~bzr-pqm/bzr/bzr.dev

# Copyright (C) 2007-2011 Canonical Ltd

#

# This program is free software; you can redistribute it and/or modify

# it under the terms of the GNU General Public License as published by

# the Free Software Foundation; either version 2 of the License, or

# (at your option) any later version.

#

# This program is distributed in the hope that it will be useful,

# but WITHOUT ANY WARRANTY; without even the implied warranty of

# MERCHANTABILITY or FITNESS FOR A PARTICULAR PURPOSE.  See the

# GNU General Public License for more details.

#

# You should have received a copy of the GNU General Public License

# along with this program; if not, write to the Free Software

# Foundation, Inc., 51 Franklin Street, Fifth Floor, Boston, MA 02110-1301 USA

from __future__ import absolute_import

import re

import sys

from bzrlib.lazy_import import lazy_import

lazy_import(globals(), """

from itertools import izip

import time

from bzrlib import (

    chk_map,

    cleanup,

    config,

    debug,

    graph,

    osutils,

    pack,

    transactions,

    tsort,

    ui,

    )

from bzrlib.index import (

    CombinedGraphIndex,

    GraphIndexPrefixAdapter,

    )

""")

from bzrlib import (

    btree_index,

    errors,

    lockable_files,

    lockdir,

    )

from bzrlib.decorators import (

    needs_read_lock,

    needs_write_lock,

    only_raises,

    )

from bzrlib.lock import LogicalLockResult

from bzrlib.repository import (

    _LazyListJoin,

    MetaDirRepository,

    RepositoryFormatMetaDir,

    RepositoryWriteLockResult,

    )

from bzrlib.vf_repository import (

    MetaDirVersionedFileRepository,

    MetaDirVersionedFileRepositoryFormat,

    VersionedFileCommitBuilder,

    VersionedFileRootCommitBuilder,

    )

from bzrlib.trace import (

    mutter,

    note,

    warning,

    )

class PackCommitBuilder(VersionedFileCommitBuilder):

    """Subclass of VersionedFileCommitBuilder to add texts with pack semantics.

    Specifically this uses one knit object rather than one knit object per

    added text, reducing memory and object pressure.

    """

    def __init__(self, repository, parents, config, timestamp=None,

                 timezone=None, committer=None, revprops=None,

                 revision_id=None, lossy=False):

        VersionedFileCommitBuilder.__init__(self, repository, parents, config,

            timestamp=timestamp, timezone=timezone, committer=committer,

            revprops=revprops, revision_id=revision_id, lossy=lossy)

        self._file_graph = graph.Graph(

            repository._pack_collection.text_index.combined_index)

    def _heads(self, file_id, revision_ids):

        keys = [(file_id, revision_id) for revision_id in revision_ids]

        return set([key[1] for key in self._file_graph.heads(keys)])

class PackRootCommitBuilder(VersionedFileRootCommitBuilder):

    """A subclass of RootCommitBuilder to add texts with pack semantics.

    Specifically this uses one knit object rather than one knit object per

    added text, reducing memory and object pressure.

    """

    def __init__(self, repository, parents, config, timestamp=None,

                 timezone=None, committer=None, revprops=None,

                 revision_id=None, lossy=False):

        super(PackRootCommitBuilder, self).__init__(repository, parents,

            config, timestamp=timestamp, timezone=timezone,

            committer=committer, revprops=revprops, revision_id=revision_id,

            lossy=lossy)

        self._file_graph = graph.Graph(

            repository._pack_collection.text_index.combined_index)

    def _heads(self, file_id, revision_ids):

        keys = [(file_id, revision_id) for revision_id in revision_ids]

        return set([key[1] for key in self._file_graph.heads(keys)])

class Pack(object):

    """An in memory proxy for a pack and its indices.

    This is a base class that is not directly used, instead the classes

    ExistingPack and NewPack are used.

    """

    # A map of index 'type' to the file extension and position in the

    # index_sizes array.

    index_definitions = {

        'chk': ('.cix', 4),

        'revision': ('.rix', 0),

        'inventory': ('.iix', 1),

        'text': ('.tix', 2),

        'signature': ('.six', 3),

        }

    def __init__(self, revision_index, inventory_index, text_index,

        signature_index, chk_index=None):

        """Create a pack instance.

        :param revision_index: A GraphIndex for determining what revisions are

            present in the Pack and accessing the locations of their texts.

        :param inventory_index: A GraphIndex for determining what inventories are

            present in the Pack and accessing the locations of their

            texts/deltas.

        :param text_index: A GraphIndex for determining what file texts

            are present in the pack and accessing the locations of their

            texts/deltas (via (fileid, revisionid) tuples).

        :param signature_index: A GraphIndex for determining what signatures are

            present in the Pack and accessing the locations of their texts.

        :param chk_index: A GraphIndex for accessing content by CHK, if the

            pack has one.

        """

        self.revision_index = revision_index

        self.inventory_index = inventory_index

        self.text_index = text_index

        self.signature_index = signature_index

        self.chk_index = chk_index

    def access_tuple(self):

        """Return a tuple (transport, name) for the pack content."""

        return self.pack_transport, self.file_name()

    def _check_references(self):

        """Make sure our external references are present.

        Packs are allowed to have deltas whose base is not in the pack, but it

        must be present somewhere in this collection.  It is not allowed to

        have deltas based on a fallback repository.

        (See <https://bugs.launchpad.net/bzr/+bug/288751>)

        """

        missing_items = {}

        for (index_name, external_refs, index) in [

            ('texts',

                self._get_external_refs(self.text_index),

                self._pack_collection.text_index.combined_index),

            ('inventories',

                self._get_external_refs(self.inventory_index),

                self._pack_collection.inventory_index.combined_index),

            ]:

            missing = external_refs.difference(

                k for (idx, k, v, r) in

                index.iter_entries(external_refs))

            if missing:

                missing_items[index_name] = sorted(list(missing))

        if missing_items:

            from pprint import pformat

            raise errors.BzrCheckError(

                "Newly created pack file %r has delta references to "

                "items not in its repository:\n%s"

                % (self, pformat(missing_items)))

    def file_name(self):

        """Get the file name for the pack on disk."""

        return self.name + '.pack'

    def get_revision_count(self):

        return self.revision_index.key_count()

    def index_name(self, index_type, name):

        """Get the disk name of an index type for pack name 'name'."""

        return name + Pack.index_definitions[index_type][0]

    def index_offset(self, index_type):

        """Get the position in a index_size array for a given index type."""

        return Pack.index_definitions[index_type][1]

    def inventory_index_name(self, name):

        """The inv index is the name + .iix."""

        return self.index_name('inventory', name)

    def revision_index_name(self, name):

        """The revision index is the name + .rix."""

        return self.index_name('revision', name)

    def signature_index_name(self, name):

        """The signature index is the name + .six."""

        return self.index_name('signature', name)

    def text_index_name(self, name):

        """The text index is the name + .tix."""

        return self.index_name('text', name)

    def _replace_index_with_readonly(self, index_type):

        unlimited_cache = False

        if index_type == 'chk':

            unlimited_cache = True

        index = self.index_class(self.index_transport,

                    self.index_name(index_type, self.name),

                    self.index_sizes[self.index_offset(index_type)],

                    unlimited_cache=unlimited_cache)

        if index_type == 'chk':

            index._leaf_factory = btree_index._gcchk_factory

        setattr(self, index_type + '_index', index)

class ExistingPack(Pack):

    """An in memory proxy for an existing .pack and its disk indices."""

    def __init__(self, pack_transport, name, revision_index, inventory_index,

        text_index, signature_index, chk_index=None):

        """Create an ExistingPack object.

        :param pack_transport: The transport where the pack file resides.

        :param name: The name of the pack on disk in the pack_transport.

        """

        Pack.__init__(self, revision_index, inventory_index, text_index,

            signature_index, chk_index)

        self.name = name

        self.pack_transport = pack_transport

        if None in (revision_index, inventory_index, text_index,

                signature_index, name, pack_transport):

            raise AssertionError()

    def __eq__(self, other):

        return self.__dict__ == other.__dict__

    def __ne__(self, other):

        return not self.__eq__(other)

    def __repr__(self):

        return "<%s.%s object at 0x%x, %s, %s" % (

            self.__class__.__module__, self.__class__.__name__, id(self),

            self.pack_transport, self.name)

class ResumedPack(ExistingPack):

    def __init__(self, name, revision_index, inventory_index, text_index,

        signature_index, upload_transport, pack_transport, index_transport,

        pack_collection, chk_index=None):

        """Create a ResumedPack object."""

        ExistingPack.__init__(self, pack_transport, name, revision_index,

            inventory_index, text_index, signature_index,

            chk_index=chk_index)

        self.upload_transport = upload_transport

        self.index_transport = index_transport

        self.index_sizes = [None, None, None, None]

        indices = [

            ('revision', revision_index),

            ('inventory', inventory_index),

            ('text', text_index),

            ('signature', signature_index),

            ]

        if chk_index is not None:

            indices.append(('chk', chk_index))

            self.index_sizes.append(None)

        for index_type, index in indices:

            offset = self.index_offset(index_type)

            self.index_sizes[offset] = index._size

        self.index_class = pack_collection._index_class

        self._pack_collection = pack_collection

        self._state = 'resumed'

        # XXX: perhaps check that the .pack file exists?

    def access_tuple(self):

        if self._state == 'finished':

            return Pack.access_tuple(self)

        elif self._state == 'resumed':

            return self.upload_transport, self.file_name()

        else:

            raise AssertionError(self._state)

    def abort(self):

        self.upload_transport.delete(self.file_name())

        indices = [self.revision_index, self.inventory_index, self.text_index,

            self.signature_index]

        if self.chk_index is not None:

            indices.append(self.chk_index)

        for index in indices:

            index._transport.delete(index._name)

    def finish(self):

        self._check_references()

        index_types = ['revision', 'inventory', 'text', 'signature']

        if self.chk_index is not None:

            index_types.append('chk')

        for index_type in index_types:

            old_name = self.index_name(index_type, self.name)

            new_name = '../indices/' + old_name

            self.upload_transport.move(old_name, new_name)

            self._replace_index_with_readonly(index_type)

        new_name = '../packs/' + self.file_name()

        self.upload_transport.move(self.file_name(), new_name)

        self._state = 'finished'

    def _get_external_refs(self, index):

        """Return compression parents for this index that are not present.

        This returns any compression parents that are referenced by this index,

        which are not contained *in* this index. They may be present elsewhere.

        """

        return index.external_references(1)

class NewPack(Pack):

    """An in memory proxy for a pack which is being created."""

    def __init__(self, pack_collection, upload_suffix='', file_mode=None):

        """Create a NewPack instance.

        :param pack_collection: A PackCollection into which this is being inserted.

        :param upload_suffix: An optional suffix to be given to any temporary

            files created during the pack creation. e.g '.autopack'

        :param file_mode: Unix permissions for newly created file.

        """

        # The relative locations of the packs are constrained, but all are

        # passed in because the caller has them, so as to avoid object churn.

        index_builder_class = pack_collection._index_builder_class

        if pack_collection.chk_index is not None:

            chk_index = index_builder_class(reference_lists=0)

        else:

            chk_index = None

        Pack.__init__(self,

            # Revisions: parents list, no text compression.

            index_builder_class(reference_lists=1),

            # Inventory: We want to map compression only, but currently the

            # knit code hasn't been updated enough to understand that, so we

            # have a regular 2-list index giving parents and compression

            # source.

            index_builder_class(reference_lists=2),

            # Texts: compression and per file graph, for all fileids - so two

            # reference lists and two elements in the key tuple.

            index_builder_class(reference_lists=2, key_elements=2),

            # Signatures: Just blobs to store, no compression, no parents

            # listing.

            index_builder_class(reference_lists=0),

            # CHK based storage - just blobs, no compression or parents.

            chk_index=chk_index

            )

        self._pack_collection = pack_collection

        # When we make readonly indices, we need this.

        self.index_class = pack_collection._index_class

        # where should the new pack be opened

        self.upload_transport = pack_collection._upload_transport

        # where are indices written out to

        self.index_transport = pack_collection._index_transport

        # where is the pack renamed to when it is finished?

        self.pack_transport = pack_collection._pack_transport

        # What file mode to upload the pack and indices with.

        self._file_mode = file_mode

        # tracks the content written to the .pack file.

        self._hash = osutils.md5()

        # a tuple with the length in bytes of the indices, once the pack

        # is finalised. (rev, inv, text, sigs, chk_if_in_use)

        self.index_sizes = None

        # How much data to cache when writing packs. Note that this is not

        # synchronised with reads, because it's not in the transport layer, so

        # is not safe unless the client knows it won't be reading from the pack

        # under creation.

        self._cache_limit = 0

        # the temporary pack file name.

        self.random_name = osutils.rand_chars(20) + upload_suffix

        # when was this pack started ?

        self.start_time = time.time()

        # open an output stream for the data added to the pack.

        self.write_stream = self.upload_transport.open_write_stream(

            self.random_name, mode=self._file_mode)

        if 'pack' in debug.debug_flags:

            mutter('%s: create_pack: pack stream open: %s%s t+%6.3fs',

                time.ctime(), self.upload_transport.base, self.random_name,

                time.time() - self.start_time)

        # A list of byte sequences to be written to the new pack, and the

        # aggregate size of them.  Stored as a list rather than separate

        # variables so that the _write_data closure below can update them.

        self._buffer = [[], 0]

        # create a callable for adding data

        #

        # robertc says- this is a closure rather than a method on the object

        # so that the variables are locals, and faster than accessing object

        # members.

        def _write_data(bytes, flush=False, _buffer=self._buffer,

            _write=self.write_stream.write, _update=self._hash.update):

            _buffer[0].append(bytes)

            _buffer[1] += len(bytes)

            # buffer cap

            if _buffer[1] > self._cache_limit or flush:

                bytes = ''.join(_buffer[0])

                _write(bytes)

                _update(bytes)

                _buffer[:] = [[], 0]

        # expose this on self, for the occasion when clients want to add data.

        self._write_data = _write_data

        # a pack writer object to serialise pack records.

        self._writer = pack.ContainerWriter(self._write_data)

        self._writer.begin()

        # what state is the pack in? (open, finished, aborted)

        self._state = 'open'

        # no name until we finish writing the content

        self.name = None

    def abort(self):

        """Cancel creating this pack."""

        self._state = 'aborted'

        self.write_stream.close()

        # Remove the temporary pack file.

        self.upload_transport.delete(self.random_name)

        # The indices have no state on disk.

    def access_tuple(self):

        """Return a tuple (transport, name) for the pack content."""

        if self._state == 'finished':

            return Pack.access_tuple(self)

        elif self._state == 'open':

            return self.upload_transport, self.random_name

        else:

            raise AssertionError(self._state)

    def data_inserted(self):

        """True if data has been added to this pack."""

        return bool(self.get_revision_count() or

            self.inventory_index.key_count() or

            self.text_index.key_count() or

            self.signature_index.key_count() or

            (self.chk_index is not None and self.chk_index.key_count()))

    def finish_content(self):

        if self.name is not None:

            return

        self._writer.end()

        if self._buffer[1]:

            self._write_data('', flush=True)

        self.name = self._hash.hexdigest()

    def finish(self, suspend=False):

        """Finish the new pack.

        This:

         - finalises the content

         - assigns a name (the md5 of the content, currently)

         - writes out the associated indices

         - renames the pack into place.

         - stores the index size tuple for the pack in the index_sizes

           attribute.

        """

        self.finish_content()

        if not suspend:

            self._check_references()

        # write indices

        # XXX: It'd be better to write them all to temporary names, then

        # rename them all into place, so that the window when only some are

        # visible is smaller.  On the other hand none will be seen until

        # they're in the names list.

        self.index_sizes = [None, None, None, None]

        self._write_index('revision', self.revision_index, 'revision',

            suspend)

        self._write_index('inventory', self.inventory_index, 'inventory',

            suspend)

        self._write_index('text', self.text_index, 'file texts', suspend)

        self._write_index('signature', self.signature_index,

            'revision signatures', suspend)

        if self.chk_index is not None:

            self.index_sizes.append(None)

            self._write_index('chk', self.chk_index,

                'content hash bytes', suspend)

        self.write_stream.close(

            want_fdatasync=self._pack_collection.config_stack.get('repository.fdatasync'))

        # Note that this will clobber an existing pack with the same name,

        # without checking for hash collisions. While this is undesirable this

        # is something that can be rectified in a subsequent release. One way

        # to rectify it may be to leave the pack at the original name, writing

        # its pack-names entry as something like 'HASH: index-sizes

        # temporary-name'. Allocate that and check for collisions, if it is

        # collision free then rename it into place. If clients know this scheme

        # they can handle missing-file errors by:

        #  - try for HASH.pack

        #  - try for temporary-name

        #  - refresh the pack-list to see if the pack is now absent

        new_name = self.name + '.pack'

        if not suspend:

            new_name = '../packs/' + new_name

        self.upload_transport.move(self.random_name, new_name)

        self._state = 'finished'

        if 'pack' in debug.debug_flags:

            # XXX: size might be interesting?

            mutter('%s: create_pack: pack finished: %s%s->%s t+%6.3fs',

                time.ctime(), self.upload_transport.base, self.random_name,

                new_name, time.time() - self.start_time)

    def flush(self):

        """Flush any current data."""

        if self._buffer[1]:

            bytes = ''.join(self._buffer[0])

            self.write_stream.write(bytes)

            self._hash.update(bytes)

            self._buffer[:] = [[], 0]

    def _get_external_refs(self, index):

        return index._external_references()

    def set_write_cache_size(self, size):

        self._cache_limit = size

    def _write_index(self, index_type, index, label, suspend=False):

        """Write out an index.

        :param index_type: The type of index to write - e.g. 'revision'.

        :param index: The index object to serialise.

        :param label: What label to give the index e.g. 'revision'.

        """

        index_name = self.index_name(index_type, self.name)

        if suspend:

            transport = self.upload_transport

        else:

            transport = self.index_transport

        index_tempfile = index.finish()

        index_bytes = index_tempfile.read()

        write_stream = transport.open_write_stream(index_name,

            mode=self._file_mode)

        write_stream.write(index_bytes)

        write_stream.close(

            want_fdatasync=self._pack_collection.config_stack.get('repository.fdatasync'))

        self.index_sizes[self.index_offset(index_type)] = len(index_bytes)

        if 'pack' in debug.debug_flags:

            # XXX: size might be interesting?

            mutter('%s: create_pack: wrote %s index: %s%s t+%6.3fs',

                time.ctime(), label, self.upload_transport.base,

                self.random_name, time.time() - self.start_time)

        # Replace the writable index on this object with a readonly,

        # presently unloaded index. We should alter

        # the index layer to make its finish() error if add_node is

        # subsequently used. RBC

        self._replace_index_with_readonly(index_type)

class AggregateIndex(object):

    """An aggregated index for the RepositoryPackCollection.

    AggregateIndex is reponsible for managing the PackAccess object,

    Index-To-Pack mapping, and all indices list for a specific type of index

    such as 'revision index'.

    A CombinedIndex provides an index on a single key space built up

    from several on-disk indices.  The AggregateIndex builds on this

    to provide a knit access layer, and allows having up to one writable

    index within the collection.

    """

    # XXX: Probably 'can be written to' could/should be separated from 'acts

    # like a knit index' -- mbp 20071024

    def __init__(self, reload_func=None, flush_func=None):

        """Create an AggregateIndex.

        :param reload_func: A function to call if we find we are missing an

            index. Should have the form reload_func() => True if the list of

            active pack files has changed.

        """

        self._reload_func = reload_func

        self.index_to_pack = {}

        self.combined_index = CombinedGraphIndex([], reload_func=reload_func)

        self.data_access = _DirectPackAccess(self.index_to_pack,

                                             reload_func=reload_func,

                                             flush_func=flush_func)

        self.add_callback = None

    def add_index(self, index, pack):

        """Add index to the aggregate, which is an index for Pack pack.

        Future searches on the aggregate index will seach this new index

        before all previously inserted indices.

        :param index: An Index for the pack.

        :param pack: A Pack instance.

        """

        # expose it to the index map

        self.index_to_pack[index] = pack.access_tuple()

        # put it at the front of the linear index list

        self.combined_index.insert_index(0, index, pack.name)

    def add_writable_index(self, index, pack):

        """Add an index which is able to have data added to it.

        There can be at most one writable index at any time.  Any

        modifications made to the knit are put into this index.

        :param index: An index from the pack parameter.

        :param pack: A Pack instance.

        """

        if self.add_callback is not None:

            raise AssertionError(

                "%s already has a writable index through %s" % \

                (self, self.add_callback))

        # allow writing: queue writes to a new index

        self.add_index(index, pack)

        # Updates the index to packs mapping as a side effect,

        self.data_access.set_writer(pack._writer, index, pack.access_tuple())

        self.add_callback = index.add_nodes

    def clear(self):

        """Reset all the aggregate data to nothing."""

        self.data_access.set_writer(None, None, (None, None))

        self.index_to_pack.clear()

        del self.combined_index._indices[:]

        del self.combined_index._index_names[:]

        self.add_callback = None

    def remove_index(self, index):

        """Remove index from the indices used to answer queries.