react-native-pdf-text-extractor

Fully offline, on-device PDF text extraction for React Native. No network calls, no cloud OCR — text is extracted locally using each platform's native PDF stack:

• iOS — PDFKit
• Android — PdfBox-Android (com.tom-roush:pdfbox-android)

This library extracts text embedded in a PDF (e.g. text layers, generated documents). It does not perform OCR — scanned/image-only pages will resolve to an empty string rather than throwing.

Installation

npm install react-native-pdf-text-extractor

iOS

cd ios && pod install

Requires iOS 15.1+.

Android

No extra setup — com.tom-roush:pdfbox-android resolves automatically from Maven Central. Minimum SDK 24.

Usage

import {
  getPageCount,
  extractText,
  extractAllText,
  extractPageText,
} from 'react-native-pdf-text-extractor';

const pageCount = await getPageCount(filePath);
// 12

const wholeDocument = await extractText(filePath);
// "Page 1 text\nPage 2 text\n..."

const perPage = await extractAllText(filePath);
// ["Page 1 text", "Page 2 text", ...]

const secondPage = await extractPageText(filePath, 1); // 0-indexed
// "Page 2 text"

API

getPageCount(filePath: string): Promise<number>

Resolves with the number of pages in the PDF.

extractText(filePath: string, options?: ExtractOptions): Promise<string>

Extracts the text of every page and concatenates it into a single string, separated by newlines.

Pass { normalize: true } to clean up common extraction artifacts (see Text normalization) before the promise resolves. Omitting options leaves the text exactly as extracted, unchanged from previous versions.

await extractText(filePath, { normalize: true });
// runs every normalization step with its default settings

await extractText(filePath, { normalize: { dehyphenate: false } });
// runs normalization, but leaves hyphenated line-wraps alone

extractAllText(filePath: string, options?: ExtractOptions): Promise<string[]>

Extracts the text of every page, returned as one string per page (index 0 is the first page). A page with no extractable text (e.g. a scanned image with no OCR layer) resolves to '' rather than throwing.

The normalize option (see Text normalization) applies to each page independently, not to the array as a whole:

await extractAllText(filePath, { normalize: true });

await extractAllText(filePath, {
  normalize: { stripInvisibleChars: true, unwrapLines: false },
});

extractPageText(filePath: string, pageIndex: number, options?: ExtractOptions): Promise<string>

Extracts the text of a single page. pageIndex is 0-indexed.

await extractPageText(filePath, 1, { normalize: true });

Text normalization

PDF text layers often carry artifacts that are harmless to read but noisy for downstream NLP — words split by a hyphen at a line-wrap, hard line breaks in the middle of a sentence, invisible control/format characters. The normalize option on extractText, extractAllText, and extractPageText (and the standalone normalizeExtractedText function) cleans these up. This is opt-in — omitting options (or passing normalize: false) leaves extracted text exactly as returned by the native extractor.

import { normalizeExtractedText } from 'react-native-pdf-text-extractor';

normalizeExtractedText('under-\nstand this'); // "understand this"

normalize accepts either true (run every step below with its default) or a partial NormalizeTextOptions object (unspecified steps still default to true):

| Option | Default | Effect | | --------------------- | ------- | --------------------------------------------------------------------------------------------------------- | | stripInvisibleChars | true | Removes invisible control/format characters (zero-width spaces, etc.), keeping \n and \t. | | dehyphenate | true | Rejoins a word split by a hyphen at a line-wrap, e.g. "under-\nstand" -> "understand". | | collapseBlankLines | true | Collapses runs of 3+ newlines down to a single blank line. | | unwrapLines | true | Replaces a single mid-paragraph line break with a space, while preserving real paragraph breaks (\n\n). |

Whitespace is always tidied (runs of spaces/tabs collapsed, trailing spaces before a newline stripped) and the result is always .trim()ed, regardless of which steps above are enabled.

filePath

All methods take a local file system path — either a bare absolute path (/data/.../document.pdf) or a file:// URL. Remote URLs and content:// URIs (as returned directly by some Android document pickers) are not supported; copy the file to a local path first (e.g. with react-native-document-picker's copyTo: 'cachesDirectory' option, or react-native-fs).

Errors

Rejections carry one of the following codes:

| Code | Meaning | | ------------------- | ----------------------------------------------------------------- | | E_FILE_NOT_FOUND | No file exists at the given path. | | E_UNSUPPORTED_URI | A content:// URI was passed instead of a local file path. | | E_INVALID_PAGE | pageIndex is out of range for the document. | | E_PDF_PROCESSING | The file could not be parsed as a PDF (corrupt, encrypted, etc.). |

filePath and pageIndex are also validated synchronously in JS before any native call is made — passing a non-string path or a non-integer/negative page index throws immediately rather than rejecting a promise.

Example app

See example/ for a runnable app that picks a PDF (or falls back to bundled sample PDFs) and displays extracted text.

Demo

🎥 Watch Demo

License

MIT