pyproject.toml

[build-system]
requires = ["hatchling"]
build-backend = "hatchling.build"

[project]
name = "wrodfreq"
version = "0.2.0"
description = "Romanian word-frequency table on the Zipf scale — drop-in compatible with wordfreq"
readme = "README.md"
requires-python = ">=3.10"
authors = [{ name = "pax", email = "pax@mioritics.ro" }]
dependencies = [
    # The shipped package's only runtime dependency (spec §10.2: "msgpack +
    # xz" for the data file; xz/lzma is stdlib). wordfreq itself depends on
    # msgpack for the same purpose.
    "msgpack>=1.0",
]

[project.optional-dependencies]
# Third-party deps needed only by the build/ pipeline scripts, never at import
# time by the shipped wrodfreq package.
build = [
    "datasets>=2.19",
    "huggingface_hub>=0.23",
    # ingest_social.py only: the Reddit per-subreddit dumps are zstandard
    # ndjson, and they need max_window_size=2**31 to decode (see that
    # module's docstring) — the stdlib has no zstd decoder before 3.14.
    "zstandard>=0.22",
]
dev = [
    "pytest>=8",
]

[tool.hatch.build]
# Hatchling's default file selection is git-tracked files only, and the data
# files are deliberately gitignored (large binaries, rebuilt by
# build/build_package.py rather than committed — see .gitignore). This has
# to be at the top level, not just under targets.wheel: `uv build`/`pip
# wheel` build the wheel FROM the sdist, so if the sdist doesn't have the
# files either, there's nothing for the wheel step to include regardless of
# its own artifacts setting — confirmed by inspecting both archives after a
# real build, not assumed from the docs.
artifacts = ["wrodfreq/data/*.msgpack.xz"]

[tool.hatch.build.targets.wheel]
packages = ["wrodfreq"]

[tool.pytest.ini_options]
testpaths = ["tests"]