pyproject.toml

[project]
name = "nameparser"
description = "Parse human names into title, given, middle, family, suffix, nickname and maiden name. Typed, zero-dependency, CJK-aware."
readme = "README.rst"
requires-python = ">=3.11"
license = "LGPL-2.1-or-later"
license-files = ["LICENSE", "AUTHORS"]
authors = [{name = "Derek Gulbranson", email = "derek73@gmail.com"}]
dynamic = ["version"]
keywords = ["names", "parser", "name parsing", "human names", "nlp", "data cleaning", "record linkage", "cjk", "i18n"]
classifiers = [
  "Intended Audience :: Developers",
  "Operating System :: OS Independent",
  "Programming Language :: Python",
  "Programming Language :: Python :: 3",
  "Programming Language :: Python :: 3 :: Only",
  "Programming Language :: Python :: 3.11",
  "Programming Language :: Python :: 3.12",
  "Programming Language :: Python :: 3.13",
  "Programming Language :: Python :: 3.14",
  "Programming Language :: Python :: 3.15",
  "Development Status :: 5 - Production/Stable",
  "Natural Language :: English",
  "Topic :: Scientific/Engineering :: Information Analysis",
  "Topic :: Sociology :: Genealogy",
  "Topic :: Software Development :: Libraries :: Python Modules",
  "Topic :: Text Processing :: Linguistic",
  "Typing :: Typed",
]
dependencies = []

[project.urls]
Documentation = "https://nameparser.readthedocs.io/en/latest/"
Changelog = "https://nameparser.readthedocs.io/en/latest/release_log.html"
Source = "https://github.com/derek73/python-nameparser"
Issues = "https://github.com/derek73/python-nameparser/issues"

[project.optional-dependencies]
# Japanese kanji-name segmentation (#272), wrapped by
# locales.ja_segmenter(). Optional and never bundled: the core stays
# zero-dependency. 0.4 is the floor its BasicNameDivider/DividedName
# API dates from.
ja = ["namedivider-python>=0.4"]

[build-system]
requires = ["setuptools (>=82.0.1)"]
build-backend = "setuptools.build_meta"

[tool.setuptools.dynamic]
version = {attr = "nameparser._version.__version__"}

[tool.setuptools.package-data]
# The per-word vocabulary ships as data for tools and later phases to
# read; the parser itself reads the compiled module, never the TOML.
nameparser = ["py.typed", "vocab/*.toml"]

[dependency-groups]
dev = [
  "pytest (>=8)",
  "dill (>=0.2.5)",
  "sphinx (>=8)",
  "furo",
  "mypy (>=2.1)",
  "ruff (>=0.15)",
  "pytest-timeout>=2.4.0",
  "pytest-cov>=6",
  "pytest-xdist>=3.6",
  "hypothesis",
]

[tool.mypy]
packages = ["nameparser", "tests"]

[[tool.mypy.overrides]]
module = [
  "dill.*",
  # optional extras, unstubbed and absent from the dev environment:
  # the ja pack imports namedivider lazily, inside the factory
  "namedivider.*",
]
ignore_missing_imports = true

[[tool.mypy.overrides]]
module = [
  "nameparser._types",
  "nameparser._lexicon",
  "nameparser._policy",
  "nameparser._render",
  "nameparser._locale",
  "nameparser._pipeline.*",
  "nameparser._parser",
]
disallow_untyped_defs = true
disallow_incomplete_defs = true
disallow_any_generics = true
warn_return_any = true
strict_equality = true

# strict_equality for the WHOLE package rather than the core modules
# above. A comparison between non-overlapping types is always False and
# is always a bug, wherever it is written, and the block above does not
# name the package's __init__, _version, config.*, locales.* or
# __main__.
# Measured 2026-09-18: clean over nameparser/ as it then stood (the
# 2.x facade layer included), so it cost nothing and catches the next
# one. tests/ keep the default,
# where the check has real hits (a test comparing an off-contract value
# to a real one is doing its job).
[[tool.mypy.overrides]]
module = ["nameparser", "nameparser.*"]
strict_equality = true

[[tool.mypy.overrides]]
module = ["tests.v2.*"]
check_untyped_defs = true

[tool.pytest.ini_options]
testpaths = ["tests", "nameparser"]
# Parallel by default: the suite is CPU-bound and took 110s serial
# against 35s on 8 workers (measured 2026-10-23). worksteal because the
# property grids are a few long tests among 23,000 short ones, and
# fixed chunks left one worker running them alone. -n0 runs serially.
addopts = "--doctest-modules -n auto --dist worksteal"
markers = [
    # deselected in a parallel run by tests/conftest.py; run them with
    # `uv run pytest -n0 -m serial`
    "serial: measures the clock, so other workers' load can fail it",
]
python_classes = ["*Tests", "*TestCase"]
# Fail if an xfail-marked test unexpectedly passes, so a fixed/regressed
# expectation surfaces instead of silently turning into an xpass.
xfail_strict = true
# Unexpected warnings are failures: warnings are load-bearing API (the
# multi-word-entry UserWarning, the construction warnings), and without
# this a warning regression is invisible to the suite -- a real bug
# (eight spurious warnings on restoring a 1.4 Constants pickle) shipped
# to review this way. Tests that EXPECT a warning assert it with
# pytest.warns; targeted ignores below need a comment justifying each.
filterwarnings = [
    "error",
]

[tool.ruff.lint]
extend-select = [
  "ANN",   # flake8-annotations
  "UP",    # pyupgrade
  "B033",  # duplicate set items: a frozenset drops them silently, so a bulk vocabulary import can carry them unseen
]
ignore = [
  "E74",     # ambiguous-{variable,class,function}-name (I have trouble believing that these are real rules)
]