Hongdown

Hongdown is a Markdown formatter that enforces Hong Minhee's Markdown style conventions. The formatter is implemented in Rust using the Comrak library for parsing. It produces consistently formatted Markdown output following a distinctive style used across multiple projects including Fedify, LogTape, and Optique.

Installation

npm

npm install -g hongdown

mise

mise use -g github:dahlia/hongdown

Cargo

cargo install hongdown

Nix

nix run github:dahlia/hongdown

Pre-built binaries

Pre-built binaries for Linux, macOS, and Windows are available on the GitHub Releases page.

Usage

Basic usage

# Format a file and print to stdout
hongdown input.md

# Format a file in place
hongdown --write input.md
hongdown -w input.md

# Format multiple files
hongdown -w *.md

# Format all Markdown files in a directory (recursive)
hongdown -w .
hongdown -w docs/

# Check if files are formatted (exit 1 if not)
hongdown --check input.md
hongdown -c input.md

# Show diff of formatting changes
hongdown --diff input.md
hongdown -d input.md

# Read from stdin (use --stdin flag or - as filename)
echo "# Hello" | hongdown --stdin
echo "# Hello" | hongdown -
hongdown --stdin < input.md
hongdown - < input.md

# Custom line width
hongdown --line-width 100 input.md

HTML comment directives

Hongdown supports special HTML comment directives to control formatting behavior within documents.

Disable formatting

You can disable formatting for specific sections:

<!-- hongdown-disable-file -->
This entire file will not be formatted.

<!-- hongdown-disable-next-line -->
This   line   preserves   its   spacing.

The next line will be formatted normally.

<!-- hongdown-disable-next-section -->
Everything here is preserved as-is
until the next heading (h1 or h2).

Next heading
------------

This section will be formatted normally.

<!-- hongdown-disable -->
This section is not formatted.
<!-- hongdown-enable -->
This section is formatted again.

Sentence case customization

When sentence case is enabled, you can define document-specific proper nouns and common nouns:

<!-- hongdown-proper-nouns: Swift, Go -->
<!-- hongdown-common-nouns: Python -->

# Using Swift And Go For Python Development

This heading will be formatted as: "Using Swift and Go for python development"

• <!-- hongdown-proper-nouns: Word1, Word2 --> – Defines proper nouns to preserve (case-sensitive)
• <!-- hongdown-common-nouns: Word1, Word2 --> – Overrides built-in proper nouns by treating them as common nouns

These directives are merged with configuration file settings.

Configuration file

Hongdown supports cascading configuration files from multiple locations. Configuration files are loaded and merged in the following order (lowest to highest priority):

1. System-wide: /etc/hongdown/config.toml (Linux/Unix only)
2. User (legacy): ~/.hongdown.toml (all platforms)
3. User (platform-specific):
   • Linux: $XDG_CONFIG_HOME/hongdown/config.toml or ~/.config/hongdown/config.toml
   • macOS: ~/Library/Application Support/hongdown/config.toml
   • Windows: %APPDATA%\hongdown\config.toml
4. Project: .hongdown.toml in the current directory or any parent directory

Settings from higher-priority configurations override those from lower-priority ones. This allows you to set global defaults at the user or system level while overriding them for specific projects.

You can also specify a configuration file explicitly with the --config option, which bypasses the cascading system and uses only that file.

Disabling configuration inheritance

To ignore all system and user configurations and use only your project config:

no_inherit = true

# Your project-specific settings
line_width = 100

When no_inherit = true, only the project config and Hongdown's defaults are used. This is useful for projects that need strict formatting control regardless of user preferences.

Configuration options

Below is an example configuration with all available options and their default values:

# File patterns (glob syntax)
include = []              # Files to format (default: none, specify on CLI)
exclude = []              # Files to skip (default: none)
git_aware = true          # Respect .gitignore and skip .git directory (default: true)

# Formatting options
line_width = 80           # Maximum line width (min: 8, default: 80)

[heading]
setext_h1 = true          # Use === underline for h1 (default: true)
setext_h2 = true          # Use --- underline for h2 (default: true)
sentence_case = false     # Convert headings to sentence case (default: false)
proper_nouns = []         # Additional proper nouns to preserve (default: [])
common_nouns = []         # Exclude built-in proper nouns (default: [])

[unordered_list]
unordered_marker = "-"    # "-", "*", or "+" (default: "-")
leading_spaces = 1        # Spaces before marker (0–3, default: 1)
trailing_spaces = 2       # Spaces after marker (0–3, default: 2)
indent_width = 4          # Indentation for nested items (min: 1, default: 4)

[ordered_list]
odd_level_marker = "."    # "." or ")" at odd nesting levels (default: ".")
even_level_marker = ")"   # "." or ")" at even nesting levels (default: ")")
pad = "start"             # "start" or "end" for number alignment (default: "start")
indent_width = 4          # Indentation for nested items (min: 1, default: 4)

[code_block]
fence_char = "~"          # "~" or "`" (default: "~")
min_fence_length = 4      # Minimum fence length (min: 3, default: 4)
space_after_fence = true  # Space between fence and language (default: true)
default_language = ""     # Default language for code blocks (default: "")

# External code formatters (see "External code formatters" section)
[code_block.formatters]
# javascript = ["deno", "fmt", "--ext=js", "-"]

[thematic_break]
# Must be valid CommonMark: at least 3 of *, -, or _ (with optional spaces)
style = "- - - - - - - - - - - - - - - - - - - - - - - - - - - - - - - - - - - - -"
leading_spaces = 3        # Leading spaces (0–3, default: 3)

[punctuation]
curly_double_quotes = true   # "text" to "text" (default: true)
curly_single_quotes = true   # 'text' to 'text' (default: true)
curly_apostrophes = false    # it's to it's (default: false)
ellipsis = true              # ... to ... (default: true)
en_dash = false              # Disabled by default (use "--" to enable)
em_dash = "--"               # -- to --- (default: "--", use false to disable)

Configuration values are validated at parse time. Invalid values will produce descriptive error messages:

$ hongdown --config invalid.toml input.md
Error: failed to parse configuration file: line_width must be at least 8, got 5.

When include patterns are configured, you can run Hongdown without specifying files:

# Format all files matching include patterns
hongdown --write

# Check all files matching include patterns
hongdown --check

CLI options override configuration file settings:

# Use config file but override line width
hongdown --line-width 100 input.md

# Use specific config file
hongdown --config /path/to/.hongdown.toml input.md

Style rules

Hongdown enforces the following conventions:

Headings

• Level 1 and 2 use Setext-style (underlined with = or -)
• Level 3+ use ATX-style (###, ####, etc.)
• Optional sentence case conversion (disabled by default)

Document Title
==============

Section
-------

### Subsection

Sentence case (optional)

When sentence_case = true is set in the configuration, Hongdown automatically converts headings to sentence case using intelligent heuristics:

# Development Commands      → Development commands
# Working With JSON APIs    → Working with JSON APIs
# Using `MyClass` In Code   → Using `MyClass` in code

The converter:

• Capitalizes only the first word
• Preserves code spans (text in backticks)
• Preserves acronyms (2+ consecutive uppercase letters, e.g., API, HTTP)
• Preserves proper nouns (built-in list + user-configured)
• Handles hyphenated words (e.g., JSON-RPC)
• Respects quoted text capitalization
• Preserves non-Latin scripts (CJK, etc.)

You can add custom proper nouns to preserve:

[heading]
sentence_case = true
proper_nouns = ["MyCompany", "MyProduct", "MyAPI"]

You can also exclude built-in proper nouns by treating them as common nouns. This is useful for words like “Go” which can be either a programming language or a common verb:

[heading]
sentence_case = true
common_nouns = ["Go", "Swift"]  # Treat these as common nouns, not proper nouns

Built-in proper nouns include ~450 entries: programming languages (JavaScript, TypeScript, Python, Rust, Go), technologies (GitHub, Docker, React, Node.js), databases (PostgreSQL, MySQL, MongoDB), countries (United States, Republic of Korea), natural languages (English, Korean, Japanese), and more.

You can also use HTML comment directives to define document-specific proper nouns and common nouns. See the “HTML comment directives” section for details.

Lists

• Unordered lists use - (space-hyphen-two spaces)
• Ordered lists use 1. format
• 4-space indentation for nested items

 -  First item
 -  Second item
     -  Nested item

Code blocks

• Fenced with four tildes (~~~~)
• Language identifier on the opening fence

~~~~ rust
fn main() {
    println!("Hello, world!");
}
~~~~

External code formatters

You can configure external formatters for code blocks in your .hongdown.toml. This allows automatic formatting of embedded code using language-specific tools.

[code_block.formatters]
# Simple format: command as array
javascript = ["deno", "fmt", "--ext=js", "-"]
typescript = ["deno", "fmt", "--ext=ts", "-"]
rust = ["rustfmt"]

# With custom timeout (default is 5 seconds)
[code_block.formatters.python]
command = ["black", "-"]
timeout = 10

Behavior:

• Language matching is exact only (javascript matches javascript, not js)
• Code is passed to the formatter via stdin, formatted output read from stdout
• If the formatter fails (non-zero exit, timeout, etc.), the original code is preserved and a warning is emitted
• External formatters are only available in CLI mode (not in WASM)

To skip formatting for a specific code block, add hongdown-no-format after the language identifier:

~~~~ python hongdown-no-format
def hello(): print("Hello, World!")
~~~~

The hongdown-no-format keyword is preserved in the output, ensuring the code block remains unformatted on subsequent runs.

For WASM builds, use the formatWithCodeFormatter function with a callback:

import { formatWithCodeFormatter } from "@hongdown/wasm";
import * as prettier from "prettier";

const { output, warnings } = await formatWithCodeFormatter(markdown, {
  codeFormatter: (language, code) => {
    if (language === "javascript" || language === "typescript") {
      return prettier.format(code, { parser: "babel" });
    }
    return null; // Keep original for other languages
  },
});

Line wrapping

• Lines wrap at approximately 80 display columns
• East Asian wide characters are counted as 2 columns
• Long words that cannot be broken are preserved

Links

• External URLs are converted to reference-style links
• References are placed at the end of each section
• Relative/local URLs remain inline

See the [documentation] for more details.

[documentation]: https://example.com/docs

Tables

• Pipes are aligned accounting for East Asian wide characters
• Minimum column width is maintained

See STYLE.md for the complete style specification, including the philosophy behind these conventions and detailed formatting rules.

Editor integrations

Zed

Add the following to your Zed settings to use Hongdown as the Markdown formatter (contributed by Lee Dogeon):

{
  "languages": {
    "Markdown": {
      "formatter": {
        "external": {
          "command": "hongdown",
          "arguments": ["-"]
        }
      }
    }
  }
}

Neovim

If you use none-ls.nvim (a community-maintained fork of null-ls.nvim), you can register Hongdown as a formatter (requires none-ls-extras.nvim, contributed by Vladimir Rubin):

local hongdown = require('none-ls.formatting.hongdown')
null_ls.register(hongdown)

Helix

If you use Helix, you can register Hongdown as a formatter (contributed by Jean Simard):

[[language]]
name = "markdown"
auto-format = true
formatter = { command = "hongdown", args = ["--stdin"] }

Library usage

Rust

Hongdown can also be used as a Rust library:

use hongdown::{format, Options};

let input = "# Hello World\nThis is a paragraph.";
let options = Options::default();
let output = format(input, &options).unwrap();
println!("{}", output);

JavaScript/TypeScript

Hongdown is available as a WebAssembly-based library for JavaScript and TypeScript:

npm install @hongdown/wasm

import { format, formatWithWarnings } from "@hongdown/wasm";

// Basic usage
const markdown = "# Hello\nWorld";
const formatted = await format(markdown);

// With options
const result = await format(markdown, {
  lineWidth: 100,
  setextH1: false,
  fenceChar: "`",
});

// Get warnings along with formatted output
const { output, warnings } = await formatWithWarnings(markdown);
if (warnings.length > 0) {
  for (const warning of warnings) {
    console.warn(`Line ${warning.line}: ${warning.message}`);
  }
}

The library works in Node.js, Bun, Deno, and web browsers. See the TypeScript type definitions for all available options.

Development

This project uses mise for task management.

Initial setup

After cloning the repository, set up the Git pre-commit hook to automatically run quality checks before each commit:

mise generate git-pre-commit --task=check --write

Quality checks

The following tasks are available:

# Run all quality checks
mise run check

# Individual checks
mise run check:clippy     # Run clippy linter
mise run check:fmt        # Check code formatting
mise run check:type       # Run Rust type checking
mise run check:markdown   # Check Markdown formatting

See AGENTS.md for detailed development guidelines including TDD practices, code style conventions, and commit message guidelines.

Etymology

The name Hongdown is a portmanteau of Hong (from Hong Minhee, the author) and Markdown. It also sounds like the Korean word hongdapda (홍답다), meaning “befitting of Hong” or “Hong-like.”

License

Distributed under the GPL-3.0-or-later. See LICENSE for more information.