yamlrocks 0.3.0

A fast, correct YAML library for Python, written in Rust

🪨 YAMLRocks

Rock-solid YAML for Python, written in Rust.

About

YAMLRocks is the rock-solid YAML library for Python: a Rust-backed extension that parses and emits YAML fast, follows the YAML 1.2 specification (with a YAML 1.1 compatibility mode), and, unlike PyYAML, round-trips documents while preserving comments, anchors, and formatting.

Rock-solid means three things: correct, secure by default, and fast, with a Rust core doing the heavy lifting. (The R in Rock is for Rust.)

The Python YAML ecosystem has long forced a trade-off. YAMLRocks refuses it:

Library	Fast	YAML 1.2	Comments / round-trip	Native includes
PyYAML	with C loader	✗ (1.1 only)	✗	✗
ruamel.yaml	✗ (pure Python)	✓	✓	✗
YAMLRocks	✓ (Rust)	✓	✓	✓

It is also fast. Release-build benchmarks (python bench/bench.py) show how many times faster YAMLRocks is:

Operation	vs PyYAML (C loader)	vs ruamel.yaml
Parse (loads)	~5-10x faster	~85-135x faster
Serialize (dumps)	~15-19x faster	~155-210x faster
Split config (!include, hundreds of files)	~18x faster	n/a

It is safe against the common YAML attack classes, free-threaded (nogil) ready, and ships with JSON Schema validation, a PyYAML-compatible shim, rich standard-library type support, and the !secret and !env_var config tags.

YAMLRocks is also tested against a reproducible real-world corpus covering Home Assistant, ESPHome, Ansible, Kubernetes, Docker Compose, GitHub Actions, CloudFormation, GitOps, Helm, OpenAPI, dbt, CircleCI, Serverless, and Tekton. Each standalone YAML file must parse and round-trip byte-for-byte; selected Home Assistant configs are additionally tested through their full !include graph. See real-world verification for the current corpus and scope.

YAMLRocks is pre-1.0 software. The core promises are already explicit: safe loading by default, YAML 1.2 semantics, reproducible real-world verification, and byte-for-byte round-trip for unmodified documents. Some advanced APIs may still change before 1.0 while the project gathers production feedback. See the stability and roadmap page for the 1.0 contract.

Installation

pip install yamlrocks

Building from source requires a Rust toolchain; see Setting up development environment below for the uv-based build.

Usage

The API stays small: loads returns native Python objects and dumps returns bytes.

import yamlrocks

# Parse YAML into native Python objects
data = yamlrocks.loads(b"key: value\nlist:\n  - 1\n  - 2")
# {'key': 'value', 'list': [1, 2]}

# Serialize back to YAML bytes (dumps() returns bytes)
yamlrocks.dumps(data)
# b'key: value\nlist:\n  - 1\n  - 2\n'

# Multiple documents
yamlrocks.loads_all(b"---\na: 1\n---\nb: 2")
# [{'a': 1}, {'b': 2}]

# Load and dump files directly
config = yamlrocks.load("config.yaml")
yamlrocks.dump(config, "config.yaml")

YAML 1.1 compatibility

YAML 1.2 is the default, so yes/no/on/off are plain strings. Opt into the 1.1 schema when you need its booleans and octals:

yamlrocks.loads(b"enabled: yes")                              # {'enabled': 'yes'}
yamlrocks.loads(b"enabled: yes", option=yamlrocks.OPT_YAML_1_1)  # {'enabled': True}

OPT_UPGRADE_1_1 reads 1.1 and always emits canonical 1.2, so a project can ease off the legacy spellings without a manual conversion step.

Structure-preserving round-trip

doc = yamlrocks.loads(content, option=yamlrocks.OPT_ROUND_TRIP)

doc["server"]["host"] = "example.com"   # deep edits write through to the AST
print(doc.to_yaml().decode())           # comments and formatting preserved

An unmodified document re-emits byte-for-byte identical; only the nodes you touch are rewritten.

Native includes (read and write back)

doc = yamlrocks.loads(
    content,
    option=yamlrocks.OPT_ROUND_TRIP | yamlrocks.OPT_INCLUDES,
    include_dir="/config",
)

doc["automation"][0]["trigger"] = "state"   # edit a value from an included file

yamlrocks.dump_includes(doc, include_dir="/config")
# Only the modified included file is rewritten; the root config is untouched.

Supported tags: !include, !include_dir_named, !include_dir_list, !include_dir_merge_named, !include_dir_merge_list.

Annotated mode (source-location tracking)

data = yamlrocks.loads(content, option=yamlrocks.OPT_ANNOTATED)
data.__line__              # 1
data["server"].__line__    # 3

YAMLRocksAnnotatedDict/YAMLRocksAnnotatedList/YAMLRocksAnnotatedStr subclass dict/list/str, so they behave exactly like the built-ins while carrying __line__, __column__, and __file__, compatible with Home Assistant's annotated YAML.

Option flags

Options compose with |, the integer bit-flag pattern:

Flag	Effect
OPT_YAML_1_1	YAML 1.1 schema (yes/no booleans, octals, etc.)
OPT_UPGRADE_1_1	Read 1.1, always emit canonical 1.2
OPT_ROUND_TRIP	Return a YAMLRocksDocument preserving comments and formatting
OPT_ANNOTATED	Return subclasses with source locations
OPT_INCLUDES	Resolve !include tags (needs include_dir)
OPT_DUPLICATE_KEYS_ERROR	Reject a repeated mapping key
OPT_INDENT_2 / OPT_INDENT_4	Indentation width for dumps
OPT_SORT_KEYS	Sort mapping keys when dumping
OPT_FLOW_STYLE	Emit flow style ({}/[])
OPT_LITERAL_STRINGS	Emit multi-line strings as literal blocks (|)
OPT_EXPLICIT_START / OPT_EXPLICIT_END	Emit --- / ... markers

See the documentation for the full option set, including the standard-library type and datetime flags.

Documentation

Full documentation lives at yaml.rocks: getting-started guides, recipes (Home Assistant, config editors), the complete API reference, security notes, and head-to-head comparisons with PyYAML and ruamel.yaml.

Changelog & Releases

This repository keeps a change log using GitHub's releases functionality. The format of the log is based on Keep a Changelog.

Releases are based on Semantic Versioning, and use the format of MAJOR.MINOR.PATCH. In a nutshell, the version will be incremented based on the following:

• MAJOR: Incompatible or major changes.
• MINOR: Backwards-compatible new features and enhancements.
• PATCH: Backwards-compatible bugfixes and package updates.

Contributing

This is an active open-source project. We are always open to people who want to use the code or contribute to it.

We've set up a separate document for our contribution guidelines.

Using AI tools to help is fine, but you must review and understand everything you submit. Please read our AI Policy first; autonomous agents are not allowed, and unreviewed AI output will be closed.

Thank you for being involved! :heart_eyes:

Setting up development environment

The easiest way to start is by opening a CodeSpace here on GitHub, or by using the Dev Container feature of Visual Studio Code.

YAMLRocks is a Rust extension built with maturin and managed with uv, which handles the Python version, the virtual environment, and every dependency from pyproject.toml. You need:

• A Rust toolchain
• uv (it installs a suitable Python 3.12+ for you)
• Node.js 22+ (only for the documentation site and some lint hooks)

To set up the environment and build the extension:

uv sync                  # create the venv and install all dev dependencies
uv run maturin develop   # build and install the extension (rerun after Rust changes)

As this repository uses the prek framework, all changes are linted and tested with each commit. You can run all checks manually:

prek run --all-files

To run just the tests (always under a memory guard during development):

timeout 120 bash -c 'ulimit -v 3000000; uv run pytest'

See CONTRIBUTING.md for the full workflow, including the just task runner and how to fetch the optional test-data submodules.

Authors & contributors

The original setup of this repository is by Franck Nijhof.

For a full list of all authors and contributors, check the contributor's page.

License

MIT License

Copyright (c) 2026 Franck Nijhof

Permission is hereby granted, free of charge, to any person obtaining a copy of this software and associated documentation files (the "Software"), to deal in the Software without restriction, including without limitation the rights to use, copy, modify, merge, publish, distribute, sublicense, and/or sell copies of the Software, and to permit persons to whom the Software is furnished to do so, subject to the following conditions:

The above copyright notice and this permission notice shall be included in all copies or substantial portions of the Software.

THE SOFTWARE IS PROVIDED "AS IS", WITHOUT WARRANTY OF ANY KIND, EXPRESS OR IMPLIED, INCLUDING BUT NOT LIMITED TO THE WARRANTIES OF MERCHANTABILITY, FITNESS FOR A PARTICULAR PURPOSE AND NONINFRINGEMENT. IN NO EVENT SHALL THE AUTHORS OR COPYRIGHT HOLDERS BE LIABLE FOR ANY CLAIM, DAMAGES OR OTHER LIABILITY, WHETHER IN AN ACTION OF CONTRACT, TORT OR OTHERWISE, ARISING FROM, OUT OF OR IN CONNECTION WITH THE SOFTWARE OR THE USE OR OTHER DEALINGS IN THE SOFTWARE.