OCR-D · kba · Dec 8, 2022 · Nov 6, 2022 · Nov 8, 2022 · Nov 14, 2022
diff --git a/ocrd/ocrd/cli/zip.py b/ocrd/ocrd/cli/zip.py
@@ -110,3 +110,22 @@ def validate(src, **kwargs):
     print(report)
     if not report.is_valid:
         sys.exit(1)
+
+# ----------------------------------------------------------------------
+# ocrd zip update
+# ----------------------------------------------------------------------
+
+@zip_cli.command('update')
+@click.argument('src', type=click.Path(dir_okay=True, readable=True, resolve_path=True), required=True)
+@click.argument('dest', type=click.Path(dir_okay=True, readable=True, writable=True, resolve_path=True), required=False)
+@click.option('-o', '--overwrite', help="overwrite bag in SRC", is_flag=True)
+def update(src, dest=None, overwrite=False):
+    """
+    Recreate files containing checksums (manifest-sha512.txt, tagmanifest-sha512.txt and
+    'Payload-Oxum' contained in bag-info.txt) of an OCRD-ZIP.
+
+    Open the bag (zip file or directory) ``src``, create or update its manifests/checksums and
+    output to (zip file or directory) ``dest``. It is also possible to output to ``src`` / overwrite
+    ``src`` in place when ``--overwrite``-flag is given.
+    """
+    WorkspaceBagger(Resolver()).recreate_checksums(src, dest=dest, overwrite=overwrite)
diff --git a/ocrd/ocrd/workspace_bagger.py b/ocrd/ocrd/workspace_bagger.py
@@ -1,12 +1,14 @@
 from datetime import datetime
 from os import makedirs, chdir, walk
 from os.path import join, isdir, basename, exists, relpath
+from pathlib import Path
 from shutil import make_archive, rmtree, copyfile, move
-from tempfile import mkdtemp
+from tempfile import mkdtemp, TemporaryDirectory
 import re
 import tempfile
 import sys
-from bagit import Bag, make_manifests, _load_tag_file  # pylint: disable=no-name-in-module
+from bagit import Bag, make_manifests, _load_tag_file, _make_tag_file, _make_tagmanifest_file  # pylint: disable=no-name-in-module
+from distutils.dir_util import copy_tree
 
 from ocrd_utils import (
     pushd_popd,
@@ -270,3 +272,56 @@ def validate(self, bag):
             - https://ocr-d.github.io/bagit-profile.json
             - https://ocr-d.github.io/bagit-profile.yml
         """
+        pass
+
+    def recreate_checksums(self, src, dest=None, overwrite=False):
+        """
+        (Re)creates the files containing the checksums of a bag
+
+        This function uses bag.py to create new files: manifest-sha512.txt and
+        tagminifest-sha512.txt for the bag. Also 'Payload-Oxum' in bag-info.txt will be set to the
+        appropriate value.
+
+        Arguments:
+            src (string):    Path to Bag. May be an zipped or unziped bagit
+            dest (string):   Path to where the result should be stored. Not needed if overwrite is
+                             set
+            overwrite(bool): Replace bag with newly created bag
+        """
+        if overwrite and dest:
+            raise Exception("Setting 'dest' and 'overwrite' is a contradiction")
+        if not overwrite and not dest:
+            raise Exception("For checksum recreation 'dest' must be provided")
+        src_path = Path(src)
+        if not src_path.exists():
+            raise Exception("Path to bag not existing")
+        is_zipped = src_path.is_file()
+
+        with TemporaryDirectory() as tempdir:
+            if is_zipped:
+                unzip_file_to_dir(src, tempdir)
+                path_to_bag = Path(tempdir)
+                if not path_to_bag.joinpath("data").exists():
+                    raise FileNotFoundError("data directory of bag not found")
+            else:
+                path_to_bag = src_path if overwrite else Path(dest)
+                if not src_path.joinpath("data").exists():
+                    raise FileNotFoundError(f"data directory of bag not found at {src}")
+                if not overwrite:
+                    path_to_bag.mkdir(parents=True, exist_ok=True)
+                    copy_tree(src, dest)
+
+            with pushd_popd(path_to_bag):
+                n_bytes, n_files = make_manifests("data", 1, ["sha512"])
+
+                bag_infos = _load_tag_file("bag-info.txt")
+                bag_infos["Payload-Oxum"] = f"{n_bytes}.{n_files}"
+                _make_tag_file("bag-info.txt", bag_infos)
+                _make_tagmanifest_file("sha512", ".")
+
+            if is_zipped:
+                name = src_path.name
+                if name.endswith(".zip"):
+                    name = name[:-4]
+                zip_path = make_archive(name, "zip", path_to_bag)
+                move(zip_path, src if overwrite else dest)
diff --git a/tests/validator/test_workspace_bagger.py b/tests/validator/test_workspace_bagger.py
@@ -1,14 +1,15 @@
 from os import makedirs
 from os.path import join, abspath, exists
-from shutil import copytree, rmtree, move
+from shutil import copytree, rmtree, move, make_archive
 from tempfile import mkdtemp
-from bagit import _load_tag_file
+from bagit import _load_tag_file, Bag
 
 from tests.base import TestCase, main, assets # pylint: disable=import-error,no-name-in-module
 
 from ocrd.workspace import Workspace
 from ocrd.workspace_bagger import WorkspaceBagger, BACKUPDIR
 from ocrd.resolver import Resolver
+from ocrd_utils import unzip_file_to_dir
 
 README_FILE = abspath('README.md')
 
@@ -141,6 +142,64 @@ def test_spill_with_changed_metsname(self):
         self.assertTrue(exists(join(spill_dest, new_metsname)),
                         "expected mets-file to be '{new_metsname}'")
 
+    def test_recreate_checksums_param_validation(self):
+        with self.assertRaisesRegex(Exception, "For checksum recreation 'dest' must be provided"):
+            self.bagger.recreate_checksums("src/path")
+        with self.assertRaisesRegex(Exception, "Setting 'dest' and 'overwrite' is a contradiction"):
+            self.bagger.recreate_checksums("src/path", "dest/path", overwrite=True)
+
+    def test_recreate_checksums_overwrite_unzipped(self):
+        # arrange
+        assert Bag(self.bagdir).is_valid(), "tests arrangements for recreate_checksums failed"
+        move(join(self.bagdir, "data", "mets.xml"), join(self.bagdir, "data", "mets-neu.xml"))
+        assert not Bag(self.bagdir).is_valid(), "tests arrangements for recreate_checksums failed"
+
+        # act
+        self.bagger.recreate_checksums(self.bagdir, overwrite=True)
+
+        # assert
+        assert Bag(self.bagdir).is_valid(), "recreate_checksums unzippd with overwrite failed"
+
+    def test_recreate_checksums_unzipped(self):
+        # arrange
+        move(join(self.bagdir, "data", "mets.xml"), join(self.bagdir, "data", "mets-neu.xml"))
+        new_bag = join(self.tempdir, "new_bag")
+
+        # act
+        self.bagger.recreate_checksums(self.bagdir, new_bag)
+
+        # assert
+        assert Bag(new_bag).is_valid(), "recreate_checksums unzipped failed"
+
+    def test_recreate_checksums_zipped_overwrite(self):
+        # arrange
+        move(join(self.bagdir, "data", "mets.xml"), join(self.bagdir, "data", "mets-neu.xml"))
+        zipped_bag = join(self.tempdir, "foo.ocrd.zip")
+        make_archive(zipped_bag.replace('.zip', ''), 'zip', self.bagdir)
+
+        # act
+        self.bagger.recreate_checksums(zipped_bag, overwrite=True)
+
+        # assert
+        bag_dest = join(self.tempdir, "new_bag")
+        unzip_file_to_dir(zipped_bag, bag_dest)
+        assert Bag(bag_dest).is_valid(), "recreate_checksums zipped with overwrite failed"
+
+    def test_recreate_checksums_zipped(self):
+        # arrange
+        move(join(self.bagdir, "data", "mets.xml"), join(self.bagdir, "data", "mets-neu.xml"))
+        zipped_bag = join(self.tempdir, "foo.ocrd.zip")
+        make_archive(zipped_bag.replace('.zip', ''), 'zip', self.bagdir)
+        zipped_bag_dest = join(self.tempdir, "foo-new.ocrd.zip")
+
+        # act
+        self.bagger.recreate_checksums(zipped_bag, zipped_bag_dest)
+
+        # assert
+        bag_dest = join(self.tempdir, "new_bag")
+        unzip_file_to_dir(zipped_bag_dest, bag_dest)
+        assert Bag(bag_dest).is_valid(), "recreate_checksums zipped failed"
+
 
 if __name__ == '__main__':
     main()