---
title: "Book cipher"
description: "Each word or letter as a place in a text both sides own. Count whole words or single letters. By number or by page and line."
canonical_url: "https://ciphers.agntn.dev/ciphers/book"
---
# Book cipher

> Each word or letter as a place in a text both sides own. Count whole words or single letters. By number or by page and line.

<cipher-facts name="book">



</cipher-facts>

The key is a whole text. A novel, a song, a founding document. Both sides own the same copy, and the ciphertext is just numbers pointing into it. Beale hid a treasure description this way, or so the pamphlet says.

```ts
const book = create("book");
const verse =
  "Ask, and it shall be given you; seek, and ye shall find; knock, and it shall be opened unto you.";
book.encode("seek and ye shall find", { book: verse }).text; // "8 2 10 4 12"
book.decode("13 3 4 20", { book: verse, pick: "letter" }).text; // "KISY"
```

That verse is Matthew 7:7, the key a spy uses in The Imitation Game. Good taste in keys.

## What's a word?

Letters and digits in a row, in any script, with the marks that sit on them. An apostrophe inside keeps it whole, so `Nature's` is one word. Anything else splits, so `self-evident` is two. Beale counted `self-evident` as two, and the dCode examples only add up with `Nature's` as one. So that's the rule.

## Words or letters

`pick: "word"` hands back the word itself, spelled as the book spells it. `pick: "letter"` takes its first letter. That's the Beale trick: word 115 is `instituted`, so it means I. Words come back with spaces between them. Letters come back uppercase and glued together. `ß` stays `ß`, since its capital is two letters.

Encoding looks up each word, or each letter with `pick: "letter"`, and ignores case. `STRAẞE` finds `straße`. Everything that isn't a letter or a digit is dropped.

Why not always the first match? Five Es with the same number is the first thing anyone attacking it looks for. So a repeat takes the next word that fits, and after the last one it starts over.

## Lines and pages

`address` sets how many numbers one place takes:

- `word` is the place in the whole book, like `305`.
- `line-word` is the line, then the word in it, like `12-3`.
- `page-line-word` is the page, the line on it and the word, like `4-12-3`.

A line without a word doesn't count. A page ends at a form feed, which is exactly what `pdftotext` puts between PDF pages. Got a PDF of the right edition? Run it through `pdftotext` and hand over the file.

```bash
ciphers decode book "4-12-3 4-12-7" --book - --address page-line-word < edition.txt
```

`start: 0` counts every word, line and page from zero. For puzzles written by programmers.

Decoding doesn't care how the numbers are dressed. `12-3-5`, `12.3.5` and `(12, 3, 5)` read the same. Every number counts, in order, and `page-line-word` takes them three at a time.

## Letters inside the book

An address can point to a letter inside the book. No need to split the book into single letter words. That's a fragile way to read a perfectly good book.

Read each address name from left to right:

- `letter` counts letters in the whole book, like `15`.
- `line-letter` counts the line, then its letter, like `2-3`.
- `page-line-letter` adds the page first, like `4-2-3`.
- `line-word-letter` counts the line, word and letter, like `2-3-1`.
- `page-line-word-letter` adds the page first, like `4-2-3-1`.

Want an Ottendorf cipher? `line-word-letter` and `page-line-letter` give you the classic shapes.

```ts
const tyger = `Tyger Tyger, burning bright,
In the forests of the night;
What immortal hand or eye,
Could frame thy fearful symmetry?`;
book.decode("1-4-1 2-2-3 3-1-3 3-2-6", { book: tyger, address: "line-word-letter" }).text; // "BEAT"
book.encode("tiger", { book: tyger, address: "line-word-letter" }).text; // "1-1-1 1-3-5 1-1-3 1-1-4 1-1-5"
book.encode("tiger", { book: tyger, address: "letter" }).text; // "1 15 3 4 5"
```

Blake spelled it `Tyger`, so the I comes from `burning`. Encoding ignores case and takes the next matching letter. After the last match, it starts over. Decoding gives uppercase letters glued together.

Letters and digits count, in any script. Spaces, punctuation and apostrophes don't. In `Don't panic`, letter 4 is T and letter 5 is P.

Marks ride along with their letters. The book is read in NFC first, so `e` plus a combining accent counts as one `é`. A mark with nothing to merge into stays behind, and you get the bare letter.

For `line-letter` and `page-line-letter`, letters start over on each line. `start: 0` counts letters from zero too. The result options include `letters`, the count for the book.

A letter address always picks a letter. Leave `pick` out or set it to `"letter"`. `pick: "word"` is an `InvalidOptionError`. A letter past the end gives a `CipherError` with its range.

## Same book, same split?

Not always. One copy has a preface, the other doesn't. One keeps the title, one starts at the first sentence. So every result says how the book split:

```ts
book.decode("8", { book: verse }).options;
// { address: "word", pick: "word", start: 1, words: 20, lines: 1, pages: 1 }
```

Compare those counts before you blame the cipher. The book itself stays out of the result. It can be a whole novel, and nobody needs it echoed back.

## Checked against

The dCode book cipher page, over the Declaration of Independence as the National Archives transcribes it, counted from "When". `15 4 12 52 7` with `pick: "letter"` is `DCODE`. `221 132 136 305` is `by of of King`, or `BOOK` in letters. And the Wikipedia page on the Beale ciphers: word 115 is `instituted`.

## Errors

A number past the end is a `CipherError` that says what it looked for and how many there are: `Word 40 is out of range: the book has 20 words, numbered from 1`. Numbers that don't split into whole addresses are a `CipherError` too. So is a word the book doesn't have. No `book` is a `MissingOptionError`. A book without a single word is an `InvalidOptionError`, and so is an `address`, `pick` or `start` it doesn't know.


## Sitemap

See the full [sitemap](https://ciphers.agntn.dev/sitemap.md) for all pages.
