Skip to main content
Pre-release

This release is a pre-release and may not be stable for production use.

RefacTrail

The Python refactorizer. A linter, formatter and verified refactoring tool with two independent engines, pure Python and Rust, built on its own parser. Run correctness checks and bounded formatting, or check structure, naming, types and documentation against a chosen profile.

Part of a family of standalone Python tools: FuncLoom (functionizer), RefacTrail (refactorizer) and RepoContour (architect, planned). Each installs and works on its own.

0.3.1a0 is an unpublished experimental alpha. Python 3.12 or newer is required. RefacTrail has its own package, CLI and VS Code extension. It installs FuncLoom automatically for its rewrite API; FuncLoom has no dependency on RefacTrail. No LLM, account or network is needed at runtime.

Quick start

pip install refactrail

That one command also installs FuncLoom and the fast Rust engine (a native wheel for Windows, macOS and Linux; on other systems a small fallback is installed and the Python engine gives the same results, so nothing ever needs compiling). Python 3.12 or newer is required. Until the first PyPI release, install from GitHub instead: pip install git+https://github.com/SAMtheROCKET/refactrail.git.

Then, in any project folder:

refactrail                     # the most useful commands
refactrail lint                # find likely bugs
refactrail format --diff       # preview formatting; --write applies it
refactrail check               # structure, naming, type hints, docstrings
refactrail fix --diff          # preview safe fixes; drop --diff to apply

Paths default to the current folder; pass files or folders to narrow it. Exit codes are 0 (clean), 1 (findings) and 2 (error), ready for CI. If your system blocks the refactrail command, use python -m refactrail.

In VS Code, install the RefacTrail extension and run RefacTrail: ... from the Command Palette. It uses the Python interpreter selected in VS Code and offers to install RefacTrail there with one click.

On every commit, add the hooks to .pre-commit-config.yaml:

repos:
  - repo: https://github.com/SAMtheROCKET/refactrail
    rev: v0.3.1a0
    hooks:
      - id: refactrail-lint
      - id: refactrail-format

refactrail-format-check (fails instead of rewriting) and refactrail-check are also available.

From source (development)

Install FuncLoom from its source folder and the local engine fallback first, then RefacTrail:

python -m pip install -e ../funcloom ./rust/fallback
python -m pip install -e .

Nothing is uploaded by these local tools.

General correctness and independent formatting

python -m refactrail lint src
python -m refactrail lint src --output-format sarif
python -m refactrail format examples/general_demo.py --diff
python -m refactrail format examples/general_demo.py --check
python -m refactrail format examples/general_demo.py --write

lint provides nine RC correctness checks without imposing personal naming rules. format uses an independent CPython AST/tokenize implementation; its default is a read-only preview. It makes bounded whitespace edits and retains literal/comment spelling, directives, BOM and original line endings. Use --line-length 79 for bracketed comma-group wrapping. Explicit Python notebooks and discovered stubs are supported; --notebooks includes notebooks in folder formatting. This is a documented style subset, not full Black parity.

python -m refactrail format examples/analysis.ipynb --line-length 79 --diff
python -m refactrail scope examples/project_demo/invoice.py
python -m refactrail index examples/project_demo --changed invoice.py
python -m refactrail rename examples/project_demo/invoice.py --function calculate_total --old amount --new base_amount_int

scope exposes partial compiler binding evidence. index inventories symbols and possible import impact under an explicit import root. rename produces a read-only, hash-linked proposal for a bounded local-variable subset; parameters, public APIs, nested scopes and dynamic namespaces are refused. See the workflow and examples.

lint --jobs 4 enables worker processes on larger batches. Lint caching hashes source bytes, rule selection, tool/interpreter version and path-sensitive policy; use --no-cache to disable it. No cache is used to authorize source changes.

See the independent engines for contracts, APIs and limitations, and the expansion roadmap for broader formatter, data-flow and repository-refactoring work. These engines do not invoke or contain Ruff or Black. The FuncLoom structural-rewrite dependency remains as documented below.

Checking and fixing

check reports line, function and main-block sizes; docstrings and their sections; missing annotations; naming conventions and verb prefixes; constants and their location; and top-level script execution. The strict profile adds generic-name and dtype-suffix checks. Defaults are 79 columns, 40 preferred / 50 maximum function lines, and 100 main-block lines.

python -m refactrail check src --profile strict --statistics
python -m refactrail check src --output-format json
python -m refactrail rules
python -m refactrail fix src --diff
python -m refactrail fix src

fix --diff previews edits without writing. fix edits files in place: selected fixes add eligible -> None annotations, create function docstring skeletons, wrap supported lines and split supported long functions. It checks compilation, rejects linked files and changed source, and replaces each file atomically. Review diffs and run your project's tests. A failed file does not roll back other files in a multi-file run.

These are structural checks, not proof of unchanged behavior. Annotations and docstrings can affect reflection. There is no automatic domain naming, public-API renaming, inferred annotation insertion beyond the bounded None case, or arbitrary repository restructuring. Missing meanings need user context. Unsupported functions can remain long. Use FuncLoom's explicit modularize command to create a separate package draft.

Exit codes: check returns 0 with no findings, 1 for findings, 2 for usage errors. --exit-zero explicitly overrides finding failures. Diff returns 1 when edits are proposed or files are skipped; fix returns 1 for skipped files. Syntax, encoding and internal errors cannot be hidden by selection or ignore lists. Suppression markers apply only inside real comment tokens.

Configuration

[tool.refactrail]
profile = "strict"
line-length = 79
function-preferred-lines = 40
function-max-lines = 50
main-max-lines = 100
select = ["RT"]
ignore = []

Widths must be 40-200; the preferred function size cannot exceed the maximum. See the rule contract and design.

The Rust engine

--engine auto selects a compatible refactrail-core when installed and otherwise uses Python. --engine python selects the reference implementation. The native core is a separate PyO3/Rayon distribution built on RefacTrail's own Rust lexer, parser, compile checks and symbol table; it contains no Ruff code. check, lint and format accept --engine, and the standalone refactrail-native binary runs check, lint, scope, format and index without Python. Outputs are byte-identical to the Python engine on the recorded corpora and fuzzed inputs, with CPython as the oracle; see the dual-engine record. Parity evidence is not a guarantee for every Python program.

Native wheels (CPython 3.12+ ABI3) are built and tested in CI for Linux x86_64 (manylinux2014), Windows x86_64 and macOS arm64; other platforms use the Python engine. Install a local wheel with pip install --no-index --no-deps WHEEL. See benchmark evidence for measured timings and their limits.

Editor and release preparation

The VS Code extension provides explicit check, diff-preview and fix commands. Install its local VSIX, select the Python environment containing these packages, and use a trusted workspace.

python scripts/verify.py
python scripts/release_check.py --output dist/0.3.1a0 --dependency-wheel path/to/funcloom-0.10.3a0-py3-none-any.whl

The release script builds and checks a wheel and source archive, installs both in fresh environments, tests installed code and checks uninstallation. CI runs the tests on Windows, Linux and macOS with Python 3.12-3.14, and the Release check workflow verifies the identical built files on all three systems. Release preparation lists the remaining launch work. No public release has been made.

MIT, copyright 2026 Sambit Supriya Dash.

Metadata

Release files for refactrail 0.3.1a0

For a detailed explanation of source distributions (sdists) and built distributions (wheels), please see the package formats documentation.

Source distribution (sdist)

Source distribution for refactrail 0.3.1a0
File Size Uploaded
refactrail-0.3.1a0.tar.gz 171.2 kB Details

Built distribution (wheel)

Table of built distributions (wheels) for refactrail 0.3.1a0
File Interpreter ABI Platform
refactrail-0.3.1a0-py3-none-any.whl Python 3 none any Details

Total release size: 273.6 kB

Release files / refactrail-0.3.1a0.tar.gz

Download URL refactrail-0.3.1a0.tar.gz
Size 171.2 kB
Tags Source
SHA-256 checksum
How to use checksums
297678e5bb0a1cb52c360fe0baddfe09885d2b89c4b51ead804c2ee17ab39ee3
BLAKE2b-256 checksum
How to use checksums
ab2c5ebe8fa81021b165a33d62134fc97d8fdc71e4383fddc5e0a7f6fdc7152f
Upload date
Uploaded using Trusted Publishing?
What is trusted publishing?
Yes
Uploaded via twine/7.0.0 CPython/3.13.14

Provenance

Provenance describes where a file came from. On PyPI, provenance is shared via attestations, which provide a verifiable record of the build or publishing details. View details, limitations and caveats.

PyPI Publish Attestation

PyPI verified that this artifact, at this checksum, originated from the publisher listed below.

Signed by GitHub Actions, verified by PyPI on Oct 5, 2026.

Transparency log

Release files / refactrail-0.3.1a0-py3-none-any.whl

Download URL refactrail-0.3.1a0-py3-none-any.whl
Size 102.4 kB
Tags Python 3
SHA-256 checksum
How to use checksums
b2bed56830db98e0641dc5174432a581093faeb6cb9fddd3dc6f062cd14af9d1
BLAKE2b-256 checksum
How to use checksums
b2b39beb70c6c1f3ff9867f71874faa87659a1fe6b94f7746685b7c81a476e4f
Upload date
Uploaded using Trusted Publishing?
What is trusted publishing?
Yes
Uploaded via twine/7.0.0 CPython/3.13.14

Provenance

Provenance describes where a file came from. On PyPI, provenance is shared via attestations, which provide a verifiable record of the build or publishing details. View details, limitations and caveats.

PyPI Publish Attestation

PyPI verified that this artifact, at this checksum, originated from the publisher listed below.

Signed by GitHub Actions, verified by PyPI on Oct 5, 2026.

Transparency log
Anthropic, PBC Visionary sponsor Bloomberg Visionary sponsor Hudson River Trading Visionary sponsor Meta Visionary sponsor NVIDIA Visionary sponsor Microsoft Sustainability sponsor Depot Continuous Integration AWS Cloud computing and Security Sponsor Datadog Monitoring Fastly CDN Google Download Analytics Sentry Error logging StatusPage Status page