them in search results.
-    //     This feature is currently a work-in-progress, please report slowdowns or issues that you might experience.<br>
-    //     PDFs being quite slow to index, <strong style="color: var(--text-accent)">it is strongly recommended to also enable "Store index in file"</strong>.<br>
-    //     <strong>Needs a restart to fully take effect.</strong>`
-    // })
-    // new Setting(containerEl)
-    //   .setName('BETA - Index PDFs')
-    //   .setDesc(indexPDFsDesc)
-    //   .addToggle(toggle =>
-    //     toggle.setValue(settings.indexPDFs).onChange(async v => {
-    //       settings.indexPDFs = v
-    //       await saveSettings(this.plugin)
-    //     })
-    //   )
+    // Index PDFs
+    const indexPDFsDesc = new DocumentFragment()
+    indexPDFsDesc.createSpan({}, span => {
+      span.innerHTML = `Omnisearch will index your PDFs, and return them in search results.
+        This feature is currently a work-in-progress, please report slowdowns or issues that you might experience.<br>
+        PDFs being quite slow to index, <strong style="color: var(--text-accent)">it is strongly recommended to also enable "Store index in file"</strong>.<br>
+        <strong>Needs a restart to fully take effect.</strong>`
+    })
+    new Setting(containerEl)
+      .setName('BETA - Index PDFs')
+      .setDesc(indexPDFsDesc)
+      .addToggle(toggle =>
+        toggle.setValue(settings.indexPDFs).onChange(async v => {
+          settings.indexPDFs = v
+          await saveSettings(this.plugin)
+        })
+      )
 
     // Store index
     const serializedIndexDesc = new DocumentFragment()

src/utils.ts

@@ -173,7 +173,7 @@ export function getCtrlKeyLabel(): 'ctrl' | '⌘' {
 }
 
 export function canIndexPDFs(): boolean {
-  return false
+  return settings.indexPDFs
 }
 
 export function isFileIndexable(path: string): boolean {