Unpack a docker image.

docker-unpack.py

#!/usr/bin/env python
import argparse
import json
import os
import sys
import tarfile

parser = argparse.ArgumentParser()
parser.add_argument(
    '--image', '-i', required=True,
    help='Read from file, instead of STDIN')
parser.add_argument(
    '--destination', '-d', default='chroots',
    help='Directory to save extracted images')


def main(tf, dest):
    manifests = json.loads(tf.extractfile('manifest.json').read().decode('utf-8'))

    for manifest in manifests:
        layers = manifest['Layers']

        repotag, *othertags = manifest["RepoTags"]
        print("Found image:", repotag)
        for othertag in othertags:
            print("        aka:", othertag)

        path = os.path.join(dest, repotag.replace('/', '-').replace(':', '-'))
        def filter_delete(m):
            if '/' in m.name:
                dirname, basename = m.name.rsplit('/', 1)
            else:
                dirname, basename = '', m.name

            # Remove files that have been deleted by this layer
            if basename.startswith('.wh.') and m.mode == 0:
                os.unlink(os.path.join(path, dirname, basename[4:]))
                return False
            return True
                
        def no_specials(m):
            if m.isdir() or m.isfile() or m.islnk() or m.issym():
                return True
            return False

        class mfilter:
            checks = [filter_delete]

            def __init__(self, members):
                self.members = iter(members)

            def __iter__(self):
                return self

            def __next__(self):
                while True:
                    m = next(self.members)
                    for check in self.checks:
                        if not check(m):
                            break
                    else:
                        return m

        def extract_layer(layer):
            with tarfile.open(fileobj=tf.extractfile(layer)) as lyr:
                try:
                    lyr.extractall(path=path, members=mfilter(lyr))
                except PermissionError:
                    mfilter.checks.append(no_specials)
                    lyr.extractall(path=path, members=mfilter(lyr))

        for layer in layers:
            print("-----> Extracting layer:", layer)
            extract_layer(layer)
        print("Extracted `{}` to `{}`".format(repotag, path))


if __name__ == '__main__':
    args = parser.parse_args()
    with tarfile.open(args.image) as tf:
        main(tf, args.destination)