Does python expose an interface for parsing package metadata?

Question

Note: I am NOT looking for a way to interrogate installed packages in the virtual environment and do not want to load packages in order to parse them.

I'm looking for a way to parse meta data from files [wheels and tar.gz] downloaded from pypi. The format isn't so complex so I'm not really against writing my own. But given that python must be parsing this info inside importlib I wanted to see if theres a way that doesn't involve re-inventing the wheel.

I can see that importlib.metadata lets me get this information for installed pacakges. But I want to do this for a lot of versions of the same packages; many of which may not be compatible with the current virtual environment or even system architecture.

Does python (>=3.10) offer any interface for parsing wheel and tar.gz metadata without actually installing it into the virtual environment?

Answer 1

This code searches for the METADATA file within the .whl file using the zipfile built-in library. The metadata is then parsed using email.parser to obtain a Message object, which can be accessed as a dictionary.

import zipfile
import email.parser

def get_metadata(path:str) -> str:
    # Open the file using zipfile
    with zipfile.ZipFile(path) as archive:
        # Find and read the METADATA file from the archive
        metadata_path = [file.filename for file in archive.filelist if "METADATA" in file.filename][0]
        metadata = archive.read(metadata_path).decode("utf-8")

    return email.parser.Parser().parsestr(metadata)


# Path to the wheel or tar.gz file
file_path = "numpy-1.25.2.whl"
METADATA = get_metadata(file_path)
print("Name: ", METADATA["name"])     
print("Version: ", METADATA["version"])   
print("Summary: ", METADATA["summary"])

Output:

Name:  numpy
Version:  1.25.2
Summary:  Fundamental package for array computing in Python