hatch-cythonize
| CI/CD | |
| Package | |
| Meta |
This is a fork of the nice hatch-cython project, aimed at providing continued maintenance and updates since the original is currently inactive.
Table of Contents
Usage
The build hook name is cython.
- pyproject.toml
# [tool.hatch.build.hooks.cython]
# or
# (hatch.toml)
# [build.targets.wheel.hooks.cython]
# or
[tool.hatch.build.targets.wheel.hooks.cython]
dependencies = ["hatch-cythonize"]
# [tool.hatch.build.hooks.cython.options]
# or
# (hatch.toml)
# [build.targets.wheel.hooks.cython.options]
# or
[tool.hatch.build.targets.wheel.hooks.cython.options]
# include .h or .cpp directories
includes = []
# include numpy headers
include_numpy = false
include_pyarrow = false
# include_{custom}
include_somelib = {
# must be included in build-dependencies
pkg = "somelib",
# somelib.gets_include() -> str
include = "gets_include",
# somelib.gets_libraries() -> list[str]
libraries = "gets_libraries",
# somelib.gets_library_dirs() -> list[str]
library_dirs = "gets_library_dirs",
# somelib.some_setup_op() before build
required_call = "some_setup_op"
}
compile_args = [
# single string
"-v",
# by platform
{ platforms = ["linux", "darwin"], arg = "-Wcpp" },
# by platform & arch
{ platforms = "darwin", arch = "x86_64", arg = "-arch x86_64" },
{ platforms = ["darwin"], arch = "arm64", arg = "-arch arm64" },
# with pep508 markers
{ platforms = ["darwin"], arch = "x86_64", arg = "-I/usr/local/opt/llvm/include", depends_path = true, marker = "python_version <= '3.10'" },
]
directives = { boundscheck = false, nonecheck = false, language_level = 3, binding = true }
compile_kwargs = { }
- hatch.toml
# [build.hooks.cython]
# or
[build.targets.wheel.hooks.cython]
dependencies = ["hatch-cythonize"]
# [build.hooks.cython.options]
# or
[build.targets.wheel.hooks.cython.options]
directives = { boundscheck = false, nonecheck = false, language_level = 3, binding = true }
compile_args = [
"-O3",
]
includes = []
include_numpy = false
# equivalent to include_pyarrow = true
include_somelib = { pkg = "pyarrow", include="get_include", libraries="get_libraries", library_dirs="get_library_dirs", required_call="create_library_symlinks" }
define_macros = [
# ["ABC"] -> ["ABC", "FOO"] | ["ABC", "DEF"]
["NPY_NO_DEPRECATED_API", "NPY_1_7_API_VERSION"],
]
Configuration Options
| Field | Type |
|---|---|
| src | str | None directory within src dir or . which aliases the package being built. e.g. package_a -> src/package_a_lib src = "package_a" |
| directives | directives to cython (see compiler-directives) |
| compile_args | str or { platforms = ["*"] | "*", arg = str }. see extensions for what args may be relevant |
| extra_link_args | str or { platforms = ["*"] | "*", arg = str }. see extensions for what args may be relevant |
| env | { env = "VAR1", arg = "VALUE", platforms = ["*"], arch = ["*"] }if flag is one of: - ARFLAGS - LDSHARED - LDFLAGS - CPPFLAGS - CFLAGS - CCSHARED the current env vars will be merged with the value (provided platform & arch applies), separated by a space. This can be enabled by adding { env = "MYVAR" ... , merges = true } to the definition. |
| includes | list str |
| include_{package} | { pkg = str, include = str, libraries = str| None, library_dirs = str | None , required_call = str | None } where all fields, but pkg, are attributes of pkg in the type of callable() -> list[str] | str | list[str] | str. pkg is a module, or loadable module object, which may be imported through import x.y.z. |
| include_numpy | include_pyarrow | include_pythran | bool 3rd party named imports. must have the respective opt in dependencies |
| parallel | bool = False if parallel, add openmp headers important: if using macos, you need the homebrew llvm vs apple's llvm in order to pass -fopenmp to clang compiler |
| compiler | compiler used at build-time. if msvc (Microsoft Visual Studio), /openmp is used as argument to compile instead of -fopenmp when parallel = true. default = false |
| compile_py | whether to include .py files when building cython exts. note, this can be enabled & you can do per file / matched file ignores as below. default = true |
| define_macros | list of list str (of len 1 or 2). len 1 == [KEY] == #define KEY FOO . len 2 == [KEY, VALUE] == #define KEY VALUE. see extensions |
| ** kwargs | keyword = value pair arguments to pass to the extension module when building. see extensions |
Files
[build.targets.wheel.hooks.cython.options.files]
exclude = [
# anything matching no_compile is ignored by cython
"*/no_compile/*",
# note - anything "*" is escaped to "([^\s]*)" (non whitespace).
# if you need an actual * for python regex, use as below:
# this excludes all pyd or pytempl extensions
"([^.]\\*).(pyd$|pytempl$)",
# only windows
{ matches = "*/windows", platforms = ["linux", "darwin", "freebsd"] },
# only darwin
{ matches = "*/darwin", platforms = ["linux", "freebsd", "windows"] },
# only linux
{ matches = "*/linux", platforms = ["darwin", "freebsd", "windows"] },
# only freebsd
{ matches = "*/freebsd", platforms = ["linux", "darwin", "windows"] }
]
aliases = {"abclib._filewithoutsuffix" = "abclib.importalias"}
Explicit Build Targets
If explicit targets are required (i.e. hatch-cythonize only builds the files specified), use options.files.targets. Specifying this option will implicly enable compile_py, in addition to checking all c, cpp, and cc files against the specified inclusions.
[build.targets.wheel.hooks.cython.options.files]
targets = [
# using a single string
"*/compile.py",
# or a match clause
{ matches = "*/windows", platforms = ["windows"] },
{ matches = "*/posix", platforms = ["darwin", "freebsd", "linux"] },
]
sdist
Sdist archives may be generated normally. hatch must be defined as the build-system build-backend in pyproject.toml. As such, hatch will automatically install hatch-cythonize, and perform the specified e.g. platform-specific adjustments to the compile-time arguments. This allows the full build-process to be respected, and generated following specifications of the developer.Note: If hatch-cythonize is specified to run outside of a wheel-step processes, the extension module is skipped. As such, the .c & .cpp, as well as templated files, may be generated and stored in the sdist should you wish. However, there is currently little purpose to this, as the extension will likely have differed compile arguments.
Templating
Cython tempita is supported for any files suffixed with .in, where the extension output is:
.pyx.in.pyd.in.pyi.inFor these files, expect the output.pyx.in->.pyx. Thus, with aliasing this would look like:
[build.targets.wheel.hooks.cython.options.files]
aliases = {"abclib._somemod" = "abclib.somemod"}
-
- Source files
somemod.pyi.in,_somemod.pyx.in
- Source files
-
- Processed templates
somemod.pyi,_somemod.pyx
- Processed templates
-
- Compiled module
abclib.somemod{.pyi,.pyx}
- Compiled module
An example of this is included in:
Template Arguments
Per-file matched namespaces are supported for templating. This follows the above platforms, arch, & marker formats, where if supplied & passing the condition the argument is passed to the template as a named series of keyword arguments.
If an index value is provided, and all other kwargs to templates are keywords for each index value. Follows FIFO priority for all keys except global, which is evaluated first and overriden if there are other matching index directives. The engine will attempt to merge the items of the keywords, roughly following:
args = {
"index": [
{"keyword": "global", ...},
{"keyword": "thisenv", ...},
],
"global": {"abc": 1, "other": 2},
"thisenv": {"other": 3},
}
merge(args) -> {"abc": 1, "other": 3}
In hatch.toml:
[build.targets.wheel.hooks.cython.options.templates]
index = [
{keyword = "global", matches = "*" },
{keyword = "templated_mac", matches = "templated.*.in", platforms = ["darwin"] },
{keyword = "templated_mac_py38", matches = "templated.*.in", platforms = ["darwin"], marker = "python == '3.8'" },
{keyword = "templated_win", matches = "templated.*.in", platforms = ["windows"] },
{keyword = "templated_win_x86_64", matches = "templated.*.in", platforms = ["windows"], arch = ["x86_64"] },
]
# these are passed as arguments for templating
# 'global' is a special directive reserved & overriden by all other matched values
global = { supported = ["int"] }
templated_mac = { supported = ["int", "float"] }
templated_mac_py38 = { supported = ["int", "float"] }
templated_win = { supported = ["int", "float", "complex"] }
# assuming numpy is cimported in the template
templated_win_x86_64 = { supported = ["int", "float", "np.double"]}
Notes
-
MacOS users with brew installed will have
brew --prefixlibs and include paths added in compilation step. Code parsing is found here -
Github Runners now run MacOS on m1 platforms. You may have ci issues if you are using MacOS m1 runners and you do not disable
macos-max-compatin hatch. e.g.# hatch.toml [build.targets.wheel] macos-max-compat = false
Development
Requirements
Scripts
- test: library & coverage
hatch run cov
- test: src structure example
task example
- test: simple structure example
task simple-structure
- commit: precommitt
task precommit
License
hatch-cythonize is distributed under the terms of the MIT license.
Metadata
Release files for hatch-cythonize 0.7.0
For a detailed explanation of source distributions (sdists) and built distributions (wheels), please see the package formats documentation.
Source distribution (sdist)
| File | Size | Uploaded | Provenance | Actions |
|---|---|---|---|---|
| hatch_cythonize-0.7.0.tar.gz | 38.3 kB | With provenance | Details for hatch_cythonize-0.7.0.tar.gz |
Built distribution (wheel)
| File | Interpreter | ABI | Platform | Actions Reset |
|---|---|---|---|---|
| hatch_cythonize-0.7.0-py3-none-any.whl | Python 3 | none | any | Details for hatch_cythonize-0.7.0-py3-none-any.whl |
Total release size: 62.7 kB
Release files / hatch_cythonize-0.7.0.tar.gz
| Download URL | hatch_cythonize-0.7.0.tar.gz |
|---|---|
| Size | 38.3 kB |
| Tags | Source |
|
SHA-256 checksum How to use checksums |
bd0f12f79798ccd45681ac93b197cf56b3514ea5a89766a71571f058aec7b7ac
|
|
BLAKE2b-256 checksum How to use checksums |
922c81b8a0aa4114375f6e86405f97dc2cf1e8e152b892703ab2e3038bd46032
|
| Upload date | |
|
Uploaded using Trusted Publishing? What is trusted publishing? |
Yes |
| Uploaded via |
twine/6.1.0 CPython/3.12.9
|
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 May 28, 2025.
Transparency logRelease files / hatch_cythonize-0.7.0-py3-none-any.whl
| Download URL | hatch_cythonize-0.7.0-py3-none-any.whl |
|---|---|
| Size | 24.4 kB |
| Tags | Python 3 |
|
SHA-256 checksum How to use checksums |
558564ff72419738292ed2fe50f9188f4a376d7fc6cfa65bf855daf231d4c005
|
|
BLAKE2b-256 checksum How to use checksums |
ee7dc0e1c41db52bb04872d9341e5a73884ef56894f0000debbd28898dd630ae
|
| Upload date | |
|
Uploaded using Trusted Publishing? What is trusted publishing? |
Yes |
| Uploaded via |
twine/6.1.0 CPython/3.12.9
|
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 May 28, 2025.
Transparency logSecurity for hatch-cythonize 0.7.0
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.
Full provenance
All 2 files in this release have been uploaded with provenance metadata.
Attestation for hatch_cythonize-0.7.0.tar.gz
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 May 28, 2025.
Transparency logA note on package security
PyPI is a repository for distributing packages, not a service for auditing them. While third-party security researchers and our staff do their best to detect and remove malicious packages, PyPI cannot guarantee the quality, security or suitability of any given package.
To learn more about protecting yourself from malicious actors, visit our package security documentation.