[agibot-world]

This PR contains the following updates:

Package	Change	Age	Confidence
pyproject-fmt (changelog)	2.11.1 → 2.30.3

---

Release Notes

tox-dev/toml-fmt (pyproject-fmt)

v2.30.3

Added

• Require 3.11 and default to 3.15. (#​480) by
  @​gaborbernat.

v2.30.2

Fixed

• Format every input after an earlier one changes. (#​479) by
  @​gaborbernat.

v2.30.1

Added

• Raise the default minimum Python to 3.11. (#​476) by
  @​gaborbernat.

v2.30.0

v2.29.4

Fixed

• Keep the classifiers a bound admits. (#​463) by
  @​gaborbernat.

v2.29.3

Fixed

• The test suite the source distribution ships passes from the unpacked sdist. The backend read the project metadata
  from beside itself, and the sdist keeps a copy of that file next to the tests, a directory below the pyproject.toml
  it belongs to, so eight tests ended on the file it could not find. It reads the metadata from either place.
  (#​454, #​457) by
  @​gaborbernat.

v2.29.2

Fixed

• A build from the source distribution runs on what that distribution carries. 2.29.0 named toml-fmt-common as a
  dependency and resolved it from PyPI, where the newest release dates from May; 2.29.1 dropped the dependency before
  the sdist carried the sources, so a build from it failed on the import. This release carries them, and its metadata
  requires nothing. (#​450,
  #​452) by @​gaborbernat.

  The wheels held their vendored copy throughout, so an install that takes a wheel was never affected.

v2.29.1

Fixed

• The source distribution no longer names toml-fmt-common as a dependency. The newest release of that package on
  PyPI dates from May, months behind this one, so a build from the sdist ran an old settings reader against the current
  formatter and failed a large part of its own test suite. (#​450,
  #​451) by @​gaborbernat.

Packaging

• This release's source distribution carries no toml-fmt-common, so a build from it fails on the import.
  2.29.2 carries it. The wheels hold their
  vendored copy and install as before. (#​452) by
  @​gaborbernat.

v2.29.0

Changes

• A string measures from the start of its key. A long key can be what pushes a value past column_width; measuring
  the value alone left lines running past the column the setting asks for.
  (#​448) by @​gaborbernat.
• Deep input comes back as an error rather than a crash. Reading a value, writing it and dropping it each walk it by
  calling themselves, so the model caps nesting at 256 levels. A 12,000-deep value used to end the process.
  (#​448) by @​gaborbernat.

Fixed

• A requirement written with a space after its operator reads as a requirement. requires-python = ">= 3.12" did
  not parse as a version bound, so the generated classifiers fell back to the configured floor and ceiling. pypa/build
  carried a 3.9 classifier it does not support. (#​448) by
  @​gaborbernat.
• The classifier window works at patch precision. requires-python = "<3.10.1" lost the whole 3.10 series, and
  claimed Programming Language :: Python :: 3 :: Only for a bound that admits Python 2.
  (#​448) by @​gaborbernat.
• A literal string sorts with the values around it. 'zz' and 'aa' held their order while the same values in
  double quotes sorted. (#​448) by
  @​gaborbernat.
• A comment written before a member's comma stays on that member's line. The comma is what says which member a
  comment belongs to: one written before it closes that member's line, one written after it leads the next member. Both
  used to end up on a line of their own, leaving the comma stranded below.
  (#​448) by @​gaborbernat.
• classifiers written as a string stays as written. Asking for generated classifiers replaced the string with an
  array, losing what the file said. (#​448) by
  @​gaborbernat.
• Folding sub-tables into their parent no longer depends on the order the file wrote them in.
  (#​448) by @​gaborbernat.
• The key order covers pyrefly's documented option names. It spells them with hyphens, the order listed only the
  underscore forms, and a file using the documented spelling fell through to alphabetical order.
  (#​448) by @​gaborbernat.
• A dependency group keeps its include-group entries where the file wrote them. An include-group pulls its group
  in at the point it sits, so moving it changes what the group resolves to.
  (#​448) by @​gaborbernat.
• Free-form license text stays as written. license = "MIT or later" came back rewritten as though it were an SPDX
  expression; the formatter now rewrites the value only once it parses as one over registered identifiers.
  (#​448) by @​gaborbernat.
• The * catch-all in a setuptools data table matches the way the file spells it. * is not a name TOML reads
  bare, so a file writes it quoted; the catch-all led the table only because a quote happens to sort before a letter.
  (#​448) by @​gaborbernat.

Performance

• The quadratic walks the old parse tree forced are gone. 32,000 interleaved root keys under two tables took 618
  seconds and now take 4.8; scaling is close to linear. (#​448) by
  @​gaborbernat.

Packaging

• This release's source distribution names a dependency on toml-fmt-common without carrying it, and the newest release
  of that package on PyPI dates from May. A build from the sdist therefore runs an older settings reader against this
  formatter. 2.29.2 carries the sources. The
  wheels hold their vendored copy and install as before. (#​450)

v2.28.2

Fixed

• A file keeps the line endings it was written with. A CRLF file came back LF, turning every line of a Windows
  checkout into a diff. (#​447) by
  @​gaborbernat.

v2.28.1

Packaging

• The wheel declares its dependencies in the METADATA header, where an installer looks for them.
  (#​444) by @​gaborbernat.

v2.28.0

Fixed

• A sub-table written under its own header reads in the same key order as the folded form.
  (#​438) by @​gaborbernat.
• [tool.cibuildwheel] overrides keep their order once folded.
  (#​437) by @​gaborbernat.
• [tool.coverage] keys follow the order the tool documents. (#​436)
  by @​gaborbernat.

Documentation

• The docs build publishes llms.txt. (#​432) by
  @​gaborbernat.

v2.27.1

Fixed

• project.version stays as written. It is the exact version published in the package metadata, and normalizing it
  rewrote what a release had already gone out as. (#​431) by
  @​gaborbernat.

v2.27.0

Added

• project.version is checked, and a value PEP 440 cannot read is named.
  (#​428) by @​gaborbernat.

v2.26.0

Added

• [tool.pyproject-fmt] orders its own keys. (#​421) by
  @​gaborbernat.

v2.25.4

Packaging

• The source distribution carries only git-tracked sources. A stray build artifact ended up in the sdist.
  (#​419) by @​gaborbernat.

v2.25.3

Fixed

• A marker written after a URL keeps the space PEP 508 requires. Without it the ; reads as part of the URI.
  (#​410) by @​gaborbernat.

v2.25.2

Fixed

• sub_table_spacing and separate_root_table honor a value of more than one blank line.
  (#​403) by @​gaborbernat.

Documentation

• The formatter itself produces every formatting example, so the docs cannot describe something the tool does not
  do. (#​397) by @​gaborbernat.

v2.25.1

Fixed

• A key the file wrote as a comment reorders and comments back out without losing what it said.
  (#​391) by @​gaborbernat.
• An array entry keeps the comments and blank lines around it when an inline table is reordered.
  (#​392) by @​gaborbernat.
• A line-ending backslash inside a multi-line basic string reads as the continuation it is.
  (#​393) by @​gaborbernat.
• base_python_file sorts with the other tox environment keys.
  (#​386) by @​edgarrmondragon.

v2.25.0

Added

• A key the file wrote as a comment formats with the table it belongs to, then comments back out. A
  # dependencies = [...] left in place for later keeps its position and its shape.
  (#​381, #​383) by
  @​gaborbernat.
• # Group: markers scope reordering. Related keys held under one marker sort among themselves, and nothing crosses
  a marker. (#​380) by @​gaborbernat.
• uv tool install pyproject-fmt works. (#​382) by
  @​bn-andrew.

Fixed

• A table the file wrote empty survives when its sub-table stays written out.
  (#​385) by @​gaborbernat.

v2.24.1

v2.24.0

Changes

• A setuptools build-system.requires drops a redundant wheel. Setuptools has bundled bdist_wheel since 70.1.
  (#​373) by @​gaborbernat.

v2.23.0

Documentation

• The formatting reference is rewritten, and now gives the key order and array policy of every tool it knows.
  (#​366) by @​gaborbernat.

v2.22.0

Added

• Thirty tool tables get a format of their own. autopep8, bandit, black, bumpversion, check-manifest,
  cibuildwheel, codespell, commitizen, deptry, djlint, docformatter, hatch, interrogate, isort,
  maturin, mypy, pdm, poetry, pylint, pyright and basedpyright, pyrefly, pytest, scikit-build,
  semantic_release, setuptools and setuptools_scm, towncrier, tox, ty, vulture, yapf. Each gets the key
  order its own documentation gives, and each array is sorted or left in place depending on whether its order carries
  meaning. (#​324 through
  #​354) by @​gaborbernat.
• sub_table_spacing and separate_root_table set the blank lines between tables.
  (#​319) by @​gaborbernat.

Fixed

• A triple-quoted literal string survives formatting. (#​356) by
  @​gaborbernat.
• _build_cli is back for consumers that called it. (#​361) by
  @​gaborbernat.

Packaging

• The wheels are self-contained, so installing one no longer pulls a matching toml-fmt-common release.
  (#​363) by @​gaborbernat.
• The source distribution carries its license file. (#​313) by
  @​gaborbernat.

v2.21.2

Added

• Free-threaded Python wheels. (#​307) by
  @​gaborbernat.

Fixed

• A table the file wrote empty no longer disappears from a lookup.
  (#​304) by @​gaborbernat.

v2.21.1

Fixed

• Sorting an array whose first value sits on the bracket line writes valid TOML.
  (#​293) by @​gaborbernat.

v2.21.0

Added

• [tool.pixi] gets a format of its own. (#​283) by
  @​gaborbernat.

Fixed

• --check and --stdout no longer ask for a writable file. Neither mode writes anything, so a read-only file
  failed on a permission neither needs. (#​278) by
  @​gaborbernat.

v2.20.0

Changes

• bandit counts as a linter, so its table sits with the other linters.
  (#​276) by @​gaborbernat.

v2.19.0

Changes

• Type checkers sort after linters, so the file reads in the order a project runs them.
  (#​274) by @​gaborbernat.

v2.18.1

Fixed

• A value that is not an array no longer panics an array rule.
  (#​266) by @​gaborbernat.

v2.18.0

Added

• Inline table keys follow the order the kind of table asks for.
  (#​264) by @​gaborbernat.

v2.17.0

Added

• One config file steers every formatter. Settings written once in a shared file reach each tool built on
  toml-fmt-common, so a project stops repeating them per tool. (#​258)
  by @​gaborbernat.

Fixed

• The parser keeps up with tombi v0.8.0. (#​259) by
  @​gaborbernat.

v2.16.2

Fixed

• A classifier the trove list does not hold no longer panics the formatter.
  (#​241) by @​Nicolaus93 and
  @​gaborbernat.
• An environment whose name holds a dot is one environment. (#​234)
  by @​gaborbernat.

v2.16.1

Changes

• authors and maintainers keep the order the file wrote them in. That order is published metadata and says who
  led the work, so sorting it changed what the package claimed. (#​228,
  #​229) by @​gaborbernat.

Documentation

• column_width documents how it wraps a string. (#​223) by
  @​gaborbernat.

v2.16.0

Fixed

• A key the file quoted reads as the one name it spells. A literal-quoted key and a basic-quoted one name the same
  key, and sorting told them apart. (#​217,
  #​219) by @​gaborbernat.

v2.15.3

Added

• skip_wrap_for_keys holds a value on one line. A key whose value carries meaning that line breaks would obscure,
  such as a command line, can now be named. (#​216) by
  @​gaborbernat.

Fixed

• Quote styles agree when keys sort. (#​215) by
  @​gaborbernat.

v2.15.2

Fixed

• A comment closing an array's last line survives a sort. (#​208) by
  @​gaborbernat.

v2.15.1

Fixed

• An array written on one line with a trailing comment stays on one line.
  (#​204) by @​gaborbernat.
• A table the file wrote empty comes back as {} rather than disappearing.
  (#​203) by @​gaborbernat.

v2.15.0

Added

• [tool.coverage] gets a key order, and its list-valued keys sort.
  (#​199) by @​gaborbernat.

Fixed

• Collapsing an array of tables carries the comments written inside it.
  (#​198) by @​gaborbernat.

v2.14.2

Fixed

• Raising test coverage to 98% turned up several defects. (#​194) by
  @​gaborbernat.

v2.14.1

Added

• [tool.uv] gets a format of its own. (#​169) by
  @​gaborbernat.

Fixed

• A table header survives formatting. (#​192) by
  @​gaborbernat.
• A comment written beside a string in an array stays with it.
  (#​191) by @​gaborbernat.
• A comment holding a double quote inside an array reads as one comment.
  (#​189) by @​gaborbernat.
• An entry with a leading comment sorts by its value. It sorted by the comment above it, so a comment decided where
  the value went. (#​188) by
  @​gaborbernat.

v2.14.0

v2.13.0

Changes

• Strings prefer double quotes, and single quotes where that saves an escape.
  (#​162) by @​gaborbernat.
• An array without a trailing comma stays on one line when it fits; a trailing comma expands it.
  (#​164) by @​gaborbernat.
• Unhandled table keys sort alphabetically. (#​164) by
  @​gaborbernat.

Fixed

• expand_tables reaches a table written any number of levels down.
  (#​146, #​160) by
  @​gaborbernat.
• A comment written above a table header stays with that table.
  (#​158) by @​gaborbernat.
• An SPDX license expression reads by the operators the spec defines.
  (#​156) by @​alexfikl.
• Sorting an array keeps comments with their values. (#​164) by
  @​gaborbernat.

Documentation

• The formatting reference is rewritten, and it corrects the array behavior the old text described.
  (#​164) by @​gaborbernat.

v2.12.1

Fixed

• expand_tables reaches a named sub-table. A setting naming one matched nothing.
  (#​148) by @​gaborbernat.

v2.12.0

Added

• table_format, expand_tables and collapse_tables choose how tables are written. A table can fold into its
  parent as dotted keys or stay under a header of its own, per project or per table.
  (#​141, #​142) by
  @​gaborbernat.
• An extra name in optional-dependencies reads the way PEP 685 spells one.
  (#​135) by @​gaborbernat.
• An SPDX operator in a license expression takes the case the spec gives it.
  (#​136) by @​gaborbernat.
• Keywords and classifiers say each entry once. (#​133) by
  @​gaborbernat.
• A PEP 794 import-names entry reads the way the spec spells one.
  (#​131) by @​gaborbernat.
• [tool.nuitka] and [tool.typos] get a format of their own.
  (#​123) by @​kdeldycke.

Changes

• Sub-tables of one tool sort among themselves, so an unrelated tool cannot land between them.
  (#​139) by @​gaborbernat.
• [[project.authors]] folds into an inline array. (#​137) by
  @​gaborbernat.
• authors and maintainers sort. (#​134) by
  @​gaborbernat.

Fixed

• A comment written before a table header stays with the table it was written for.
  (#​140) by @​gaborbernat.
• A multi-line string using a line continuation no longer crashes the formatter.
  (#​132) by @​gaborbernat.
• A literal string holding what looks like an invalid escape survives. 'C:\path' came back rewritten.
  (#​130) by @​gaborbernat.
• requires-python with the ~= operator reads. (#​128) by
  @​gaborbernat.
• A build requirement survives where two entries name the same package. One of them was dropped.
  (#​127) by @​gaborbernat.

Documentation

• The docs spell out the formatting principles and every normalization, and README.rst comes from them so the two
  cannot drift. (#​144,
  #​145) by @​gaborbernat.

---

Configuration

📅 Schedule: Branch creation - At any time (no schedule defined), Automerge - At any time (no schedule defined).

🚦 Automerge: Disabled by config. Please merge this manually once you are satisfied.

♻ Rebasing: Whenever PR is behind base branch, or you tick the rebase/retry checkbox.

🔕 Ignore: Close this PR and you won't be reminded about this update again.

---

• If you want to rebase/retry this PR, check this box

---

This PR has been generated by Renovate Bot.

[codecov]

Codecov Report

✅ All modified and coverable lines are covered by tests.

📢 Thoughts on this report? Let us know!