1
# Copyright (C) 2005, 2006, 2007, 2008 Canonical Ltd
3
# This program is free software; you can redistribute it and/or modify
4
# it under the terms of the GNU General Public License as published by
5
# the Free Software Foundation; either version 2 of the License, or
6
# (at your option) any later version.
8
# This program is distributed in the hope that it will be useful,
9
# but WITHOUT ANY WARRANTY; without even the implied warranty of
10
# MERCHANTABILITY or FITNESS FOR A PARTICULAR PURPOSE. See the
11
# GNU General Public License for more details.
13
# You should have received a copy of the GNU General Public License
14
# along with this program; if not, write to the Free Software
15
# Foundation, Inc., 59 Temple Place, Suite 330, Boston, MA 02111-1307 USA
17
from bzrlib.lazy_import import lazy_import
18
lazy_import(globals(), """
19
from itertools import izip
30
from bzrlib.index import (
35
GraphIndexPrefixAdapter,
37
from bzrlib.knit import (
43
from bzrlib.osutils import rand_chars, split_lines
44
from bzrlib.pack import ContainerWriter
45
from bzrlib.store import revision
46
from bzrlib import tsort
61
from bzrlib.decorators import needs_read_lock, needs_write_lock
62
from bzrlib.repofmt.knitrepo import KnitRepository
63
from bzrlib.repository import (
67
MetaDirRepositoryFormat,
71
import bzrlib.revision as _mod_revision
72
from bzrlib.store.versioned import VersionedFileStore
73
from bzrlib.trace import (
81
class PackCommitBuilder(CommitBuilder):
82
"""A subclass of CommitBuilder to add texts with pack semantics.
84
Specifically this uses one knit object rather than one knit object per
85
added text, reducing memory and object pressure.
88
def __init__(self, repository, parents, config, timestamp=None,
89
timezone=None, committer=None, revprops=None,
91
CommitBuilder.__init__(self, repository, parents, config,
92
timestamp=timestamp, timezone=timezone, committer=committer,
93
revprops=revprops, revision_id=revision_id)
94
self._file_graph = graph.Graph(
95
repository._pack_collection.text_index.combined_index)
97
def _heads(self, file_id, revision_ids):
98
keys = [(file_id, revision_id) for revision_id in revision_ids]
99
return set([key[1] for key in self._file_graph.heads(keys)])
102
class PackRootCommitBuilder(RootCommitBuilder):
103
"""A subclass of RootCommitBuilder to add texts with pack semantics.
105
Specifically this uses one knit object rather than one knit object per
106
added text, reducing memory and object pressure.
109
def __init__(self, repository, parents, config, timestamp=None,
110
timezone=None, committer=None, revprops=None,
112
CommitBuilder.__init__(self, repository, parents, config,
113
timestamp=timestamp, timezone=timezone, committer=committer,
114
revprops=revprops, revision_id=revision_id)
115
self._file_graph = graph.Graph(
116
repository._pack_collection.text_index.combined_index)
118
def _heads(self, file_id, revision_ids):
119
keys = [(file_id, revision_id) for revision_id in revision_ids]
120
return set([key[1] for key in self._file_graph.heads(keys)])
124
"""An in memory proxy for a pack and its indices.
126
This is a base class that is not directly used, instead the classes
127
ExistingPack and NewPack are used.
130
def __init__(self, revision_index, inventory_index, text_index,
132
"""Create a pack instance.
134
:param revision_index: A GraphIndex for determining what revisions are
135
present in the Pack and accessing the locations of their texts.
136
:param inventory_index: A GraphIndex for determining what inventories are
137
present in the Pack and accessing the locations of their
139
:param text_index: A GraphIndex for determining what file texts
140
are present in the pack and accessing the locations of their
141
texts/deltas (via (fileid, revisionid) tuples).
142
:param signature_index: A GraphIndex for determining what signatures are
143
present in the Pack and accessing the locations of their texts.
145
self.revision_index = revision_index
146
self.inventory_index = inventory_index
147
self.text_index = text_index
148
self.signature_index = signature_index
150
def access_tuple(self):
151
"""Return a tuple (transport, name) for the pack content."""
152
return self.pack_transport, self.file_name()
155
"""Get the file name for the pack on disk."""
156
return self.name + '.pack'
158
def get_revision_count(self):
159
return self.revision_index.key_count()
161
def inventory_index_name(self, name):
162
"""The inv index is the name + .iix."""
163
return self.index_name('inventory', name)
165
def revision_index_name(self, name):
166
"""The revision index is the name + .rix."""
167
return self.index_name('revision', name)
169
def signature_index_name(self, name):
170
"""The signature index is the name + .six."""
171
return self.index_name('signature', name)
173
def text_index_name(self, name):
174
"""The text index is the name + .tix."""
175
return self.index_name('text', name)
177
def _external_compression_parents_of_texts(self):
180
for node in self.text_index.iter_all_entries():
182
refs.update(node[3][1])
186
class ExistingPack(Pack):
187
"""An in memory proxy for an existing .pack and its disk indices."""
189
def __init__(self, pack_transport, name, revision_index, inventory_index,
190
text_index, signature_index):
191
"""Create an ExistingPack object.
193
:param pack_transport: The transport where the pack file resides.
194
:param name: The name of the pack on disk in the pack_transport.
196
Pack.__init__(self, revision_index, inventory_index, text_index,
199
self.pack_transport = pack_transport
200
if None in (revision_index, inventory_index, text_index,
201
signature_index, name, pack_transport):
202
raise AssertionError()
204
def __eq__(self, other):
205
return self.__dict__ == other.__dict__
207
def __ne__(self, other):
208
return not self.__eq__(other)
211
return "<bzrlib.repofmt.pack_repo.Pack object at 0x%x, %s, %s" % (
212
id(self), self.pack_transport, self.name)
216
"""An in memory proxy for a pack which is being created."""
218
# A map of index 'type' to the file extension and position in the
220
index_definitions = {
221
'revision': ('.rix', 0),
222
'inventory': ('.iix', 1),
224
'signature': ('.six', 3),
227
def __init__(self, upload_transport, index_transport, pack_transport,
228
upload_suffix='', file_mode=None):
229
"""Create a NewPack instance.
231
:param upload_transport: A writable transport for the pack to be
232
incrementally uploaded to.
233
:param index_transport: A writable transport for the pack's indices to
234
be written to when the pack is finished.
235
:param pack_transport: A writable transport for the pack to be renamed
236
to when the upload is complete. This *must* be the same as
237
upload_transport.clone('../packs').
238
:param upload_suffix: An optional suffix to be given to any temporary
239
files created during the pack creation. e.g '.autopack'
240
:param file_mode: An optional file mode to create the new files with.
242
# The relative locations of the packs are constrained, but all are
243
# passed in because the caller has them, so as to avoid object churn.
245
# Revisions: parents list, no text compression.
246
InMemoryGraphIndex(reference_lists=1),
247
# Inventory: We want to map compression only, but currently the
248
# knit code hasn't been updated enough to understand that, so we
249
# have a regular 2-list index giving parents and compression
251
InMemoryGraphIndex(reference_lists=2),
252
# Texts: compression and per file graph, for all fileids - so two
253
# reference lists and two elements in the key tuple.
254
InMemoryGraphIndex(reference_lists=2, key_elements=2),
255
# Signatures: Just blobs to store, no compression, no parents
257
InMemoryGraphIndex(reference_lists=0),
259
# where should the new pack be opened
260
self.upload_transport = upload_transport
261
# where are indices written out to
262
self.index_transport = index_transport
263
# where is the pack renamed to when it is finished?
264
self.pack_transport = pack_transport
265
# What file mode to upload the pack and indices with.
266
self._file_mode = file_mode
267
# tracks the content written to the .pack file.
268
self._hash = md5.new()
269
# a four-tuple with the length in bytes of the indices, once the pack
270
# is finalised. (rev, inv, text, sigs)
271
self.index_sizes = None
272
# How much data to cache when writing packs. Note that this is not
273
# synchronised with reads, because it's not in the transport layer, so
274
# is not safe unless the client knows it won't be reading from the pack
276
self._cache_limit = 0
277
# the temporary pack file name.
278
self.random_name = rand_chars(20) + upload_suffix
279
# when was this pack started ?
280
self.start_time = time.time()
281
# open an output stream for the data added to the pack.
282
self.write_stream = self.upload_transport.open_write_stream(
283
self.random_name, mode=self._file_mode)
284
if 'pack' in debug.debug_flags:
285
mutter('%s: create_pack: pack stream open: %s%s t+%6.3fs',
286
time.ctime(), self.upload_transport.base, self.random_name,
287
time.time() - self.start_time)
288
# A list of byte sequences to be written to the new pack, and the
289
# aggregate size of them. Stored as a list rather than separate
290
# variables so that the _write_data closure below can update them.
291
self._buffer = [[], 0]
292
# create a callable for adding data
294
# robertc says- this is a closure rather than a method on the object
295
# so that the variables are locals, and faster than accessing object
297
def _write_data(bytes, flush=False, _buffer=self._buffer,
298
_write=self.write_stream.write, _update=self._hash.update):
299
_buffer[0].append(bytes)
300
_buffer[1] += len(bytes)
302
if _buffer[1] > self._cache_limit or flush:
303
bytes = ''.join(_buffer[0])
307
# expose this on self, for the occasion when clients want to add data.
308
self._write_data = _write_data
309
# a pack writer object to serialise pack records.
310
self._writer = pack.ContainerWriter(self._write_data)
312
# what state is the pack in? (open, finished, aborted)
316
"""Cancel creating this pack."""
317
self._state = 'aborted'
318
self.write_stream.close()
319
# Remove the temporary pack file.
320
self.upload_transport.delete(self.random_name)
321
# The indices have no state on disk.
323
def access_tuple(self):
324
"""Return a tuple (transport, name) for the pack content."""
325
if self._state == 'finished':
326
return Pack.access_tuple(self)
327
elif self._state == 'open':
328
return self.upload_transport, self.random_name
330
raise AssertionError(self._state)
332
def data_inserted(self):
333
"""True if data has been added to this pack."""
334
return bool(self.get_revision_count() or
335
self.inventory_index.key_count() or
336
self.text_index.key_count() or
337
self.signature_index.key_count())
340
"""Finish the new pack.
343
- finalises the content
344
- assigns a name (the md5 of the content, currently)
345
- writes out the associated indices
346
- renames the pack into place.
347
- stores the index size tuple for the pack in the index_sizes
352
self._write_data('', flush=True)
353
self.name = self._hash.hexdigest()
355
# XXX: It'd be better to write them all to temporary names, then
356
# rename them all into place, so that the window when only some are
357
# visible is smaller. On the other hand none will be seen until
358
# they're in the names list.
359
self.index_sizes = [None, None, None, None]
360
self._write_index('revision', self.revision_index, 'revision')
361
self._write_index('inventory', self.inventory_index, 'inventory')
362
self._write_index('text', self.text_index, 'file texts')
363
self._write_index('signature', self.signature_index,
364
'revision signatures')
365
self.write_stream.close()
366
# Note that this will clobber an existing pack with the same name,
367
# without checking for hash collisions. While this is undesirable this
368
# is something that can be rectified in a subsequent release. One way
369
# to rectify it may be to leave the pack at the original name, writing
370
# its pack-names entry as something like 'HASH: index-sizes
371
# temporary-name'. Allocate that and check for collisions, if it is
372
# collision free then rename it into place. If clients know this scheme
373
# they can handle missing-file errors by:
374
# - try for HASH.pack
375
# - try for temporary-name
376
# - refresh the pack-list to see if the pack is now absent
377
self.upload_transport.rename(self.random_name,
378
'../packs/' + self.name + '.pack')
379
self._state = 'finished'
380
if 'pack' in debug.debug_flags:
381
# XXX: size might be interesting?
382
mutter('%s: create_pack: pack renamed into place: %s%s->%s%s t+%6.3fs',
383
time.ctime(), self.upload_transport.base, self.random_name,
384
self.pack_transport, self.name,
385
time.time() - self.start_time)
388
"""Flush any current data."""
390
bytes = ''.join(self._buffer[0])
391
self.write_stream.write(bytes)
392
self._hash.update(bytes)
393
self._buffer[:] = [[], 0]
395
def index_name(self, index_type, name):
396
"""Get the disk name of an index type for pack name 'name'."""
397
return name + NewPack.index_definitions[index_type][0]
399
def index_offset(self, index_type):
400
"""Get the position in a index_size array for a given index type."""
401
return NewPack.index_definitions[index_type][1]
403
def _replace_index_with_readonly(self, index_type):
404
setattr(self, index_type + '_index',
405
GraphIndex(self.index_transport,
406
self.index_name(index_type, self.name),
407
self.index_sizes[self.index_offset(index_type)]))
409
def set_write_cache_size(self, size):
410
self._cache_limit = size
412
def _write_index(self, index_type, index, label):
413
"""Write out an index.
415
:param index_type: The type of index to write - e.g. 'revision'.
416
:param index: The index object to serialise.
417
:param label: What label to give the index e.g. 'revision'.
419
index_name = self.index_name(index_type, self.name)
420
self.index_sizes[self.index_offset(index_type)] = \
421
self.index_transport.put_file(index_name, index.finish(),
422
mode=self._file_mode)
423
if 'pack' in debug.debug_flags:
424
# XXX: size might be interesting?
425
mutter('%s: create_pack: wrote %s index: %s%s t+%6.3fs',
426
time.ctime(), label, self.upload_transport.base,
427
self.random_name, time.time() - self.start_time)
428
# Replace the writable index on this object with a readonly,
429
# presently unloaded index. We should alter
430
# the index layer to make its finish() error if add_node is
431
# subsequently used. RBC
432
self._replace_index_with_readonly(index_type)
435
class AggregateIndex(object):
436
"""An aggregated index for the RepositoryPackCollection.
438
AggregateIndex is reponsible for managing the PackAccess object,
439
Index-To-Pack mapping, and all indices list for a specific type of index
440
such as 'revision index'.
442
A CombinedIndex provides an index on a single key space built up
443
from several on-disk indices. The AggregateIndex builds on this
444
to provide a knit access layer, and allows having up to one writable
445
index within the collection.
447
# XXX: Probably 'can be written to' could/should be separated from 'acts
448
# like a knit index' -- mbp 20071024
451
"""Create an AggregateIndex."""
452
self.index_to_pack = {}
453
self.combined_index = CombinedGraphIndex([])
454
self.data_access = _DirectPackAccess(self.index_to_pack)
455
self.add_callback = None
457
def replace_indices(self, index_to_pack, indices):
458
"""Replace the current mappings with fresh ones.
460
This should probably not be used eventually, rather incremental add and
461
removal of indices. It has been added during refactoring of existing
464
:param index_to_pack: A mapping from index objects to
465
(transport, name) tuples for the pack file data.
466
:param indices: A list of indices.
468
# refresh the revision pack map dict without replacing the instance.
469
self.index_to_pack.clear()
470
self.index_to_pack.update(index_to_pack)
471
# XXX: API break - clearly a 'replace' method would be good?
472
self.combined_index._indices[:] = indices
473
# the current add nodes callback for the current writable index if
475
self.add_callback = None
477
def add_index(self, index, pack):
478
"""Add index to the aggregate, which is an index for Pack pack.
480
Future searches on the aggregate index will seach this new index
481
before all previously inserted indices.
483
:param index: An Index for the pack.
484
:param pack: A Pack instance.
486
# expose it to the index map
487
self.index_to_pack[index] = pack.access_tuple()
488
# put it at the front of the linear index list
489
self.combined_index.insert_index(0, index)
491
def add_writable_index(self, index, pack):
492
"""Add an index which is able to have data added to it.
494
There can be at most one writable index at any time. Any
495
modifications made to the knit are put into this index.
497
:param index: An index from the pack parameter.
498
:param pack: A Pack instance.
500
if self.add_callback is not None:
501
raise AssertionError(
502
"%s already has a writable index through %s" % \
503
(self, self.add_callback))
504
# allow writing: queue writes to a new index
505
self.add_index(index, pack)
506
# Updates the index to packs mapping as a side effect,
507
self.data_access.set_writer(pack._writer, index, pack.access_tuple())
508
self.add_callback = index.add_nodes
511
"""Reset all the aggregate data to nothing."""
512
self.data_access.set_writer(None, None, (None, None))
513
self.index_to_pack.clear()
514
del self.combined_index._indices[:]
515
self.add_callback = None
517
def remove_index(self, index, pack):
518
"""Remove index from the indices used to answer queries.
520
:param index: An index from the pack parameter.
521
:param pack: A Pack instance.
523
del self.index_to_pack[index]
524
self.combined_index._indices.remove(index)
525
if (self.add_callback is not None and
526
getattr(index, 'add_nodes', None) == self.add_callback):
527
self.add_callback = None
528
self.data_access.set_writer(None, None, (None, None))
531
class Packer(object):
532
"""Create a pack from packs."""
534
def __init__(self, pack_collection, packs, suffix, revision_ids=None):
537
:param pack_collection: A RepositoryPackCollection object where the
538
new pack is being written to.
539
:param packs: The packs to combine.
540
:param suffix: The suffix to use on the temporary files for the pack.
541
:param revision_ids: Revision ids to limit the pack to.
545
self.revision_ids = revision_ids
546
# The pack object we are creating.
548
self._pack_collection = pack_collection
549
# The index layer keys for the revisions being copied. None for 'all
551
self._revision_keys = None
552
# What text keys to copy. None for 'all texts'. This is set by
553
# _copy_inventory_texts
554
self._text_filter = None
557
def _extra_init(self):
558
"""A template hook to allow extending the constructor trivially."""
560
def pack(self, pb=None):
561
"""Create a new pack by reading data from other packs.
563
This does little more than a bulk copy of data. One key difference
564
is that data with the same item key across multiple packs is elided
565
from the output. The new pack is written into the current pack store
566
along with its indices, and the name added to the pack names. The
567
source packs are not altered and are not required to be in the current
570
:param pb: An optional progress bar to use. A nested bar is created if
572
:return: A Pack object, or None if nothing was copied.
574
# open a pack - using the same name as the last temporary file
575
# - which has already been flushed, so its safe.
576
# XXX: - duplicate code warning with start_write_group; fix before
577
# considering 'done'.
578
if self._pack_collection._new_pack is not None:
579
raise errors.BzrError('call to create_pack_from_packs while '
580
'another pack is being written.')
581
if self.revision_ids is not None:
582
if len(self.revision_ids) == 0:
583
# silly fetch request.
586
self.revision_ids = frozenset(self.revision_ids)
587
self.revision_keys = frozenset((revid,) for revid in
590
self.pb = ui.ui_factory.nested_progress_bar()
594
return self._create_pack_from_packs()
600
"""Open a pack for the pack we are creating."""
601
return NewPack(self._pack_collection._upload_transport,
602
self._pack_collection._index_transport,
603
self._pack_collection._pack_transport, upload_suffix=self.suffix,
604
file_mode=self._pack_collection.repo.bzrdir._get_file_mode())
606
def _copy_revision_texts(self):
607
"""Copy revision data to the new pack."""
609
if self.revision_ids:
610
revision_keys = [(revision_id,) for revision_id in self.revision_ids]
613
# select revision keys
614
revision_index_map = self._pack_collection._packs_list_to_pack_map_and_index_list(
615
self.packs, 'revision_index')[0]
616
revision_nodes = self._pack_collection._index_contents(revision_index_map, revision_keys)
617
# copy revision keys and adjust values
618
self.pb.update("Copying revision texts", 1)
619
total_items, readv_group_iter = self._revision_node_readv(revision_nodes)
620
list(self._copy_nodes_graph(revision_index_map, self.new_pack._writer,
621
self.new_pack.revision_index, readv_group_iter, total_items))
622
if 'pack' in debug.debug_flags:
623
mutter('%s: create_pack: revisions copied: %s%s %d items t+%6.3fs',
624
time.ctime(), self._pack_collection._upload_transport.base,
625
self.new_pack.random_name,
626
self.new_pack.revision_index.key_count(),
627
time.time() - self.new_pack.start_time)
628
self._revision_keys = revision_keys
630
def _copy_inventory_texts(self):
631
"""Copy the inventory texts to the new pack.
633
self._revision_keys is used to determine what inventories to copy.
635
Sets self._text_filter appropriately.
637
# select inventory keys
638
inv_keys = self._revision_keys # currently the same keyspace, and note that
639
# querying for keys here could introduce a bug where an inventory item
640
# is missed, so do not change it to query separately without cross
641
# checking like the text key check below.
642
inventory_index_map = self._pack_collection._packs_list_to_pack_map_and_index_list(
643
self.packs, 'inventory_index')[0]
644
inv_nodes = self._pack_collection._index_contents(inventory_index_map, inv_keys)
645
# copy inventory keys and adjust values
646
# XXX: Should be a helper function to allow different inv representation
648
self.pb.update("Copying inventory texts", 2)
649
total_items, readv_group_iter = self._least_readv_node_readv(inv_nodes)
650
# Only grab the output lines if we will be processing them
651
output_lines = bool(self.revision_ids)
652
inv_lines = self._copy_nodes_graph(inventory_index_map,
653
self.new_pack._writer, self.new_pack.inventory_index,
654
readv_group_iter, total_items, output_lines=output_lines)
655
if self.revision_ids:
656
self._process_inventory_lines(inv_lines)
658
# eat the iterator to cause it to execute.
660
self._text_filter = None
661
if 'pack' in debug.debug_flags:
662
mutter('%s: create_pack: inventories copied: %s%s %d items t+%6.3fs',
663
time.ctime(), self._pack_collection._upload_transport.base,
664
self.new_pack.random_name,
665
self.new_pack.inventory_index.key_count(),
666
time.time() - self.new_pack.start_time)
668
def _copy_text_texts(self):
670
text_index_map, text_nodes = self._get_text_nodes()
671
if self._text_filter is not None:
672
# We could return the keys copied as part of the return value from
673
# _copy_nodes_graph but this doesn't work all that well with the
674
# need to get line output too, so we check separately, and as we're
675
# going to buffer everything anyway, we check beforehand, which
676
# saves reading knit data over the wire when we know there are
678
text_nodes = set(text_nodes)
679
present_text_keys = set(_node[1] for _node in text_nodes)
680
missing_text_keys = set(self._text_filter) - present_text_keys
681
if missing_text_keys:
682
# TODO: raise a specific error that can handle many missing
684
a_missing_key = missing_text_keys.pop()
685
raise errors.RevisionNotPresent(a_missing_key[1],
687
# copy text keys and adjust values
688
self.pb.update("Copying content texts", 3)
689
total_items, readv_group_iter = self._least_readv_node_readv(text_nodes)
690
list(self._copy_nodes_graph(text_index_map, self.new_pack._writer,
691
self.new_pack.text_index, readv_group_iter, total_items))
692
self._log_copied_texts()
694
def _check_references(self):
695
"""Make sure our external refereneces are present."""
696
external_refs = self.new_pack._external_compression_parents_of_texts()
698
index = self._pack_collection.text_index.combined_index
699
found_items = list(index.iter_entries(external_refs))
700
if len(found_items) != len(external_refs):
701
found_keys = set(k for idx, k, refs, value in found_items)
702
missing_items = external_refs - found_keys
703
missing_file_id, missing_revision_id = missing_items.pop()
704
raise errors.RevisionNotPresent(missing_revision_id,
707
def _create_pack_from_packs(self):
708
self.pb.update("Opening pack", 0, 5)
709
self.new_pack = self.open_pack()
710
new_pack = self.new_pack
711
# buffer data - we won't be reading-back during the pack creation and
712
# this makes a significant difference on sftp pushes.
713
new_pack.set_write_cache_size(1024*1024)
714
if 'pack' in debug.debug_flags:
715
plain_pack_list = ['%s%s' % (a_pack.pack_transport.base, a_pack.name)
716
for a_pack in self.packs]
717
if self.revision_ids is not None:
718
rev_count = len(self.revision_ids)
721
mutter('%s: create_pack: creating pack from source packs: '
722
'%s%s %s revisions wanted %s t=0',
723
time.ctime(), self._pack_collection._upload_transport.base, new_pack.random_name,
724
plain_pack_list, rev_count)
725
self._copy_revision_texts()
726
self._copy_inventory_texts()
727
self._copy_text_texts()
728
# select signature keys
729
signature_filter = self._revision_keys # same keyspace
730
signature_index_map = self._pack_collection._packs_list_to_pack_map_and_index_list(
731
self.packs, 'signature_index')[0]
732
signature_nodes = self._pack_collection._index_contents(signature_index_map,
734
# copy signature keys and adjust values
735
self.pb.update("Copying signature texts", 4)
736
self._copy_nodes(signature_nodes, signature_index_map, new_pack._writer,
737
new_pack.signature_index)
738
if 'pack' in debug.debug_flags:
739
mutter('%s: create_pack: revision signatures copied: %s%s %d items t+%6.3fs',
740
time.ctime(), self._pack_collection._upload_transport.base, new_pack.random_name,
741
new_pack.signature_index.key_count(),
742
time.time() - new_pack.start_time)
743
self._check_references()
744
if not self._use_pack(new_pack):
747
self.pb.update("Finishing pack", 5)
749
self._pack_collection.allocate(new_pack)
752
def _copy_nodes(self, nodes, index_map, writer, write_index):
753
"""Copy knit nodes between packs with no graph references."""
754
pb = ui.ui_factory.nested_progress_bar()
756
return self._do_copy_nodes(nodes, index_map, writer,
761
def _do_copy_nodes(self, nodes, index_map, writer, write_index, pb):
762
# for record verification
763
knit = KnitVersionedFiles(None, None)
764
# plan a readv on each source pack:
766
nodes = sorted(nodes)
767
# how to map this into knit.py - or knit.py into this?
768
# we don't want the typical knit logic, we want grouping by pack
769
# at this point - perhaps a helper library for the following code
770
# duplication points?
772
for index, key, value in nodes:
773
if index not in request_groups:
774
request_groups[index] = []
775
request_groups[index].append((key, value))
777
pb.update("Copied record", record_index, len(nodes))
778
for index, items in request_groups.iteritems():
779
pack_readv_requests = []
780
for key, value in items:
781
# ---- KnitGraphIndex.get_position
782
bits = value[1:].split(' ')
783
offset, length = int(bits[0]), int(bits[1])
784
pack_readv_requests.append((offset, length, (key, value[0])))
785
# linear scan up the pack
786
pack_readv_requests.sort()
788
transport, path = index_map[index]
789
reader = pack.make_readv_reader(transport, path,
790
[offset[0:2] for offset in pack_readv_requests])
791
for (names, read_func), (_1, _2, (key, eol_flag)) in \
792
izip(reader.iter_records(), pack_readv_requests):
793
raw_data = read_func(None)
794
# check the header only
795
df, _ = knit._parse_record_header(key, raw_data)
797
pos, size = writer.add_bytes_record(raw_data, names)
798
write_index.add_node(key, eol_flag + "%d %d" % (pos, size))
799
pb.update("Copied record", record_index)
802
def _copy_nodes_graph(self, index_map, writer, write_index,
803
readv_group_iter, total_items, output_lines=False):
804
"""Copy knit nodes between packs.
806
:param output_lines: Return lines present in the copied data as
807
an iterator of line,version_id.
809
pb = ui.ui_factory.nested_progress_bar()
811
for result in self._do_copy_nodes_graph(index_map, writer,
812
write_index, output_lines, pb, readv_group_iter, total_items):
815
# Python 2.4 does not permit try:finally: in a generator.
821
def _do_copy_nodes_graph(self, index_map, writer, write_index,
822
output_lines, pb, readv_group_iter, total_items):
823
# for record verification
824
knit = KnitVersionedFiles(None, None)
825
# for line extraction when requested (inventories only)
827
factory = KnitPlainFactory()
829
pb.update("Copied record", record_index, total_items)
830
for index, readv_vector, node_vector in readv_group_iter:
832
transport, path = index_map[index]
833
reader = pack.make_readv_reader(transport, path, readv_vector)
834
for (names, read_func), (key, eol_flag, references) in \
835
izip(reader.iter_records(), node_vector):
836
raw_data = read_func(None)
838
# read the entire thing
839
content, _ = knit._parse_record(key[-1], raw_data)
840
if len(references[-1]) == 0:
841
line_iterator = factory.get_fulltext_content(content)
843
line_iterator = factory.get_linedelta_content(content)
844
for line in line_iterator:
847
# check the header only
848
df, _ = knit._parse_record_header(key, raw_data)
850
pos, size = writer.add_bytes_record(raw_data, names)
851
write_index.add_node(key, eol_flag + "%d %d" % (pos, size), references)
852
pb.update("Copied record", record_index)
855
def _get_text_nodes(self):
856
text_index_map = self._pack_collection._packs_list_to_pack_map_and_index_list(
857
self.packs, 'text_index')[0]
858
return text_index_map, self._pack_collection._index_contents(text_index_map,
861
def _least_readv_node_readv(self, nodes):
862
"""Generate request groups for nodes using the least readv's.
864
:param nodes: An iterable of graph index nodes.
865
:return: Total node count and an iterator of the data needed to perform
866
readvs to obtain the data for nodes. Each item yielded by the
867
iterator is a tuple with:
868
index, readv_vector, node_vector. readv_vector is a list ready to
869
hand to the transport readv method, and node_vector is a list of
870
(key, eol_flag, references) for the the node retrieved by the
871
matching readv_vector.
873
# group by pack so we do one readv per pack
874
nodes = sorted(nodes)
877
for index, key, value, references in nodes:
878
if index not in request_groups:
879
request_groups[index] = []
880
request_groups[index].append((key, value, references))
882
for index, items in request_groups.iteritems():
883
pack_readv_requests = []
884
for key, value, references in items:
885
# ---- KnitGraphIndex.get_position
886
bits = value[1:].split(' ')
887
offset, length = int(bits[0]), int(bits[1])
888
pack_readv_requests.append(
889
((offset, length), (key, value[0], references)))
890
# linear scan up the pack to maximum range combining.
891
pack_readv_requests.sort()
892
# split out the readv and the node data.
893
pack_readv = [readv for readv, node in pack_readv_requests]
894
node_vector = [node for readv, node in pack_readv_requests]
895
result.append((index, pack_readv, node_vector))
898
def _log_copied_texts(self):
899
if 'pack' in debug.debug_flags:
900
mutter('%s: create_pack: file texts copied: %s%s %d items t+%6.3fs',
901
time.ctime(), self._pack_collection._upload_transport.base,
902
self.new_pack.random_name,
903
self.new_pack.text_index.key_count(),
904
time.time() - self.new_pack.start_time)
906
def _process_inventory_lines(self, inv_lines):
907
"""Use up the inv_lines generator and setup a text key filter."""
908
repo = self._pack_collection.repo
909
fileid_revisions = repo._find_file_ids_from_xml_inventory_lines(
910
inv_lines, self.revision_keys)
912
for fileid, file_revids in fileid_revisions.iteritems():
913
text_filter.extend([(fileid, file_revid) for file_revid in file_revids])
914
self._text_filter = text_filter
916
def _revision_node_readv(self, revision_nodes):
917
"""Return the total revisions and the readv's to issue.
919
:param revision_nodes: The revision index contents for the packs being
920
incorporated into the new pack.
921
:return: As per _least_readv_node_readv.
923
return self._least_readv_node_readv(revision_nodes)
925
def _use_pack(self, new_pack):
926
"""Return True if new_pack should be used.
928
:param new_pack: The pack that has just been created.
929
:return: True if the pack should be used.
931
return new_pack.data_inserted()
934
class OptimisingPacker(Packer):
935
"""A packer which spends more time to create better disk layouts."""
937
def _revision_node_readv(self, revision_nodes):
938
"""Return the total revisions and the readv's to issue.
940
This sort places revisions in topological order with the ancestors
943
:param revision_nodes: The revision index contents for the packs being
944
incorporated into the new pack.
945
:return: As per _least_readv_node_readv.
947
# build an ancestors dict
950
for index, key, value, references in revision_nodes:
951
ancestors[key] = references[0]
952
by_key[key] = (index, value, references)
953
order = tsort.topo_sort(ancestors)
955
# Single IO is pathological, but it will work as a starting point.
957
for key in reversed(order):
958
index, value, references = by_key[key]
959
# ---- KnitGraphIndex.get_position
960
bits = value[1:].split(' ')
961
offset, length = int(bits[0]), int(bits[1])
963
(index, [(offset, length)], [(key, value[0], references)]))
964
# TODO: combine requests in the same index that are in ascending order.
965
return total, requests
968
class ReconcilePacker(Packer):
969
"""A packer which regenerates indices etc as it copies.
971
This is used by ``bzr reconcile`` to cause parent text pointers to be
975
def _extra_init(self):
976
self._data_changed = False
978
def _process_inventory_lines(self, inv_lines):
979
"""Generate a text key reference map rather for reconciling with."""
980
repo = self._pack_collection.repo
981
refs = repo._find_text_key_references_from_xml_inventory_lines(
983
self._text_refs = refs
984
# during reconcile we:
985
# - convert unreferenced texts to full texts
986
# - correct texts which reference a text not copied to be full texts
987
# - copy all others as-is but with corrected parents.
988
# - so at this point we don't know enough to decide what becomes a full
990
self._text_filter = None
992
def _copy_text_texts(self):
993
"""generate what texts we should have and then copy."""
994
self.pb.update("Copying content texts", 3)
995
# we have three major tasks here:
996
# 1) generate the ideal index
997
repo = self._pack_collection.repo
998
ancestors = dict([(key[0], tuple(ref[0] for ref in refs[0])) for
1000
self.new_pack.revision_index.iter_all_entries()])
1001
ideal_index = repo._generate_text_key_index(self._text_refs, ancestors)
1002
# 2) generate a text_nodes list that contains all the deltas that can
1003
# be used as-is, with corrected parents.
1006
discarded_nodes = []
1007
NULL_REVISION = _mod_revision.NULL_REVISION
1008
text_index_map, text_nodes = self._get_text_nodes()
1009
for node in text_nodes:
1015
ideal_parents = tuple(ideal_index[node[1]])
1017
discarded_nodes.append(node)
1018
self._data_changed = True
1020
if ideal_parents == (NULL_REVISION,):
1022
if ideal_parents == node[3][0]:
1024
ok_nodes.append(node)
1025
elif ideal_parents[0:1] == node[3][0][0:1]:
1026
# the left most parent is the same, or there are no parents
1027
# today. Either way, we can preserve the representation as
1028
# long as we change the refs to be inserted.
1029
self._data_changed = True
1030
ok_nodes.append((node[0], node[1], node[2],
1031
(ideal_parents, node[3][1])))
1032
self._data_changed = True
1034
# Reinsert this text completely
1035
bad_texts.append((node[1], ideal_parents))
1036
self._data_changed = True
1037
# we're finished with some data.
1040
# 3) bulk copy the ok data
1041
total_items, readv_group_iter = self._least_readv_node_readv(ok_nodes)
1042
list(self._copy_nodes_graph(text_index_map, self.new_pack._writer,
1043
self.new_pack.text_index, readv_group_iter, total_items))
1044
# 4) adhoc copy all the other texts.
1045
# We have to topologically insert all texts otherwise we can fail to
1046
# reconcile when parts of a single delta chain are preserved intact,
1047
# and other parts are not. E.g. Discarded->d1->d2->d3. d1 will be
1048
# reinserted, and if d3 has incorrect parents it will also be
1049
# reinserted. If we insert d3 first, d2 is present (as it was bulk
1050
# copied), so we will try to delta, but d2 is not currently able to be
1051
# extracted because it's basis d1 is not present. Topologically sorting
1052
# addresses this. The following generates a sort for all the texts that
1053
# are being inserted without having to reference the entire text key
1054
# space (we only topo sort the revisions, which is smaller).
1055
topo_order = tsort.topo_sort(ancestors)
1056
rev_order = dict(zip(topo_order, range(len(topo_order))))
1057
bad_texts.sort(key=lambda key:rev_order[key[0][1]])
1058
transaction = repo.get_transaction()
1059
file_id_index = GraphIndexPrefixAdapter(
1060
self.new_pack.text_index,
1062
add_nodes_callback=self.new_pack.text_index.add_nodes)
1063
data_access = _DirectPackAccess(
1064
{self.new_pack.text_index:self.new_pack.access_tuple()})
1065
data_access.set_writer(self.new_pack._writer, self.new_pack.text_index,
1066
self.new_pack.access_tuple())
1067
output_texts = KnitVersionedFiles(
1068
_KnitGraphIndex(self.new_pack.text_index,
1069
add_callback=self.new_pack.text_index.add_nodes,
1070
deltas=True, parents=True, is_locked=repo.is_locked),
1071
data_access=data_access, max_delta_chain=200)
1072
for key, parent_keys in bad_texts:
1073
# We refer to the new pack to delta data being output.
1074
# A possible improvement would be to catch errors on short reads
1075
# and only flush then.
1076
self.new_pack.flush()
1078
for parent_key in parent_keys:
1079
if parent_key[0] != key[0]:
1080
# Graph parents must match the fileid
1081
raise errors.BzrError('Mismatched key parent %r:%r' %
1083
parents.append(parent_key[1])
1084
text_lines = split_lines(repo.texts.get_record_stream(
1085
[key], 'unordered', True).next().get_bytes_as('fulltext'))
1086
output_texts.add_lines(key, parent_keys, text_lines,
1087
random_id=True, check_content=False)
1088
# 5) check that nothing inserted has a reference outside the keyspace.
1089
missing_text_keys = self.new_pack._external_compression_parents_of_texts()
1090
if missing_text_keys:
1091
raise errors.BzrError('Reference to missing compression parents %r'
1092
% (missing_text_keys,))
1093
self._log_copied_texts()
1095
def _use_pack(self, new_pack):
1096
"""Override _use_pack to check for reconcile having changed content."""
1097
# XXX: we might be better checking this at the copy time.
1098
original_inventory_keys = set()
1099
inv_index = self._pack_collection.inventory_index.combined_index
1100
for entry in inv_index.iter_all_entries():
1101
original_inventory_keys.add(entry[1])
1102
new_inventory_keys = set()
1103
for entry in new_pack.inventory_index.iter_all_entries():
1104
new_inventory_keys.add(entry[1])
1105
if new_inventory_keys != original_inventory_keys:
1106
self._data_changed = True
1107
return new_pack.data_inserted() and self._data_changed
1110
class RepositoryPackCollection(object):
1111
"""Management of packs within a repository.
1113
:ivar _names: map of {pack_name: (index_size,)}
1116
def __init__(self, repo, transport, index_transport, upload_transport,
1118
"""Create a new RepositoryPackCollection.
1120
:param transport: Addresses the repository base directory
1121
(typically .bzr/repository/).
1122
:param index_transport: Addresses the directory containing indices.
1123
:param upload_transport: Addresses the directory into which packs are written
1124
while they're being created.
1125
:param pack_transport: Addresses the directory of existing complete packs.
1128
self.transport = transport
1129
self._index_transport = index_transport
1130
self._upload_transport = upload_transport
1131
self._pack_transport = pack_transport
1132
self._suffix_offsets = {'.rix': 0, '.iix': 1, '.tix': 2, '.six': 3}
1135
self._packs_by_name = {}
1136
# the previous pack-names content
1137
self._packs_at_load = None
1138
# when a pack is being created by this object, the state of that pack.
1139
self._new_pack = None
1140
# aggregated revision index data
1141
self.revision_index = AggregateIndex()
1142
self.inventory_index = AggregateIndex()
1143
self.text_index = AggregateIndex()
1144
self.signature_index = AggregateIndex()
1146
def add_pack_to_memory(self, pack):
1147
"""Make a Pack object available to the repository to satisfy queries.
1149
:param pack: A Pack object.
1151
if pack.name in self._packs_by_name:
1152
raise AssertionError()
1153
self.packs.append(pack)
1154
self._packs_by_name[pack.name] = pack
1155
self.revision_index.add_index(pack.revision_index, pack)
1156
self.inventory_index.add_index(pack.inventory_index, pack)
1157
self.text_index.add_index(pack.text_index, pack)
1158
self.signature_index.add_index(pack.signature_index, pack)
1160
def all_packs(self):
1161
"""Return a list of all the Pack objects this repository has.
1163
Note that an in-progress pack being created is not returned.
1165
:return: A list of Pack objects for all the packs in the repository.
1168
for name in self.names():
1169
result.append(self.get_pack_by_name(name))
1173
"""Pack the pack collection incrementally.
1175
This will not attempt global reorganisation or recompression,
1176
rather it will just ensure that the total number of packs does
1177
not grow without bound. It uses the _max_pack_count method to
1178
determine if autopacking is needed, and the pack_distribution
1179
method to determine the number of revisions in each pack.
1181
If autopacking takes place then the packs name collection will have
1182
been flushed to disk - packing requires updating the name collection
1183
in synchronisation with certain steps. Otherwise the names collection
1186
:return: True if packing took place.
1188
# XXX: Should not be needed when the management of indices is sane.
1189
total_revisions = self.revision_index.combined_index.key_count()
1190
total_packs = len(self._names)
1191
if self._max_pack_count(total_revisions) >= total_packs:
1193
# XXX: the following may want to be a class, to pack with a given
1195
mutter('Auto-packing repository %s, which has %d pack files, '
1196
'containing %d revisions into %d packs.', self, total_packs,
1197
total_revisions, self._max_pack_count(total_revisions))
1198
# determine which packs need changing
1199
pack_distribution = self.pack_distribution(total_revisions)
1201
for pack in self.all_packs():
1202
revision_count = pack.get_revision_count()
1203
if revision_count == 0:
1204
# revision less packs are not generated by normal operation,
1205
# only by operations like sign-my-commits, and thus will not
1206
# tend to grow rapdily or without bound like commit containing
1207
# packs do - leave them alone as packing them really should
1208
# group their data with the relevant commit, and that may
1209
# involve rewriting ancient history - which autopack tries to
1210
# avoid. Alternatively we could not group the data but treat
1211
# each of these as having a single revision, and thus add
1212
# one revision for each to the total revision count, to get
1213
# a matching distribution.
1215
existing_packs.append((revision_count, pack))
1216
pack_operations = self.plan_autopack_combinations(
1217
existing_packs, pack_distribution)
1218
self._execute_pack_operations(pack_operations)
1221
def _execute_pack_operations(self, pack_operations, _packer_class=Packer):
1222
"""Execute a series of pack operations.
1224
:param pack_operations: A list of [revision_count, packs_to_combine].
1225
:param _packer_class: The class of packer to use (default: Packer).
1228
for revision_count, packs in pack_operations:
1229
# we may have no-ops from the setup logic
1232
_packer_class(self, packs, '.autopack').pack()
1234
self._remove_pack_from_memory(pack)
1235
# record the newly available packs and stop advertising the old
1237
self._save_pack_names(clear_obsolete_packs=True)
1238
# Move the old packs out of the way now they are no longer referenced.
1239
for revision_count, packs in pack_operations:
1240
self._obsolete_packs(packs)
1242
def lock_names(self):
1243
"""Acquire the mutex around the pack-names index.
1245
This cannot be used in the middle of a read-only transaction on the
1248
self.repo.control_files.lock_write()
1251
"""Pack the pack collection totally."""
1252
self.ensure_loaded()
1253
total_packs = len(self._names)
1255
# This is arguably wrong because we might not be optimal, but for
1256
# now lets leave it in. (e.g. reconcile -> one pack. But not
1259
total_revisions = self.revision_index.combined_index.key_count()
1260
# XXX: the following may want to be a class, to pack with a given
1262
mutter('Packing repository %s, which has %d pack files, '
1263
'containing %d revisions into 1 packs.', self, total_packs,
1265
# determine which packs need changing
1266
pack_distribution = [1]
1267
pack_operations = [[0, []]]
1268
for pack in self.all_packs():
1269
pack_operations[-1][0] += pack.get_revision_count()
1270
pack_operations[-1][1].append(pack)
1271
self._execute_pack_operations(pack_operations, OptimisingPacker)
1273
def plan_autopack_combinations(self, existing_packs, pack_distribution):
1274
"""Plan a pack operation.
1276
:param existing_packs: The packs to pack. (A list of (revcount, Pack)
1278
:param pack_distribution: A list with the number of revisions desired
1281
if len(existing_packs) <= len(pack_distribution):
1283
existing_packs.sort(reverse=True)
1284
pack_operations = [[0, []]]
1285
# plan out what packs to keep, and what to reorganise
1286
while len(existing_packs):
1287
# take the largest pack, and if its less than the head of the
1288
# distribution chart we will include its contents in the new pack for
1289
# that position. If its larger, we remove its size from the
1290
# distribution chart
1291
next_pack_rev_count, next_pack = existing_packs.pop(0)
1292
if next_pack_rev_count >= pack_distribution[0]:
1293
# this is already packed 'better' than this, so we can
1294
# not waste time packing it.
1295
while next_pack_rev_count > 0:
1296
next_pack_rev_count -= pack_distribution[0]
1297
if next_pack_rev_count >= 0:
1299
del pack_distribution[0]
1301
# didn't use that entire bucket up
1302
pack_distribution[0] = -next_pack_rev_count
1304
# add the revisions we're going to add to the next output pack
1305
pack_operations[-1][0] += next_pack_rev_count
1306
# allocate this pack to the next pack sub operation
1307
pack_operations[-1][1].append(next_pack)
1308
if pack_operations[-1][0] >= pack_distribution[0]:
1309
# this pack is used up, shift left.
1310
del pack_distribution[0]
1311
pack_operations.append([0, []])
1313
return pack_operations
1315
def ensure_loaded(self):
1316
# NB: if you see an assertion error here, its probably access against
1317
# an unlocked repo. Naughty.
1318
if not self.repo.is_locked():
1319
raise errors.ObjectNotLocked(self.repo)
1320
if self._names is None:
1322
self._packs_at_load = set()
1323
for index, key, value in self._iter_disk_pack_index():
1325
self._names[name] = self._parse_index_sizes(value)
1326
self._packs_at_load.add((key, value))
1327
# populate all the metadata.
1330
def _parse_index_sizes(self, value):
1331
"""Parse a string of index sizes."""
1332
return tuple([int(digits) for digits in value.split(' ')])
1334
def get_pack_by_name(self, name):
1335
"""Get a Pack object by name.
1337
:param name: The name of the pack - e.g. '123456'
1338
:return: A Pack object.
1341
return self._packs_by_name[name]
1343
rev_index = self._make_index(name, '.rix')
1344
inv_index = self._make_index(name, '.iix')
1345
txt_index = self._make_index(name, '.tix')
1346
sig_index = self._make_index(name, '.six')
1347
result = ExistingPack(self._pack_transport, name, rev_index,
1348
inv_index, txt_index, sig_index)
1349
self.add_pack_to_memory(result)
1352
def allocate(self, a_new_pack):
1353
"""Allocate name in the list of packs.
1355
:param a_new_pack: A NewPack instance to be added to the collection of
1356
packs for this repository.
1358
self.ensure_loaded()
1359
if a_new_pack.name in self._names:
1360
raise errors.BzrError(
1361
'Pack %r already exists in %s' % (a_new_pack.name, self))
1362
self._names[a_new_pack.name] = tuple(a_new_pack.index_sizes)
1363
self.add_pack_to_memory(a_new_pack)
1365
def _iter_disk_pack_index(self):
1366
"""Iterate over the contents of the pack-names index.
1368
This is used when loading the list from disk, and before writing to
1369
detect updates from others during our write operation.
1370
:return: An iterator of the index contents.
1372
return GraphIndex(self.transport, 'pack-names', None
1373
).iter_all_entries()
1375
def _make_index(self, name, suffix):
1376
size_offset = self._suffix_offsets[suffix]
1377
index_name = name + suffix
1378
index_size = self._names[name][size_offset]
1380
self._index_transport, index_name, index_size)
1382
def _max_pack_count(self, total_revisions):
1383
"""Return the maximum number of packs to use for total revisions.
1385
:param total_revisions: The total number of revisions in the
1388
if not total_revisions:
1390
digits = str(total_revisions)
1392
for digit in digits:
1393
result += int(digit)
1397
"""Provide an order to the underlying names."""
1398
return sorted(self._names.keys())
1400
def _obsolete_packs(self, packs):
1401
"""Move a number of packs which have been obsoleted out of the way.
1403
Each pack and its associated indices are moved out of the way.
1405
Note: for correctness this function should only be called after a new
1406
pack names index has been written without these pack names, and with
1407
the names of packs that contain the data previously available via these
1410
:param packs: The packs to obsolete.
1411
:param return: None.
1414
pack.pack_transport.rename(pack.file_name(),
1415
'../obsolete_packs/' + pack.file_name())
1416
# TODO: Probably needs to know all possible indices for this pack
1417
# - or maybe list the directory and move all indices matching this
1418
# name whether we recognize it or not?
1419
for suffix in ('.iix', '.six', '.tix', '.rix'):
1420
self._index_transport.rename(pack.name + suffix,
1421
'../obsolete_packs/' + pack.name + suffix)
1423
def pack_distribution(self, total_revisions):
1424
"""Generate a list of the number of revisions to put in each pack.
1426
:param total_revisions: The total number of revisions in the
1429
if total_revisions == 0:
1431
digits = reversed(str(total_revisions))
1433
for exponent, count in enumerate(digits):
1434
size = 10 ** exponent
1435
for pos in range(int(count)):
1437
return list(reversed(result))
1439
def _pack_tuple(self, name):
1440
"""Return a tuple with the transport and file name for a pack name."""
1441
return self._pack_transport, name + '.pack'
1443
def _remove_pack_from_memory(self, pack):
1444
"""Remove pack from the packs accessed by this repository.
1446
Only affects memory state, until self._save_pack_names() is invoked.
1448
self._names.pop(pack.name)
1449
self._packs_by_name.pop(pack.name)
1450
self._remove_pack_indices(pack)
1452
def _remove_pack_indices(self, pack):
1453
"""Remove the indices for pack from the aggregated indices."""
1454
self.revision_index.remove_index(pack.revision_index, pack)
1455
self.inventory_index.remove_index(pack.inventory_index, pack)
1456
self.text_index.remove_index(pack.text_index, pack)
1457
self.signature_index.remove_index(pack.signature_index, pack)
1460
"""Clear all cached data."""
1461
# cached revision data
1462
self.repo._revision_knit = None
1463
self.revision_index.clear()
1464
# cached signature data
1465
self.repo._signature_knit = None
1466
self.signature_index.clear()
1467
# cached file text data
1468
self.text_index.clear()
1469
self.repo._text_knit = None
1470
# cached inventory data
1471
self.inventory_index.clear()
1472
# remove the open pack
1473
self._new_pack = None
1474
# information about packs.
1477
self._packs_by_name = {}
1478
self._packs_at_load = None
1480
def _make_index_map(self, index_suffix):
1481
"""Return information on existing indices.
1483
:param suffix: Index suffix added to pack name.
1485
:returns: (pack_map, indices) where indices is a list of GraphIndex
1486
objects, and pack_map is a mapping from those objects to the
1487
pack tuple they describe.
1489
# TODO: stop using this; it creates new indices unnecessarily.
1490
self.ensure_loaded()
1491
suffix_map = {'.rix': 'revision_index',
1492
'.six': 'signature_index',
1493
'.iix': 'inventory_index',
1494
'.tix': 'text_index',
1496
return self._packs_list_to_pack_map_and_index_list(self.all_packs(),
1497
suffix_map[index_suffix])
1499
def _packs_list_to_pack_map_and_index_list(self, packs, index_attribute):
1500
"""Convert a list of packs to an index pack map and index list.
1502
:param packs: The packs list to process.
1503
:param index_attribute: The attribute that the desired index is found
1505
:return: A tuple (map, list) where map contains the dict from
1506
index:pack_tuple, and lsit contains the indices in the same order
1512
index = getattr(pack, index_attribute)
1513
indices.append(index)
1514
pack_map[index] = (pack.pack_transport, pack.file_name())
1515
return pack_map, indices
1517
def _index_contents(self, pack_map, key_filter=None):
1518
"""Get an iterable of the index contents from a pack_map.
1520
:param pack_map: A map from indices to pack details.
1521
:param key_filter: An optional filter to limit the
1524
indices = [index for index in pack_map.iterkeys()]
1525
all_index = CombinedGraphIndex(indices)
1526
if key_filter is None:
1527
return all_index.iter_all_entries()
1529
return all_index.iter_entries(key_filter)
1531
def _unlock_names(self):
1532
"""Release the mutex around the pack-names index."""
1533
self.repo.control_files.unlock()
1535
def _save_pack_names(self, clear_obsolete_packs=False):
1536
"""Save the list of packs.
1538
This will take out the mutex around the pack names list for the
1539
duration of the method call. If concurrent updates have been made, a
1540
three-way merge between the current list and the current in memory list
1543
:param clear_obsolete_packs: If True, clear out the contents of the
1544
obsolete_packs directory.
1548
builder = GraphIndexBuilder()
1549
# load the disk nodes across
1551
for index, key, value in self._iter_disk_pack_index():
1552
disk_nodes.add((key, value))
1553
# do a two-way diff against our original content
1554
current_nodes = set()
1555
for name, sizes in self._names.iteritems():
1557
((name, ), ' '.join(str(size) for size in sizes)))
1558
deleted_nodes = self._packs_at_load - current_nodes
1559
new_nodes = current_nodes - self._packs_at_load
1560
disk_nodes.difference_update(deleted_nodes)
1561
disk_nodes.update(new_nodes)
1562
# TODO: handle same-name, index-size-changes here -
1563
# e.g. use the value from disk, not ours, *unless* we're the one
1565
for key, value in disk_nodes:
1566
builder.add_node(key, value)
1567
self.transport.put_file('pack-names', builder.finish(),
1568
mode=self.repo.bzrdir._get_file_mode())
1569
# move the baseline forward
1570
self._packs_at_load = disk_nodes
1571
if clear_obsolete_packs:
1572
self._clear_obsolete_packs()
1574
self._unlock_names()
1575
# synchronise the memory packs list with what we just wrote:
1576
new_names = dict(disk_nodes)
1577
# drop no longer present nodes
1578
for pack in self.all_packs():
1579
if (pack.name,) not in new_names:
1580
self._remove_pack_from_memory(pack)
1581
# add new nodes/refresh existing ones
1582
for key, value in disk_nodes:
1584
sizes = self._parse_index_sizes(value)
1585
if name in self._names:
1587
if sizes != self._names[name]:
1588
# the pack for name has had its indices replaced - rare but
1589
# important to handle. XXX: probably can never happen today
1590
# because the three-way merge code above does not handle it
1591
# - you may end up adding the same key twice to the new
1592
# disk index because the set values are the same, unless
1593
# the only index shows up as deleted by the set difference
1594
# - which it may. Until there is a specific test for this,
1595
# assume its broken. RBC 20071017.
1596
self._remove_pack_from_memory(self.get_pack_by_name(name))
1597
self._names[name] = sizes
1598
self.get_pack_by_name(name)
1601
self._names[name] = sizes
1602
self.get_pack_by_name(name)
1604
def _clear_obsolete_packs(self):
1605
"""Delete everything from the obsolete-packs directory.
1607
obsolete_pack_transport = self.transport.clone('obsolete_packs')
1608
for filename in obsolete_pack_transport.list_dir('.'):
1610
obsolete_pack_transport.delete(filename)
1611
except (errors.PathError, errors.TransportError), e:
1612
warning("couldn't delete obsolete pack, skipping it:\n%s" % (e,))
1614
def _start_write_group(self):
1615
# Do not permit preparation for writing if we're not in a 'write lock'.
1616
if not self.repo.is_write_locked():
1617
raise errors.NotWriteLocked(self)
1618
self._new_pack = NewPack(self._upload_transport, self._index_transport,
1619
self._pack_transport, upload_suffix='.pack',
1620
file_mode=self.repo.bzrdir._get_file_mode())
1621
# allow writing: queue writes to a new index
1622
self.revision_index.add_writable_index(self._new_pack.revision_index,
1624
self.inventory_index.add_writable_index(self._new_pack.inventory_index,
1626
self.text_index.add_writable_index(self._new_pack.text_index,
1628
self.signature_index.add_writable_index(self._new_pack.signature_index,
1631
self.repo.inventories._index._add_callback = self.inventory_index.add_callback
1632
self.repo.revisions._index._add_callback = self.revision_index.add_callback
1633
self.repo.signatures._index._add_callback = self.signature_index.add_callback
1634
self.repo.texts._index._add_callback = self.text_index.add_callback
1636
def _abort_write_group(self):
1637
# FIXME: just drop the transient index.
1638
# forget what names there are
1639
if self._new_pack is not None:
1640
self._new_pack.abort()
1641
self._remove_pack_indices(self._new_pack)
1642
self._new_pack = None
1643
self.repo._text_knit = None
1645
def _commit_write_group(self):
1646
self._remove_pack_indices(self._new_pack)
1647
if self._new_pack.data_inserted():
1648
# get all the data to disk and read to use
1649
self._new_pack.finish()
1650
self.allocate(self._new_pack)
1651
self._new_pack = None
1652
if not self.autopack():
1653
# when autopack takes no steps, the names list is still
1655
self._save_pack_names()
1657
self._new_pack.abort()
1658
self._new_pack = None
1659
self.repo._text_knit = None
1662
class KnitPackRepository(KnitRepository):
1663
"""Repository with knit objects stored inside pack containers.
1665
The layering for a KnitPackRepository is:
1667
Graph | HPSS | Repository public layer |
1668
===================================================
1669
Tuple based apis below, string based, and key based apis above
1670
---------------------------------------------------
1672
Provides .texts, .revisions etc
1673
This adapts the N-tuple keys to physical knit records which only have a
1674
single string identifier (for historical reasons), which in older formats
1675
was always the revision_id, and in the mapped code for packs is always
1676
the last element of key tuples.
1677
---------------------------------------------------
1679
A separate GraphIndex is used for each of the
1680
texts/inventories/revisions/signatures contained within each individual
1681
pack file. The GraphIndex layer works in N-tuples and is unaware of any
1683
===================================================
1687
def __init__(self, _format, a_bzrdir, control_files, _commit_builder_class,
1689
KnitRepository.__init__(self, _format, a_bzrdir, control_files,
1690
_commit_builder_class, _serializer)
1691
index_transport = self._transport.clone('indices')
1692
self._pack_collection = RepositoryPackCollection(self, self._transport,
1694
self._transport.clone('upload'),
1695
self._transport.clone('packs'))
1696
self.inventories = KnitVersionedFiles(
1697
_KnitGraphIndex(self._pack_collection.inventory_index.combined_index,
1698
add_callback=self._pack_collection.inventory_index.add_callback,
1699
deltas=True, parents=True, is_locked=self.is_locked),
1700
data_access=self._pack_collection.inventory_index.data_access,
1701
max_delta_chain=200)
1702
self.revisions = KnitVersionedFiles(
1703
_KnitGraphIndex(self._pack_collection.revision_index.combined_index,
1704
add_callback=self._pack_collection.revision_index.add_callback,
1705
deltas=False, parents=True, is_locked=self.is_locked),
1706
data_access=self._pack_collection.revision_index.data_access,
1708
self.signatures = KnitVersionedFiles(
1709
_KnitGraphIndex(self._pack_collection.signature_index.combined_index,
1710
add_callback=self._pack_collection.signature_index.add_callback,
1711
deltas=False, parents=False, is_locked=self.is_locked),
1712
data_access=self._pack_collection.signature_index.data_access,
1714
self.texts = KnitVersionedFiles(
1715
_KnitGraphIndex(self._pack_collection.text_index.combined_index,
1716
add_callback=self._pack_collection.text_index.add_callback,
1717
deltas=True, parents=True, is_locked=self.is_locked),
1718
data_access=self._pack_collection.text_index.data_access,
1719
max_delta_chain=200)
1720
# True when the repository object is 'write locked' (as opposed to the
1721
# physical lock only taken out around changes to the pack-names list.)
1722
# Another way to represent this would be a decorator around the control
1723
# files object that presents logical locks as physical ones - if this
1724
# gets ugly consider that alternative design. RBC 20071011
1725
self._write_lock_count = 0
1726
self._transaction = None
1728
self._reconcile_does_inventory_gc = True
1729
self._reconcile_fixes_text_parents = True
1730
self._reconcile_backsup_inventory = False
1731
self._fetch_order = 'unsorted'
1733
def _warn_if_deprecated(self):
1734
# This class isn't deprecated
1737
def _abort_write_group(self):
1738
self._pack_collection._abort_write_group()
1740
def _find_inconsistent_revision_parents(self):
1741
"""Find revisions with incorrectly cached parents.
1743
:returns: an iterator yielding tuples of (revison-id, parents-in-index,
1744
parents-in-revision).
1746
if not self.is_locked():
1747
raise errors.ObjectNotLocked(self)
1748
pb = ui.ui_factory.nested_progress_bar()
1751
revision_nodes = self._pack_collection.revision_index \
1752
.combined_index.iter_all_entries()
1753
index_positions = []
1754
# Get the cached index values for all revisions, and also the location
1755
# in each index of the revision text so we can perform linear IO.
1756
for index, key, value, refs in revision_nodes:
1757
pos, length = value[1:].split(' ')
1758
index_positions.append((index, int(pos), key[0],
1759
tuple(parent[0] for parent in refs[0])))
1760
pb.update("Reading revision index.", 0, 0)
1761
index_positions.sort()
1762
batch_count = len(index_positions) / 1000 + 1
1763
pb.update("Checking cached revision graph.", 0, batch_count)
1764
for offset in xrange(batch_count):
1765
pb.update("Checking cached revision graph.", offset)
1766
to_query = index_positions[offset * 1000:(offset + 1) * 1000]
1769
rev_ids = [item[2] for item in to_query]
1770
revs = self.get_revisions(rev_ids)
1771
for revision, item in zip(revs, to_query):
1772
index_parents = item[3]
1773
rev_parents = tuple(revision.parent_ids)
1774
if index_parents != rev_parents:
1775
result.append((revision.revision_id, index_parents, rev_parents))
1780
@symbol_versioning.deprecated_method(symbol_versioning.one_one)
1781
def get_parents(self, revision_ids):
1782
"""See graph._StackedParentsProvider.get_parents."""
1783
parent_map = self.get_parent_map(revision_ids)
1784
return [parent_map.get(r, None) for r in revision_ids]
1786
def _make_parents_provider(self):
1787
return graph.CachingParentsProvider(self)
1789
def _refresh_data(self):
1790
if self._write_lock_count == 1 or (
1791
self.control_files._lock_count == 1 and
1792
self.control_files._lock_mode == 'r'):
1793
# forget what names there are
1794
self._pack_collection.reset()
1795
# XXX: Better to do an in-memory merge when acquiring a new lock -
1796
# factor out code from _save_pack_names.
1797
self._pack_collection.ensure_loaded()
1799
def _start_write_group(self):
1800
self._pack_collection._start_write_group()
1802
def _commit_write_group(self):
1803
return self._pack_collection._commit_write_group()
1805
def get_transaction(self):
1806
if self._write_lock_count:
1807
return self._transaction
1809
return self.control_files.get_transaction()
1811
def is_locked(self):
1812
return self._write_lock_count or self.control_files.is_locked()
1814
def is_write_locked(self):
1815
return self._write_lock_count
1817
def lock_write(self, token=None):
1818
if not self._write_lock_count and self.is_locked():
1819
raise errors.ReadOnlyError(self)
1820
self._write_lock_count += 1
1821
if self._write_lock_count == 1:
1822
from bzrlib import transactions
1823
self._transaction = transactions.WriteTransaction()
1824
for repo in self._fallback_repositories:
1825
# Writes don't affect fallback repos
1827
self._refresh_data()
1829
def lock_read(self):
1830
if self._write_lock_count:
1831
self._write_lock_count += 1
1833
self.control_files.lock_read()
1834
for repo in self._fallback_repositories:
1835
# Writes don't affect fallback repos
1837
self._refresh_data()
1839
def leave_lock_in_place(self):
1840
# not supported - raise an error
1841
raise NotImplementedError(self.leave_lock_in_place)
1843
def dont_leave_lock_in_place(self):
1844
# not supported - raise an error
1845
raise NotImplementedError(self.dont_leave_lock_in_place)
1849
"""Compress the data within the repository.
1851
This will pack all the data to a single pack. In future it may
1852
recompress deltas or do other such expensive operations.
1854
self._pack_collection.pack()
1857
def reconcile(self, other=None, thorough=False):
1858
"""Reconcile this repository."""
1859
from bzrlib.reconcile import PackReconciler
1860
reconciler = PackReconciler(self, thorough=thorough)
1861
reconciler.reconcile()
1865
if self._write_lock_count == 1 and self._write_group is not None:
1866
self.abort_write_group()
1867
self._transaction = None
1868
self._write_lock_count = 0
1869
raise errors.BzrError(
1870
'Must end write group before releasing write lock on %s'
1872
if self._write_lock_count:
1873
self._write_lock_count -= 1
1874
if not self._write_lock_count:
1875
transaction = self._transaction
1876
self._transaction = None
1877
transaction.finish()
1878
for repo in self._fallback_repositories:
1881
self.control_files.unlock()
1882
for repo in self._fallback_repositories:
1886
class RepositoryFormatPack(MetaDirRepositoryFormat):
1887
"""Format logic for pack structured repositories.
1889
This repository format has:
1890
- a list of packs in pack-names
1891
- packs in packs/NAME.pack
1892
- indices in indices/NAME.{iix,six,tix,rix}
1893
- knit deltas in the packs, knit indices mapped to the indices.
1894
- thunk objects to support the knits programming API.
1895
- a format marker of its own
1896
- an optional 'shared-storage' flag
1897
- an optional 'no-working-trees' flag
1901
# Set this attribute in derived classes to control the repository class
1902
# created by open and initialize.
1903
repository_class = None
1904
# Set this attribute in derived classes to control the
1905
# _commit_builder_class that the repository objects will have passed to
1906
# their constructor.
1907
_commit_builder_class = None
1908
# Set this attribute in derived clases to control the _serializer that the
1909
# repository objects will have passed to their constructor.
1911
# External references are not supported in pack repositories yet.
1912
supports_external_lookups = False
1914
def initialize(self, a_bzrdir, shared=False):
1915
"""Create a pack based repository.
1917
:param a_bzrdir: bzrdir to contain the new repository; must already
1919
:param shared: If true the repository will be initialized as a shared
1922
mutter('creating repository in %s.', a_bzrdir.transport.base)
1923
dirs = ['indices', 'obsolete_packs', 'packs', 'upload']
1924
builder = GraphIndexBuilder()
1925
files = [('pack-names', builder.finish())]
1926
utf8_files = [('format', self.get_format_string())]
1928
self._upload_blank_content(a_bzrdir, dirs, files, utf8_files, shared)
1929
return self.open(a_bzrdir=a_bzrdir, _found=True)
1931
def open(self, a_bzrdir, _found=False, _override_transport=None):
1932
"""See RepositoryFormat.open().
1934
:param _override_transport: INTERNAL USE ONLY. Allows opening the
1935
repository at a slightly different url
1936
than normal. I.e. during 'upgrade'.
1939
format = RepositoryFormat.find_format(a_bzrdir)
1940
if _override_transport is not None:
1941
repo_transport = _override_transport
1943
repo_transport = a_bzrdir.get_repository_transport(None)
1944
control_files = lockable_files.LockableFiles(repo_transport,
1945
'lock', lockdir.LockDir)
1946
return self.repository_class(_format=self,
1948
control_files=control_files,
1949
_commit_builder_class=self._commit_builder_class,
1950
_serializer=self._serializer)
1953
class RepositoryFormatKnitPack1(RepositoryFormatPack):
1954
"""A no-subtrees parameterized Pack repository.
1956
This format was introduced in 0.92.
1959
repository_class = KnitPackRepository
1960
_commit_builder_class = PackCommitBuilder
1961
_serializer = xml5.serializer_v5
1963
def _get_matching_bzrdir(self):
1964
return bzrdir.format_registry.make_bzrdir('pack-0.92')
1966
def _ignore_setting_bzrdir(self, format):
1969
_matchingbzrdir = property(_get_matching_bzrdir, _ignore_setting_bzrdir)
1971
def get_format_string(self):
1972
"""See RepositoryFormat.get_format_string()."""
1973
return "Bazaar pack repository format 1 (needs bzr 0.92)\n"
1975
def get_format_description(self):
1976
"""See RepositoryFormat.get_format_description()."""
1977
return "Packs containing knits without subtree support"
1979
def check_conversion_target(self, target_format):
1983
class RepositoryFormatPack(MetaDirRepositoryFormat):
1984
"""Format logic for pack structured repositories.
1986
This repository format has:
1987
- a list of packs in pack-names
1988
- packs in packs/NAME.pack
1989
- indices in indices/NAME.{iix,six,tix,rix}
1990
- knit deltas in the packs, knit indices mapped to the indices.
1991
- thunk objects to support the knits programming API.
1992
- a format marker of its own
1993
- an optional 'shared-storage' flag
1994
- an optional 'no-working-trees' flag
1998
# Set this attribute in derived classes to control the repository class
1999
# created by open and initialize.
2000
repository_class = None
2001
# Set this attribute in derived classes to control the
2002
# _commit_builder_class that the repository objects will have passed to
2003
# their constructor.
2004
_commit_builder_class = None
2005
# Set this attribute in derived clases to control the _serializer that the
2006
# repository objects will have passed to their constructor.
2008
# External references are not supported in pack repositories yet.
2009
supports_external_lookups = False
2011
def initialize(self, a_bzrdir, shared=False):
2012
"""Create a pack based repository.
2014
:param a_bzrdir: bzrdir to contain the new repository; must already
2016
:param shared: If true the repository will be initialized as a shared
2019
mutter('creating repository in %s.', a_bzrdir.transport.base)
2020
dirs = ['indices', 'obsolete_packs', 'packs', 'upload']
2021
builder = GraphIndexBuilder()
2022
files = [('pack-names', builder.finish())]
2023
utf8_files = [('format', self.get_format_string())]
2025
self._upload_blank_content(a_bzrdir, dirs, files, utf8_files, shared)
2026
return self.open(a_bzrdir=a_bzrdir, _found=True)
2028
def open(self, a_bzrdir, _found=False, _override_transport=None):
2029
"""See RepositoryFormat.open().
2031
:param _override_transport: INTERNAL USE ONLY. Allows opening the
2032
repository at a slightly different url
2033
than normal. I.e. during 'upgrade'.
2036
format = RepositoryFormat.find_format(a_bzrdir)
2037
if _override_transport is not None:
2038
repo_transport = _override_transport
2040
repo_transport = a_bzrdir.get_repository_transport(None)
2041
control_files = lockable_files.LockableFiles(repo_transport,
2042
'lock', lockdir.LockDir)
2043
return self.repository_class(_format=self,
2045
control_files=control_files,
2046
_commit_builder_class=self._commit_builder_class,
2047
_serializer=self._serializer)
2050
class RepositoryFormatKnitPack1(RepositoryFormatPack):
2051
"""A no-subtrees parameterized Pack repository.
2053
This format was introduced in 0.92.
2056
repository_class = KnitPackRepository
2057
_commit_builder_class = PackCommitBuilder
2058
_serializer = xml5.serializer_v5
2060
def _get_matching_bzrdir(self):
2061
return bzrdir.format_registry.make_bzrdir('pack-0.92')
2063
def _ignore_setting_bzrdir(self, format):
2066
_matchingbzrdir = property(_get_matching_bzrdir, _ignore_setting_bzrdir)
2068
def get_format_string(self):
2069
"""See RepositoryFormat.get_format_string()."""
2070
return "Bazaar pack repository format 1 (needs bzr 0.92)\n"
2072
def get_format_description(self):
2073
"""See RepositoryFormat.get_format_description()."""
2074
return "Packs containing knits without subtree support"
2076
def check_conversion_target(self, target_format):
2080
class RepositoryFormatKnitPack3(RepositoryFormatPack):
2081
"""A subtrees parameterized Pack repository.
2083
This repository format uses the xml7 serializer to get:
2084
- support for recording full info about the tree root
2085
- support for recording tree-references
2087
This format was introduced in 0.92.
2090
repository_class = KnitPackRepository
2091
_commit_builder_class = PackRootCommitBuilder
2092
rich_root_data = True
2093
supports_tree_reference = True
2094
_serializer = xml7.serializer_v7
2096
def _get_matching_bzrdir(self):
2097
return bzrdir.format_registry.make_bzrdir(
2098
'pack-0.92-subtree')
2100
def _ignore_setting_bzrdir(self, format):
2103
_matchingbzrdir = property(_get_matching_bzrdir, _ignore_setting_bzrdir)
2105
def check_conversion_target(self, target_format):
2106
if not target_format.rich_root_data:
2107
raise errors.BadConversionTarget(
2108
'Does not support rich root data.', target_format)
2109
if not getattr(target_format, 'supports_tree_reference', False):
2110
raise errors.BadConversionTarget(
2111
'Does not support nested trees', target_format)
2113
def get_format_string(self):
2114
"""See RepositoryFormat.get_format_string()."""
2115
return "Bazaar pack repository format 1 with subtree support (needs bzr 0.92)\n"
2117
def get_format_description(self):
2118
"""See RepositoryFormat.get_format_description()."""
2119
return "Packs containing knits with subtree support\n"
2122
class RepositoryFormatKnitPack4(RepositoryFormatPack):
2123
"""A rich-root, no subtrees parameterized Pack repository.
2125
This repository format uses the xml6 serializer to get:
2126
- support for recording full info about the tree root
2128
This format was introduced in 1.0.
2131
repository_class = KnitPackRepository
2132
_commit_builder_class = PackRootCommitBuilder
2133
rich_root_data = True
2134
supports_tree_reference = False
2135
_serializer = xml6.serializer_v6
2137
def _get_matching_bzrdir(self):
2138
return bzrdir.format_registry.make_bzrdir(
2141
def _ignore_setting_bzrdir(self, format):
2144
_matchingbzrdir = property(_get_matching_bzrdir, _ignore_setting_bzrdir)
2146
def check_conversion_target(self, target_format):
2147
if not target_format.rich_root_data:
2148
raise errors.BadConversionTarget(
2149
'Does not support rich root data.', target_format)
2151
def get_format_string(self):
2152
"""See RepositoryFormat.get_format_string()."""
2153
return ("Bazaar pack repository format 1 with rich root"
2154
" (needs bzr 1.0)\n")
2156
def get_format_description(self):
2157
"""See RepositoryFormat.get_format_description()."""
2158
return "Packs containing knits with rich root support\n"
2161
class RepositoryFormatKnitPack5(RepositoryFormatPack):
2162
"""Repository that supports external references to allow stacking.
2166
Supports external lookups, which results in non-truncated ghosts after
2167
reconcile compared to pack-0.92 formats.
2170
repository_class = KnitPackRepository
2171
_commit_builder_class = PackCommitBuilder
2172
_serializer = xml5.serializer_v5
2173
supports_external_lookups = True
2175
def _get_matching_bzrdir(self):
2176
return bzrdir.format_registry.make_bzrdir('development1')
2178
def _ignore_setting_bzrdir(self, format):
2181
_matchingbzrdir = property(_get_matching_bzrdir, _ignore_setting_bzrdir)
2183
def get_format_string(self):
2184
"""See RepositoryFormat.get_format_string()."""
2185
return "Bazaar RepositoryFormatKnitPack5 (bzr 1.6)\n"
2187
def get_format_description(self):
2188
"""See RepositoryFormat.get_format_description()."""
2191
def check_conversion_target(self, target_format):
2195
class RepositoryFormatKnitPack5RichRoot(RepositoryFormatPack):
2196
"""A repository with subtrees and external references.
2200
Supports external lookups, which results in non-truncated ghosts after
2201
reconcile compared to pack-0.92 formats.
2204
repository_class = KnitPackRepository
2205
_commit_builder_class = PackRootCommitBuilder
2206
rich_root_data = True
2207
supports_tree_reference = False # no subtrees
2208
_serializer = xml7.serializer_v7
2210
supports_external_lookups = True
2212
def _get_matching_bzrdir(self):
2213
return bzrdir.format_registry.make_bzrdir(
2214
'development1-subtree')
2216
def _ignore_setting_bzrdir(self, format):
2219
_matchingbzrdir = property(_get_matching_bzrdir, _ignore_setting_bzrdir)
2221
def check_conversion_target(self, target_format):
2222
if not target_format.rich_root_data:
2223
raise errors.BadConversionTarget(
2224
'Does not support rich root data.', target_format)
2226
def get_format_string(self):
2227
"""See RepositoryFormat.get_format_string()."""
2228
return "Bazaar RepositoryFormatKnitPack5RichRoot (bzr 1.6)\n"
2230
def get_format_description(self):
2231
"""See RepositoryFormat.get_format_description()."""
2235
class RepositoryFormatPackDevelopment0(RepositoryFormatPack):
2236
"""A no-subtrees development repository.
2238
This format should be retained until the second release after bzr 1.0.
2240
No changes to the disk behaviour from pack-0.92.
2243
repository_class = KnitPackRepository
2244
_commit_builder_class = PackCommitBuilder
2245
_serializer = xml5.serializer_v5
2247
def _get_matching_bzrdir(self):
2248
return bzrdir.format_registry.make_bzrdir('development0')
2250
def _ignore_setting_bzrdir(self, format):
2253
_matchingbzrdir = property(_get_matching_bzrdir, _ignore_setting_bzrdir)
2255
def get_format_string(self):
2256
"""See RepositoryFormat.get_format_string()."""
2257
return "Bazaar development format 0 (needs bzr.dev from before 1.3)\n"
2259
def get_format_description(self):
2260
"""See RepositoryFormat.get_format_description()."""
2261
return ("Development repository format, currently the same as "
2264
def check_conversion_target(self, target_format):
2268
class RepositoryFormatPackDevelopment0Subtree(RepositoryFormatPack):
2269
"""A subtrees development repository.
2271
This format should be retained until the second release after bzr 1.0.
2273
No changes to the disk behaviour from pack-0.92-subtree.
2276
repository_class = KnitPackRepository
2277
_commit_builder_class = PackRootCommitBuilder
2278
rich_root_data = True
2279
supports_tree_reference = True
2280
_serializer = xml7.serializer_v7
2282
def _get_matching_bzrdir(self):
2283
return bzrdir.format_registry.make_bzrdir(
2284
'development0-subtree')
2286
def _ignore_setting_bzrdir(self, format):
2289
_matchingbzrdir = property(_get_matching_bzrdir, _ignore_setting_bzrdir)
2291
def check_conversion_target(self, target_format):
2292
if not target_format.rich_root_data:
2293
raise errors.BadConversionTarget(
2294
'Does not support rich root data.', target_format)
2295
if not getattr(target_format, 'supports_tree_reference', False):
2296
raise errors.BadConversionTarget(
2297
'Does not support nested trees', target_format)
2299
def get_format_string(self):
2300
"""See RepositoryFormat.get_format_string()."""
2301
return ("Bazaar development format 0 with subtree support "
2302
"(needs bzr.dev from before 1.3)\n")
2304
def get_format_description(self):
2305
"""See RepositoryFormat.get_format_description()."""
2306
return ("Development repository format, currently the same as "
2307
"pack-0.92-subtree\n")
2310
class RepositoryFormatPackDevelopment1(RepositoryFormatPackDevelopment0):
2311
"""A no-subtrees development repository.
2313
This format should be retained until the second release after bzr 1.5.
2315
Supports external lookups, which results in non-truncated ghosts after
2316
reconcile compared to pack-0.92 formats.
2319
supports_external_lookups = True
2321
def _get_matching_bzrdir(self):
2322
return bzrdir.format_registry.make_bzrdir('development1')
2324
def _ignore_setting_bzrdir(self, format):
2327
_matchingbzrdir = property(_get_matching_bzrdir, _ignore_setting_bzrdir)
2329
def get_format_string(self):
2330
"""See RepositoryFormat.get_format_string()."""
2331
return "Bazaar development format 1 (needs bzr.dev from before 1.6)\n"
2333
def get_format_description(self):
2334
"""See RepositoryFormat.get_format_description()."""
2335
return ("Development repository format, currently the same as "
2336
"pack-0.92 with external reference support.\n")
2338
def check_conversion_target(self, target_format):
2342
class RepositoryFormatPackDevelopment1Subtree(RepositoryFormatPackDevelopment0Subtree):
2343
"""A subtrees development repository.
2345
This format should be retained until the second release after bzr 1.5.
2347
Supports external lookups, which results in non-truncated ghosts after
2348
reconcile compared to pack-0.92 formats.
2351
supports_external_lookups = True
2353
def _get_matching_bzrdir(self):
2354
return bzrdir.format_registry.make_bzrdir(
2355
'development1-subtree')
2357
def _ignore_setting_bzrdir(self, format):
2360
_matchingbzrdir = property(_get_matching_bzrdir, _ignore_setting_bzrdir)
2362
def check_conversion_target(self, target_format):
2363
if not target_format.rich_root_data:
2364
raise errors.BadConversionTarget(
2365
'Does not support rich root data.', target_format)
2366
if not getattr(target_format, 'supports_tree_reference', False):
2367
raise errors.BadConversionTarget(
2368
'Does not support nested trees', target_format)
2370
def get_format_string(self):
2371
"""See RepositoryFormat.get_format_string()."""
2372
return ("Bazaar development format 1 with subtree support "
2373
"(needs bzr.dev from before 1.6)\n")
2375
def get_format_description(self):
2376
"""See RepositoryFormat.get_format_description()."""
2377
return ("Development repository format, currently the same as "
2378
"pack-0.92-subtree with external reference support.\n")
added
removed
bzrlib/repofmt/pack_repo.py
