pyproject-metadata

Dataclass for pyproject.toml [project] metadata with support for [core > metadata] generation

This project does not implement the parsing of pyproject.toml containing pyproject.toml [project] metadata.

Instead, given a Python data structure representing pyproject.toml [project] metadata (already parsed), it will validate this input and generate a PEP 643-compliant metadata file (e.g. PKG-INFO).

Usage

After installing pyproject-metadata, you can use it as a library in your scripts and programs:

from pyproject_metadata import StandardMetadata

parsed_pyproject = {...}  # you can use parsers like `tomli` to obtain this dict
metadata = StandardMetadata.from_pyproject(parsed_pyproject, allow_extra_keys=False)
# Same fields as defined in pyproject.toml [project] metadata
print(metadata.entrypoints)

pkg_info = metadata.as_rfc822()
print(str(pkg_info))  # core metadata

SPDX licenses (METADATA 2.4+)

If project.license is a string or project.license-files is present, then METADATA 2.4+ will be used. A user is expected to validate and normalize metadata.license with an SPDX validation tool, such as the one being added to packaging. Add something like this (requires packaging 24.2+):

if isinstance(metadata.license, str):
    metadata.license = packaging.licenses.canonicalize_license_expression(
        metadata.license
    )

A backend is also expected to copy entries from project.license_files, which are paths relative to the project directory, into the dist-info/licenses folder, preserving the original source structure.

Dynamic Metadata (METADATA 2.2+)

Pyproject-metadata supports dynamic metadata. To use it, specify your METADATA fields in dynamic_metadata. If you want to convert pyproject.toml field names to METADATA field(s), use pyproject_metadata.pyproject_to_metadata("field-name"), which will return a frozenset of metadata names that are touched by that field.

Adding extra fields

You can add extra fields to the Message returned by to_rfc822(), as long as they are valid metadata entries.

Collecting multiple errors

You can use the all_errors argument to from_pyproject to show all errors in the metadata parse at once, instead of raising an exception on the first one. The exception type will be pyproject_metadata.errors.ExceptionGroup (which is just ExceptionGroup on Python 3.11+).

Validating extra fields

By default, a warning (pyproject_metadata.errors.ExtraKeyWarning) will be issued for extra fields at the project table. You can pass allow_extra_keys= to either avoid the check (True) or hard error (False). If you want to detect extra keys, you can get them with pyproject_metadata.extra_top_level and pyproject_metadata.extra_build_system. It is recommended that build systems only warn on failures with these extra keys.

Validating classifiers

If you want to validate classifiers, then install the trove_classifiers library (the canonical source for classifiers), and run:

import trove_classifiers

metadata_classifieres = {
    c for c in metadata.classifiers if not c.startswith("Private ::")
}
invalid_classifiers = set(metadata.classifiers) - trove_classifiers.classifiers

# Also the deprecated dict if you want it
dep_names = set(metadata.classifiers) & set(trove_classifiers.deprecated_classifiers)
deprecated_classifiers = {
    k: trove_classifiers.deprecated_classifiers[k] for k in dep_names
}

If you are writing a build backend, you should not validate classifiers with a Private :: prefix; these are only restricted for upload to PyPI (such as Private :: Do Not Upload).

Since classifiers are a moving target, it is probably best for build backends (which may be shipped by third party distributors like Debian or Fedora) to either ignore or have optional classifier validation.