Arid 1.2.0 — Performance & Portability

Released: August 19, 2026

Arid 1.2.0 is focused on making Arid faster to operate, easier to install across common Linux environments, and harder to integrate incorrectly—without changing what Arid considers duplicate code.

The release adds opt-in automatic parallelism, Linux ARM64 distribution, published machine-contract schemas, and stronger release validation while preserving serial execution as the default.

Highlights

• Add --workers auto for bounded automatic file-preparation parallelism.
• Keep normal arid . execution serial by default.
• Preserve explicit numeric --workers N behavior.
• Publish Linux x86_64 and ARM64 wheels and standalone archives.
• Target manylinux_2_17 / glibc 2.17 for Linux release artifacts.
• Publish JSON Schema documents for report schema v3 and baseline schema v1.
• Harden release packaging and native Linux ARM64 verification.

Automatic parallelism

Arid still runs serially unless you opt in:

arid .

For an explicit worker count:

arid . --workers 4

Or let Arid choose conservatively:

arid . --workers auto

auto is capped at 4 workers and is further bounded by available parallelism and the number of discovered Python files. Parallelism applies to file preparation—reading, parsing, and normalization—and does not change findings, metrics, report ordering, or exit status.

Broader Linux support

Arid 1.2 publishes Linux release artifacts for:

• x86_64
• ARM64 (aarch64)

Existing macOS x86_64, macOS ARM64, and Windows x86_64 release targets remain supported.

Published machine contracts

Arid now publishes versioned JSON Schema documents for its existing machine-readable contracts:

• schemas/report-v3.schema.json
• schemas/baseline-v1.schema.json

These schemas document the contracts that already existed; they do not introduce a report or baseline format break.

Performance

The qualified 1.2.0-rc.1 benchmark campaign measured Arid at approximately:

• 215.82x faster than Pylint on the pinned Pydantic corpus.
• 255.95x faster than Pylint on the pinned Polaris corpus.

Profiling showed Python parsing through Ruff as the dominant cost. Arid's own post-parse duplicate analysis was already inexpensive, so 1.2 deliberately avoided adding algorithmic complexity that the measurements did not justify.

Compatibility

Arid 1.2.0 is backward compatible with 1.1.

In particular:

• duplicate-detection semantics are unchanged
• serial execution remains the default
• existing numeric worker counts retain their meaning
• report schema remains version 3
• baseline schema remains version 1
• existing baselines remain valid
• text, JSON, Markdown, and SARIF contracts remain compatible

No migration is required when upgrading from 1.1.0.

Install or upgrade

With uv:

uv tool upgrade arid

With pip:

python -m pip install --upgrade arid

Documentation

• README
• Report schema v3
• Baseline schema v1
• Performance report
• v1.2 release roadmap