---
title: "Reading text"
description: "Get the text of a page, part of a page, or a whole document."
source: "https://www.embedpdf.com/docs/engine/text/extraction"
---

# Reading text

Get a page's text with `page.text.get()`:

```ts
const page = doc.page(pageRef);
const { text } = await page.text.get();
```

`text` is the page's text in reading order, as a normal JavaScript string. Lines are separated
by `\r\n`.

## Part of a page

Characters on a page are numbered from 0. To read just some of them (the text a user selected, or
a search match), pass a range:

```ts
const selected = await page.text.slice({ start: 120, count: 42 });
```

A range is always `{ start, count }`: `start` is the first character, `count` how many. Search
matches and selections are ranges too, so you can pass them straight in:

```ts
const { matches } = await doc.search.query({ text: 'total' });
const found = await doc.page(matches[0].page).text.slice(matches[0]);
```

## A whole document

There's no single call for the whole document; read the pages and join them:

```ts
const { pages } = await doc.pages.list();

const texts = await Promise.all(pages.map((page) => doc.page(page.ref).text.get()));
const fullText = texts.map(({ text }) => text).join('\n\n');
```

The engine reads pages one at a time in the background, so this is fine for long documents too.

## Permissions

Reading text needs the `doc.text.copy` permission, because text you read can leave the document.
Without it, both calls fail with `Forbidden`.

A user without `doc.text.copy` can still [search](https://www.embedpdf.com/docs/engine/text/search) (without snippets)
and [select](https://www.embedpdf.com/docs/engine/text/selection-and-geometry) text: both need only where the text is,
not what it says. That lets you allow highlighting a PDF while stopping its text from being
copied.

> **Stop text from being copied, for real**
>
> The local engine runs in the user's browser, which already holds the whole file, so a permission
> there keeps your interface honest but can't keep the text from a user who opens the file another
> way. With CloudPDF the file stays on your server: a user without `doc.text.copy` receives page
> images and highlight positions, never the text as characters they can copy.
>
> [How CloudPDF enforces permissions](https://www.cloudpdf.com/docs/engine/concepts/permissions)

## Good to know

### Characters and string positions

The engine numbers characters the way the PDF stores them. Most of the time that matches the
string: character 5 is `text[5]`. Two things make them differ:

- **Characters outside the Basic Multilingual Plane** (most emoji, some CJK characters) are one
  character in the PDF but two positions in a JavaScript string.
- **Characters with no text of their own** (a glyph the PDF maps to no Unicode character) have a
  number but add nothing to the string.

That's why you pass ranges to `slice()` instead of cutting `text` yourself: `slice()` converts for
you. `get()` also returns `charCount` (how many characters the page has) and, when the two
numberings differ, a `charMap` that records where. You only need `charMap` if you convert
positions yourself, for example to run your own search over `text`:

```ts
import { charRangeForTextOffsets } from '@embedpdf/engine';

const snapshot = await page.text.get();
const at = snapshot.text.indexOf('Total');
const range = charRangeForTextOffsets(snapshot, { start: at, count: 'Total'.length });
// range is { start, count } in characters: ready for slice(), layout.segments() or a highlight.
```