v0.4.1

v0.4.1

This release focuses on lower allocation overhead in key folding, faster direct bytes input from Python, and stronger release-profile optimization.

Highlights

• Reduced key-folding overhead in the Rust encoder by removing the temporary sibling HashSet allocation. Collision checks now reuse sonic_rs::Object lookups directly.
• Avoided an unnecessary memory copy in etoon.dumps() when the input is already bytes. bytearray inputs still copy once into immutable bytes before entering the Rust binding.
• Enabled full LTO for release builds (lto = true) to improve cross-crate optimization.

Performance

• PR benchmark notes reported the release-profile change improving encode time from about 8.26us to 7.91us in the tested case.
• The key-folding path now avoids per-object sibling-set allocation when folding is enabled.

CI And Release

• OpenSSF Scorecard now runs on main branch pushes and weekly schedule, matching Scorecard's expected trigger model.
• Release artifacts are built for Linux, macOS, and Windows, with Python wheels for CPython 3.10-3.13.

Install

cargo install etoon

pip install etoon

Verification

• cargo test
• uv run pytest tests/test_spec_fixtures.py

Release Assets

• etoon-linux-x86_64
• etoon-linux-aarch64
• etoon-macos-x86_64
• etoon-macos-aarch64
• etoon-windows-x86_64.exe
• SHA256SUMS.txt

VirusTotal Scan Results

• etoon-linux-aarch64
• etoon-linux-x86_64
• etoon-macos-aarch64
• etoon-macos-x86_64
• etoon-windows-x86_64.exe

Full changelog: v0.3.0...v0.4.1