bagit.py

#!/usr/bin/env python
# -*- coding: utf-8 -*-

from __future__ import absolute_import, division, print_function, unicode_literals

import argparse
import codecs
import gettext
import hashlib
import logging
import multiprocessing
import os
import re
import signal
import sys
import tempfile
import unicodedata
import warnings
from collections import defaultdict
from datetime import date
from functools import partial
from os.path import abspath, isdir, isfile, join

from pkg_resources import DistributionNotFound, get_distribution

try:
    from urllib.parse import urlparse
except ImportError:
    from urlparse import urlparse


def find_locale_dir():
    for prefix in (os.path.dirname(__file__), sys.prefix):
        locale_dir = os.path.join(prefix, "locale")
        if os.path.isdir(locale_dir):
            return locale_dir


TRANSLATION_CATALOG = gettext.translation(
    "bagit-python", localedir=find_locale_dir(), fallback=True
)
if sys.version_info < (3,):
    _ = TRANSLATION_CATALOG.ugettext
else:
    _ = TRANSLATION_CATALOG.gettext

MODULE_NAME = "bagit" if __name__ == "__main__" else __name__

LOGGER = logging.getLogger(MODULE_NAME)

try:
    VERSION = get_distribution(MODULE_NAME).version
except DistributionNotFound:
    VERSION = "0.0.dev0"

PROJECT_URL = "https://github.com/LibraryOfCongress/bagit-python"

__doc__ = (
    _(
        """
BagIt is a directory, filename convention for bundling an arbitrary set of
files with a manifest, checksums, and additional metadata. More about BagIt
can be found at:

    http://purl.org/net/bagit

bagit.py is a pure python drop in library and command line tool for creating,
and working with BagIt directories.


Command-Line Usage:

Basic usage is to give bagit.py a directory to bag up:

    $ bagit.py my_directory

This does a bag-in-place operation where the current contents will be moved
into the appropriate BagIt structure and the metadata files will be created.

You can bag multiple directories if you wish:

    $ bagit.py directory1 directory2

Optionally you can provide metadata which will be stored in bag-info.txt:

    $ bagit.py --source-organization "Library of Congress" directory

You can also select which manifest algorithms will be used:

    $ bagit.py --sha1 --md5 --sha256 --sha512 directory


Using BagIt from your Python code:

    import bagit
    bag = bagit.make_bag('example-directory', {'Contact-Name': 'Ed Summers'})
    print(bag.entries)

For more information or to contribute to bagit-python's development, please
visit %(PROJECT_URL)s
"""
    )
    % globals()
)

# standard bag-info.txt metadata
STANDARD_BAG_INFO_HEADERS = [
    "Source-Organization",
    "Organization-Address",
    "Contact-Name",
    "Contact-Phone",
    "Contact-Email",
    "External-Description",
    "External-Identifier",
    "Bag-Size",
    "Bag-Group-Identifier",
    "Bag-Count",
    "Internal-Sender-Identifier",
    "Internal-Sender-Description",
    "BagIt-Profile-Identifier",
    # Bagging-Date is autogenerated
    # Payload-Oxum is autogenerated
]

CHECKSUM_ALGOS = hashlib.algorithms_guaranteed
DEFAULT_CHECKSUMS = ["sha256", "sha512"]

#: Block size used when reading files for hashing:
HASH_BLOCK_SIZE = 512 * 1024

#: Convenience function used everywhere we want to open a file to read text
#: rather than undecoded bytes:
open_text_file = partial(codecs.open, encoding="utf-8", errors="strict")

# This is the same as decoding the byte values in codecs.BOM:
UNICODE_BYTE_ORDER_MARK = "\uFEFF"


def make_bag(
    bag_dir, bag_info=None, processes=1, checksums=None, checksum=None, encoding="utf-8"
):
    """
    Convert a given directory into a bag. You can pass in arbitrary
    key/value pairs to put into the bag-info.txt metadata file as
    the bag_info dictionary.
    """

    if checksum is not None:
        warnings.warn(
            _(
                "The `checksum` argument for `make_bag` should be replaced with `checksums`"
            ),
            DeprecationWarning,
        )
        checksums = checksum

    if checksums is None:
        checksums = DEFAULT_CHECKSUMS

    bag_dir = os.path.abspath(bag_dir)
    cwd = os.path.abspath(os.path.curdir)

    if cwd.startswith(bag_dir) and cwd != bag_dir:
        raise RuntimeError(
            _("Bagging a parent of the current directory is not supported")
        )

    LOGGER.info(_("Creating bag for directory %s"), bag_dir)

    if not os.path.isdir(bag_dir):
        LOGGER.error(_("Bag directory %s does not exist"), bag_dir)
        raise RuntimeError(_("Bag directory %s does not exist") % bag_dir)

    # FIXME: we should do the permissions checks before changing directories
    old_dir = os.path.abspath(os.path.curdir)

    try:
        # TODO: These two checks are currently redundant since an unreadable directory will also
        #       often be unwritable, and this code will require review when we add the option to
        #       bag to a destination other than the source. It would be nice if we could avoid
        #       walking the directory tree more than once even if most filesystems will cache it

        unbaggable = _can_bag(bag_dir)

        if unbaggable:
            LOGGER.error(
                _("Unable to write to the following directories and files:\n%s"),
                unbaggable,
            )
            raise BagError(_("Missing permissions to move all files and directories"))

        unreadable_dirs, unreadable_files = _can_read(bag_dir)

        if unreadable_dirs or unreadable_files:
            if unreadable_dirs:
                LOGGER.error(
                    _("The following directories do not have read permissions:\n%s"),
                    unreadable_dirs,
                )
            if unreadable_files:
                LOGGER.error(
                    _("The following files do not have read permissions:\n%s"),
                    unreadable_files,
                )
            raise BagError(
                _("Read permissions are required to calculate file fixities")
            )
        else:
            LOGGER.info(_("Creating data directory"))

            # FIXME: if we calculate full paths we won't need to deal with changing directories
            os.chdir(bag_dir)
            cwd = os.getcwd()
            temp_data = tempfile.mkdtemp(dir=cwd)

            for f in os.listdir("."):
                if os.path.abspath(f) == temp_data:
                    continue
                new_f = os.path.join(temp_data, f)
                LOGGER.info(
                    _("Moving %(source)s to %(destination)s"),
                    {"source": f, "destination": new_f},
                )
                os.rename(f, new_f)

            LOGGER.info(
                _("Moving %(source)s to %(destination)s"),
                {"source": temp_data, "destination": "data"},
            )
            os.rename(temp_data, "data")

            # permissions for the payload directory should match those of the
            # original directory
            os.chmod("data", os.stat(cwd).st_mode)

            total_bytes, total_files = make_manifests(
                "data", processes, algorithms=checksums, encoding=encoding
            )

            LOGGER.info(_("Creating bagit.txt"))
            txt = """BagIt-Version: 0.97\nTag-File-Character-Encoding: UTF-8\n"""
            with open_text_file("bagit.txt", "w") as bagit_file:
                bagit_file.write(txt)

            LOGGER.info(_("Creating bag-info.txt"))
            if bag_info is None:
                bag_info = {}

            # allow 'Bagging-Date' and 'Bag-Software-Agent' to be overidden
            if "Bagging-Date" not in bag_info:
                bag_info["Bagging-Date"] = date.strftime(date.today(), "%Y-%m-%d")
            if "Bag-Software-Agent" not in bag_info:
                bag_info["Bag-Software-Agent"] = "bagit.py v%s <%s>" % (
                    VERSION,
                    PROJECT_URL,
                )

            bag_info["Payload-Oxum"] = "%s.%s" % (total_bytes, total_files)
            _make_tag_file("bag-info.txt", bag_info)

            for c in checksums:
                _make_tagmanifest_file(c, bag_dir, encoding="utf-8")
    except Exception:
        LOGGER.exception(_("An error occurred creating a bag in %s"), bag_dir)
        raise
    finally:
        os.chdir(old_dir)

    return Bag(bag_dir)


class Bag(object):
    """A representation of a bag."""

    valid_files = ["bagit.txt", "fetch.txt"]
    valid_directories = ["data"]

    def __init__(self, path=None):
        super(Bag, self).__init__()
        self.tags = {}
        self.info = {}
        #: Dictionary of manifest entries and the checksum values for each
        #: algorithm:
        self.entries = {}

        # To reliably handle Unicode normalization differences, we maintain
        # lookup dictionaries in both directions for the filenames read from
        # the filesystem and the manifests so we can handle cases where the
        # normalization form changed between the bag being created and read.
        # See https://github.com/LibraryOfCongress/bagit-python/issues/51.

        #: maps Unicode-normalized values to the raw value from the filesystem
        self.normalized_filesystem_names = {}

        #: maps Unicode-normalized values to the raw value in the manifest
        self.normalized_manifest_names = {}

        self.algorithms = []
        self.tag_file_name = None
        self.path = abspath(path)
        if path:
            # if path ends in a path separator, strip it off
            if path[-1] == os.sep:
                self.path = path[:-1]
            self._open()

    def __str__(self):
        # FIXME: develop a more informative string representation for a Bag
        return self.path

    @property
    def algs(self):
        warnings.warn(_("Use Bag.algorithms instead of Bag.algs"), DeprecationWarning)
        return self.algorithms

    @property
    def version(self):
        warnings.warn(
            _("Use the Bag.version_info tuple instead of Bag.version"),
            DeprecationWarning,
        )
        return self._version

    def _open(self):
        # Open the bagit.txt file, and load any tags from it, including
        # the required version and encoding.
        bagit_file_path = os.path.join(self.path, "bagit.txt")

        if not isfile(bagit_file_path):
            raise BagError(_("Expected bagit.txt does not exist: %s") % bagit_file_path)

        self.tags = tags = _load_tag_file(bagit_file_path)

        required_tags = ("BagIt-Version", "Tag-File-Character-Encoding")
        missing_tags = [i for i in required_tags if i not in tags]
        if missing_tags:
            raise BagError(
                _("Missing required tag in bagit.txt: %s") % ", ".join(missing_tags)
            )

        # To avoid breaking existing code we'll leave self.version as the string
        # and parse it into a numeric version_info tuple. In version 2.0 we can
        # break that.

        self._version = tags["BagIt-Version"]

        try:
            self.version_info = tuple(int(i) for i in self._version.split(".", 1))
        except ValueError:
            raise BagError(
                _("Bag version numbers must be MAJOR.MINOR numbers, not %s")
                % self._version
            )

        if (0, 93) <= self.version_info <= (0, 95):
            self.tag_file_name = "package-info.txt"
        elif (0, 96) <= self.version_info < (2,):
            self.tag_file_name = "bag-info.txt"
        else:
            raise BagError(_("Unsupported bag version: %s") % self._version)

        self.encoding = tags["Tag-File-Character-Encoding"]

        try:
            codecs.lookup(self.encoding)
        except LookupError:
            raise BagValidationError(_("Unsupported encoding: %s") % self.encoding)

        info_file_path = os.path.join(self.path, self.tag_file_name)
        if os.path.exists(info_file_path):
            self.info = _load_tag_file(info_file_path, encoding=self.encoding)

        self._load_manifests()

    def manifest_files(self):
        for filename in ["manifest-%s.txt" % a for a in CHECKSUM_ALGOS]:
            f = os.path.join(self.path, filename)
            if isfile(f):
                yield f

    def tagmanifest_files(self):
        for filename in ["tagmanifest-%s.txt" % a for a in CHECKSUM_ALGOS]:
            f = os.path.join(self.path, filename)
            if isfile(f):
                yield f

    def compare_manifests_with_fs(self):
        """
        Compare the filenames in the manifests to the filenames present on the
        local filesystem and returns two lists of the files which are only
        present in the manifests and the files which are only present on the
        local filesystem, respectively.
        """

        # We compare the filenames after Unicode normalization so we can
        # reliably detect normalization changes after bag creation:
        files_on_fs = set(normalize_unicode(i) for i in self.payload_files())
        files_in_manifest = set(
            normalize_unicode(i) for i in self.payload_entries().keys()
        )

        if self.version_info >= (0, 97):
            files_in_manifest.update(self.missing_optional_tagfiles())

        only_on_fs = list()
        only_in_manifest = list()

        for i in files_on_fs.difference(files_in_manifest):
            only_on_fs.append(self.normalized_filesystem_names[i])

        for i in files_in_manifest.difference(files_on_fs):
            only_in_manifest.append(self.normalized_manifest_names[i])

        return only_in_manifest, only_on_fs

    def compare_fetch_with_fs(self):
        """Compares the fetch entries with the files actually
           in the payload, and returns a list of all the files
           that still need to be fetched.
        """

        files_on_fs = set(self.payload_files())
        files_in_fetch = set(self.files_to_be_fetched())

        return list(files_in_fetch - files_on_fs)

    def payload_files(self):
        """Returns a list of filenames which are present on the local filesystem"""
        payload_dir = os.path.join(self.path, "data")

        for dirpath, _, filenames in os.walk(payload_dir):
            for f in filenames:
                # Jump through some hoops here to make the payload files are
                # returned with the directory structure relative to the base
                # directory rather than the
                normalized_f = os.path.normpath(f)
                rel_path = os.path.relpath(
                    os.path.join(dirpath, normalized_f), start=self.path
                )

                self.normalized_filesystem_names[normalize_unicode(rel_path)] = rel_path
                yield rel_path

    def payload_entries(self):
        """Return a dictionary of items """
        # Don't use dict comprehension (compatibility with Python < 2.7)
        return dict(
            (key, value)
            for (key, value) in self.entries.items()
            if key.startswith("data" + os.sep)
        )

    def save(self, processes=1, manifests=False):
        """
        save will persist any changes that have been made to the bag
        metadata (self.info).

        If you have modified the payload of the bag (added, modified,
        removed files in the data directory) and want to regenerate manifests
        set the manifests parameter to True. The default is False since you
        wouldn't want a save to accidentally create a new manifest for
        a corrupted bag.

        If you want to control the number of processes that are used when
        recalculating checksums use the processes parameter.
        """
        # Error checking
        if not self.path:
            raise BagError(_("Bag.save() called before setting the path!"))

        if not os.access(self.path, os.R_OK | os.W_OK | os.X_OK):
            raise BagError(
                _("Cannot save bag to non-existent or inaccessible directory %s")
                % self.path
            )

        unbaggable = _can_bag(self.path)
        if unbaggable:
            LOGGER.error(
                _(
                    "Missing write permissions for the following directories and files:\n%s"
                ),
                unbaggable,
            )
            raise BagError(_("Missing permissions to move all files and directories"))

        unreadable_dirs, unreadable_files = _can_read(self.path)
        if unreadable_dirs or unreadable_files:
            if unreadable_dirs:
                LOGGER.error(
                    _("The following directories do not have read permissions:\n%s"),
                    unreadable_dirs,
                )
            if unreadable_files:
                LOGGER.error(
                    _("The following files do not have read permissions:\n%s"),
                    unreadable_files,
                )
            raise BagError(
                _("Read permissions are required to calculate file fixities")
            )

        # Change working directory to bag directory so helper functions work
        old_dir = os.path.abspath(os.path.curdir)
        os.chdir(self.path)

        # Generate new manifest files
        if manifests:
            total_bytes, total_files = make_manifests(
                "data", processes, algorithms=self.algorithms, encoding=self.encoding
            )

            # Update Payload-Oxum
            LOGGER.info(_("Updating Payload-Oxum in %s"), self.tag_file_name)
            self.info["Payload-Oxum"] = "%s.%s" % (total_bytes, total_files)

        _make_tag_file(self.tag_file_name, self.info)

        # Update tag-manifest for changes to manifest & bag-info files
        for alg in self.algorithms:
            _make_tagmanifest_file(alg, self.path, encoding=self.encoding)

        # Reload the manifests
        self._load_manifests()

        os.chdir(old_dir)

    def tagfile_entries(self):
        return dict(
            (key, value)
            for (key, value) in self.entries.items()
            if not key.startswith("data" + os.sep)
        )

    def missing_optional_tagfiles(self):
        """
        From v0.97 we need to validate any tagfiles listed
        in the optional tagmanifest(s). As there is no mandatory
        directory structure for additional tagfiles we can
        only check for entries with missing files (not missing
        entries for existing files).
        """
        for tagfilepath in self.tagfile_entries().keys():
            if not os.path.isfile(os.path.join(self.path, tagfilepath)):
                yield tagfilepath

    def fetch_entries(self):
        """Load fetch.txt if present and iterate over its contents

        yields (url, size, filename) tuples

        raises BagError for errors such as an unsafe filename referencing
        data outside of the bag directory
        """

        fetch_file_path = os.path.join(self.path, "fetch.txt")

        if isfile(fetch_file_path):
            with open_text_file(
                fetch_file_path, "r", encoding=self.encoding
            ) as fetch_file:
                for line in fetch_file:
                    url, file_size, filename = line.strip().split(None, 2)

                    if self._path_is_dangerous(filename):
                        raise BagError(
                            _('Path "%(payload_file)s" in "%(source_file)s" is unsafe')
                            % {
                                "payload_file": filename,
                                "source_file": os.path.join(self.path, "fetch.txt"),
                            }
                        )

                    yield url, file_size, filename

    def files_to_be_fetched(self):
        """
        Convenience wrapper for fetch_entries which returns only the
        local filename
        """

        for url, file_size, filename in self.fetch_entries():
            yield filename

    def has_oxum(self):
        return "Payload-Oxum" in self.info

    def validate(self, processes=1, fast=False, completeness_only=False):
        """Checks the structure and contents are valid.

        If you supply the parameter fast=True the Payload-Oxum (if present) will
        be used to check that the payload files are present and accounted for,
        instead of re-calculating fixities and comparing them against the
        manifest. By default validate() will re-calculate fixities (fast=False).
        """

        self._validate_structure()
        self._validate_bagittxt()

        self.validate_fetch()

        self._validate_contents(
            processes=processes, fast=fast, completeness_only=completeness_only
        )

        return True

    def is_valid(self, fast=False, completeness_only=False):
        """Returns validation success or failure as boolean.
        Optional fast parameter passed directly to validate().
        """

        try:
            self.validate(fast=fast, completeness_only=completeness_only)
        except BagError:
            return False

        return True

    def _load_manifests(self):
        self.entries = {}
        manifests = list(self.manifest_files())

        if self.version_info >= (0, 97):
            # v0.97+ requires that optional tagfiles are verified.
            manifests += list(self.tagmanifest_files())

        for manifest_filename in manifests:
            if not manifest_filename.find("tagmanifest-") is -1:
                search = "tagmanifest-"
            else:
                search = "manifest-"
            alg = (
                os.path.basename(manifest_filename)
                .replace(search, "")
                .replace(".txt", "")
            )
            if alg not in self.algorithms:
                self.algorithms.append(alg)

            with open_text_file(
                manifest_filename, "r", encoding=self.encoding
            ) as manifest_file:
                if manifest_file.encoding.startswith("UTF"):
                    # We'll check the first character to see if it's a BOM:
                    if manifest_file.read(1) == UNICODE_BYTE_ORDER_MARK:
                        # We'll skip it either way by letting line decoding
                        # happen at the new offset but we will issue a warning
                        # for UTF-8 since the presence of a BOM  is contrary to
                        # the BagIt specification:
                        if manifest_file.encoding == "UTF-8":
                            LOGGER.warning(
                                _(
                                    "%s is encoded using UTF-8 but contains an unnecessary"
                                    " byte-order mark, which is not in compliance with the"
                                    " BagIt RFC"
                                ),
                                manifest_file.name,
                            )
                    else:
                        manifest_file.seek(0)  # Pretend the first read never happened

                for line in manifest_file:
                    line = line.strip()

                    # Ignore blank lines and comments.
                    if line == "" or line.startswith("#"):
                        continue

                    entry = line.split(None, 1)

                    # Format is FILENAME *CHECKSUM
                    if len(entry) != 2:
                        LOGGER.error(
                            _(
                                "%(bag)s: Invalid %(algorithm)s manifest entry: %(line)s"
                            ),
                            {"bag": self, "algorithm": alg, "line": line},
                        )
                        continue

                    entry_hash = entry[0]
                    entry_path = os.path.normpath(entry[1].lstrip("*"))
                    entry_path = _decode_filename(entry_path)

                    if self._path_is_dangerous(entry_path):
                        raise BagError(
                            _(
                                'Path "%(payload_file)s" in manifest "%(manifest_file)s" is unsafe'
                            )
                            % {
                                "payload_file": entry_path,
                                "manifest_file": manifest_file.name,
                            }
                        )

                    entry_hashes = self.entries.setdefault(entry_path, {})

                    if alg in entry_hashes:
                        warning_ctx = {
                            "bag": self,
                            "algorithm": alg,
                            "filename": entry_path,
                        }
                        if entry_hashes[alg] == entry_hash:
                            msg = _(
                                "%(bag)s: %(algorithm)s manifest lists %(filename)s"
                                " multiple times with the same value"
                            )
                            if self.version_info >= (1,):
                                raise BagError(msg % warning_ctx)
                            else:
                                LOGGER.warning(msg, warning_ctx)
                        else:
                            raise BagError(
                                _(
                                    "%(bag)s: %(algorithm)s manifest lists %(filename)s"
                                    " multiple times with conflicting values"
                                )
                                % warning_ctx
                            )

                    entry_hashes[alg] = entry_hash

        self.normalized_manifest_names.update(
            (normalize_unicode(i), i) for i in self.entries.keys()
        )

    def _validate_structure(self):
        """
        Checks the structure of the bag to determine whether it conforms to the
        BagIt spec. Returns true on success, otherwise it will raise a
        BagValidationError exception.
        """

        self._validate_structure_payload_directory()
        self._validate_structure_tag_files()

    def _validate_structure_payload_directory(self):
        data_dir_path = os.path.join(self.path, "data")

        if not isdir(data_dir_path):
            raise BagValidationError(
                _("Expected data directory %s does not exist") % data_dir_path
            )

    def _validate_structure_tag_files(self):
        # Note: we deviate somewhat from v0.96 of the spec in that it allows
        # other files and directories to be present in the base directory

        if not list(self.manifest_files()):
            raise BagValidationError(_("No manifest files found"))
        if "bagit.txt" not in os.listdir(self.path):
            raise BagValidationError(
                _('Expected %s to contain "bagit.txt"') % self.path
            )

    def validate_fetch(self):
        """Validate the fetch.txt file

        Raises `BagError` for errors and otherwise returns no value
        """

        for url, file_size, filename in self.fetch_entries():
            # fetch_entries will raise a BagError for unsafe filenames
            # so at this point we will check only that the URL is minimally
            # well formed:
            parsed_url = urlparse(url)

            if not all((parsed_url.scheme, parsed_url.netloc)):
                raise BagError(_("Malformed URL in fetch.txt: %s") % url)

    def _validate_contents(self, processes=1, fast=False, completeness_only=False):
        if fast and not self.has_oxum():
            raise BagValidationError(
                _("Fast validation requires bag-info.txt to include Payload-Oxum")
            )

        # Perform the fast file count + size check so we can fail early:
        self._validate_oxum()

        if fast:
            return

        self._validate_completeness()

        if completeness_only:
            return

        self._validate_entries(processes)

    def _validate_oxum(self):
        oxum = self.info.get("Payload-Oxum")

        if oxum is None:
            return

        # If multiple Payload-Oxum tags (bad idea)
        # use the first listed in bag-info.txt
        if isinstance(oxum, list):
            LOGGER.warning(_("bag-info.txt defines multiple Payload-Oxum values!"))
            oxum = oxum[0]

        oxum_byte_count, oxum_file_count = oxum.split(".", 1)

        if not oxum_byte_count.isdigit() or not oxum_file_count.isdigit():
            raise BagError(_("Malformed Payload-Oxum value: %s") % oxum)

        oxum_byte_count = int(oxum_byte_count)
        oxum_file_count = int(oxum_file_count)
        total_bytes = 0
        total_files = 0

        for payload_file in self.payload_files():
            payload_file = os.path.join(self.path, payload_file)
            total_bytes += os.stat(payload_file).st_size
            total_files += 1

        if oxum_file_count != total_files or oxum_byte_count != total_bytes:
            raise BagValidationError(
                _(
                    "Payload-Oxum validation failed."
                    " Expected %(oxum_file_count)d files and %(oxum_byte_count)d bytes"
                    " but found %(found_file_count)d files and %(found_byte_count)d bytes"
                )
                % {
                    "found_file_count": total_files,
                    "found_byte_count": total_bytes,
                    "oxum_file_count": oxum_file_count,
                    "oxum_byte_count": oxum_byte_count,
                }
            )

    def _validate_completeness(self):
        """
        Verify that the actual file manifests match the files in the data directory
        """
        errors = list()

        # First we'll make sure there's no mismatch between the filesystem
        # and the list of files in the manifest(s)
        only_in_manifests, only_on_fs = self.compare_manifests_with_fs()
        for path in only_in_manifests:
            e = FileMissing(path)
            LOGGER.warning(force_unicode(e))
            errors.append(e)
        for path in only_on_fs:
            e = UnexpectedFile(path)
            LOGGER.warning(force_unicode(e))
            errors.append(e)

        if errors:
            raise BagValidationError(_("Bag validation failed"), errors)

    def _validate_entries(self, processes):
        """
        Verify that the actual file contents match the recorded hashes stored in the manifest files
        """
        errors = list()

        if os.name == "posix":
            worker_init = posix_multiprocessing_worker_initializer
        else:
            worker_init = None

        args = (
            (
                self.path,
                self.normalized_filesystem_names.get(rel_path, rel_path),
                hashes,
                self.algorithms,
            )
            for rel_path, hashes in self.entries.items()
        )

        try:
            if processes == 1:
                hash_results = [_calc_hashes(i) for i in args]
            else:
                try:
                    pool = multiprocessing.Pool(
                        processes if processes else None, initializer=worker_init
                    )
                    hash_results = pool.map(_calc_hashes, args)
                finally:
                    pool.terminate()

        # Any unhandled exceptions are probably fatal
        except:
            LOGGER.exception(_("Unable to calculate file hashes for %s"), self)
            raise

        for rel_path, f_hashes, hashes in hash_results:
            for alg, computed_hash in f_hashes.items():
                stored_hash = hashes[alg]
                if stored_hash.lower() != computed_hash:
                    e = ChecksumMismatch(
                        rel_path, alg, stored_hash.lower(), computed_hash
                    )
                    LOGGER.warning(force_unicode(e))
                    errors.append(e)

        if errors:
            raise BagValidationError(_("Bag validation failed"), errors)

    def _validate_bagittxt(self):
        """
        Verify that bagit.txt conforms to specification
        """
        bagit_file_path = os.path.join(self.path, "bagit.txt")

        # Note that we are intentionally opening this file in binary mode so we can confirm
        # that it does not start with the UTF-8 byte-order-mark
        with open(bagit_file_path, "rb") as bagit_file:
            first_line = bagit_file.read(4)
            if first_line.startswith(codecs.BOM_UTF8):
                raise BagValidationError(
                    _("bagit.txt must not contain a byte-order mark")
                )

    def _path_is_dangerous(self, path):
        """
        Return true if path looks dangerous, i.e. potentially operates
        outside the bagging directory structure, e.g. ~/.bashrc, ../../../secrets.json,
            \\?\c:\, D:\sys32\cmd.exe
        """
        if os.path.isabs(path):
            return True
        if os.path.expanduser(path) != path:
            return True
        if os.path.expandvars(path) != path:
            return True
        real_path = os.path.realpath(os.path.join(self.path, path))
        real_path = os.path.normpath(real_path)
        bag_path = os.path.realpath(self.path)
        bag_path = os.path.normpath(bag_path)
        common = os.path.commonprefix((bag_path, real_path))
        return not (common == bag_path)


class BagError(Exception):
    pass


class BagValidationError(BagError):
    def __init__(self, message, details=None):
        super(BagValidationError, self).__init__()

        if details is None:
            details = []

        self.message = message
        self.details = details

    def __str__(self):
        if len(self.details) > 0:
            details = "; ".join([force_unicode(e) for e in self.details])
            return "%s: %s" % (self.message, details)
        return self.message


class ManifestErrorDetail(BagError):
    def __init__(self, path):
        super(ManifestErrorDetail, self).__init__()

        self.path = path


class ChecksumMismatch(ManifestErrorDetail):
    def __init__(self, path, algorithm=None, expected=None, found=None):
        super(ChecksumMismatch, self).__init__(path)

        self.path = path
        self.algorithm = algorithm
        self.expected = expected
        self.found = found

    def __str__(self):
        return _(
            '%(path)s %(algorithm)s validation failed: expected="%(expected)s" found="%(found)s"'
        ) % {
            "path": force_unicode(self.path),
            "algorithm": self.algorithm,
            "expected": self.expected,
            "found": self.found,
        }


class FileMissing(ManifestErrorDetail):
    def __str__(self):
        return _(
            "%s exists in manifest but was not found on filesystem"
        ) % force_unicode(self.path)


class UnexpectedFile(ManifestErrorDetail):
    def __str__(self):
        return _("%s exists on filesystem but is not in the manifest") % self.path


class FileNormalizationConflict(BagError):
    """
    Exception raised when two files differ only in normalization and thus
    are not safely portable
    """

    def __init__(self, file_a, file_b):
        super(FileNormalizationConflict, self).__init__()

        self.file_a = file_a
        self.file_b = file_b

    def __str__(self):
        return _(
            'Unicode normalization conflict for file "%(file_a)s" and "%(file_b)s"'
        ) % {"file_a": self.file_a, "file_b": self.file_b}


def posix_multiprocessing_worker_initializer():
    """Ignore SIGINT in multiprocessing workers on POSIX systems"""
    signal.signal(signal.SIGINT, signal.SIG_IGN)


# The Unicode normalization form used here doesn't matter – all we care about
# is consistency since the input value will be preserved:


def normalize_unicode_py3(s):
    return unicodedata.normalize("NFC", s)


def normalize_unicode_py2(s):
    if isinstance(s, str):
        s = s.decode("utf-8")
    return unicodedata.normalize("NFC", s)


if sys.version_info > (3, 0):
    normalize_unicode = normalize_unicode_py3
else:
    normalize_unicode = normalize_unicode_py2


def build_unicode_normalized_lookup_dict(filenames):
    """
    Return a dictionary mapping unicode-normalized filenames to as-encoded
    values to efficiently detect conflicts between the filesystem and manifests.

    This is necessary because some filesystems and utilities may automatically
    apply a different Unicode normalization form to filenames than was applied
    when the bag was originally created.

    The best known example of this is when a bag is created using a
    normalization form other than NFD and then transferred to a Mac where the
    HFS+ filesystem will transparently normalize filenames to a variant of NFD
    for every call:

    https://developer.apple.com/legacy/library/technotes/tn/tn1150.html#UnicodeSubtleties

    Windows is documented as storing filenames exactly as provided:

    https://msdn.microsoft.com/en-us/library/windows/desktop/aa365247%28v=vs.85%29.aspx

    Linux performs no normalization in the kernel but it is technically
    valid for a filesystem to perform normalization, such as when an HFS+
    volume is mounted.

    See http://www.unicode.org/reports/tr15/ for a full discussion of
    equivalence and normalization in Unicode.
    """

    output = dict()

    for filename in filenames:
        normalized_filename = normalize_unicode(filename)
        if normalized_filename in output:
            raise FileNormalizationConflict(filename, output[normalized_filename])
        else:
            output[normalized_filename] = filename

    return output


def get_hashers(algorithms):
    """
    Given a list of algorithm names, return a dictionary of hasher instances

    This avoids redundant code between the creation and validation code where in
    both cases we want to avoid reading the same file more than once. The
    intended use is a simple for loop:

        for block in file:
            for hasher in hashers.values():
                hasher.update(block)
    """

    hashers = {}

    for alg in algorithms:
        try:
            hasher = hashlib.new(alg)
        except ValueError:
            LOGGER.warning(
                _("Disabling requested hash algorithm %s: hashlib does not support it"),
                alg,
            )
            continue

        hashers[alg] = hasher

    if not hashers:
        raise ValueError(
            _(
                "Unable to continue: hashlib does not support any of the requested algorithms!"
            )
        )

    return hashers


def _calc_hashes(args):
    # auto unpacking of sequences illegal in Python3
    (base_path, rel_path, hashes, algorithms) = args
    full_path = os.path.join(base_path, rel_path)

    # Create a clone of the default empty hash objects:
    f_hashers = dict((alg, hashlib.new(alg)) for alg in hashes if alg in algorithms)

    try:
        f_hashes = _calculate_file_hashes(full_path, f_hashers)
    except BagValidationError as e:
        f_hashes = dict((alg, force_unicode(e)) for alg in f_hashers.keys())

    return rel_path, f_hashes, hashes


def _calculate_file_hashes(full_path, f_hashers):
    """
    Returns a dictionary of (algorithm, hexdigest) values for the provided
    filename
    """
    LOGGER.info(_("Verifying checksum for file %s"), full_path)

    try:
        with open(full_path, "rb") as f:
            while True:
                block = f.read(HASH_BLOCK_SIZE)
                if not block:
                    break
                for i in f_hashers.values():
                    i.update(block)
    except (OSError, IOError) as e:
        raise BagValidationError(
            _("Could not read %(filename)s: %(error)s")
            % {"filename": full_path, "error": force_unicode(e)}
        )

    return dict((alg, h.hexdigest()) for alg, h in f_hashers.items())


def _load_tag_file(tag_file_name, encoding="utf-8-sig"):
    with open_text_file(tag_file_name, "r", encoding=encoding) as tag_file:
        # Store duplicate tags as list of vals
        # in order of parsing under the same key.
        tags = {}
        for name, value in _parse_tags(tag_file):
            if name not in tags:
                tags[name] = value
                continue

            if not isinstance(tags[name], list):
                tags[name] = [tags[name], value]
            else:
                tags[name].append(value)

        return tags


def _parse_tags(tag_file):
    """Parses a tag file, according to RFC 2822.  This
       includes line folding, permitting extra-long
       field values.

       See http://www.faqs.org/rfcs/rfc2822.html for
       more information.
    """

    tag_name = None
    tag_value = None

    # Line folding is handled by yielding values only after we encounter
    # the start of a new tag, or if we pass the EOF.
    for num, line in enumerate(tag_file):
        # Skip over any empty or blank lines.
        if len(line) == 0 or line.isspace():
            continue
        elif line[0].isspace() and tag_value is not None:  # folded line
            tag_value += line
        else:
            # Starting a new tag; yield the last one.
            if tag_name:
                yield (tag_name, tag_value.strip())

            if ":" not in line:
                raise BagValidationError(
                    _("%(filename)s contains invalid tag: %(line)s")
                    % {
                        "line": line.strip(),
                        "filename": os.path.basename(tag_file.name),
                    }
                )

            parts = line.strip().split(":", 1)
            tag_name = parts[0].strip()
            tag_value = parts[1]

    # Passed the EOF.  All done after this.
    if tag_name:
        yield (tag_name, tag_value.strip())


def _make_tag_file(bag_info_path, bag_info):
    headers = sorted(bag_info.keys())
    with open_text_file(bag_info_path, "w") as f:
        for h in headers:
            values = bag_info[h]
            if not isinstance(values, list):
                values = [values]
            for txt in values:
                # strip CR, LF and CRLF so they don't mess up the tag file
                txt = re.sub(r"\n|\r|(\r\n)", "", force_unicode(txt))
                f.write("%s: %s\n" % (h, txt))


def make_manifests(data_dir, processes, algorithms=DEFAULT_CHECKSUMS, encoding="utf-8"):
    LOGGER.info(
        _("Using %(process_count)d processes to generate manifests: %(algorithms)s"),
        {"process_count": processes, "algorithms": ", ".join(algorithms)},
    )

    manifest_line_generator = partial(generate_manifest_lines, algorithms=algorithms)

    if processes > 1:
        pool = multiprocessing.Pool(processes=processes)
        checksums = pool.map(manifest_line_generator, _walk(data_dir))
        pool.close()
        pool.join()
    else:
        checksums = [manifest_line_generator(i) for i in _walk(data_dir)]

    # At this point we have a list of tuples which start with the algorithm name:
    manifest_data = {}
    for batch in checksums:
        for entry in batch:
            manifest_data.setdefault(entry[0], []).append(entry[1:])

    # These will be keyed on the algorithm name so we can perform sanity checks
    # below to catch failures in the hashing process:
    num_files = defaultdict(lambda: 0)
    total_bytes = defaultdict(lambda: 0)

    for algorithm, values in manifest_data.items():
        manifest_filename = "manifest-%s.txt" % algorithm

        with open_text_file(manifest_filename, "w", encoding=encoding) as manifest:
            for digest, filename, byte_count in values:
                manifest.write("%s  %s\n" % (digest, _encode_filename(filename)))
                num_files[algorithm] += 1
                total_bytes[algorithm] += byte_count

    # We'll use sets of the values for the error checks and eventually return the payload oxum values:
    byte_value_set = set(total_bytes.values())
    file_count_set = set(num_files.values())

    # allow a bag with an empty payload
    if not byte_value_set and not file_count_set:
        return 0, 0

    if len(file_count_set) != 1:
        raise RuntimeError(_("Expected the same number of files for each checksum"))

    if len(byte_value_set) != 1:
        raise RuntimeError(_("Expected the same number of bytes for each checksums"))

    return byte_value_set.pop(), file_count_set.pop()


def _make_tagmanifest_file(alg, bag_dir, encoding="utf-8"):
    tagmanifest_file = join(bag_dir, "tagmanifest-%s.txt" % alg)
    LOGGER.info(_("Creating %s"), tagmanifest_file)

    checksums = []
    for f in _find_tag_files(bag_dir):
        if re.match(r"^tagmanifest-.+\.txt$", f):
            continue
        with open(join(bag_dir, f), "rb") as fh:
            m = hashlib.new(alg)
            while True:
                block = fh.read(HASH_BLOCK_SIZE)
                if not block:
                    break
                m.update(block)
            checksums.append((m.hexdigest(), f))

    with open_text_file(
        join(bag_dir, tagmanifest_file), mode="w", encoding=encoding
    ) as tagmanifest:
        for digest, filename in checksums:
            tagmanifest.write("%s %s\n" % (digest, filename))


def _find_tag_files(bag_dir):
    for dir in os.listdir(bag_dir):
        if dir != "data":
            if os.path.isfile(dir) and not dir.startswith("tagmanifest-"):
                yield dir
            for dir_name, _, filenames in os.walk(dir):
                for filename in filenames:
                    if filename.startswith("tagmanifest-"):
                        continue
                    # remove everything up to the bag_dir directory
                    p = join(dir_name, filename)
                    yield os.path.relpath(p, bag_dir)


def _walk(data_dir):
    for dirpath, dirnames, filenames in os.walk(data_dir):
        # if we don't sort here the order of entries is non-deterministic
        # which makes it hard to test the fixity of tagmanifest-md5.txt
        filenames.sort()
        dirnames.sort()
        for fn in filenames:
            path = os.path.join(dirpath, fn)
            # BagIt spec requires manifest to always use '/' as path separator
            if os.path.sep != "/":
                parts = path.split(os.path.sep)
                path = "/".join(parts)
            yield path


def _can_bag(test_dir):
    """Scan the provided directory for files which cannot be bagged due to insufficient permissions"""
    unbaggable = []

    if not os.access(test_dir, os.R_OK):
        # We cannot continue without permission to read the source directory
        unbaggable.append(test_dir)
        return unbaggable

    if not os.access(test_dir, os.W_OK):
        unbaggable.append(test_dir)

    for dirpath, dirnames, filenames in os.walk(test_dir):
        for directory in dirnames:
            full_path = os.path.join(dirpath, directory)
            if not os.access(full_path, os.W_OK):
                unbaggable.append(full_path)

    return unbaggable


def _can_read(test_dir):
    """
    returns ((unreadable_dirs), (unreadable_files))
    """
    unreadable_dirs = []
    unreadable_files = []

    if not os.access(test_dir, os.R_OK):
        unreadable_dirs.append(test_dir)
    else:
        for dirpath, dirnames, filenames in os.walk(test_dir):
            for dn in dirnames:
                full_path = os.path.join(dirpath, dn)
                if not os.access(full_path, os.R_OK):
                    unreadable_dirs.append(full_path)
            for fn in filenames:
                full_path = os.path.join(dirpath, fn)
                if not os.access(full_path, os.R_OK):
                    unreadable_files.append(full_path)
    return (tuple(unreadable_dirs), tuple(unreadable_files))


def generate_manifest_lines(filename, algorithms=DEFAULT_CHECKSUMS):
    LOGGER.info(_("Generating manifest lines for file %s"), filename)

    # For performance we'll read the file only once and pass it block
    # by block to every requested hash algorithm:
    hashers = get_hashers(algorithms)

    total_bytes = 0

    with open(filename, "rb") as f:
        while True:
            block = f.read(HASH_BLOCK_SIZE)

            if not block:
                break

            total_bytes += len(block)
            for hasher in hashers.values():
                hasher.update(block)

    decoded_filename = _decode_filename(filename)

    # We'll generate a list of results in roughly manifest format but prefixed with the algorithm:
    results = [
        (alg, hasher.hexdigest(), decoded_filename, total_bytes)
        for alg, hasher in hashers.items()
    ]

    return results


def _encode_filename(s):
    s = s.replace("\r", "%0D")
    s = s.replace("\n", "%0A")
    return s


def _decode_filename(s):
    s = re.sub(r"%0D", "\r", s, re.IGNORECASE)
    s = re.sub(r"%0A", "\n", s, re.IGNORECASE)
    return s


def force_unicode_py2(s):
    """Reliably return a Unicode string given a possible unicode or byte string"""
    if isinstance(s, str):
        return s.decode("utf-8")
    else:
        return unicode(s)


if sys.version_info > (3, 0):
    force_unicode = str
else:
    force_unicode = force_unicode_py2

# following code is used for command line program


class BagArgumentParser(argparse.ArgumentParser):
    def __init__(self, *args, **kwargs):
        argparse.ArgumentParser.__init__(self, *args, **kwargs)
        self.set_defaults(bag_info={})


class BagHeaderAction(argparse.Action):
    def __call__(self, parser, namespace, values, option_string=None):
        opt = option_string.lstrip("--")
        opt_caps = "-".join([o.capitalize() for o in opt.split("-")])
        namespace.bag_info[opt_caps] = values


def _make_parser():
    parser = BagArgumentParser(
        formatter_class=argparse.RawDescriptionHelpFormatter,
        description="bagit-python version %s\n\n%s\n" % (VERSION, __doc__.strip()),
    )
    parser.add_argument(
        "--processes",
        type=int,
        dest="processes",
        default=1,
        help=_(
            "Use multiple processes to calculate checksums faster (default: %(default)s)"
        ),
    )
    parser.add_argument("--log", help=_("The name of the log file (default: stdout)"))
    parser.add_argument(
        "--quiet",
        action="store_true",
        help=_("Suppress all progress information other than errors"),
    )
    parser.add_argument(
        "--validate",
        action="store_true",
        help=_(
            "Validate existing bags in the provided directories instead of"
            " creating new ones"
        ),
    )
    parser.add_argument(
        "--fast",
        action="store_true",
        help=_(
            "Modify --validate behaviour to only test whether the bag directory"
            " has the number of files and total size specified in Payload-Oxum"
            " without performing checksum validation to detect corruption."
        ),
    )
    parser.add_argument(
        "--completeness-only",
        action="store_true",
        help=_(
            "Modify --validate behaviour to test whether the bag directory"
            " has the expected payload specified in the checksum manifests"
            " without performing checksum validation to detect corruption."
        ),
    )

    checksum_args = parser.add_argument_group(
        _("Checksum Algorithms"),
        _(
            "Select the manifest algorithms to be used when creating bags"
            " (default=%s)"
        )
        % ", ".join(DEFAULT_CHECKSUMS),
    )

    for i in CHECKSUM_ALGOS:
        alg_name = re.sub(r"^([A-Z]+)(\d+)$", r"\1-\2", i.upper())
        checksum_args.add_argument(
            "--%s" % i,
            action="append_const",
            dest="checksums",
            const=i,
            help=_("Generate %s manifest when creating a bag") % alg_name,
        )

    metadata_args = parser.add_argument_group(_("Optional Bag Metadata"))
    for header in STANDARD_BAG_INFO_HEADERS:
        metadata_args.add_argument(
            "--%s" % header.lower(), type=str, action=BagHeaderAction, default=argparse.SUPPRESS
        )

    parser.add_argument(
        "directory",
        nargs="+",
        help=_(
            "Directory which will be converted into a bag in place"
            " by moving any existing files into the BagIt structure"
            " and creating the manifests and other metadata."
        ),
    )

    return parser


def _configure_logging(opts):
    log_format = "%(asctime)s - %(levelname)s - %(message)s"
    if opts.quiet:
        level = logging.ERROR
    else:
        level = logging.INFO
    if opts.log:
        logging.basicConfig(filename=opts.log, level=level, format=log_format)
    else:
        logging.basicConfig(level=level, format=log_format)


def main():
    if "--version" in sys.argv:
        print(_("bagit-python version %s") % VERSION)
        sys.exit(0)

    parser = _make_parser()
    args = parser.parse_args()

    if args.processes < 0:
        parser.error(_("The number of processes must be 0 or greater"))

    if args.fast and not args.validate:
        parser.error(_("--fast is only allowed as an option for --validate!"))

    _configure_logging(args)

    rc = 0
    for bag_dir in args.directory:
        # validate the bag
        if args.validate:
            try:
                bag = Bag(bag_dir)
                # validate throws a BagError or BagValidationError
                bag.validate(
                    processes=args.processes,
                    fast=args.fast,
                    completeness_only=args.completeness_only,
                )
                if args.fast:
                    LOGGER.info(_("%s valid according to Payload-Oxum"), bag_dir)
                else:
                    LOGGER.info(_("%s is valid"), bag_dir)
            except BagError as e:
                LOGGER.error(
                    _("%(bag)s is invalid: %(error)s"), {"bag": bag_dir, "error": e}
                )
                rc = 1

        # make the bag
        else:
            try:
                make_bag(
                    bag_dir,
                    bag_info=args.bag_info,
                    processes=args.processes,
                    checksums=args.checksums,
                )
            except Exception as exc:
                LOGGER.error(
                    _("Failed to create bag in %(bag_directory)s: %(error)s"),
                    {"bag_directory": bag_dir, "error": exc},
                    exc_info=True,
                )
                rc = 1

    sys.exit(rc)


if __name__ == "__main__":
    main()