Transfer to fortran-lang?

[awvwgk]

See discussion at fortran-lang/fpm#748

This project could be moved to @fortran-lang to officially provide a schema for the fpm package manifest. Alternatively, the files could be checked in as part of another project, e.g. the upstream fpm repository. Maintaining them here is of course also an option.

cc @milancurcic @certik @LKedward @gnikit

[gnikit]

I think it would make things a bit easier to maintain if it was part of fortran-lang

[milancurcic]

Thanks both. This seems like a useful package and fine with me to maintain under fortran-lang.

Consider a rename: pyfpm suggests to me a Python binding to fpm. Perhaps a more descriptive name like fpm-package-schema or fpm-package-model?

[awvwgk]

Maybe fpm-metadata, similar to the former pep621 project which is now named python-pyproject-metadata.

[certik]

Fine to move. Is it because Python is easier than Fortran? :)

[awvwgk]

Is it because Python is easier than Fortran?

The original idea was to provide Python projects a possibility to access the fpm package manifest. Candidates would be fprettify or ford should they decide to store their configuration in fpm.toml.

Once we decide on how to implement a schema library in Fortran the export of the schema can be provided directly by fpm (see discussion in https://fortran-lang.discourse.group/t/3963). Now that TOML Fortran has a JSON frontend there is the possibility to implement a schema library which is immediately usable by fpm.

In fact this would be preferable because the schema produced by the pydantic model is imperfect and does not 100% reflect the currently allowed rules in the package manifest (the biggest issue is the equivalence between underscored and dashed keywords imposed by pydantic):

>>> from fpm.metadata import load_manifest
>>> load_manifest("name = 'hello'\nbuild.auto_executables = true")
Manifest(name='hello', version='0', license=None, maintainer=[], author=[], copyright=None, description=None, categories=[], keywords=[], executable=[], example=[], test=[], dependencies={}, dev_dependencies={}, extra={}, build=Build(auto_tests=True, auto_executables=True, auto_examples=True, link=[], external_modules=[]), install=Install(library=False), library=Library(source_dir='src', include_dir='include'))

[gnikit]

FYI I am happy with the schema generator to be in any language we find easiest to maintain, the only important thing for vscode is the final product i.e. the JSON file. I will create a pre-release of the extension using the schema for fpm validation and see how that goes.

[certik]

No, I like Python, especially with LPython now. :)

[awvwgk]

Done

[awvwgk]

If anyone has a PyPI account and needs access to the project, let me know. I also made sure (with some trial and error), that tagging a new release will automatically upload release artifacts for GitHub and PyPI.

[gnikit]

Do you think we can add fortran-lang's PyPi account as a maintainer?