xml/setup.py

from setuptools import setup
from setuptools.command.build_py import build_py

import os
import subprocess

# Typically in files like this we'd use find_packages() to traverse directories
# for any static packages. However, we are trying to add data to a package that
# will actually exist in another repository. Therefore, we have to explicitly
# list out the package name, directory, and data information.

# We are building data for the following module:
package_name = 'pel.hwdiags'

# Since we are not using find_packages() we have to provide a package directory,
# but in this case nothing exists because there are no static package
# directories. Therefore, we will just use the empty string.
package_dir = ''

# Split the package data directory into its components.
data_dir_components = [ *package_name.split('.'), 'data' ]

# It is important to note that '/' must be used as the path separator, even on
# Windows. Setuptools will automatically convert the slashes where appropriate.
package_data_glob = '/'.join( data_dir_components )

# This is a custom build class that is used to dynamically build the data files.
class my_build_py(build_py):
    def run(self):
        if not self.dry_run: # honor --dry-run flag

            # Make sure the build directory for the data exists.
            # Yes, os.path.join() is necessary in this case, which is different
            # that what is stated above regarding package_data_glob.
            data_dir = os.path.join(self.build_lib, *data_dir_components)
            self.mkpath(data_dir)

            # Generate the PEL parser data JSON from the Chip Data XML.
            # TODO: The current tool to build the PEL parser JSON files is a
            #       perl script. Eventually, it will be converted to a python
            #       module.
            # TODO: The list of data file directories will need to be
            #       configurable via the package config in the bitbake recipes.
            for chip in ('p10', 'explorer'):
                subprocess.run([ './parse_chip_data_xml', '--json',
                                 '-i', chip, '-o', data_dir ])

        # Call the superclass run() to ensure everything else builds.
        super().run()


setup(
    name         = 'openpower-hw-diags-pel-parser-data',
    version      = '0.1',
    classifiers  = [ 'License :: OSI Approved :: Apache Software License' ],
    cmdclass     = { 'build_py': my_build_py }, # register custom build class
    packages     = [ package_name ],
    package_dir  = { package_name: package_dir },
    package_data = { package_name: [ package_data_glob ] },
)