GitHub - blopker/codebook: Spell Checker for Code

13 min read Original article ↗

Logo

CODEBOOK

Usage

Example

No configuration needed. Codebook will automatically detect the language you are editing and mark issues for you. Codebook will try to only mark issues for words that you create, where they are initially defined.

Please give a ⭐ if you find Codebook useful!

Supported platforms: Prebuilt archives are published for macOS (x86_64, aarch64), Linux (x86_64, aarch64), and Windows (x86_64, arm64).

Integrations

Zed

Codebook is the most popular spell checker for Zed! To install, go to the Extension tab in Zed and look for "Codebook". Done!

Note: The version that Zed displays in the extension menus is for the Zed Extension, and not the LSP version (this repo). The extension will automatically update the LSP. If that updater is broken for some reason, try uninstalling the extension and reinstalling.

If quickfix code actions are not showing up for specific languages, ensure your settings.json file includes the special "...", or "codebook", value in any language_servers values defined:

"languages": {
  "Python": {
    "language_servers": ["pyright", "ruff", "..."],
    // OR
    "language_servers": ["pyright", "ruff", "codebook"],
    "format_on_save": "on"
  }
},

Helix

Codebook can also be enabled for the Helix editor by adding the LSP to the language.toml configuration file.

First, install Codebook, and ensure that codebook-lsp is installed into your $PATH.

Then, add into the Helix config file:

[language-server.codebook]
command = "codebook-lsp"
args = ["serve"]

# Example use in markdown:
[[language]]
name = "markdown"
language-servers = ["codebook"]

This can be verified with:

Suggestions will appear in files opened, and space-mode a key binding can be used to accept suggestions.

Neovim

nvim-lspconfig includes a configuration for Codebook.

First, install Codebook, and ensure that codebook-lsp is installed into your $PATH.

Install nvim-lspconfig if you have not already. Then, add the following to your Neovim configuration:

vim.lsp.enable('codebook')

VS Code (Unreleased)

A VS Code extension lives in editors/vscode. The extension manages the codebook-lsp binary for you, starts it with the right flags, and exposes a few configuration toggles (codebook.binaryPath, codebook.enablePrerelease, and codebook.logLevel).

To try it locally:

cd editors/vscode
bun install       # or npm install
bun run build
bun run package   # or npm run package
code --install-extension codebook-vscode-*.vsix

Once the extension is installed it will activate automatically for every supported language.

Note: This extension is a work in progress and is not in the Marketplace yet. If you try it out, we'd love feedback!

Other Editors

Any editor that implements the Language Server Protocol should be compatible with Codebook. To get started, follow the installation instructions, then consult your editor's documentation to learn how to configure and enable a new language server. For your reference, the following command starts the server such that it listens on STDIN and emits on STDOUT:

About

Codebook is a spell checker for code. It binds together the venerable Tree Sitter and the fast spell checker Spellbook. Included is a Language Server for use in (theoretically) any editor. Everything is done in Rust to keep response times snappy and memory usage low.

However, if you are looking for a traditional spell checker for prose, Codebook may not be what you are looking for. For example, capitalization issues are handled loosely and grammar checking is out of scope.

To see the motivations behind Codebook, read this blog post.

Status

Codebook is in active development. As better dictionaries are added, words that were previously marked as misspelled (or spelled correctly), might change over time to improve correctness.

Supported Programming Languages

Language Status
C ✅
C++ ⚠️
CSS ⚠️
Elixir ⚠️
Go ✅
HTML ⚠️
Haskell ⚠️
Java ✅
JavaScript ✅
LaTeX ⚠️
Lua ✅
Markdown ✅
PHP ⚠️
Plain Text ✅
Python ✅
Ruby ✅
Rust ✅
TOML ✅
TypeScript ✅
Typst ⚠️
Zig ✅
C# ⚠️

✅ = Good to go. ⚠️ = Supported, but needs more testing. Help us improve! ❌ = Work has started, but there are issues.

If Codebook is not marking issues you think it should, please file a GitHub issue!

Installation

If you are a Zed user, you may skip this step and consult the Zed section of this document. Otherwise, you will need to install the codebook-lsp binary and make it available on your $PATH. You have a number of options to do this.

Manual

  1. Download the latest release for your architecture from the releases page.
    • Prebuilt archives are published for macOS (x86_64, aarch64), Linux (x86_64, aarch64), and Windows (x86_64, arm64).
    • Windows artifacts are provided as .zip files; macOS and Linux artifacts are .tar.gz.
  2. Extract the binary from the archive, and move it somewhere on your system $PATH.
  • ~/.local/bin/codebook-lsp
  • /usr/bin/codebook-lsp
  • Etc...

Eget Installation

You can install the latest release using eget:

Arch Linux

You can install the Codebook LSP using pacman:

Cargo Install

You can also install the Codebook LSP using Cargo:

cargo install codebook-lsp

To install directly from the GitHub repository:

cargo install --git https://github.com/blopker/codebook codebook-lsp

From Source

You may also build codebook from source by cloning the repository and running make build.

Configuration

Codebook supports both global and project-specific configuration. Configuration files use the TOML format, with project settings overriding global ones.

Global Configuration

The global configuration applies to all projects by default. Location depends on your operating system:

  • Linux/macOS: $XDG_CONFIG_HOME/codebook/codebook.toml or ~/.config/codebook/codebook.toml
  • Windows: %APPDATA%\codebook\codebook.toml or %APPDATA%\Roaming\codebook\codebook.toml

You can override this location if you sync your config elsewhere by providing initializationOptions.globalConfigPath from your LSP client. When no override is provided, the OS-specific default above is used.

Project Configuration

Project-specific configuration is loaded from either codebook.toml or .codebook.toml in the project root. Codebook searches for this file starting from the current directory and moving up to parent directories.

Note: Codebook picks which config to use on startup. If a config file is manually created or renamed (like switching between codebook.toml and .codebook.toml), restart your editor (or the LSP server) for the new file to be recognized.

Configuration Options

# List of dictionaries to use for spell checking
# Default: ["en_us"]
# Available dictionaries:
#  - English: "en_us", "en_gb"
#  - German: "de", "de_at", "de_ch"
#  - Dutch: "nl_nl"
#  - Spanish: "es"
#  - French: "fr"
#  - Italian: "it"
#  - Portuguese (Brazil): "pt_br"
#  - Russian: "ru"
#  - Swedish: "sv"
#  - Danish: "da"
dictionaries = ["en_us", "en_gb"]

# Custom allowlist of words to ignore (case-insensitive)
# Codebook will add words here when you select "Add to dictionary"
# Default: []
words = ["codebook", "rustc"]

# Words that should always be flagged as incorrect
# Default: []
flag_words = ["todo", "fixme"]

# List of glob patterns for paths to ignore when spell checking
# Default: []
ignore_paths = ["target/**/*", "**/*.json", ".git/**/*"]

# List of regex patterns to ignore when spell checking
# Patterns are matched against each line of text, not individual words
# Useful for domain-specific strings or patterns
# Note: Backslashes must be escaped in TOML (use \\ instead of \)
# Default: []
ignore_patterns = [
    "\\b[ATCG]+\\b",             # DNA sequences
    "\\d{3}-\\d{2}-\\d{4}",      # Social Security Number format
    "^[A-Z]{2,}$",               # All caps words like "HTML", "CSS"
    "https?://[^\\s]+"           # URLs
]

# Minimum word length to check (words shorter than this are ignored)
# Default: 3
# Set to 0 to check all words including single letters
# Set to 2 to check words with 2 or more characters
min_word_length = 3

# Whether to use global configuration (project config only)
# Set to false to completely ignore global settings
# Default: true
use_global = true

Configuration Precedence

  1. Project configuration overrides global configuration
  2. If use_global = false in project config, global settings are ignored entirely
  3. If no project config exists, global config is used
  4. If neither exists, default settings are used

Working with Configurations

  • Words added with "Add to dictionary" are stored in the project configuration
  • Words added with "Add to global dictionary" are stored in the global configuration file
  • Project settings are saved automatically when words are added
  • Configuration files are automatically reloaded when they change

User-Defined Regex Patterns

The ignore_patterns configuration allows you to define custom regex patterns to skip during spell checking. Here are important details about how they work:

Default Patterns: Codebook already includes built-in regex patterns for common technical strings, so you don't need to define these yourself:

  • URLs: https?://[^\\s]+
  • Hex colors: #[0-9a-fA-F]{3,8} (like #deadbeef, #fff)
  • Email addresses: [a-zA-Z0-9._%+-]+@[a-zA-Z0-9.-]+\\.[a-zA-Z]{2,}
  • File paths: /[^\\s]* (Unix) and [A-Za-z]:\\\\[^\\s]* (Windows)
  • UUIDs: [0-9a-fA-F]{8}-[0-9a-fA-F]{4}-[0-9a-fA-F]{4}-[0-9a-fA-F]{4}-[0-9a-fA-F]{12}
  • Base64 strings: [A-Za-z0-9+/]{20,}={0,2} (20+ characters)
  • Git commit hashes: \\b[0-9a-fA-F]{7,40}\\b
  • Markdown links: \\[([^\\]]+)\\]\\(([^)]+)\\)

Line-by-Line Matching: Regex patterns are applied to each line of text, not individual words. This means your patterns should account for the line context.

TOML Escaping: Since configuration files use TOML format, backslashes in regex patterns must be escaped by doubling them:

  • Use \\b for word boundaries (not \b)
  • Use \\d for digits (not \d)
  • Use \\\\ for literal backslashes (not \\)

Examples:

ignore_patterns = [
    "\\b[ATCG]+\\b",           # DNA sequences with word boundaries
    "^\\s*//.*$",              # Comment lines starting with //
    "https?://[^\\s]+",        # URLs (note the escaped \s)
    "\\$[a-zA-Z_][a-zA-Z0-9_]*", # Variables starting with $
]

Migration Note: If you're upgrading from an older version, patterns that used ^ and $ anchors may need adjustment since matching now occurs line-by-line rather than word-by-word.

LSP Initialization Options

Editors can pass initializationOptions when starting the Codebook LSP for LSP-specific options. Refer to your editor's documentation for how to apply these options. All values are optional, omit them for the default behavior:

  • logLevel ("trace" | "debug" | "info" | "warn" | "error", default "info"): sets the verbosity of logs.
  • globalConfigPath (string): overrides the auto-detected global codebook.toml path, useful if you sync configs from another location. On macOS and Linux, the ~/ prefix for the current user's home directory is supported.
  • checkWhileTyping (bool, default true): when false, spelling diagnostics are only published on save instead of each keystroke. This is useful for example if performance is a problem, or the real-time diagnostics are annoying (sorry!).
  • diagnosticSeverity ("error" | "warning" | "information" | "hint", default "information"): sets the severity of spell check diagnostics.

Example payload:

{
  "logLevel": "debug",
  "globalConfigPath": "~/dotfiles/codebook.toml",
  "checkWhileTyping": false,
  "diagnosticSeverity": "information"
}

Goals

Spell checking is complicated and opinions about how it should be done, especially with code, differs. This section is about the trade offs that steer decisions.

Privacy

No remote calls for spell checking or analytics. Once dictionaries are cached, Codebook needs to be usable offline. Codebook will never send the contents of files to a remote server.

Don't be annoying

Codebook should have high signal and low noise. It should only highlight words that users have control over. For example, a misspelled word in an imported function should not be highlighted as the user can't do anything about it.

As a consequence, Codebook only marks issues for terms where they are defined, and not where they are used. Correcting both the definition (flagged by Codebook) and all subsequent uses should be done by other language specific tooling - usually available as a LSP for that language.

Efficient

All features will be weighed against their impact on CPU and memory. Codebook should be fast enough to spell check on every keystroke on even low-end hardware.

Features

Code-aware spell checking

Codebook will only check the parts of your code where a normal linter wouldn't. Comments, string literals and variable definitions for example. Codebook knows how to split camel case and snake case variables, and makes suggestions in the original case.

Language Server

Codebook comes with a language server. Originally developed for the Zed editor, this language server can be integrated into any editor that supports the language server protocol.

Dictionary Management

Codebook comes with a dictionary manager, which will automatically download and cache dictionaries.

Hierarchical Configuration

Codebook uses a hierarchical configuration system with global (user-level) and project-specific settings, giving you flexibility to set defaults and override them as needed per project.

Adding a New Dictionary

Dictionaries in Codebook are currently hardcoded in the dictionary repository file at crates/codebook/src/dictionaries/repo.rs.

To add a new Hunspell-compatible dictionary:

  1. Open crates/codebook/src/dictionaries/repo.rs

  2. Locate the HUNSPELL_DICTIONARIES static vector (for Hunspell dictionaries) or TEXT_DICTIONARIES (for plain text word lists)

  3. Add a new entry using the appropriate constructor. For Hunspell dictionaries:

    HunspellRepo::new(
    "nl_nl",  // Dictionary name in snake_case
    "https://raw.githubusercontent.com/wooorm/dictionaries/refs/heads/main/dictionaries/nl/index.aff",
    "https://raw.githubusercontent.com/wooorm/dictionaries/refs/heads/main/dictionaries/nl/index.dic",
),

  4. Follow the naming convention:

    • Use snake_case format (e.g., en_us, nl_nl, pt_br)
    • Language codes should be lowercase
    • Names must be unique across all dictionaries

  5. Find dictionary sources:

  6. (Optional) Run the tests to verify your addition:

    cargo test test_dictionary_names_unique_and_snake_case

    This test ensures dictionary names are unique and follow the snake_case convention.

For plain text dictionaries, use TextRepo::new() instead and add to TEXT_DICTIONARIES.

Adding New Programming Language Support

Codebook uses Tree-sitter support additional programming languages. Here's how to add support for a new language:

1. Create a Tree-sitter Query

Each language needs a Tree-sitter query file that defines which parts of the code should be checked for spelling issues. The query needs to capture:

  • Identifiers (variable names, function names, class names, etc.)
  • String literals
  • Comments

Create a new .scm file in codebook/crates/codebook/src/queries/ named after your language (e.g., java.scm).

2. Understand the Language's AST

To write an effective query, you need to understand the Abstract Syntax Tree (AST) structure of your language. Use these tools:

A good approach is to:

  1. Write sample code with identifiers, strings, and comments
  2. Paste it into the playground/visualizer
  3. Observe the node types used for each element
  4. Create capture patterns that target only definition nodes, not usages

3. Update the Language Settings

Add your language to codebook/crates/codebook/src/queries.rs:

  1. Add a new variant to the LanguageType enum
  2. Add a new entry to the LANGUAGE_SETTINGS array with:
    • The language type
    • File extensions for your language
    • Language identifiers
    • Path to your query file

4. Add the Tree-sitter Grammar

Make sure the appropriate Tree-sitter grammar is added as a dependency in Cargo.toml and update the language() function in queries.rs to return the correct language parser.

5. Test Your Implementation

Run the tests to ensure your query is valid:

cargo test -p codebook queries::tests::test_all_queries_are_valid

Additional language tests should go in codebook/tests. There are many example tests to copy.

You can also test with real code files to verify that Codebook correctly identifies spelling issues in your language. Example files should go in examples/ and contain at least one spelling error to pass integration tests.

Tips for Writing Effective Queries

  • Focus on capturing definitions, not usages
  • Include only nodes that contain user-defined text (not keywords)
  • Test with representative code samples
  • Start simple and add complexity as needed
  • Look at existing language queries for patterns

If you've successfully added support for a new language, please consider contributing it back to Codebook with a pull request!

Running Tests

Run test with make test after cloning. Integration tests are also available with make integration_test, but requires BunJS to run.

Acknowledgments

Release

To publish a new version:

  1. Update and commit changelog with new version number
  2. Run make release-lsp
  3. Follow instructions
  4. Wait for Actions to finish
  5. Go to GitHub Releases
  6. Un-mark "prerelease" and publish
  7. Run make publish_crates to upload to crates.io