~bzr-pqm/bzr/bzr.dev

# Copyright (C) 2005, 2006, 2007, 2008 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., 59 Temple Place, Suite 330, Boston, MA  02111-1307  USA

import sys

from bzrlib.lazy_import import lazy_import

lazy_import(globals(), """

from itertools import izip

import time

from bzrlib import (

    debug,

    graph,

    osutils,

    pack,

    transactions,

    ui,

    xml5,

    xml6,

    xml7,

    )

from bzrlib.index import (

    CombinedGraphIndex,

    GraphIndex,

    GraphIndexBuilder,

    GraphIndexPrefixAdapter,

    InMemoryGraphIndex,

    )

from bzrlib.knit import (

    KnitPlainFactory,

    KnitVersionedFiles,

    _KnitGraphIndex,

    _DirectPackAccess,

    )

from bzrlib import tsort

""")

from bzrlib import (

    bzrdir,

    errors,

    lockable_files,

    lockdir,

    symbol_versioning,

    )

from bzrlib.decorators import needs_write_lock

from bzrlib.btree_index import (

    BTreeGraphIndex,

    BTreeBuilder,

    )

from bzrlib.index import (

    GraphIndex,

    InMemoryGraphIndex,

    )

from bzrlib.repofmt.knitrepo import KnitRepository

from bzrlib.repository import (

    CommitBuilder,

    MetaDirRepositoryFormat,

    RepositoryFormat,

    RootCommitBuilder,

    )

import bzrlib.revision as _mod_revision

from bzrlib.trace import (

    mutter,

    warning,

    )

class PackCommitBuilder(CommitBuilder):

    """A subclass of CommitBuilder 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):

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

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

            revprops=revprops, revision_id=revision_id)

        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(RootCommitBuilder):

    """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):

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

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

            revprops=revprops, revision_id=revision_id)

        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.

    """

    def __init__(self, revision_index, inventory_index, text_index,

        signature_index):

        """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.

        """

        self.revision_index = revision_index

        self.inventory_index = inventory_index

        self.text_index = text_index

        self.signature_index = signature_index

    def access_tuple(self):

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

        return self.pack_transport, self.file_name()

    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 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 _external_compression_parents_of_texts(self):

        keys = set()

        refs = set()

        for node in self.text_index.iter_all_entries():

            keys.add(node[1])

            refs.update(node[3][1])

        return refs - keys

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):

        """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)

        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 "<bzrlib.repofmt.pack_repo.Pack object at 0x%x, %s, %s" % (

            id(self), self.pack_transport, self.name)

class NewPack(Pack):

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

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

    # index_sizes array.

    index_definitions = {

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

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

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

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

        }

    def __init__(self, upload_transport, index_transport, pack_transport,

        upload_suffix='', file_mode=None, index_builder_class=None,

        index_class=None):

        """Create a NewPack instance.

        :param upload_transport: A writable transport for the pack to be

            incrementally uploaded to.

        :param index_transport: A writable transport for the pack's indices to

            be written to when the pack is finished.

        :param pack_transport: A writable transport for the pack to be renamed

            to when the upload is complete. This *must* be the same as

            upload_transport.clone('../packs').

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

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

        :param file_mode: An optional file mode to create the new files with.

        :param index_builder_class: Required keyword parameter - the class of

            index builder to use.

        :param index_class: Required keyword parameter - the class of index

            object to use.

        """

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

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

        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),

            )

        # When we make readonly indices, we need this.

        self.index_class = index_class

        # where should the new pack be opened

        self.upload_transport = upload_transport

        # where are indices written out to

        self.index_transport = index_transport

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

        self.pack_transport = 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 four-tuple with the length in bytes of the indices, once the pack

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

        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'

    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())

    def finish(self):

        """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._writer.end()

        if self._buffer[1]:

            self._write_data('', flush=True)

        self.name = self._hash.hexdigest()

        # 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')

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

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

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

            'revision signatures')

        self.write_stream.close()

        # 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

        self.upload_transport.rename(self.random_name,

                '../packs/' + self.name + '.pack')

        self._state = 'finished'

        if 'pack' in debug.debug_flags:

            # XXX: size might be interesting?

            mutter('%s: create_pack: pack renamed into place: %s%s->%s%s t+%6.3fs',

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

                self.pack_transport, self.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 index_name(self, index_type, name):

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

        return name + NewPack.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 NewPack.index_definitions[index_type][1]

    def _replace_index_with_readonly(self, index_type):

        setattr(self, index_type + '_index',

            self.index_class(self.index_transport,

                self.index_name(index_type, self.name),

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

    def set_write_cache_size(self, size):

        self._cache_limit = size

    def _write_index(self, index_type, index, label):

        """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)

        self.index_sizes[self.index_offset(index_type)] = \

            self.index_transport.put_file(index_name, index.finish(),

            mode=self._file_mode)

        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):

        """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/False to indicate

            if reloading actually changed anything.

        """

        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)

        self.add_callback = None

    def replace_indices(self, index_to_pack, indices):

        """Replace the current mappings with fresh ones.

        This should probably not be used eventually, rather incremental add and

        removal of indices. It has been added during refactoring of existing

        code.

        :param index_to_pack: A mapping from index objects to

            (transport, name) tuples for the pack file data.

        :param indices: A list of indices.

        """

        # refresh the revision pack map dict without replacing the instance.

        self.index_to_pack.clear()

        self.index_to_pack.update(index_to_pack)

        # XXX: API break - clearly a 'replace' method would be good?

        self.combined_index._indices[:] = indices

        # the current add nodes callback for the current writable index if

        # there is one.

        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)

    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[:]

        self.add_callback = None

    def remove_index(self, index, pack):

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

        :param index: An index from the pack parameter.

        :param pack: A Pack instance.

        """

        del self.index_to_pack[index]

        self.combined_index._indices.remove(index)

        if (self.add_callback is not None and

            getattr(index, 'add_nodes', None) == self.add_callback):

            self.add_callback = None

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

class Packer(object):

    """Create a pack from packs."""

    def __init__(self, pack_collection, packs, suffix, revision_ids=None,

                 reload_func=None):

        """Create a Packer.

        :param pack_collection: A RepositoryPackCollection object where the

            new pack is being written to.

        :param packs: The packs to combine.

        :param suffix: The suffix to use on the temporary files for the pack.

        :param revision_ids: Revision ids to limit the pack to.

        :param reload_func: A function to call if a pack file/index goes

            missing. The side effect of calling this function should be to

            update self.packs. See also AggregateIndex

        """

        self.packs = packs

        self.suffix = suffix

        self.revision_ids = revision_ids

        # The pack object we are creating.

        self.new_pack = None

        self._pack_collection = pack_collection

        self._reload_func = reload_func

        # The index layer keys for the revisions being copied. None for 'all

        # objects'.

        self._revision_keys = None

        # What text keys to copy. None for 'all texts'. This is set by

        # _copy_inventory_texts

        self._text_filter = None

        self._extra_init()

    def _extra_init(self):

        """A template hook to allow extending the constructor trivially."""

    def pack(self, pb=None):

        """Create a new pack by reading data from other packs.

        This does little more than a bulk copy of data. One key difference

        is that data with the same item key across multiple packs is elided

        from the output. The new pack is written into the current pack store

        along with its indices, and the name added to the pack names. The 

        source packs are not altered and are not required to be in the current

        pack collection.

        :param pb: An optional progress bar to use. A nested bar is created if

            this is None.

        :return: A Pack object, or None if nothing was copied.

        """

        # open a pack - using the same name as the last temporary file

        # - which has already been flushed, so its safe.

        # XXX: - duplicate code warning with start_write_group; fix before

        #      considering 'done'.

        if self._pack_collection._new_pack is not None:

            raise errors.BzrError('call to %s.pack() while another pack is'

                                  ' being written.'

                                  % (self.__class__.__name__,))

        if self.revision_ids is not None:

            if len(self.revision_ids) == 0:

                # silly fetch request.

                return None

            else:

                self.revision_ids = frozenset(self.revision_ids)

                self.revision_keys = frozenset((revid,) for revid in

                    self.revision_ids)

        if pb is None:

            self.pb = ui.ui_factory.nested_progress_bar()

        else:

            self.pb = pb

        try:

            return self._create_pack_from_packs()

        finally:

            if pb is None:

                self.pb.finished()

    def open_pack(self):

        """Open a pack for the pack we are creating."""

        return NewPack(self._pack_collection._upload_transport,

            self._pack_collection._index_transport,

            self._pack_collection._pack_transport, upload_suffix=self.suffix,

            file_mode=self._pack_collection.repo.bzrdir._get_file_mode(),

            index_builder_class=self._pack_collection._index_builder_class,

            index_class=self._pack_collection._index_class)

    def _copy_revision_texts(self):

        """Copy revision data to the new pack."""

        # select revisions

        if self.revision_ids:

            revision_keys = [(revision_id,) for revision_id in self.revision_ids]

        else:

            revision_keys = None

        completed_keys = []

        while True:

            try:

                # select revision keys

                revision_index_map = self._pack_collection._packs_list_to_pack_map_and_index_list(

                    self.packs, 'revision_index')[0]

                revision_nodes = self._pack_collection._index_contents(

                                    revision_index_map, revision_keys)

                # copy revision keys and adjust values

                self.pb.update("Copying revision texts", 1)

                total_items, readv_group_iter = self._revision_node_readv(revision_nodes)

                list(self._copy_nodes_graph(revision_index_map, self.new_pack._writer,

                    self.new_pack.revision_index, readv_group_iter,

                    total_items, completed_keys=completed_keys))

                break

            except errors.NoSuchFile:

                if self._reload_func is None:

                    raise

                # A pack file went missing, try reloading in case it was just

                # someone else repacking the repo.

                if not self._reload_func():

                    raise

        if 'pack' in debug.debug_flags:

            mutter('%s: create_pack: revisions copied: %s%s %d items t+%6.3fs',

                time.ctime(), self._pack_collection._upload_transport.base,

                self.new_pack.random_name,

                self.new_pack.revision_index.key_count(),

                time.time() - self.new_pack.start_time)

        self._revision_keys = revision_keys

    def _copy_inventory_texts(self):

        """Copy the inventory texts to the new pack.

        self._revision_keys is used to determine what inventories to copy.

        Sets self._text_filter appropriately.

        """

        # select inventory keys

        inv_keys = self._revision_keys # currently the same keyspace, and note that

        # querying for keys here could introduce a bug where an inventory item

        # is missed, so do not change it to query separately without cross

        # checking like the text key check below.

        inventory_index_map = self._pack_collection._packs_list_to_pack_map_and_index_list(

            self.packs, 'inventory_index')[0]

        inv_nodes = self._pack_collection._index_contents(inventory_index_map, inv_keys)

        # copy inventory keys and adjust values