dulwich/dulwich/object_store.py

# object_store.py -- Object store for git objects
# Copyright (C) 2008-2013 Jelmer Vernooij <jelmer@jelmer.uk>
#                         and others
#
# Dulwich is dual-licensed under the Apache License, Version 2.0 and the GNU
# General Public License as public by the Free Software Foundation; version 2.0
# or (at your option) any later version. You can redistribute it and/or
# modify it under the terms of either of these two licenses.
#
# Unless required by applicable law or agreed to in writing, software
# distributed under the License is distributed on an "AS IS" BASIS,
# WITHOUT WARRANTIES OR CONDITIONS OF ANY KIND, either express or implied.
# See the License for the specific language governing permissions and
# limitations under the License.
#
# You should have received a copy of the licenses; if not, see
# <http://www.gnu.org/licenses/> for a copy of the GNU General Public License
# and <http://www.apache.org/licenses/LICENSE-2.0> for a copy of the Apache
# License, Version 2.0.
#


"""Git object store interfaces and implementation."""

from io import BytesIO
import errno
from itertools import chain
import os
import stat
import sys
import tempfile
import time

from dulwich.diff_tree import (
    tree_changes,
    walk_trees,
    )
from dulwich.errors import (
    NotTreeError,
    )
from dulwich.file import GitFile
from dulwich.objects import (
    Commit,
    ShaFile,
    Tag,
    Tree,
    ZERO_SHA,
    hex_to_sha,
    sha_to_hex,
    hex_to_filename,
    S_ISGITLINK,
    object_class,
    )
from dulwich.pack import (
    Pack,
    PackData,
    PackInflater,
    iter_sha1,
    pack_objects_to_data,
    write_pack_header,
    write_pack_index_v2,
    write_pack_data,
    write_pack_object,
    compute_file_sha,
    PackIndexer,
    PackStreamCopier,
    )

INFODIR = 'info'
PACKDIR = 'pack'


class BaseObjectStore(object):
    """Object store interface."""

    def determine_wants_all(self, refs):
        return [sha for (ref, sha) in refs.items()
                if sha not in self and not ref.endswith(b"^{}") and
                not sha == ZERO_SHA]

    def iter_shas(self, shas):
        """Iterate over the objects for the specified shas.

        :param shas: Iterable object with SHAs
        :return: Object iterator
        """
        return ObjectStoreIterator(self, shas)

    def contains_loose(self, sha):
        """Check if a particular object is present by SHA1 and is loose."""
        raise NotImplementedError(self.contains_loose)

    def contains_packed(self, sha):
        """Check if a particular object is present by SHA1 and is packed."""
        raise NotImplementedError(self.contains_packed)

    def __contains__(self, sha):
        """Check if a particular object is present by SHA1.

        This method makes no distinction between loose and packed objects.
        """
        return self.contains_packed(sha) or self.contains_loose(sha)

    @property
    def packs(self):
        """Iterable of pack objects."""
        raise NotImplementedError

    def get_raw(self, name):
        """Obtain the raw text for an object.

        :param name: sha for the object.
        :return: tuple with numeric type and object contents.
        """
        raise NotImplementedError(self.get_raw)

    def __getitem__(self, sha):
        """Obtain an object by SHA1."""
        type_num, uncomp = self.get_raw(sha)
        return ShaFile.from_raw_string(type_num, uncomp, sha=sha)

    def __iter__(self):
        """Iterate over the SHAs that are present in this store."""
        raise NotImplementedError(self.__iter__)

    def add_object(self, obj):
        """Add a single object to this object store.

        """
        raise NotImplementedError(self.add_object)

    def add_objects(self, objects, progress=None):
        """Add a set of objects to this object store.

        :param objects: Iterable over a list of (object, path) tuples
        """
        raise NotImplementedError(self.add_objects)

    def add_pack_data(self, count, pack_data, progress=None):
        """Add pack data to this object store.

        :param num_items: Number of items to add
        :param pack_data: Iterator over pack data tuples
        """
        if count == 0:
            # Don't bother writing an empty pack file
            return
        f, commit, abort = self.add_pack()
        try:
            write_pack_data(f, count, pack_data, progress)
        except BaseException:
            abort()
            raise
        else:
            return commit()

    def tree_changes(self, source, target, want_unchanged=False,
                     include_trees=False, change_type_same=False):
        """Find the differences between the contents of two trees

        :param source: SHA1 of the source tree
        :param target: SHA1 of the target tree
        :param want_unchanged: Whether unchanged files should be reported
        :param include_trees: Whether to include trees
        :param change_type_same: Whether to report files changing
            type in the same entry.
        :return: Iterator over tuples with
            (oldpath, newpath), (oldmode, newmode), (oldsha, newsha)
        """
        for change in tree_changes(self, source, target,
                                   want_unchanged=want_unchanged,
                                   include_trees=include_trees,
                                   change_type_same=change_type_same):
            yield ((change.old.path, change.new.path),
                   (change.old.mode, change.new.mode),
                   (change.old.sha, change.new.sha))

    def iter_tree_contents(self, tree_id, include_trees=False):
        """Iterate the contents of a tree and all subtrees.

        Iteration is depth-first pre-order, as in e.g. os.walk.

        :param tree_id: SHA1 of the tree.
        :param include_trees: If True, include tree objects in the iteration.
        :return: Iterator over TreeEntry namedtuples for all the objects in a
            tree.
        """
        for entry, _ in walk_trees(self, tree_id, None):
            if ((entry.mode is not None and
                 not stat.S_ISDIR(entry.mode)) or include_trees):
                yield entry

    def find_missing_objects(self, haves, wants, progress=None,
                             get_tagged=None,
                             get_parents=lambda commit: commit.parents):
        """Find the missing objects required for a set of revisions.

        :param haves: Iterable over SHAs already in common.
        :param wants: Iterable over SHAs of objects to fetch.
        :param progress: Simple progress function that will be called with
            updated progress strings.
        :param get_tagged: Function that returns a dict of pointed-to sha ->
            tag sha for including tags.
        :param get_parents: Optional function for getting the parents of a
            commit.
        :return: Iterator over (sha, path) pairs.
        """
        finder = MissingObjectFinder(self, haves, wants, progress, get_tagged,
                                     get_parents=get_parents)
        return iter(finder.next, None)

    def find_common_revisions(self, graphwalker):
        """Find which revisions this store has in common using graphwalker.

        :param graphwalker: A graphwalker object.
        :return: List of SHAs that are in common
        """
        haves = []
        sha = next(graphwalker)
        while sha:
            if sha in self:
                haves.append(sha)
                graphwalker.ack(sha)
            sha = next(graphwalker)
        return haves

    def generate_pack_contents(self, have, want, progress=None):
        """Iterate over the contents of a pack file.

        :param have: List of SHA1s of objects that should not be sent
        :param want: List of SHA1s of objects that should be sent
        :param progress: Optional progress reporting method
        """
        return self.iter_shas(self.find_missing_objects(have, want, progress))

    def generate_pack_data(self, have, want, progress=None, ofs_delta=True):
        """Generate pack data objects for a set of wants/haves.

        :param have: List of SHA1s of objects that should not be sent
        :param want: List of SHA1s of objects that should be sent
        :param ofs_delta: Whether OFS deltas can be included
        :param progress: Optional progress reporting method
        """
        # TODO(jelmer): More efficient implementation
        return pack_objects_to_data(
            self.generate_pack_contents(have, want, progress))

    def peel_sha(self, sha):
        """Peel all tags from a SHA.

        :param sha: The object SHA to peel.
        :return: The fully-peeled SHA1 of a tag object, after peeling all
            intermediate tags; if the original ref does not point to a tag,
            this will equal the original SHA1.
        """
        obj = self[sha]
        obj_class = object_class(obj.type_name)
        while obj_class is Tag:
            obj_class, sha = obj.object
            obj = self[sha]
        return obj

    def _collect_ancestors(self, heads, common=set(),
                           get_parents=lambda commit: commit.parents):
        """Collect all ancestors of heads up to (excluding) those in common.

        :param heads: commits to start from
        :param common: commits to end at, or empty set to walk repository
            completely
        :param get_parents: Optional function for getting the parents of a
            commit.
        :return: a tuple (A, B) where A - all commits reachable
            from heads but not present in common, B - common (shared) elements
            that are directly reachable from heads
        """
        bases = set()
        commits = set()
        queue = []
        queue.extend(heads)
        while queue:
            e = queue.pop(0)
            if e in common:
                bases.add(e)
            elif e not in commits:
                commits.add(e)
                cmt = self[e]
                queue.extend(get_parents(cmt))
        return (commits, bases)

    def close(self):
        """Close any files opened by this object store."""
        # Default implementation is a NO-OP


class PackBasedObjectStore(BaseObjectStore):

    def __init__(self):
        self._pack_cache = {}

    @property
    def alternates(self):
        return []

    def contains_packed(self, sha):
        """Check if a particular object is present by SHA1 and is packed.

        This does not check alternates.
        """
        for pack in self.packs:
            if sha in pack:
                return True
        return False

    def __contains__(self, sha):
        """Check if a particular object is present by SHA1.

        This method makes no distinction between loose and packed objects.
        """
        if self.contains_packed(sha) or self.contains_loose(sha):
            return True
        for alternate in self.alternates:
            if sha in alternate:
                return True
        return False

    def _pack_cache_stale(self):
        """Check whether the pack cache is stale."""
        raise NotImplementedError(self._pack_cache_stale)

    def _add_known_pack(self, base_name, pack):
        """Add a newly appeared pack to the cache by path.

        """
        prev_pack = self._pack_cache.get(base_name)
        if prev_pack is not pack:
            self._pack_cache[base_name] = pack
            if prev_pack:
                prev_pack.close()

    def _flush_pack_cache(self):
        pack_cache = self._pack_cache
        self._pack_cache = {}
        while pack_cache:
            (name, pack) = pack_cache.popitem()
            pack.close()

    def close(self):
        self._flush_pack_cache()

    @property
    def packs(self):
        """List with pack objects."""
        if self._pack_cache is None or self._pack_cache_stale():
            self._update_pack_cache()

        return self._pack_cache.values()

    def _iter_alternate_objects(self):
        """Iterate over the SHAs of all the objects in alternate stores."""
        for alternate in self.alternates:
            for alternate_object in alternate:
                yield alternate_object

    def _iter_loose_objects(self):
        """Iterate over the SHAs of all loose objects."""
        raise NotImplementedError(self._iter_loose_objects)

    def _get_loose_object(self, sha):
        raise NotImplementedError(self._get_loose_object)

    def _remove_loose_object(self, sha):
        raise NotImplementedError(self._remove_loose_object)

    def _remove_pack(self, name):
        raise NotImplementedError(self._remove_pack)

    def pack_loose_objects(self):
        """Pack loose objects.

        :return: Number of objects packed
        """
        objects = set()
        for sha in self._iter_loose_objects():
            objects.add((self._get_loose_object(sha), None))
        self.add_objects(list(objects))
        for obj, path in objects:
            self._remove_loose_object(obj.id)
        return len(objects)

    def repack(self):
        """Repack the packs in this repository.

        Note that this implementation is fairly naive and currently keeps all
        objects in memory while it repacks.
        """
        loose_objects = set()
        for sha in self._iter_loose_objects():
            loose_objects.add(self._get_loose_object(sha))
        objects = {(obj, None) for obj in loose_objects}
        old_packs = {p.name(): p for p in self.packs}
        for name, pack in old_packs.items():
            objects.update((obj, None) for obj in pack.iterobjects())
        self._flush_pack_cache()

        # The name of the consolidated pack might match the name of a
        # pre-existing pack. Take care not to remove the newly created
        # consolidated pack.

        consolidated = self.add_objects(objects)
        old_packs.pop(consolidated.name(), None)

        for obj in loose_objects:
            self._remove_loose_object(obj.id)
        for name, pack in old_packs.items():
            self._remove_pack(pack)
        self._update_pack_cache()
        return len(objects)

    def __iter__(self):
        """Iterate over the SHAs that are present in this store."""
        iterables = (list(self.packs) + [self._iter_loose_objects()] +
                     [self._iter_alternate_objects()])
        return chain(*iterables)

    def contains_loose(self, sha):
        """Check if a particular object is present by SHA1 and is loose.

        This does not check alternates.
        """
        return self._get_loose_object(sha) is not None

    def get_raw(self, name):
        """Obtain the raw fulltext for an object.

        :param name: sha for the object.
        :return: tuple with numeric type and object contents.
        """
        if len(name) == 40:
            sha = hex_to_sha(name)
            hexsha = name
        elif len(name) == 20:
            sha = name
            hexsha = None
        else:
            raise AssertionError("Invalid object name %r" % name)
        for pack in self.packs:
            try:
                return pack.get_raw(sha)
            except KeyError:
                pass
        if hexsha is None:
            hexsha = sha_to_hex(name)
        ret = self._get_loose_object(hexsha)
        if ret is not None:
            return ret.type_num, ret.as_raw_string()
        for alternate in self.alternates:
            try:
                return alternate.get_raw(hexsha)
            except KeyError:
                pass
        raise KeyError(hexsha)

    def add_objects(self, objects, progress=None):
        """Add a set of objects to this object store.

        :param objects: Iterable over (object, path) tuples, should support
            __len__.
        :return: Pack object of the objects written.
        """
        return self.add_pack_data(
                *pack_objects_to_data(objects),
                progress=progress)


class DiskObjectStore(PackBasedObjectStore):
    """Git-style object store that exists on disk."""

    def __init__(self, path):
        """Open an object store.

        :param path: Path of the object store.
        """
        super(DiskObjectStore, self).__init__()
        self.path = path
        self.pack_dir = os.path.join(self.path, PACKDIR)
        self._pack_cache_time = 0
        self._pack_cache = {}
        self._alternates = None

    def __repr__(self):
        return "<%s(%r)>" % (self.__class__.__name__, self.path)

    @property
    def alternates(self):
        if self._alternates is not None:
            return self._alternates
        self._alternates = []
        for path in self._read_alternate_paths():
            self._alternates.append(DiskObjectStore(path))
        return self._alternates

    def _read_alternate_paths(self):
        try:
            f = GitFile(os.path.join(self.path, INFODIR, "alternates"), 'rb')
        except (OSError, IOError) as e:
            if e.errno == errno.ENOENT:
                return
            raise
        with f:
            for line in f.readlines():
                line = line.rstrip(b"\n")
                if line[0] == b"#":
                    continue
                if os.path.isabs(line):
                    yield line.decode(sys.getfilesystemencoding())
                else:
                    yield os.path.join(self.path, line).decode(
                        sys.getfilesystemencoding())

    def add_alternate_path(self, path):
        """Add an alternate path to this object store.
        """
        try:
            os.mkdir(os.path.join(self.path, INFODIR))
        except OSError as e:
            if e.errno != errno.EEXIST:
                raise
        alternates_path = os.path.join(self.path, INFODIR, "alternates")
        with GitFile(alternates_path, 'wb') as f:
            try:
                orig_f = open(alternates_path, 'rb')
            except (OSError, IOError) as e:
                if e.errno != errno.ENOENT:
                    raise
            else:
                with orig_f:
                    f.write(orig_f.read())
            f.write(path.encode(sys.getfilesystemencoding()) + b"\n")

        if not os.path.isabs(path):
            path = os.path.join(self.path, path)
        self.alternates.append(DiskObjectStore(path))

    def _update_pack_cache(self):
        try:
            pack_dir_contents = os.listdir(self.pack_dir)
        except OSError as e:
            if e.errno == errno.ENOENT:
                self._pack_cache_time = 0
                self.close()
                return
            raise
        self._pack_cache_time = max(
                os.stat(self.pack_dir).st_mtime, time.time())
        pack_files = set()
        for name in pack_dir_contents:
            if name.startswith("pack-") and name.endswith(".pack"):
                # verify that idx exists first (otherwise the pack was not yet
                # fully written)
                idx_name = os.path.splitext(name)[0] + ".idx"
                if idx_name in pack_dir_contents:
                    pack_name = name[:-len(".pack")]
                    pack_files.add(pack_name)

        # Open newly appeared pack files
        for f in pack_files:
            if f not in self._pack_cache:
                self._pack_cache[f] = Pack(os.path.join(self.pack_dir, f))
        # Remove disappeared pack files
        for f in set(self._pack_cache) - pack_files:
            self._pack_cache.pop(f).close()

    def _pack_cache_stale(self):
        try:
            return os.stat(self.pack_dir).st_mtime >= self._pack_cache_time
        except OSError as e:
            if e.errno == errno.ENOENT:
                return True
            raise

    def _get_shafile_path(self, sha):
        # Check from object dir
        return hex_to_filename(self.path, sha)

    def _iter_loose_objects(self):
        for base in os.listdir(self.path):
            if len(base) != 2:
                continue
            for rest in os.listdir(os.path.join(self.path, base)):
                yield (base+rest).encode(sys.getfilesystemencoding())

    def _get_loose_object(self, sha):
        path = self._get_shafile_path(sha)
        try:
            return ShaFile.from_path(path)
        except (OSError, IOError) as e:
            if e.errno == errno.ENOENT:
                return None
            raise

    def _remove_loose_object(self, sha):
        os.remove(self._get_shafile_path(sha))

    def _remove_pack(self, pack):
        os.remove(pack.data.path)
        os.remove(pack.index.path)

    def _get_pack_basepath(self, entries):
        suffix = iter_sha1(entry[0] for entry in entries)
        # TODO: Handle self.pack_dir being bytes
        suffix = suffix.decode('ascii')
        return os.path.join(self.pack_dir, "pack-" + suffix)

    def _complete_thin_pack(self, f, path, copier, indexer):
        """Move a specific file containing a pack into the pack directory.

        :note: The file should be on the same file system as the
            packs directory.

        :param f: Open file object for the pack.
        :param path: Path to the pack file.
        :param copier: A PackStreamCopier to use for writing pack data.
        :param indexer: A PackIndexer for indexing the pack.
        """
        entries = list(indexer)

        # Update the header with the new number of objects.
        f.seek(0)
        write_pack_header(f, len(entries) + len(indexer.ext_refs()))

        # Must flush before reading (http://bugs.python.org/issue3207)
        f.flush()

        # Rescan the rest of the pack, computing the SHA with the new header.
        new_sha = compute_file_sha(f, end_ofs=-20)

        # Must reposition before writing (http://bugs.python.org/issue3207)
        f.seek(0, os.SEEK_CUR)

        # Complete the pack.
        for ext_sha in indexer.ext_refs():
            assert len(ext_sha) == 20
            type_num, data = self.get_raw(ext_sha)
            offset = f.tell()
            crc32 = write_pack_object(f, type_num, data, sha=new_sha)
            entries.append((ext_sha, offset, crc32))
        pack_sha = new_sha.digest()
        f.write(pack_sha)
        f.close()

        # Move the pack in.
        entries.sort()
        pack_base_name = self._get_pack_basepath(entries)
        target_pack = pack_base_name + '.pack'
        if sys.platform == 'win32':
            # Windows might have the target pack file lingering. Attempt
            # removal, silently passing if the target does not exist.
            try:
                os.remove(target_pack)
            except (IOError, OSError) as e:
                if e.errno != errno.ENOENT:
                    raise
        os.rename(path, target_pack)

        # Write the index.
        index_file = GitFile(pack_base_name + '.idx', 'wb')
        try:
            write_pack_index_v2(index_file, entries, pack_sha)
            index_file.close()
        finally:
            index_file.abort()

        # Add the pack to the store and return it.
        final_pack = Pack(pack_base_name)
        final_pack.check_length_and_checksum()
        self._add_known_pack(pack_base_name, final_pack)
        return final_pack

    def add_thin_pack(self, read_all, read_some):
        """Add a new thin pack to this object store.

        Thin packs are packs that contain deltas with parents that exist
        outside the pack. They should never be placed in the object store
        directly, and always indexed and completed as they are copied.

        :param read_all: Read function that blocks until the number of
            requested bytes are read.
        :param read_some: Read function that returns at least one byte, but may
            not return the number of bytes requested.
        :return: A Pack object pointing at the now-completed thin pack in the
            objects/pack directory.
        """
        fd, path = tempfile.mkstemp(dir=self.path, prefix='tmp_pack_')
        with os.fdopen(fd, 'w+b') as f:
            indexer = PackIndexer(f, resolve_ext_ref=self.get_raw)
            copier = PackStreamCopier(read_all, read_some, f,
                                      delta_iter=indexer)
            copier.verify()
            return self._complete_thin_pack(f, path, copier, indexer)

    def move_in_pack(self, path):
        """Move a specific file containing a pack into the pack directory.

        :note: The file should be on the same file system as the
            packs directory.

        :param path: Path to the pack file.
        """
        with PackData(path) as p:
            entries = p.sorted_entries()
            basename = self._get_pack_basepath(entries)
            with GitFile(basename+".idx", "wb") as f:
                write_pack_index_v2(f, entries, p.get_stored_checksum())
        if self._pack_cache is None or self._pack_cache_stale():
            self._update_pack_cache()
        try:
            return self._pack_cache[basename]
        except KeyError:
            pass
        target_pack = basename + '.pack'
        if sys.platform == 'win32':
            # Windows might have the target pack file lingering. Attempt
            # removal, silently passing if the target does not exist.
            try:
                os.remove(target_pack)
            except (IOError, OSError) as e:
                if e.errno != errno.ENOENT:
                    raise
        os.rename(path, target_pack)
        final_pack = Pack(basename)
        self._add_known_pack(basename, final_pack)
        return final_pack

    def add_pack(self):
        """Add a new pack to this object store.

        :return: Fileobject to write to, a commit function to
            call when the pack is finished and an abort
            function.
        """
        fd, path = tempfile.mkstemp(dir=self.pack_dir, suffix=".pack")
        f = os.fdopen(fd, 'wb')

        def commit():
            f.flush()
            os.fsync(fd)
            f.close()
            if os.path.getsize(path) > 0:
                return self.move_in_pack(path)
            else:
                os.remove(path)
                return None

        def abort():
            f.close()
            os.remove(path)
        return f, commit, abort

    def add_object(self, obj):
        """Add a single object to this object store.

        :param obj: Object to add
        """
        path = self._get_shafile_path(obj.id)
        dir = os.path.dirname(path)
        try:
            os.mkdir(dir)
        except OSError as e:
            if e.errno != errno.EEXIST:
                raise
        if os.path.exists(path):
            return  # Already there, no need to write again
        with GitFile(path, 'wb') as f:
            f.write(obj.as_legacy_object())

    @classmethod
    def init(cls, path):
        try:
            os.mkdir(path)
        except OSError as e:
            if e.errno != errno.EEXIST:
                raise
        os.mkdir(os.path.join(path, "info"))
        os.mkdir(os.path.join(path, PACKDIR))
        return cls(path)


class MemoryObjectStore(BaseObjectStore):
    """Object store that keeps all objects in memory."""

    def __init__(self):
        super(MemoryObjectStore, self).__init__()
        self._data = {}

    def _to_hexsha(self, sha):
        if len(sha) == 40:
            return sha
        elif len(sha) == 20:
            return sha_to_hex(sha)
        else:
            raise ValueError("Invalid sha %r" % (sha,))

    def contains_loose(self, sha):
        """Check if a particular object is present by SHA1 and is loose."""
        return self._to_hexsha(sha) in self._data

    def contains_packed(self, sha):
        """Check if a particular object is present by SHA1 and is packed."""
        return False

    def __iter__(self):
        """Iterate over the SHAs that are present in this store."""
        return iter(self._data.keys())

    @property
    def packs(self):
        """List with pack objects."""
        return []

    def get_raw(self, name):
        """Obtain the raw text for an object.

        :param name: sha for the object.
        :return: tuple with numeric type and object contents.
        """
        obj = self[self._to_hexsha(name)]
        return obj.type_num, obj.as_raw_string()

    def __getitem__(self, name):
        return self._data[self._to_hexsha(name)].copy()

    def __delitem__(self, name):
        """Delete an object from this store, for testing only."""
        del self._data[self._to_hexsha(name)]

    def add_object(self, obj):
        """Add a single object to this object store.

        """
        self._data[obj.id] = obj.copy()

    def add_objects(self, objects, progress=None):
        """Add a set of objects to this object store.

        :param objects: Iterable over a list of (object, path) tuples
        """
        for obj, path in objects:
            self.add_object(obj)

    def add_pack(self):
        """Add a new pack to this object store.

        Because this object store doesn't support packs, we extract and add the
        individual objects.

        :return: Fileobject to write to and a commit function to
            call when the pack is finished.
        """
        f = BytesIO()

        def commit():
            p = PackData.from_file(BytesIO(f.getvalue()), f.tell())
            f.close()
            for obj in PackInflater.for_pack_data(p, self.get_raw):
                self.add_object(obj)

        def abort():
            pass
        return f, commit, abort

    def _complete_thin_pack(self, f, indexer):
        """Complete a thin pack by adding external references.

        :param f: Open file object for the pack.
        :param indexer: A PackIndexer for indexing the pack.
        """
        entries = list(indexer)

        # Update the header with the new number of objects.
        f.seek(0)
        write_pack_header(f, len(entries) + len(indexer.ext_refs()))

        # Rescan the rest of the pack, computing the SHA with the new header.
        new_sha = compute_file_sha(f, end_ofs=-20)

        # Complete the pack.
        for ext_sha in indexer.ext_refs():
            assert len(ext_sha) == 20
            type_num, data = self.get_raw(ext_sha)
            write_pack_object(f, type_num, data, sha=new_sha)
        pack_sha = new_sha.digest()
        f.write(pack_sha)

    def add_thin_pack(self, read_all, read_some):
        """Add a new thin pack to this object store.

        Thin packs are packs that contain deltas with parents that exist
        outside the pack. Because this object store doesn't support packs, we
        extract and add the individual objects.

        :param read_all: Read function that blocks until the number of
            requested bytes are read.
        :param read_some: Read function that returns at least one byte, but may
            not return the number of bytes requested.
        """
        f, commit, abort = self.add_pack()
        try:
            indexer = PackIndexer(f, resolve_ext_ref=self.get_raw)
            copier = PackStreamCopier(read_all, read_some, f,
                                      delta_iter=indexer)
            copier.verify()
            self._complete_thin_pack(f, indexer)
        except BaseException:
            abort()
            raise
        else:
            commit()


class ObjectIterator(object):
    """Interface for iterating over objects."""

    def iterobjects(self):
        raise NotImplementedError(self.iterobjects)


class ObjectStoreIterator(ObjectIterator):
    """ObjectIterator that works on top of an ObjectStore."""

    def __init__(self, store, sha_iter):
        """Create a new ObjectIterator.

        :param store: Object store to retrieve from
        :param sha_iter: Iterator over (sha, path) tuples
        """
        self.store = store
        self.sha_iter = sha_iter
        self._shas = []

    def __iter__(self):
        """Yield tuple with next object and path."""
        for sha, path in self.itershas():
            yield self.store[sha], path

    def iterobjects(self):
        """Iterate over just the objects."""
        for o, path in self:
            yield o

    def itershas(self):
        """Iterate over the SHAs."""
        for sha in self._shas:
            yield sha
        for sha in self.sha_iter:
            self._shas.append(sha)
            yield sha

    def __contains__(self, needle):
        """Check if an object is present.

        :note: This checks if the object is present in
            the underlying object store, not if it would
            be yielded by the iterator.

        :param needle: SHA1 of the object to check for
        """
        return needle in self.store

    def __getitem__(self, key):
        """Find an object by SHA1.

        :note: This retrieves the object from the underlying
            object store. It will also succeed if the object would
            not be returned by the iterator.
        """
        return self.store[key]

    def __len__(self):
        """Return the number of objects."""
        return len(list(self.itershas()))

    def empty(self):
        iter = self.itershas()
        try:
            iter()
        except StopIteration:
            return True
        else:
            return False

    def __bool__(self):
        """Indicate whether this object has contents."""
        return not self.empty()


def tree_lookup_path(lookup_obj, root_sha, path):
    """Look up an object in a Git tree.

    :param lookup_obj: Callback for retrieving object by SHA1
    :param root_sha: SHA1 of the root tree
    :param path: Path to lookup
    :return: A tuple of (mode, SHA) of the resulting path.
    """
    tree = lookup_obj(root_sha)
    if not isinstance(tree, Tree):
        raise NotTreeError(root_sha)
    return tree.lookup_path(lookup_obj, path)


def _collect_filetree_revs(obj_store, tree_sha, kset):
    """Collect SHA1s of files and directories for specified tree.

    :param obj_store: Object store to get objects by SHA from
    :param tree_sha: tree reference to walk
    :param kset: set to fill with references to files and directories
    """
    filetree = obj_store[tree_sha]
    for name, mode, sha in filetree.iteritems():
        if not S_ISGITLINK(mode) and sha not in kset:
            kset.add(sha)
            if stat.S_ISDIR(mode):
                _collect_filetree_revs(obj_store, sha, kset)


def _split_commits_and_tags(obj_store, lst, ignore_unknown=False):
    """Split object id list into three lists with commit, tag, and other SHAs.

    Commits referenced by tags are included into commits
    list as well. Only SHA1s known in this repository will get
    through, and unless ignore_unknown argument is True, KeyError
    is thrown for SHA1 missing in the repository

    :param obj_store: Object store to get objects by SHA1 from
    :param lst: Collection of commit and tag SHAs
    :param ignore_unknown: True to skip SHA1 missing in the repository
        silently.
    :return: A tuple of (commits, tags, others) SHA1s
    """
    commits = set()
    tags = set()
    others = set()
    for e in lst:
        try:
            o = obj_store[e]
        except KeyError:
            if not ignore_unknown:
                raise
        else:
            if isinstance(o, Commit):
                commits.add(e)
            elif isinstance(o, Tag):
                tags.add(e)
                tagged = o.object[1]
                c, t, o = _split_commits_and_tags(
                    obj_store, [tagged], ignore_unknown=ignore_unknown)
                commits |= c
                tags |= t
                others |= o
            else:
                others.add(e)
    return (commits, tags, others)


class MissingObjectFinder(object):
    """Find the objects missing from another object store.

    :param object_store: Object store containing at least all objects to be
        sent
    :param haves: SHA1s of commits not to send (already present in target)
    :param wants: SHA1s of commits to send
    :param progress: Optional function to report progress to.
    :param get_tagged: Function that returns a dict of pointed-to sha -> tag
        sha for including tags.
    :param get_parents: Optional function for getting the parents of a commit.
    :param tagged: dict of pointed-to sha -> tag sha for including tags
    """

    def __init__(self, object_store, haves, wants, progress=None,
                 get_tagged=None, get_parents=lambda commit: commit.parents):
        self.object_store = object_store
        self._get_parents = get_parents
        # process Commits and Tags differently
        # Note, while haves may list commits/tags not available locally,
        # and such SHAs would get filtered out by _split_commits_and_tags,
        # wants shall list only known SHAs, and otherwise
        # _split_commits_and_tags fails with KeyError
        have_commits, have_tags, have_others = (
            _split_commits_and_tags(object_store, haves, True))
        want_commits, want_tags, want_others = (
            _split_commits_and_tags(object_store, wants, False))
        # all_ancestors is a set of commits that shall not be sent
        # (complete repository up to 'haves')
        all_ancestors = object_store._collect_ancestors(
            have_commits, get_parents=self._get_parents)[0]
        # all_missing - complete set of commits between haves and wants
        # common - commits from all_ancestors we hit into while
        # traversing parent hierarchy of wants
        missing_commits, common_commits = object_store._collect_ancestors(
            want_commits, all_ancestors, get_parents=self._get_parents)
        self.sha_done = set()
        # Now, fill sha_done with commits and revisions of
        # files and directories known to be both locally
        # and on target. Thus these commits and files
        # won't get selected for fetch
        for h in common_commits:
            self.sha_done.add(h)
            cmt = object_store[h]
            _collect_filetree_revs(object_store, cmt.tree, self.sha_done)
        # record tags we have as visited, too
        for t in have_tags:
            self.sha_done.add(t)

        missing_tags = want_tags.difference(have_tags)
        missing_others = want_others.difference(have_others)
        # in fact, what we 'want' is commits, tags, and others
        # we've found missing
        wants = missing_commits.union(missing_tags)
        wants = wants.union(missing_others)

        self.objects_to_send = set([(w, None, False) for w in wants])

        if progress is None:
            self.progress = lambda x: None
        else:
            self.progress = progress
        self._tagged = get_tagged and get_tagged() or {}

    def add_todo(self, entries):
        self.objects_to_send.update([e for e in entries
                                     if not e[0] in self.sha_done])

    def next(self):
        while True:
            if not self.objects_to_send:
                return None
            (sha, name, leaf) = self.objects_to_send.pop()
            if sha not in self.sha_done:
                break
        if not leaf:
            o = self.object_store[sha]
            if isinstance(o, Commit):
                self.add_todo([(o.tree, "", False)])
            elif isinstance(o, Tree):
                self.add_todo([(s, n, not stat.S_ISDIR(m))
                               for n, m, s in o.iteritems()
                               if not S_ISGITLINK(m)])
            elif isinstance(o, Tag):
                self.add_todo([(o.object[1], None, False)])
        if sha in self._tagged:
            self.add_todo([(self._tagged[sha], None, True)])
        self.sha_done.add(sha)
        self.progress(("counting objects: %d\r" %
                       len(self.sha_done)).encode('ascii'))
        return (sha, name)

    __next__ = next


class ObjectStoreGraphWalker(object):
    """Graph walker that finds what commits are missing from an object store.

    :ivar heads: Revisions without descendants in the local repo
    :ivar get_parents: Function to retrieve parents in the local repo
    """

    def __init__(self, local_heads, get_parents):
        """Create a new instance.

        :param local_heads: Heads to start search with
        :param get_parents: Function for finding the parents of a SHA1.
        """
        self.heads = set(local_heads)
        self.get_parents = get_parents
        self.parents = {}

    def ack(self, sha):
        """Ack that a revision and its ancestors are present in the source."""
        if len(sha) != 40:
            raise ValueError("unexpected sha %r received" % sha)
        ancestors = set([sha])

        # stop if we run out of heads to remove
        while self.heads:
            for a in ancestors:
                if a in self.heads:
                    self.heads.remove(a)

            # collect all ancestors
            new_ancestors = set()
            for a in ancestors:
                ps = self.parents.get(a)
                if ps is not None:
                    new_ancestors.update(ps)
                self.parents[a] = None

            # no more ancestors; stop
            if not new_ancestors:
                break

            ancestors = new_ancestors

    def next(self):
        """Iterate over ancestors of heads in the target."""
        if self.heads:
            ret = self.heads.pop()
            ps = self.get_parents(ret)
            self.parents[ret] = ps
            self.heads.update(
                [p for p in ps if p not in self.parents])
            return ret
        return None

    __next__ = next


def commit_tree_changes(object_store, tree, changes):
    """Commit a specified set of changes to a tree structure.

    This will apply a set of changes on top of an existing tree, storing new
    objects in object_store.

    changes are a list of tuples with (path, mode, object_sha).
    Paths can be both blobs and trees. See the mode and
    object sha to None deletes the path.

    This method works especially well if there are only a small
    number of changes to a big tree. For a large number of changes
    to a large tree, use e.g. commit_tree.

    :param object_store: Object store to store new objects in
        and retrieve old ones from.
    :param tree: Original tree root
    :param changes: changes to apply
    :return: New tree root object
    """
    # TODO(jelmer): Save up the objects and add them using .add_objects
    # rather than with individual calls to .add_object.
    nested_changes = {}
    for (path, new_mode, new_sha) in changes:
        try:
            (dirname, subpath) = path.split(b'/', 1)
        except ValueError:
            if new_sha is None:
                del tree[path]
            else:
                tree[path] = (new_mode, new_sha)
        else:
            nested_changes.setdefault(dirname, []).append(
                (subpath, new_mode, new_sha))
    for name, subchanges in nested_changes.items():
        try:
            orig_subtree = object_store[tree[name][1]]
        except KeyError:
            orig_subtree = Tree()
        subtree = commit_tree_changes(object_store, orig_subtree, subchanges)
        if len(subtree) == 0:
            del tree[name]
        else:
            tree[name] = (stat.S_IFDIR, subtree.id)
    object_store.add_object(tree)
    return tree


class OverlayObjectStore(BaseObjectStore):
    """Object store that can overlay multiple object stores."""

    def __init__(self, bases, add_store=None):
        self.bases = bases
        self.add_store = add_store

    def add_object(self, object):
        if self.add_store is None:
            raise NotImplementedError(self.add_object)
        return self.add_store.add_object(object)

    def add_objects(self, objects, progress=None):
        if self.add_store is None:
            raise NotImplementedError(self.add_object)
        return self.add_store.add_objects(objects, progress)

    @property
    def packs(self):
        ret = []
        for b in self.bases:
            ret.extend(b.packs)
        return ret

    def __iter__(self):
        done = set()
        for b in self.bases:
            for o_id in b:
                if o_id not in done:
                    yield o_id
                    done.add(o_id)

    def get_raw(self, sha_id):
        for b in self.bases:
            try:
                return b.get_raw(sha_id)
            except KeyError:
                pass
        else:
            raise KeyError(sha_id)

    def contains_packed(self, sha):
        for b in self.bases:
            if b.contains_packed(sha):
                return True
        else:
            return False

    def contains_loose(self, sha):
        for b in self.bases:
            if b.contains_loose(sha):
                return True
        else:
            return False