hatch-conda-build 0.1.2

A Hatch plugin to enable building a conda package from your pyproject.toml without having to write a separate conda build recipe meta.yaml file

hatch-conda-build

Conda package builder for Hatch. Hatch is modern, extensible Python project manager.

Usage

Add hatch-conda-build within the build-system.requires field in your pyproject.toml file.

[build-system]
requires = ["hatchling", "hatch-conda-build"]
build-backend = "hatchling.build"

Additionally conda-build must be in your current path when running a hatch build.

hatch-conda-build uses grayskull to translate the requirements section of your pyproject.toml, which specifies dependencies on PyPi, to appropriate conda package names.

Building Conda Package

The builder plugin defines the conda target.

To start build process, run hatch build -t conda:

$ hatch build -t conda
[conda]
...

Installing the local package

The ouptut package is written to the dist/conda directory. This directory is a fully-indexed conda channel and can be used to install the package.

$ conda install -c ./dist/conda <package-name>

Conda-build Options

Additional builder configuration can be set in the following TOML header.

[tool.hatch.build.targets.conda]
...

Following table contains available customization of builder behavior.

Option	Type	Default	Description
channels	list[str]	['conda-forge']	Channels used for package build and testing
default_numpy_version	str	None	numpy version, otherwise use conda-build default

Modifying the generated recipe

You can specify recipe metadata keys in following TOML header

[tool.hatch.build.targets.conda.recipe]
...

Recipe metadata is defined in TOML syntax. The metadata is merged using deepmerge with the following strategies

recipe_merger = Merger(
    type_strategies=[
        (dict, ["merge"]),
        (list, ["append"])
    ],
    fallback_strategies=["override"],
    type_conflict_strategies=["override"]
)

Most importantly, this will append list values if the key already exists. For example you can use nested key TOML syntax:

[tool.hatch.build.targets.conda.recipe]
requirements.run = ["anaconda-anon-usage"]
test.imports = ["my_package"]

will append the package name anaconda-anon-usage to the autogenerated run requirements and inject the test section into the recipe before it is built.

License

Plugin hatch-conda-build is distributed under the terms of the MIT license.