Skip to main content

Fast TOML parser for Python with SIMD optimizations

Project description

FastTOML

Fast TOML parser for Python with SIMD optimizations.

Features

  • Fast: 5-10x faster than existing TOML parsers
  • 🎯 SIMD Optimized: Uses AVX2/SSE4.2 for maximum performance
  • 🔧 Drop-in Replacement: Compatible API with existing TOML libraries
  • 📦 Full TOML v1.0.0 Support: Complete specification implementation
  • 🚀 C++17 Backend: High-performance native implementation

Installation

pip install fasttoml

Usage

import fasttoml

# Parse TOML string
toml_str = '''
title = "TOML Example"
[owner]
name = "Tom Preston-Werner"
age = 42
'''

data = fasttoml.loads(toml_str)
print(data)

# Parse TOML file
data = fasttoml.load('config.toml')

# Serialize dict to TOML string
toml_out = fasttoml.dumps(data)

# Write dict to a TOML file
fasttoml.dump(data, 'output.toml')

Performance

FastTOML is designed for maximum performance using SIMD optimizations. Benchmarks compare against tomli (and optionally toml).

Run benchmarks:

pip install -e .[dev]
pytest tests/test_benchmark.py -v --benchmark-only

Example results (vs tomli): ~6–9× faster depending on document size (small/medium/large/real-world TOML). Results are grouped by payload size; install tomli (or use Python 3.11+ with built-in tomllib) for comparison.

Status and limitations

  • Status: Alpha. Suitable for parsing configs and tests; run the test suite before relying in production.
  • API: loads(s), load(fp), dumps(obj), and dump(obj, fp) are provided. Serialization (dumps/dump) is implemented in Python.
  • Types: Offset datetimes (with Z or +/-HH:MM) are returned as timezone-aware datetime (UTC). Local datetime (no offset, e.g. 1979-05-27T07:32:00) is returned as a string for toml-test/tagged-JSON compatibility. Date-only and time-only TOML values are returned as strings ("YYYY-MM-DD", "HH:MM:SS").
  • Invalid TOML: Invalid input raises ValueError with an error message; the parser does not crash on malformed data.

Requirements

  • Python 3.10+
  • CMake 3.15+
  • C++17 compatible compiler

Development

# Install development dependencies
pip install -e .[dev]

# Build
python setup.py build_ext --inplace

# Run tests (excluding long benchmarks)
pytest tests/ --ignore=tests/test_benchmark.py

# Run benchmarks vs tomli
pytest tests/test_benchmark.py -v --benchmark-only

# Run toml-test suite (optional)
# Option A: use the official toml-test binary (requires Go)
#   go install github.com/toml-lang/toml-test/v2/cmd/toml-test@latest
#   python scripts/run_toml_test.py
# Option B: pytest over cloned toml-test (clone once, then)
#   pytest tests/test_toml_test_suite.py -v
#   (uses .toml-test/ if present, else clones from GitHub; by default 80 valid + 10 invalid; TOML_TEST_FULL=1 runs all valid with 9 decoder-format skips, invalid still limited to 10)

License

MIT

Project details


Download files

Download the file for your platform. If you're not sure which to choose, learn more about installing packages.

Source Distribution

fasttoml-0.2.0b1.tar.gz (27.8 kB view details)

Uploaded Source

Built Distribution

If you're not sure about the file name format, learn more about wheel file names.

fasttoml-0.2.0b1-cp310-cp310-win_amd64.whl (92.9 kB view details)

Uploaded CPython 3.10Windows x86-64

File details

Details for the file fasttoml-0.2.0b1.tar.gz.

File metadata

  • Download URL: fasttoml-0.2.0b1.tar.gz
  • Upload date:
  • Size: 27.8 kB
  • Tags: Source
  • Uploaded using Trusted Publishing? No
  • Uploaded via: twine/6.2.0 CPython/3.10.11

File hashes

Hashes for fasttoml-0.2.0b1.tar.gz
Algorithm Hash digest
SHA256 555ef07573c1471bac48fd8866c1b0153b2a519c6ecb102657fd4ff6fcc79871
MD5 81ad569a8c8a851ed7f9064ca9301117
BLAKE2b-256 28988e69a486631d921435afbc71460c1c18a333978201298a06e1e4453e41df

See more details on using hashes here.

File details

Details for the file fasttoml-0.2.0b1-cp310-cp310-win_amd64.whl.

File metadata

  • Download URL: fasttoml-0.2.0b1-cp310-cp310-win_amd64.whl
  • Upload date:
  • Size: 92.9 kB
  • Tags: CPython 3.10, Windows x86-64
  • Uploaded using Trusted Publishing? No
  • Uploaded via: twine/6.2.0 CPython/3.10.11

File hashes

Hashes for fasttoml-0.2.0b1-cp310-cp310-win_amd64.whl
Algorithm Hash digest
SHA256 e9b50bc4373f0c787a5f3ee4e52b03d726c9b8c649f768f1c69042c1f8b38a95
MD5 97e06dfa1d7fddf539555dae92f6d194
BLAKE2b-256 0fea4064b095381dff7bba555e726c7d87352ee1237ed48820c191a4dd416466

See more details on using hashes here.

Supported by

AWS Cloud computing and Security Sponsor Datadog Monitoring Depot Continuous Integration Fastly CDN Google Download Analytics Pingdom Monitoring Sentry Error logging StatusPage Status page