psf / black · Getting started

The metadata also records concrete facts: the primary language is Python, and almost all tracked language bytes are Python — 5,440,093 of about 5.46 million — with small Vim Script, Jupyter Notebook and Dockerfile shares. The license is recorded as MIT, with origin given as both the GitHub API and a repository file. The latest release is 26.5.1, published 2026-05-18, and the repository counted 41,844 stargazers on 2026-09-14. The analyzer's maturity signals list 482 analyzed files, 326 test-like files, 7 manifest files, 9,339 static relation edges and 176 bounded execution paths.

If your question is what to install, this packet cannot answer it. It contains no dependency records; dependency records come only from requirements-style manifests parsed by analyzer v0.10, and none were parsed for this repository. pyproject.toml dependency tables are not parsed, so declared dependencies and version constraints are not listed — even though pyproject.toml is present and classified as a project-level important file. The maturity signals count 7 manifest files in the inventory, but with no parsed-manifest records their contents are not reflected here. Because install, run and build inference is partial and installation commands are not verified, the asset also cannot name packages or versions to install.

Eighteen entrypoint records are listed: seventeen whose excerpts show a Python __main__ execution guard — the excerpt fields of those records are where you can see what such a guard looks like — plus one filename-heuristic candidate, action/main.py (the analysis labels it a filename-heuristic candidate, based on the filename rather than a guard excerpt). In the src tree, the package inits src/black/__init__.py and src/blackd/__init__.py each guard a call to patched_main(). Under src/blib2to3/pgen2, driver.py ends with sys.exit(int(not main())), while literals.py and tokenize.py carry smaller, test-oriented guards. In scripts/, diff_shades_gha_helper.py, generate_schema.py and make_width_table.py guard a call to main(); release.py ends with sys.exit(main()) and release_tests.py calls unittest.main(). scripts/fuzz.py's guard runs test_idempotent_any_syntactically_valid_python() and, when the Atheris fuzzing library is available, coverage-guided fuzzing. Guard-bearing records also include tests/test_tokenize.py, whose guard reads stdin, tokenizes it and prints an assert_tokenizes assertion, and paths under tests/data/cases/. The reconstruction tallies 18 entrypoints and 176 bounded execution paths; the five bounded paths listed for action/main.py are execution-path records, and their steps show them terminating immediately at unresolved external boundaries such as Path and os.getenv. The analysis's limitations list records the general point behind that last term: dynamic behaviour of several kinds is not resolved by this static analysis. Runtime behaviour beyond these guard and bounded-path records, including how the package is invoked as an installed command, is not established by this analysis.

README.md, pyproject.toml and LICENSE are classified as project-level important files. The analyzer's learning-path claim advises starting from manifests and important symbols, then tracing the bounded execution path and inspecting unresolved boundaries before changing code; this packet cannot show the manifest contents that advice presupposes, so the advice is not actionable from this packet alone. Its modify-guide claim says to begin modification analysis at the detected entrypoint action/main.py, then verify downstream effects manually. By analyzed file count the top-level tree is tests (326), docs (48), src (48), .github (22), scripts (9), profiling (7), 19 files at the root, and action, plugin and autoload with one file each. src holds the guard-bearing package inits src/black/__init__.py and src/blackd/__init__.py and the src/blib2to3/pgen2 modules named in the entrypoint records. The 326 test-like files are roughly two thirds of all 482 analyzed files; the counts alone do not show what those files contain. docs/ spans 48 files, and entrypoint excerpts show two scripts opening CHANGES.md and docs files such as docs/usage_and_configuration/the_basics.md and docs/integrations/source_version_control.md.

This is a static analysis, and the packet documents the boundary itself. Installation, build and run commands are not verified, and install/run/build inference is partial, which is why this asset contains no commands. Only requirements-style manifests feed dependency records and pyproject.toml dependency tables are not parsed, so the dependency picture here is empty. README-derived statements are extracted line-by-line and may capture code lines instead of prose, so any README claim is author_claimed at best. Reflection, runtime dependency injection, dynamic imports, monkey-patching, generated code, framework runtime wiring and dynamic dispatch are not resolved. Of 9,339 static relation edges, the reconstruction counts only 813 resolved call relations, so most edges remain unresolved or external. The packet also contradicts itself: one teaching claim says no bounded static execution path is available, yet five bounded paths are listed for action/main.py and 176 overall, and the summary prose names scripts/check_pre_commit_rev_in_example.py as the main starting place while the teaching claims point to action/main.py. How the project is meant to be invoked day-to-day is not established.

The repository metadata describes the project as 'The uncompromising Python code formatter'; treat that as the authors' self-description, not an analysis finding. The homepage field points to a Black documentation site on Read the Docs, stable channel. Attached topics include formatter, codeformatter, python, pre-commit-hook, autopep8, yapf, gofmt, hacktoberfest and code.