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
33
from bzrlib.index import (
37
GraphIndexPrefixAdapter,
40
from bzrlib.knit import (
46
from bzrlib import tsort
56
from bzrlib.decorators import needs_write_lock
57
from bzrlib.btree_index import (
61
from bzrlib.index import (
65
from bzrlib.repofmt.knitrepo import KnitRepository
66
from bzrlib.repository import (
68
MetaDirRepositoryFormat,
72
import bzrlib.revision as _mod_revision
73
from bzrlib.trace import (
79
class PackCommitBuilder(CommitBuilder):
80
"""A subclass of CommitBuilder to add texts with pack semantics.
82
Specifically this uses one knit object rather than one knit object per
83
added text, reducing memory and object pressure.
86
def __init__(self, repository, parents, config, timestamp=None,
87
timezone=None, committer=None, revprops=None,
89
CommitBuilder.__init__(self, repository, parents, config,
90
timestamp=timestamp, timezone=timezone, committer=committer,
91
revprops=revprops, revision_id=revision_id)
92
self._file_graph = graph.Graph(
93
repository._pack_collection.text_index.combined_index)
95
def _heads(self, file_id, revision_ids):
96
keys = [(file_id, revision_id) for revision_id in revision_ids]
97
return set([key[1] for key in self._file_graph.heads(keys)])
100
class PackRootCommitBuilder(RootCommitBuilder):
101
"""A subclass of RootCommitBuilder to add texts with pack semantics.
103
Specifically this uses one knit object rather than one knit object per
104
added text, reducing memory and object pressure.
107
def __init__(self, repository, parents, config, timestamp=None,
108
timezone=None, committer=None, revprops=None,
110
CommitBuilder.__init__(self, repository, parents, config,
111
timestamp=timestamp, timezone=timezone, committer=committer,
112
revprops=revprops, revision_id=revision_id)
113
self._file_graph = graph.Graph(
114
repository._pack_collection.text_index.combined_index)
116
def _heads(self, file_id, revision_ids):
117
keys = [(file_id, revision_id) for revision_id in revision_ids]
118
return set([key[1] for key in self._file_graph.heads(keys)])
122
"""An in memory proxy for a pack and its indices.
124
This is a base class that is not directly used, instead the classes
125
ExistingPack and NewPack are used.
128
def __init__(self, revision_index, inventory_index, text_index,
130
"""Create a pack instance.
132
:param revision_index: A GraphIndex for determining what revisions are
133
present in the Pack and accessing the locations of their texts.
134
:param inventory_index: A GraphIndex for determining what inventories are
135
present in the Pack and accessing the locations of their
137
:param text_index: A GraphIndex for determining what file texts
138
are present in the pack and accessing the locations of their
139
texts/deltas (via (fileid, revisionid) tuples).
140
:param signature_index: A GraphIndex for determining what signatures are
141
present in the Pack and accessing the locations of their texts.
143
self.revision_index = revision_index
144
self.inventory_index = inventory_index
145
self.text_index = text_index
146
self.signature_index = signature_index
148
def access_tuple(self):
149
"""Return a tuple (transport, name) for the pack content."""
150
return self.pack_transport, self.file_name()
153
"""Get the file name for the pack on disk."""
154
return self.name + '.pack'
156
def get_revision_count(self):
157
return self.revision_index.key_count()
159
def inventory_index_name(self, name):
160
"""The inv index is the name + .iix."""
161
return self.index_name('inventory', name)
163
def revision_index_name(self, name):
164
"""The revision index is the name + .rix."""
165
return self.index_name('revision', name)
167
def signature_index_name(self, name):
168
"""The signature index is the name + .six."""
169
return self.index_name('signature', name)
171
def text_index_name(self, name):
172
"""The text index is the name + .tix."""
173
return self.index_name('text', name)
175
def _external_compression_parents_of_texts(self):
178
for node in self.text_index.iter_all_entries():
180
refs.update(node[3][1])
184
class ExistingPack(Pack):
185
"""An in memory proxy for an existing .pack and its disk indices."""
187
def __init__(self, pack_transport, name, revision_index, inventory_index,
188
text_index, signature_index):
189
"""Create an ExistingPack object.
191
:param pack_transport: The transport where the pack file resides.
192
:param name: The name of the pack on disk in the pack_transport.
194
Pack.__init__(self, revision_index, inventory_index, text_index,
197
self.pack_transport = pack_transport
198
if None in (revision_index, inventory_index, text_index,
199
signature_index, name, pack_transport):
200
raise AssertionError()
202
def __eq__(self, other):
203
return self.__dict__ == other.__dict__
205
def __ne__(self, other):
206
return not self.__eq__(other)
209
return "<bzrlib.repofmt.pack_repo.Pack object at 0x%x, %s, %s" % (
210
id(self), self.pack_transport, self.name)
214
"""An in memory proxy for a pack which is being created."""
216
# A map of index 'type' to the file extension and position in the
218
index_definitions = {
219
'revision': ('.rix', 0),
220
'inventory': ('.iix', 1),
222
'signature': ('.six', 3),
225
def __init__(self, upload_transport, index_transport, pack_transport,
226
upload_suffix='', file_mode=None, index_builder_class=None,
228
"""Create a NewPack instance.
230
:param upload_transport: A writable transport for the pack to be
231
incrementally uploaded to.
232
:param index_transport: A writable transport for the pack's indices to
233
be written to when the pack is finished.
234
:param pack_transport: A writable transport for the pack to be renamed
235
to when the upload is complete. This *must* be the same as
236
upload_transport.clone('../packs').
237
:param upload_suffix: An optional suffix to be given to any temporary
238
files created during the pack creation. e.g '.autopack'
239
:param file_mode: An optional file mode to create the new files with.
240
:param index_builder_class: Required keyword parameter - the class of
241
index builder to use.
242
:param index_class: Required keyword parameter - the class of index
245
# The relative locations of the packs are constrained, but all are
246
# passed in because the caller has them, so as to avoid object churn.
248
# Revisions: parents list, no text compression.
249
index_builder_class(reference_lists=1),
250
# Inventory: We want to map compression only, but currently the
251
# knit code hasn't been updated enough to understand that, so we
252
# have a regular 2-list index giving parents and compression
254
index_builder_class(reference_lists=2),
255
# Texts: compression and per file graph, for all fileids - so two
256
# reference lists and two elements in the key tuple.
257
index_builder_class(reference_lists=2, key_elements=2),
258
# Signatures: Just blobs to store, no compression, no parents
260
index_builder_class(reference_lists=0),
262
# When we make readonly indices, we need this.
263
self.index_class = index_class
264
# where should the new pack be opened
265
self.upload_transport = upload_transport
266
# where are indices written out to
267
self.index_transport = index_transport
268
# where is the pack renamed to when it is finished?
269
self.pack_transport = pack_transport
270
# What file mode to upload the pack and indices with.
271
self._file_mode = file_mode
272
# tracks the content written to the .pack file.
273
self._hash = osutils.md5()
274
# a four-tuple with the length in bytes of the indices, once the pack
275
# is finalised. (rev, inv, text, sigs)
276
self.index_sizes = None
277
# How much data to cache when writing packs. Note that this is not
278
# synchronised with reads, because it's not in the transport layer, so
279
# is not safe unless the client knows it won't be reading from the pack
281
self._cache_limit = 0
282
# the temporary pack file name.
283
self.random_name = osutils.rand_chars(20) + upload_suffix
284
# when was this pack started ?
285
self.start_time = time.time()
286
# open an output stream for the data added to the pack.
287
self.write_stream = self.upload_transport.open_write_stream(
288
self.random_name, mode=self._file_mode)
289
if 'pack' in debug.debug_flags:
290
mutter('%s: create_pack: pack stream open: %s%s t+%6.3fs',
291
time.ctime(), self.upload_transport.base, self.random_name,
292
time.time() - self.start_time)
293
# A list of byte sequences to be written to the new pack, and the
294
# aggregate size of them. Stored as a list rather than separate
295
# variables so that the _write_data closure below can update them.
296
self._buffer = [[], 0]
297
# create a callable for adding data
299
# robertc says- this is a closure rather than a method on the object
300
# so that the variables are locals, and faster than accessing object
302
def _write_data(bytes, flush=False, _buffer=self._buffer,
303
_write=self.write_stream.write, _update=self._hash.update):
304
_buffer[0].append(bytes)
305
_buffer[1] += len(bytes)
307
if _buffer[1] > self._cache_limit or flush:
308
bytes = ''.join(_buffer[0])
312
# expose this on self, for the occasion when clients want to add data.
313
self._write_data = _write_data
314
# a pack writer object to serialise pack records.
315
self._writer = pack.ContainerWriter(self._write_data)
317
# what state is the pack in? (open, finished, aborted)
321
"""Cancel creating this pack."""
322
self._state = 'aborted'
323
self.write_stream.close()
324
# Remove the temporary pack file.
325
self.upload_transport.delete(self.random_name)
326
# The indices have no state on disk.
328
def access_tuple(self):
329
"""Return a tuple (transport, name) for the pack content."""
330
if self._state == 'finished':
331
return Pack.access_tuple(self)
332
elif self._state == 'open':
333
return self.upload_transport, self.random_name
335
raise AssertionError(self._state)
337
def data_inserted(self):
338
"""True if data has been added to this pack."""
339
return bool(self.get_revision_count() or
340
self.inventory_index.key_count() or
341
self.text_index.key_count() or
342
self.signature_index.key_count())
345
"""Finish the new pack.
348
- finalises the content
349
- assigns a name (the md5 of the content, currently)
350
- writes out the associated indices
351
- renames the pack into place.
352
- stores the index size tuple for the pack in the index_sizes
357
self._write_data('', flush=True)
358
self.name = self._hash.hexdigest()
360
# XXX: It'd be better to write them all to temporary names, then
361
# rename them all into place, so that the window when only some are
362
# visible is smaller. On the other hand none will be seen until
363
# they're in the names list.
364
self.index_sizes = [None, None, None, None]
365
self._write_index('revision', self.revision_index, 'revision')
366
self._write_index('inventory', self.inventory_index, 'inventory')
367
self._write_index('text', self.text_index, 'file texts')
368
self._write_index('signature', self.signature_index,
369
'revision signatures')
370
self.write_stream.close()
371
# Note that this will clobber an existing pack with the same name,
372
# without checking for hash collisions. While this is undesirable this
373
# is something that can be rectified in a subsequent release. One way
374
# to rectify it may be to leave the pack at the original name, writing
375
# its pack-names entry as something like 'HASH: index-sizes
376
# temporary-name'. Allocate that and check for collisions, if it is
377
# collision free then rename it into place. If clients know this scheme
378
# they can handle missing-file errors by:
379
# - try for HASH.pack
380
# - try for temporary-name
381
# - refresh the pack-list to see if the pack is now absent
382
self.upload_transport.rename(self.random_name,
383
'../packs/' + self.name + '.pack')
384
self._state = 'finished'
385
if 'pack' in debug.debug_flags:
386
# XXX: size might be interesting?
387
mutter('%s: create_pack: pack renamed into place: %s%s->%s%s t+%6.3fs',
388
time.ctime(), self.upload_transport.base, self.random_name,
389
self.pack_transport, self.name,
390
time.time() - self.start_time)
393
"""Flush any current data."""
395
bytes = ''.join(self._buffer[0])
396
self.write_stream.write(bytes)
397
self._hash.update(bytes)
398
self._buffer[:] = [[], 0]
400
def index_name(self, index_type, name):
401
"""Get the disk name of an index type for pack name 'name'."""
402
return name + NewPack.index_definitions[index_type][0]
404
def index_offset(self, index_type):
405
"""Get the position in a index_size array for a given index type."""
406
return NewPack.index_definitions[index_type][1]
408
def _replace_index_with_readonly(self, index_type):
409
setattr(self, index_type + '_index',
410
self.index_class(self.index_transport,
411
self.index_name(index_type, self.name),
412
self.index_sizes[self.index_offset(index_type)]))
414
def set_write_cache_size(self, size):
415
self._cache_limit = size
417
def _write_index(self, index_type, index, label):
418
"""Write out an index.
420
:param index_type: The type of index to write - e.g. 'revision'.
421
:param index: The index object to serialise.
422
:param label: What label to give the index e.g. 'revision'.
424
index_name = self.index_name(index_type, self.name)
425
self.index_sizes[self.index_offset(index_type)] = \
426
self.index_transport.put_file(index_name, index.finish(),
427
mode=self._file_mode)
428
if 'pack' in debug.debug_flags:
429
# XXX: size might be interesting?
430
mutter('%s: create_pack: wrote %s index: %s%s t+%6.3fs',
431
time.ctime(), label, self.upload_transport.base,
432
self.random_name, time.time() - self.start_time)
433
# Replace the writable index on this object with a readonly,
434
# presently unloaded index. We should alter
435
# the index layer to make its finish() error if add_node is
436
# subsequently used. RBC
437
self._replace_index_with_readonly(index_type)
440
class AggregateIndex(object):
441
"""An aggregated index for the RepositoryPackCollection.
443
AggregateIndex is reponsible for managing the PackAccess object,
444
Index-To-Pack mapping, and all indices list for a specific type of index
445
such as 'revision index'.
447
A CombinedIndex provides an index on a single key space built up
448
from several on-disk indices. The AggregateIndex builds on this
449
to provide a knit access layer, and allows having up to one writable
450
index within the collection.
452
# XXX: Probably 'can be written to' could/should be separated from 'acts
453
# like a knit index' -- mbp 20071024
455
def __init__(self, reload_func=None):
456
"""Create an AggregateIndex.
458
:param reload_func: A function to call if we find we are missing an
459
index. Should have the form reload_func() => True if the list of
460
active pack files has changed.
462
self._reload_func = reload_func
463
self.index_to_pack = {}
464
self.combined_index = CombinedGraphIndex([], reload_func=reload_func)
465
self.data_access = _DirectPackAccess(self.index_to_pack)
466
self.add_callback = None
468
def replace_indices(self, index_to_pack, indices):
469
"""Replace the current mappings with fresh ones.
471
This should probably not be used eventually, rather incremental add and
472
removal of indices. It has been added during refactoring of existing
475
:param index_to_pack: A mapping from index objects to
476
(transport, name) tuples for the pack file data.
477
:param indices: A list of indices.
479
# refresh the revision pack map dict without replacing the instance.
480
self.index_to_pack.clear()
481
self.index_to_pack.update(index_to_pack)
482
# XXX: API break - clearly a 'replace' method would be good?
483
self.combined_index._indices[:] = indices
484
# the current add nodes callback for the current writable index if
486
self.add_callback = None
488
def add_index(self, index, pack):
489
"""Add index to the aggregate, which is an index for Pack pack.
491
Future searches on the aggregate index will seach this new index
492
before all previously inserted indices.
494
:param index: An Index for the pack.
495
:param pack: A Pack instance.
497
# expose it to the index map
498
self.index_to_pack[index] = pack.access_tuple()
499
# put it at the front of the linear index list
500
self.combined_index.insert_index(0, index)
502
def add_writable_index(self, index, pack):
503
"""Add an index which is able to have data added to it.
505
There can be at most one writable index at any time. Any
506
modifications made to the knit are put into this index.
508
:param index: An index from the pack parameter.
509
:param pack: A Pack instance.
511
if self.add_callback is not None:
512
raise AssertionError(
513
"%s already has a writable index through %s" % \
514
(self, self.add_callback))
515
# allow writing: queue writes to a new index
516
self.add_index(index, pack)
517
# Updates the index to packs mapping as a side effect,
518
self.data_access.set_writer(pack._writer, index, pack.access_tuple())
519
self.add_callback = index.add_nodes
522
"""Reset all the aggregate data to nothing."""
523
self.data_access.set_writer(None, None, (None, None))
524
self.index_to_pack.clear()
525
del self.combined_index._indices[:]
526
self.add_callback = None
528
def remove_index(self, index, pack):
529
"""Remove index from the indices used to answer queries.
531
:param index: An index from the pack parameter.
532
:param pack: A Pack instance.
534
del self.index_to_pack[index]
535
self.combined_index._indices.remove(index)
536
if (self.add_callback is not None and
537
getattr(index, 'add_nodes', None) == self.add_callback):
538
self.add_callback = None
539
self.data_access.set_writer(None, None, (None, None))
542
class Packer(object):
543
"""Create a pack from packs."""
545
def __init__(self, pack_collection, packs, suffix, revision_ids=None):
548
:param pack_collection: A RepositoryPackCollection object where the
549
new pack is being written to.
550
:param packs: The packs to combine.
551
:param suffix: The suffix to use on the temporary files for the pack.
552
:param revision_ids: Revision ids to limit the pack to.
556
self.revision_ids = revision_ids
557
# The pack object we are creating.
559
self._pack_collection = pack_collection
560
# The index layer keys for the revisions being copied. None for 'all
562
self._revision_keys = None
563
# What text keys to copy. None for 'all texts'. This is set by
564
# _copy_inventory_texts
565
self._text_filter = None
568
def _extra_init(self):
569
"""A template hook to allow extending the constructor trivially."""
571
def pack(self, pb=None):
572
"""Create a new pack by reading data from other packs.
574
This does little more than a bulk copy of data. One key difference
575
is that data with the same item key across multiple packs is elided
576
from the output. The new pack is written into the current pack store
577
along with its indices, and the name added to the pack names. The
578
source packs are not altered and are not required to be in the current
581
:param pb: An optional progress bar to use. A nested bar is created if
583
:return: A Pack object, or None if nothing was copied.
585
# open a pack - using the same name as the last temporary file
586
# - which has already been flushed, so its safe.
587
# XXX: - duplicate code warning with start_write_group; fix before
588
# considering 'done'.
589
if self._pack_collection._new_pack is not None:
590
raise errors.BzrError('call to create_pack_from_packs while '
591
'another pack is being written.')
592
if self.revision_ids is not None:
593
if len(self.revision_ids) == 0:
594
# silly fetch request.
597
self.revision_ids = frozenset(self.revision_ids)
598
self.revision_keys = frozenset((revid,) for revid in
601
self.pb = ui.ui_factory.nested_progress_bar()
605
return self._create_pack_from_packs()
611
"""Open a pack for the pack we are creating."""
612
return NewPack(self._pack_collection._upload_transport,
613
self._pack_collection._index_transport,
614
self._pack_collection._pack_transport, upload_suffix=self.suffix,
615
file_mode=self._pack_collection.repo.bzrdir._get_file_mode(),
616
index_builder_class=self._pack_collection._index_builder_class,
617
index_class=self._pack_collection._index_class)
619
def _copy_revision_texts(self):
620
"""Copy revision data to the new pack."""
622
if self.revision_ids:
623
revision_keys = [(revision_id,) for revision_id in self.revision_ids]
626
# select revision keys
627
revision_index_map = self._pack_collection._packs_list_to_pack_map_and_index_list(
628
self.packs, 'revision_index')[0]
629
revision_nodes = self._pack_collection._index_contents(revision_index_map, revision_keys)
630
# copy revision keys and adjust values
631
self.pb.update("Copying revision texts", 1)
632
total_items, readv_group_iter = self._revision_node_readv(revision_nodes)
633
list(self._copy_nodes_graph(revision_index_map, self.new_pack._writer,
634
self.new_pack.revision_index, readv_group_iter, total_items))
635
if 'pack' in debug.debug_flags:
636
mutter('%s: create_pack: revisions copied: %s%s %d items t+%6.3fs',
637
time.ctime(), self._pack_collection._upload_transport.base,
638
self.new_pack.random_name,
639
self.new_pack.revision_index.key_count(),
640
time.time() - self.new_pack.start_time)
641
self._revision_keys = revision_keys
643
def _copy_inventory_texts(self):
644
"""Copy the inventory texts to the new pack.
646
self._revision_keys is used to determine what inventories to copy.
648
Sets self._text_filter appropriately.
650
# select inventory keys
651
inv_keys = self._revision_keys # currently the same keyspace, and note that
652
# querying for keys here could introduce a bug where an inventory item
653
# is missed, so do not change it to query separately without cross
654
# checking like the text key check below.
655
inventory_index_map = self._pack_collection._packs_list_to_pack_map_and_index_list(
656
self.packs, 'inventory_index')[0]
657
inv_nodes = self._pack_collection._index_contents(inventory_index_map, inv_keys)
658
# copy inventory keys and adjust values
659
# XXX: Should be a helper function to allow different inv representation
661
self.pb.update("Copying inventory texts", 2)
662
total_items, readv_group_iter = self._least_readv_node_readv(inv_nodes)
663
# Only grab the output lines if we will be processing them
664
output_lines = bool(self.revision_ids)
665
inv_lines = self._copy_nodes_graph(inventory_index_map,
666
self.new_pack._writer, self.new_pack.inventory_index,
667
readv_group_iter, total_items, output_lines=output_lines)
668
if self.revision_ids:
669
self._process_inventory_lines(inv_lines)
671
# eat the iterator to cause it to execute.
673
self._text_filter = None
674
if 'pack' in debug.debug_flags:
675
mutter('%s: create_pack: inventories copied: %s%s %d items t+%6.3fs',
676
time.ctime(), self._pack_collection._upload_transport.base,
677
self.new_pack.random_name,
678
self.new_pack.inventory_index.key_count(),
679
time.time() - self.new_pack.start_time)
681
def _copy_text_texts(self):
683
text_index_map, text_nodes = self._get_text_nodes()
684
if self._text_filter is not None:
685
# We could return the keys copied as part of the return value from
686
# _copy_nodes_graph but this doesn't work all that well with the
687
# need to get line output too, so we check separately, and as we're
688
# going to buffer everything anyway, we check beforehand, which
689
# saves reading knit data over the wire when we know there are
691
text_nodes = set(text_nodes)
692
present_text_keys = set(_node[1] for _node in text_nodes)
693
missing_text_keys = set(self._text_filter) - present_text_keys
694
if missing_text_keys:
695
# TODO: raise a specific error that can handle many missing
697
a_missing_key = missing_text_keys.pop()
698
raise errors.RevisionNotPresent(a_missing_key[1],
700
# copy text keys and adjust values
701
self.pb.update("Copying content texts", 3)
702
total_items, readv_group_iter = self._least_readv_node_readv(text_nodes)
703
list(self._copy_nodes_graph(text_index_map, self.new_pack._writer,
704
self.new_pack.text_index, readv_group_iter, total_items))
705
self._log_copied_texts()
707
def _check_references(self):
708
"""Make sure our external refereneces are present."""
709
external_refs = self.new_pack._external_compression_parents_of_texts()
711
index = self._pack_collection.text_index.combined_index
712
found_items = list(index.iter_entries(external_refs))
713
if len(found_items) != len(external_refs):
714
found_keys = set(k for idx, k, refs, value in found_items)
715
missing_items = external_refs - found_keys
716
missing_file_id, missing_revision_id = missing_items.pop()
717
raise errors.RevisionNotPresent(missing_revision_id,
720
def _create_pack_from_packs(self):
721
self.pb.update("Opening pack", 0, 5)
722
self.new_pack = self.open_pack()
723
new_pack = self.new_pack
724
# buffer data - we won't be reading-back during the pack creation and
725
# this makes a significant difference on sftp pushes.
726
new_pack.set_write_cache_size(1024*1024)
727
if 'pack' in debug.debug_flags:
728
plain_pack_list = ['%s%s' % (a_pack.pack_transport.base, a_pack.name)
729
for a_pack in self.packs]
730
if self.revision_ids is not None:
731
rev_count = len(self.revision_ids)
734
mutter('%s: create_pack: creating pack from source packs: '
735
'%s%s %s revisions wanted %s t=0',
736
time.ctime(), self._pack_collection._upload_transport.base, new_pack.random_name,
737
plain_pack_list, rev_count)
738
self._copy_revision_texts()
739
self._copy_inventory_texts()
740
self._copy_text_texts()
741
# select signature keys
742
signature_filter = self._revision_keys # same keyspace
743
signature_index_map = self._pack_collection._packs_list_to_pack_map_and_index_list(
744
self.packs, 'signature_index')[0]
745
signature_nodes = self._pack_collection._index_contents(signature_index_map,
747
# copy signature keys and adjust values
748
self.pb.update("Copying signature texts", 4)
749
self._copy_nodes(signature_nodes, signature_index_map, new_pack._writer,
750
new_pack.signature_index)
751
if 'pack' in debug.debug_flags:
752
mutter('%s: create_pack: revision signatures copied: %s%s %d items t+%6.3fs',
753
time.ctime(), self._pack_collection._upload_transport.base, new_pack.random_name,
754
new_pack.signature_index.key_count(),
755
time.time() - new_pack.start_time)
756
self._check_references()
757
if not self._use_pack(new_pack):
760
self.pb.update("Finishing pack", 5)
762
self._pack_collection.allocate(new_pack)
765
def _copy_nodes(self, nodes, index_map, writer, write_index):
766
"""Copy knit nodes between packs with no graph references."""
767
pb = ui.ui_factory.nested_progress_bar()
769
return self._do_copy_nodes(nodes, index_map, writer,
774
def _do_copy_nodes(self, nodes, index_map, writer, write_index, pb):
775
# for record verification
776
knit = KnitVersionedFiles(None, None)
777
# plan a readv on each source pack:
779
nodes = sorted(nodes)
780
# how to map this into knit.py - or knit.py into this?
781
# we don't want the typical knit logic, we want grouping by pack
782
# at this point - perhaps a helper library for the following code
783
# duplication points?
785
for index, key, value in nodes:
786
if index not in request_groups:
787
request_groups[index] = []
788
request_groups[index].append((key, value))
790
pb.update("Copied record", record_index, len(nodes))
791
for index, items in request_groups.iteritems():
792
pack_readv_requests = []
793
for key, value in items:
794
# ---- KnitGraphIndex.get_position
795
bits = value[1:].split(' ')
796
offset, length = int(bits[0]), int(bits[1])
797
pack_readv_requests.append((offset, length, (key, value[0])))
798
# linear scan up the pack
799
pack_readv_requests.sort()
801
transport, path = index_map[index]
802
reader = pack.make_readv_reader(transport, path,
803
[offset[0:2] for offset in pack_readv_requests])
804
for (names, read_func), (_1, _2, (key, eol_flag)) in \
805
izip(reader.iter_records(), pack_readv_requests):
806
raw_data = read_func(None)
807
# check the header only
808
df, _ = knit._parse_record_header(key, raw_data)
810
pos, size = writer.add_bytes_record(raw_data, names)
811
write_index.add_node(key, eol_flag + "%d %d" % (pos, size))
812
pb.update("Copied record", record_index)
815
def _copy_nodes_graph(self, index_map, writer, write_index,
816
readv_group_iter, total_items, output_lines=False):
817
"""Copy knit nodes between packs.
819
:param output_lines: Return lines present in the copied data as
820
an iterator of line,version_id.
822
pb = ui.ui_factory.nested_progress_bar()
824
for result in self._do_copy_nodes_graph(index_map, writer,
825
write_index, output_lines, pb, readv_group_iter, total_items):
828
# Python 2.4 does not permit try:finally: in a generator.
834
def _do_copy_nodes_graph(self, index_map, writer, write_index,
835
output_lines, pb, readv_group_iter, total_items):
836
# for record verification
837
knit = KnitVersionedFiles(None, None)
838
# for line extraction when requested (inventories only)
840
factory = KnitPlainFactory()
842
pb.update("Copied record", record_index, total_items)
843
for index, readv_vector, node_vector in readv_group_iter:
845
transport, path = index_map[index]
846
reader = pack.make_readv_reader(transport, path, readv_vector)
847
for (names, read_func), (key, eol_flag, references) in \
848
izip(reader.iter_records(), node_vector):
849
raw_data = read_func(None)
851
# read the entire thing
852
content, _ = knit._parse_record(key[-1], raw_data)
853
if len(references[-1]) == 0:
854
line_iterator = factory.get_fulltext_content(content)
856
line_iterator = factory.get_linedelta_content(content)
857
for line in line_iterator:
860
# check the header only
861
df, _ = knit._parse_record_header(key, raw_data)
863
pos, size = writer.add_bytes_record(raw_data, names)
864
write_index.add_node(key, eol_flag + "%d %d" % (pos, size), references)
865
pb.update("Copied record", record_index)
868
def _get_text_nodes(self):
869
text_index_map = self._pack_collection._packs_list_to_pack_map_and_index_list(
870
self.packs, 'text_index')[0]
871
return text_index_map, self._pack_collection._index_contents(text_index_map,
874
def _least_readv_node_readv(self, nodes):
875
"""Generate request groups for nodes using the least readv's.
877
:param nodes: An iterable of graph index nodes.
878
:return: Total node count and an iterator of the data needed to perform
879
readvs to obtain the data for nodes. Each item yielded by the
880
iterator is a tuple with:
881
index, readv_vector, node_vector. readv_vector is a list ready to
882
hand to the transport readv method, and node_vector is a list of
883
(key, eol_flag, references) for the the node retrieved by the
884
matching readv_vector.
886
# group by pack so we do one readv per pack
887
nodes = sorted(nodes)
890
for index, key, value, references in nodes:
891
if index not in request_groups:
892
request_groups[index] = []
893
request_groups[index].append((key, value, references))
895
for index, items in request_groups.iteritems():
896
pack_readv_requests = []
897
for key, value, references in items:
898
# ---- KnitGraphIndex.get_position
899
bits = value[1:].split(' ')
900
offset, length = int(bits[0]), int(bits[1])
901
pack_readv_requests.append(
902
((offset, length), (key, value[0], references)))
903
# linear scan up the pack to maximum range combining.
904
pack_readv_requests.sort()
905
# split out the readv and the node data.
906
pack_readv = [readv for readv, node in pack_readv_requests]
907
node_vector = [node for readv, node in pack_readv_requests]
908
result.append((index, pack_readv, node_vector))
911
def _log_copied_texts(self):
912
if 'pack' in debug.debug_flags:
913
mutter('%s: create_pack: file texts copied: %s%s %d items t+%6.3fs',
914
time.ctime(), self._pack_collection._upload_transport.base,
915
self.new_pack.random_name,
916
self.new_pack.text_index.key_count(),
917
time.time() - self.new_pack.start_time)
919
def _process_inventory_lines(self, inv_lines):
920
"""Use up the inv_lines generator and setup a text key filter."""
921
repo = self._pack_collection.repo
922
fileid_revisions = repo._find_file_ids_from_xml_inventory_lines(
923
inv_lines, self.revision_keys)
925
for fileid, file_revids in fileid_revisions.iteritems():
926
text_filter.extend([(fileid, file_revid) for file_revid in file_revids])
927
self._text_filter = text_filter
929
def _revision_node_readv(self, revision_nodes):
930
"""Return the total revisions and the readv's to issue.
932
:param revision_nodes: The revision index contents for the packs being
933
incorporated into the new pack.
934
:return: As per _least_readv_node_readv.
936
return self._least_readv_node_readv(revision_nodes)
938
def _use_pack(self, new_pack):
939
"""Return True if new_pack should be used.
941
:param new_pack: The pack that has just been created.
942
:return: True if the pack should be used.
944
return new_pack.data_inserted()
947
class OptimisingPacker(Packer):
948
"""A packer which spends more time to create better disk layouts."""
950
def _revision_node_readv(self, revision_nodes):
951
"""Return the total revisions and the readv's to issue.
953
This sort places revisions in topological order with the ancestors
956
:param revision_nodes: The revision index contents for the packs being
957
incorporated into the new pack.
958
:return: As per _least_readv_node_readv.
960
# build an ancestors dict
963
for index, key, value, references in revision_nodes:
964
ancestors[key] = references[0]
965
by_key[key] = (index, value, references)
966
order = tsort.topo_sort(ancestors)
968
# Single IO is pathological, but it will work as a starting point.
970
for key in reversed(order):
971
index, value, references = by_key[key]
972
# ---- KnitGraphIndex.get_position
973
bits = value[1:].split(' ')
974
offset, length = int(bits[0]), int(bits[1])
976
(index, [(offset, length)], [(key, value[0], references)]))
977
# TODO: combine requests in the same index that are in ascending order.
978
return total, requests
981
"""Open a pack for the pack we are creating."""
982
new_pack = super(OptimisingPacker, self).open_pack()
983
# Turn on the optimization flags for all the index builders.
984
new_pack.revision_index.set_optimize(for_size=True)
985
new_pack.inventory_index.set_optimize(for_size=True)
986
new_pack.text_index.set_optimize(for_size=True)
987
new_pack.signature_index.set_optimize(for_size=True)
991
class ReconcilePacker(Packer):
992
"""A packer which regenerates indices etc as it copies.
994
This is used by ``bzr reconcile`` to cause parent text pointers to be
998
def _extra_init(self):
999
self._data_changed = False
1001
def _process_inventory_lines(self, inv_lines):
1002
"""Generate a text key reference map rather for reconciling with."""
1003
repo = self._pack_collection.repo
1004
refs = repo._find_text_key_references_from_xml_inventory_lines(
1006
self._text_refs = refs
1007
# during reconcile we:
1008
# - convert unreferenced texts to full texts
1009
# - correct texts which reference a text not copied to be full texts
1010
# - copy all others as-is but with corrected parents.
1011
# - so at this point we don't know enough to decide what becomes a full
1013
self._text_filter = None
1015
def _copy_text_texts(self):
1016
"""generate what texts we should have and then copy."""
1017
self.pb.update("Copying content texts", 3)
1018
# we have three major tasks here:
1019
# 1) generate the ideal index
1020
repo = self._pack_collection.repo
1021
ancestors = dict([(key[0], tuple(ref[0] for ref in refs[0])) for
1022
_1, key, _2, refs in
1023
self.new_pack.revision_index.iter_all_entries()])
1024
ideal_index = repo._generate_text_key_index(self._text_refs, ancestors)
1025
# 2) generate a text_nodes list that contains all the deltas that can
1026
# be used as-is, with corrected parents.
1029
discarded_nodes = []
1030
NULL_REVISION = _mod_revision.NULL_REVISION
1031
text_index_map, text_nodes = self._get_text_nodes()
1032
for node in text_nodes:
1038
ideal_parents = tuple(ideal_index[node[1]])
1040
discarded_nodes.append(node)
1041
self._data_changed = True
1043
if ideal_parents == (NULL_REVISION,):
1045
if ideal_parents == node[3][0]:
1047
ok_nodes.append(node)
1048
elif ideal_parents[0:1] == node[3][0][0:1]:
1049
# the left most parent is the same, or there are no parents
1050
# today. Either way, we can preserve the representation as
1051
# long as we change the refs to be inserted.
1052
self._data_changed = True
1053
ok_nodes.append((node[0], node[1], node[2],
1054
(ideal_parents, node[3][1])))
1055
self._data_changed = True
1057
# Reinsert this text completely
1058
bad_texts.append((node[1], ideal_parents))
1059
self._data_changed = True
1060
# we're finished with some data.
1063
# 3) bulk copy the ok data
1064
total_items, readv_group_iter = self._least_readv_node_readv(ok_nodes)
1065
list(self._copy_nodes_graph(text_index_map, self.new_pack._writer,
1066
self.new_pack.text_index, readv_group_iter, total_items))
1067
# 4) adhoc copy all the other texts.
1068
# We have to topologically insert all texts otherwise we can fail to
1069
# reconcile when parts of a single delta chain are preserved intact,
1070
# and other parts are not. E.g. Discarded->d1->d2->d3. d1 will be
1071
# reinserted, and if d3 has incorrect parents it will also be
1072
# reinserted. If we insert d3 first, d2 is present (as it was bulk
1073
# copied), so we will try to delta, but d2 is not currently able to be
1074
# extracted because it's basis d1 is not present. Topologically sorting
1075
# addresses this. The following generates a sort for all the texts that
1076
# are being inserted without having to reference the entire text key
1077
# space (we only topo sort the revisions, which is smaller).
1078
topo_order = tsort.topo_sort(ancestors)
1079
rev_order = dict(zip(topo_order, range(len(topo_order))))
1080
bad_texts.sort(key=lambda key:rev_order[key[0][1]])
1081
transaction = repo.get_transaction()
1082
file_id_index = GraphIndexPrefixAdapter(
1083
self.new_pack.text_index,
1085
add_nodes_callback=self.new_pack.text_index.add_nodes)
1086
data_access = _DirectPackAccess(
1087
{self.new_pack.text_index:self.new_pack.access_tuple()})
1088
data_access.set_writer(self.new_pack._writer, self.new_pack.text_index,
1089
self.new_pack.access_tuple())
1090
output_texts = KnitVersionedFiles(
1091
_KnitGraphIndex(self.new_pack.text_index,
1092
add_callback=self.new_pack.text_index.add_nodes,
1093
deltas=True, parents=True, is_locked=repo.is_locked),
1094
data_access=data_access, max_delta_chain=200)
1095
for key, parent_keys in bad_texts:
1096
# We refer to the new pack to delta data being output.
1097
# A possible improvement would be to catch errors on short reads
1098
# and only flush then.
1099
self.new_pack.flush()
1101
for parent_key in parent_keys:
1102
if parent_key[0] != key[0]:
1103
# Graph parents must match the fileid
1104
raise errors.BzrError('Mismatched key parent %r:%r' %
1106
parents.append(parent_key[1])
1107
text_lines = osutils.split_lines(repo.texts.get_record_stream(
1108
[key], 'unordered', True).next().get_bytes_as('fulltext'))
1109
output_texts.add_lines(key, parent_keys, text_lines,
1110
random_id=True, check_content=False)
1111
# 5) check that nothing inserted has a reference outside the keyspace.
1112
missing_text_keys = self.new_pack._external_compression_parents_of_texts()
1113
if missing_text_keys:
1114
raise errors.BzrError('Reference to missing compression parents %r'
1115
% (missing_text_keys,))
1116
self._log_copied_texts()
1118
def _use_pack(self, new_pack):
1119
"""Override _use_pack to check for reconcile having changed content."""
1120
# XXX: we might be better checking this at the copy time.
1121
original_inventory_keys = set()
1122
inv_index = self._pack_collection.inventory_index.combined_index
1123
for entry in inv_index.iter_all_entries():
1124
original_inventory_keys.add(entry[1])
1125
new_inventory_keys = set()
1126
for entry in new_pack.inventory_index.iter_all_entries():
1127
new_inventory_keys.add(entry[1])
1128
if new_inventory_keys != original_inventory_keys:
1129
self._data_changed = True
1130
return new_pack.data_inserted() and self._data_changed
1133
class RepositoryPackCollection(object):
1134
"""Management of packs within a repository.
1136
:ivar _names: map of {pack_name: (index_size,)}
1139
def __init__(self, repo, transport, index_transport, upload_transport,
1140
pack_transport, index_builder_class, index_class):
1141
"""Create a new RepositoryPackCollection.
1143
:param transport: Addresses the repository base directory
1144
(typically .bzr/repository/).
1145
:param index_transport: Addresses the directory containing indices.
1146
:param upload_transport: Addresses the directory into which packs are written
1147
while they're being created.
1148
:param pack_transport: Addresses the directory of existing complete packs.
1149
:param index_builder_class: The index builder class to use.
1150
:param index_class: The index class to use.
1153
self.transport = transport
1154
self._index_transport = index_transport
1155
self._upload_transport = upload_transport
1156
self._pack_transport = pack_transport
1157
self._index_builder_class = index_builder_class
1158
self._index_class = index_class
1159
self._suffix_offsets = {'.rix': 0, '.iix': 1, '.tix': 2, '.six': 3}
1162
self._packs_by_name = {}
1163
# the previous pack-names content
1164
self._packs_at_load = None
1165
# when a pack is being created by this object, the state of that pack.
1166
self._new_pack = None
1167
# aggregated revision index data
1168
self.revision_index = AggregateIndex(self.reload_pack_names)
1169
self.inventory_index = AggregateIndex(self.reload_pack_names)
1170
self.text_index = AggregateIndex(self.reload_pack_names)
1171
self.signature_index = AggregateIndex(self.reload_pack_names)
1173
def add_pack_to_memory(self, pack):
1174
"""Make a Pack object available to the repository to satisfy queries.
1176
:param pack: A Pack object.
1178
if pack.name in self._packs_by_name:
1179
raise AssertionError()
1180
self.packs.append(pack)
1181
self._packs_by_name[pack.name] = pack
1182
self.revision_index.add_index(pack.revision_index, pack)
1183
self.inventory_index.add_index(pack.inventory_index, pack)
1184
self.text_index.add_index(pack.text_index, pack)
1185
self.signature_index.add_index(pack.signature_index, pack)
1187
def all_packs(self):
1188
"""Return a list of all the Pack objects this repository has.
1190
Note that an in-progress pack being created is not returned.
1192
:return: A list of Pack objects for all the packs in the repository.
1195
for name in self.names():
1196
result.append(self.get_pack_by_name(name))
1200
"""Pack the pack collection incrementally.
1202
This will not attempt global reorganisation or recompression,
1203
rather it will just ensure that the total number of packs does
1204
not grow without bound. It uses the _max_pack_count method to
1205
determine if autopacking is needed, and the pack_distribution
1206
method to determine the number of revisions in each pack.
1208
If autopacking takes place then the packs name collection will have
1209
been flushed to disk - packing requires updating the name collection
1210
in synchronisation with certain steps. Otherwise the names collection
1213
:return: True if packing took place.
1215
# XXX: Should not be needed when the management of indices is sane.
1216
total_revisions = self.revision_index.combined_index.key_count()
1217
total_packs = len(self._names)
1218
if self._max_pack_count(total_revisions) >= total_packs:
1220
# XXX: the following may want to be a class, to pack with a given
1222
mutter('Auto-packing repository %s, which has %d pack files, '
1223
'containing %d revisions into %d packs.', self, total_packs,
1224
total_revisions, self._max_pack_count(total_revisions))
1225
# determine which packs need changing
1226
pack_distribution = self.pack_distribution(total_revisions)
1228
for pack in self.all_packs():
1229
revision_count = pack.get_revision_count()
1230
if revision_count == 0:
1231
# revision less packs are not generated by normal operation,
1232
# only by operations like sign-my-commits, and thus will not
1233
# tend to grow rapdily or without bound like commit containing
1234
# packs do - leave them alone as packing them really should
1235
# group their data with the relevant commit, and that may
1236
# involve rewriting ancient history - which autopack tries to
1237
# avoid. Alternatively we could not group the data but treat
1238
# each of these as having a single revision, and thus add
1239
# one revision for each to the total revision count, to get
1240
# a matching distribution.
1242
existing_packs.append((revision_count, pack))
1243
pack_operations = self.plan_autopack_combinations(
1244
existing_packs, pack_distribution)
1245
self._execute_pack_operations(pack_operations)
1248
def _execute_pack_operations(self, pack_operations, _packer_class=Packer):
1249
"""Execute a series of pack operations.
1251
:param pack_operations: A list of [revision_count, packs_to_combine].
1252
:param _packer_class: The class of packer to use (default: Packer).
1255
for revision_count, packs in pack_operations:
1256
# we may have no-ops from the setup logic
1259
_packer_class(self, packs, '.autopack').pack()
1261
self._remove_pack_from_memory(pack)
1262
# record the newly available packs and stop advertising the old
1264
self._save_pack_names(clear_obsolete_packs=True)
1265
# Move the old packs out of the way now they are no longer referenced.
1266
for revision_count, packs in pack_operations:
1267
self._obsolete_packs(packs)
1269
def lock_names(self):
1270
"""Acquire the mutex around the pack-names index.
1272
This cannot be used in the middle of a read-only transaction on the
1275
self.repo.control_files.lock_write()
1278
"""Pack the pack collection totally."""
1279
self.ensure_loaded()
1280
total_packs = len(self._names)
1282
# This is arguably wrong because we might not be optimal, but for
1283
# now lets leave it in. (e.g. reconcile -> one pack. But not
1286
total_revisions = self.revision_index.combined_index.key_count()
1287
# XXX: the following may want to be a class, to pack with a given
1289
mutter('Packing repository %s, which has %d pack files, '
1290
'containing %d revisions into 1 packs.', self, total_packs,
1292
# determine which packs need changing
1293
pack_distribution = [1]
1294
pack_operations = [[0, []]]
1295
for pack in self.all_packs():
1296
pack_operations[-1][0] += pack.get_revision_count()
1297
pack_operations[-1][1].append(pack)
1298
self._execute_pack_operations(pack_operations, OptimisingPacker)
1300
def plan_autopack_combinations(self, existing_packs, pack_distribution):
1301
"""Plan a pack operation.
1303
:param existing_packs: The packs to pack. (A list of (revcount, Pack)
1305
:param pack_distribution: A list with the number of revisions desired
1308
if len(existing_packs) <= len(pack_distribution):
1310
existing_packs.sort(reverse=True)
1311
pack_operations = [[0, []]]
1312
# plan out what packs to keep, and what to reorganise
1313
while len(existing_packs):
1314
# take the largest pack, and if its less than the head of the
1315
# distribution chart we will include its contents in the new pack
1316
# for that position. If its larger, we remove its size from the
1317
# distribution chart
1318
next_pack_rev_count, next_pack = existing_packs.pop(0)
1319
if next_pack_rev_count >= pack_distribution[0]:
1320
# this is already packed 'better' than this, so we can
1321
# not waste time packing it.
1322
while next_pack_rev_count > 0:
1323
next_pack_rev_count -= pack_distribution[0]
1324
if next_pack_rev_count >= 0:
1326
del pack_distribution[0]
1328
# didn't use that entire bucket up
1329
pack_distribution[0] = -next_pack_rev_count
1331
# add the revisions we're going to add to the next output pack
1332
pack_operations[-1][0] += next_pack_rev_count
1333
# allocate this pack to the next pack sub operation
1334
pack_operations[-1][1].append(next_pack)
1335
if pack_operations[-1][0] >= pack_distribution[0]:
1336
# this pack is used up, shift left.
1337
del pack_distribution[0]
1338
pack_operations.append([0, []])
1339
# Now that we know which pack files we want to move, shove them all
1340
# into a single pack file.
1342
final_pack_list = []
1343
for num_revs, pack_files in pack_operations:
1344
final_rev_count += num_revs
1345
final_pack_list.extend(pack_files)
1346
if len(final_pack_list) == 1:
1347
raise AssertionError('We somehow generated an autopack with a'
1348
' single pack file being moved.')
1350
return [[final_rev_count, final_pack_list]]
1352
def ensure_loaded(self):
1353
# NB: if you see an assertion error here, its probably access against
1354
# an unlocked repo. Naughty.
1355
if not self.repo.is_locked():
1356
raise errors.ObjectNotLocked(self.repo)
1357
if self._names is None:
1359
self._packs_at_load = set()
1360
for index, key, value in self._iter_disk_pack_index():
1362
self._names[name] = self._parse_index_sizes(value)
1363
self._packs_at_load.add((key, value))
1364
# populate all the metadata.
1367
def _parse_index_sizes(self, value):
1368
"""Parse a string of index sizes."""
1369
return tuple([int(digits) for digits in value.split(' ')])
1371
def get_pack_by_name(self, name):
1372
"""Get a Pack object by name.
1374
:param name: The name of the pack - e.g. '123456'
1375
:return: A Pack object.
1378
return self._packs_by_name[name]
1380
rev_index = self._make_index(name, '.rix')
1381
inv_index = self._make_index(name, '.iix')
1382
txt_index = self._make_index(name, '.tix')
1383
sig_index = self._make_index(name, '.six')
1384
result = ExistingPack(self._pack_transport, name, rev_index,
1385
inv_index, txt_index, sig_index)
1386
self.add_pack_to_memory(result)
1389
def allocate(self, a_new_pack):
1390
"""Allocate name in the list of packs.
1392
:param a_new_pack: A NewPack instance to be added to the collection of
1393
packs for this repository.
1395
self.ensure_loaded()
1396
if a_new_pack.name in self._names:
1397
raise errors.BzrError(
1398
'Pack %r already exists in %s' % (a_new_pack.name, self))
1399
self._names[a_new_pack.name] = tuple(a_new_pack.index_sizes)
1400
self.add_pack_to_memory(a_new_pack)
1402
def _iter_disk_pack_index(self):
1403
"""Iterate over the contents of the pack-names index.
1405
This is used when loading the list from disk, and before writing to
1406
detect updates from others during our write operation.
1407
:return: An iterator of the index contents.
1409
return self._index_class(self.transport, 'pack-names', None
1410
).iter_all_entries()
1412
def _make_index(self, name, suffix):
1413
size_offset = self._suffix_offsets[suffix]
1414
index_name = name + suffix
1415
index_size = self._names[name][size_offset]
1416
return self._index_class(
1417
self._index_transport, index_name, index_size)
1419
def _max_pack_count(self, total_revisions):
1420
"""Return the maximum number of packs to use for total revisions.
1422
:param total_revisions: The total number of revisions in the
1425
if not total_revisions:
1427
digits = str(total_revisions)
1429
for digit in digits:
1430
result += int(digit)
1434
"""Provide an order to the underlying names."""
1435
return sorted(self._names.keys())
1437
def _obsolete_packs(self, packs):
1438
"""Move a number of packs which have been obsoleted out of the way.
1440
Each pack and its associated indices are moved out of the way.
1442
Note: for correctness this function should only be called after a new
1443
pack names index has been written without these pack names, and with
1444
the names of packs that contain the data previously available via these
1447
:param packs: The packs to obsolete.
1448
:param return: None.
1451
pack.pack_transport.rename(pack.file_name(),
1452
'../obsolete_packs/' + pack.file_name())
1453
# TODO: Probably needs to know all possible indices for this pack
1454
# - or maybe list the directory and move all indices matching this
1455
# name whether we recognize it or not?
1456
for suffix in ('.iix', '.six', '.tix', '.rix'):
1457
self._index_transport.rename(pack.name + suffix,
1458
'../obsolete_packs/' + pack.name + suffix)
1460
def pack_distribution(self, total_revisions):
1461
"""Generate a list of the number of revisions to put in each pack.
1463
:param total_revisions: The total number of revisions in the
1466
if total_revisions == 0:
1468
digits = reversed(str(total_revisions))
1470
for exponent, count in enumerate(digits):
1471
size = 10 ** exponent
1472
for pos in range(int(count)):
1474
return list(reversed(result))
1476
def _pack_tuple(self, name):
1477
"""Return a tuple with the transport and file name for a pack name."""
1478
return self._pack_transport, name + '.pack'
1480
def _remove_pack_from_memory(self, pack):
1481
"""Remove pack from the packs accessed by this repository.
1483
Only affects memory state, until self._save_pack_names() is invoked.
1485
self._names.pop(pack.name)
1486
self._packs_by_name.pop(pack.name)
1487
self._remove_pack_indices(pack)
1488
self.packs.remove(pack)
1490
def _remove_pack_indices(self, pack):
1491
"""Remove the indices for pack from the aggregated indices."""
1492
self.revision_index.remove_index(pack.revision_index, pack)
1493
self.inventory_index.remove_index(pack.inventory_index, pack)
1494
self.text_index.remove_index(pack.text_index, pack)
1495
self.signature_index.remove_index(pack.signature_index, pack)
1498
"""Clear all cached data."""
1499
# cached revision data
1500
self.repo._revision_knit = None
1501
self.revision_index.clear()
1502
# cached signature data
1503
self.repo._signature_knit = None
1504
self.signature_index.clear()
1505
# cached file text data
1506
self.text_index.clear()
1507
self.repo._text_knit = None
1508
# cached inventory data
1509
self.inventory_index.clear()
1510
# remove the open pack
1511
self._new_pack = None
1512
# information about packs.
1515
self._packs_by_name = {}
1516
self._packs_at_load = None
1518
def _make_index_map(self, index_suffix):
1519
"""Return information on existing indices.
1521
:param suffix: Index suffix added to pack name.
1523
:returns: (pack_map, indices) where indices is a list of GraphIndex
1524
objects, and pack_map is a mapping from those objects to the
1525
pack tuple they describe.
1527
# TODO: stop using this; it creates new indices unnecessarily.
1528
self.ensure_loaded()
1529
suffix_map = {'.rix': 'revision_index',
1530
'.six': 'signature_index',
1531
'.iix': 'inventory_index',
1532
'.tix': 'text_index',
1534
return self._packs_list_to_pack_map_and_index_list(self.all_packs(),
1535
suffix_map[index_suffix])
1537
def _packs_list_to_pack_map_and_index_list(self, packs, index_attribute):
1538
"""Convert a list of packs to an index pack map and index list.
1540
:param packs: The packs list to process.
1541
:param index_attribute: The attribute that the desired index is found
1543
:return: A tuple (map, list) where map contains the dict from
1544
index:pack_tuple, and lsit contains the indices in the same order
1550
index = getattr(pack, index_attribute)
1551
indices.append(index)
1552
pack_map[index] = (pack.pack_transport, pack.file_name())
1553
return pack_map, indices
1555
def _index_contents(self, pack_map, key_filter=None):
1556
"""Get an iterable of the index contents from a pack_map.
1558
:param pack_map: A map from indices to pack details.
1559
:param key_filter: An optional filter to limit the
1562
indices = [index for index in pack_map.iterkeys()]
1563
all_index = CombinedGraphIndex(indices)
1564
if key_filter is None:
1565
return all_index.iter_all_entries()
1567
return all_index.iter_entries(key_filter)
1569
def _unlock_names(self):
1570
"""Release the mutex around the pack-names index."""
1571
self.repo.control_files.unlock()
1573
def _diff_pack_names(self):
1574
"""Read the pack names from disk, and compare it to the one in memory.
1576
:return: (disk_nodes, deleted_nodes, new_nodes)
1577
disk_nodes The final set of nodes that should be referenced
1578
deleted_nodes Nodes which have been removed from when we started
1579
new_nodes Nodes that are newly introduced
1581
# load the disk nodes across
1583
for index, key, value in self._iter_disk_pack_index():
1584
disk_nodes.add((key, value))
1586
# do a two-way diff against our original content
1587
current_nodes = set()
1588
for name, sizes in self._names.iteritems():
1590
((name, ), ' '.join(str(size) for size in sizes)))
1592
# Packs no longer present in the repository, which were present when we
1593
# locked the repository
1594
deleted_nodes = self._packs_at_load - current_nodes
1595
# Packs which this process is adding
1596
new_nodes = current_nodes - self._packs_at_load
1598
# Update the disk_nodes set to include the ones we are adding, and
1599
# remove the ones which were removed by someone else
1600
disk_nodes.difference_update(deleted_nodes)
1601
disk_nodes.update(new_nodes)
1603
return disk_nodes, deleted_nodes, new_nodes
1605
def _syncronize_pack_names_from_disk_nodes(self, disk_nodes):
1606
"""Given the correct set of pack files, update our saved info.
1608
:return: (removed, added, modified)
1609
removed pack names removed from self._names
1610
added pack names added to self._names
1611
modified pack names that had changed value
1616
## self._packs_at_load = disk_nodes
1617
new_names = dict(disk_nodes)
1618
# drop no longer present nodes
1619
for pack in self.all_packs():
1620
if (pack.name,) not in new_names:
1621
removed.append(pack.name)
1622
self._remove_pack_from_memory(pack)
1623
# add new nodes/refresh existing ones
1624
for key, value in disk_nodes:
1626
sizes = self._parse_index_sizes(value)
1627
if name in self._names:
1629
if sizes != self._names[name]:
1630
# the pack for name has had its indices replaced - rare but
1631
# important to handle. XXX: probably can never happen today
1632
# because the three-way merge code above does not handle it
1633
# - you may end up adding the same key twice to the new
1634
# disk index because the set values are the same, unless
1635
# the only index shows up as deleted by the set difference
1636
# - which it may. Until there is a specific test for this,
1637
# assume its broken. RBC 20071017.
1638
self._remove_pack_from_memory(self.get_pack_by_name(name))
1639
self._names[name] = sizes
1640
self.get_pack_by_name(name)
1641
modified.append(name)
1644
self._names[name] = sizes
1645
self.get_pack_by_name(name)
1647
return removed, added, modified
1649
def _save_pack_names(self, clear_obsolete_packs=False):
1650
"""Save the list of packs.
1652
This will take out the mutex around the pack names list for the
1653
duration of the method call. If concurrent updates have been made, a
1654
three-way merge between the current list and the current in memory list
1657
:param clear_obsolete_packs: If True, clear out the contents of the
1658
obsolete_packs directory.
1662
builder = self._index_builder_class()
1663
disk_nodes, deleted_nodes, new_nodes = self._diff_pack_names()
1664
# TODO: handle same-name, index-size-changes here -
1665
# e.g. use the value from disk, not ours, *unless* we're the one
1667
for key, value in disk_nodes:
1668
builder.add_node(key, value)
1669
self.transport.put_file('pack-names', builder.finish(),
1670
mode=self.repo.bzrdir._get_file_mode())
1671
# move the baseline forward
1672
self._packs_at_load = disk_nodes
1673
if clear_obsolete_packs:
1674
self._clear_obsolete_packs()
1676
self._unlock_names()
1677
# synchronise the memory packs list with what we just wrote:
1678
self._syncronize_pack_names_from_disk_nodes(disk_nodes)
1680
def reload_pack_names(self):
1681
"""Sync our pack listing with what is present in the repository.
1683
This should be called when we find out that something we thought was
1684
present is now missing. This happens when another process re-packs the
1687
# This is functionally similar to _save_pack_names, but we don't write
1688
# out the new value.
1689
disk_nodes, _, _ = self._diff_pack_names()
1690
self._packs_at_load = disk_nodes
1692
modified) = self._syncronize_pack_names_from_disk_nodes(disk_nodes)
1693
if removed or added or modified:
1697
def _clear_obsolete_packs(self):
1698
"""Delete everything from the obsolete-packs directory.
1700
obsolete_pack_transport = self.transport.clone('obsolete_packs')
1701
for filename in obsolete_pack_transport.list_dir('.'):
1703
obsolete_pack_transport.delete(filename)
1704
except (errors.PathError, errors.TransportError), e:
1705
warning("couldn't delete obsolete pack, skipping it:\n%s" % (e,))
1707
def _start_write_group(self):
1708
# Do not permit preparation for writing if we're not in a 'write lock'.
1709
if not self.repo.is_write_locked():
1710
raise errors.NotWriteLocked(self)
1711
self._new_pack = NewPack(self._upload_transport, self._index_transport,
1712
self._pack_transport, upload_suffix='.pack',
1713
file_mode=self.repo.bzrdir._get_file_mode(),
1714
index_builder_class=self._index_builder_class,
1715
index_class=self._index_class)
1716
# allow writing: queue writes to a new index
1717
self.revision_index.add_writable_index(self._new_pack.revision_index,
1719
self.inventory_index.add_writable_index(self._new_pack.inventory_index,
1721
self.text_index.add_writable_index(self._new_pack.text_index,
1723
self.signature_index.add_writable_index(self._new_pack.signature_index,
1726
self.repo.inventories._index._add_callback = self.inventory_index.add_callback
1727
self.repo.revisions._index._add_callback = self.revision_index.add_callback
1728
self.repo.signatures._index._add_callback = self.signature_index.add_callback
1729
self.repo.texts._index._add_callback = self.text_index.add_callback
1731
def _abort_write_group(self):
1732
# FIXME: just drop the transient index.
1733
# forget what names there are
1734
if self._new_pack is not None:
1736
self._new_pack.abort()
1738
self._remove_pack_indices(self._new_pack)
1739
self._new_pack = None
1740
self.repo._text_knit = None
1742
def _commit_write_group(self):
1743
self._remove_pack_indices(self._new_pack)
1744
if self._new_pack.data_inserted():
1745
# get all the data to disk and read to use
1746
self._new_pack.finish()
1747
self.allocate(self._new_pack)
1748
self._new_pack = None
1749
if not self.autopack():
1750
# when autopack takes no steps, the names list is still
1752
self._save_pack_names()
1754
self._new_pack.abort()
1755
self._new_pack = None
1756
self.repo._text_knit = None
1759
class KnitPackRepository(KnitRepository):
1760
"""Repository with knit objects stored inside pack containers.
1762
The layering for a KnitPackRepository is:
1764
Graph | HPSS | Repository public layer |
1765
===================================================
1766
Tuple based apis below, string based, and key based apis above
1767
---------------------------------------------------
1769
Provides .texts, .revisions etc
1770
This adapts the N-tuple keys to physical knit records which only have a
1771
single string identifier (for historical reasons), which in older formats
1772
was always the revision_id, and in the mapped code for packs is always
1773
the last element of key tuples.
1774
---------------------------------------------------
1776
A separate GraphIndex is used for each of the
1777
texts/inventories/revisions/signatures contained within each individual
1778
pack file. The GraphIndex layer works in N-tuples and is unaware of any
1780
===================================================
1784
def __init__(self, _format, a_bzrdir, control_files, _commit_builder_class,
1786
KnitRepository.__init__(self, _format, a_bzrdir, control_files,
1787
_commit_builder_class, _serializer)
1788
index_transport = self._transport.clone('indices')
1789
self._pack_collection = RepositoryPackCollection(self, self._transport,
1791
self._transport.clone('upload'),
1792
self._transport.clone('packs'),
1793
_format.index_builder_class,
1794
_format.index_class)
1795
self.inventories = KnitVersionedFiles(
1796
_KnitGraphIndex(self._pack_collection.inventory_index.combined_index,
1797
add_callback=self._pack_collection.inventory_index.add_callback,
1798
deltas=True, parents=True, is_locked=self.is_locked),
1799
data_access=self._pack_collection.inventory_index.data_access,
1800
max_delta_chain=200)
1801
self.revisions = KnitVersionedFiles(
1802
_KnitGraphIndex(self._pack_collection.revision_index.combined_index,
1803
add_callback=self._pack_collection.revision_index.add_callback,
1804
deltas=False, parents=True, is_locked=self.is_locked),
1805
data_access=self._pack_collection.revision_index.data_access,
1807
self.signatures = KnitVersionedFiles(
1808
_KnitGraphIndex(self._pack_collection.signature_index.combined_index,
1809
add_callback=self._pack_collection.signature_index.add_callback,
1810
deltas=False, parents=False, is_locked=self.is_locked),
1811
data_access=self._pack_collection.signature_index.data_access,
1813
self.texts = KnitVersionedFiles(
1814
_KnitGraphIndex(self._pack_collection.text_index.combined_index,
1815
add_callback=self._pack_collection.text_index.add_callback,
1816
deltas=True, parents=True, is_locked=self.is_locked),
1817
data_access=self._pack_collection.text_index.data_access,
1818
max_delta_chain=200)
1819
# True when the repository object is 'write locked' (as opposed to the
1820
# physical lock only taken out around changes to the pack-names list.)
1821
# Another way to represent this would be a decorator around the control
1822
# files object that presents logical locks as physical ones - if this
1823
# gets ugly consider that alternative design. RBC 20071011
1824
self._write_lock_count = 0
1825
self._transaction = None
1827
self._reconcile_does_inventory_gc = True
1828
self._reconcile_fixes_text_parents = True
1829
self._reconcile_backsup_inventory = False
1830
self._fetch_order = 'unordered'
1832
def _warn_if_deprecated(self):
1833
# This class isn't deprecated, but one sub-format is
1834
if isinstance(self._format, RepositoryFormatKnitPack5RichRootBroken):
1835
from bzrlib import repository
1836
if repository._deprecation_warning_done:
1838
repository._deprecation_warning_done = True
1839
warning("Format %s for %s is deprecated - please use"
1840
" 'bzr upgrade --1.6.1-rich-root'"
1841
% (self._format, self.bzrdir.transport.base))
1843
def _abort_write_group(self):
1844
self._pack_collection._abort_write_group()
1846
def _find_inconsistent_revision_parents(self):
1847
"""Find revisions with incorrectly cached parents.
1849
:returns: an iterator yielding tuples of (revison-id, parents-in-index,
1850
parents-in-revision).
1852
if not self.is_locked():
1853
raise errors.ObjectNotLocked(self)
1854
pb = ui.ui_factory.nested_progress_bar()
1857
revision_nodes = self._pack_collection.revision_index \
1858
.combined_index.iter_all_entries()
1859
index_positions = []
1860
# Get the cached index values for all revisions, and also the location
1861
# in each index of the revision text so we can perform linear IO.
1862
for index, key, value, refs in revision_nodes:
1863
pos, length = value[1:].split(' ')
1864
index_positions.append((index, int(pos), key[0],
1865
tuple(parent[0] for parent in refs[0])))
1866
pb.update("Reading revision index.", 0, 0)
1867
index_positions.sort()
1868
batch_count = len(index_positions) / 1000 + 1
1869
pb.update("Checking cached revision graph.", 0, batch_count)
1870
for offset in xrange(batch_count):
1871
pb.update("Checking cached revision graph.", offset)
1872
to_query = index_positions[offset * 1000:(offset + 1) * 1000]
1875
rev_ids = [item[2] for item in to_query]
1876
revs = self.get_revisions(rev_ids)
1877
for revision, item in zip(revs, to_query):
1878
index_parents = item[3]
1879
rev_parents = tuple(revision.parent_ids)
1880
if index_parents != rev_parents:
1881
result.append((revision.revision_id, index_parents, rev_parents))
1886
@symbol_versioning.deprecated_method(symbol_versioning.one_one)
1887
def get_parents(self, revision_ids):
1888
"""See graph._StackedParentsProvider.get_parents."""
1889
parent_map = self.get_parent_map(revision_ids)
1890
return [parent_map.get(r, None) for r in revision_ids]
1892
def _make_parents_provider(self):
1893
return graph.CachingParentsProvider(self)
1895
def _refresh_data(self):
1896
if self._write_lock_count == 1 or (
1897
self.control_files._lock_count == 1 and
1898
self.control_files._lock_mode == 'r'):
1899
# forget what names there are
1900
self._pack_collection.reset()
1901
# XXX: Better to do an in-memory merge when acquiring a new lock -
1902
# factor out code from _save_pack_names.
1903
self._pack_collection.ensure_loaded()
1905
def _start_write_group(self):
1906
self._pack_collection._start_write_group()
1908
def _commit_write_group(self):
1909
return self._pack_collection._commit_write_group()
1911
def get_transaction(self):
1912
if self._write_lock_count:
1913
return self._transaction
1915
return self.control_files.get_transaction()
1917
def is_locked(self):
1918
return self._write_lock_count or self.control_files.is_locked()
1920
def is_write_locked(self):
1921
return self._write_lock_count
1923
def lock_write(self, token=None):
1924
if not self._write_lock_count and self.is_locked():
1925
raise errors.ReadOnlyError(self)
1926
self._write_lock_count += 1
1927
if self._write_lock_count == 1:
1928
self._transaction = transactions.WriteTransaction()
1929
for repo in self._fallback_repositories:
1930
# Writes don't affect fallback repos
1932
self._refresh_data()
1934
def lock_read(self):
1935
if self._write_lock_count:
1936
self._write_lock_count += 1
1938
self.control_files.lock_read()
1939
for repo in self._fallback_repositories:
1940
# Writes don't affect fallback repos
1942
self._refresh_data()
1944
def leave_lock_in_place(self):
1945
# not supported - raise an error
1946
raise NotImplementedError(self.leave_lock_in_place)
1948
def dont_leave_lock_in_place(self):
1949
# not supported - raise an error
1950
raise NotImplementedError(self.dont_leave_lock_in_place)
1954
"""Compress the data within the repository.
1956
This will pack all the data to a single pack. In future it may
1957
recompress deltas or do other such expensive operations.
1959
self._pack_collection.pack()
1962
def reconcile(self, other=None, thorough=False):
1963
"""Reconcile this repository."""
1964
from bzrlib.reconcile import PackReconciler
1965
reconciler = PackReconciler(self, thorough=thorough)
1966
reconciler.reconcile()
1970
if self._write_lock_count == 1 and self._write_group is not None:
1971
self.abort_write_group()
1972
self._transaction = None
1973
self._write_lock_count = 0
1974
raise errors.BzrError(
1975
'Must end write group before releasing write lock on %s'
1977
if self._write_lock_count:
1978
self._write_lock_count -= 1
1979
if not self._write_lock_count:
1980
transaction = self._transaction
1981
self._transaction = None
1982
transaction.finish()
1983
for repo in self._fallback_repositories:
1986
self.control_files.unlock()
1987
for repo in self._fallback_repositories:
1991
class RepositoryFormatPack(MetaDirRepositoryFormat):
1992
"""Format logic for pack structured repositories.
1994
This repository format has:
1995
- a list of packs in pack-names
1996
- packs in packs/NAME.pack
1997
- indices in indices/NAME.{iix,six,tix,rix}
1998
- knit deltas in the packs, knit indices mapped to the indices.
1999
- thunk objects to support the knits programming API.
2000
- a format marker of its own
2001
- an optional 'shared-storage' flag
2002
- an optional 'no-working-trees' flag
2006
# Set this attribute in derived classes to control the repository class
2007
# created by open and initialize.
2008
repository_class = None
2009
# Set this attribute in derived classes to control the
2010
# _commit_builder_class that the repository objects will have passed to
2011
# their constructor.
2012
_commit_builder_class = None
2013
# Set this attribute in derived clases to control the _serializer that the
2014
# repository objects will have passed to their constructor.
2016
# External references are not supported in pack repositories yet.
2017
supports_external_lookups = False
2018
# What index classes to use
2019
index_builder_class = None
2022
def initialize(self, a_bzrdir, shared=False):
2023
"""Create a pack based repository.
2025
:param a_bzrdir: bzrdir to contain the new repository; must already
2027
:param shared: If true the repository will be initialized as a shared
2030
mutter('creating repository in %s.', a_bzrdir.transport.base)
2031
dirs = ['indices', 'obsolete_packs', 'packs', 'upload']
2032
builder = self.index_builder_class()
2033
files = [('pack-names', builder.finish())]
2034
utf8_files = [('format', self.get_format_string())]
2036
self._upload_blank_content(a_bzrdir, dirs, files, utf8_files, shared)
2037
return self.open(a_bzrdir=a_bzrdir, _found=True)
2039
def open(self, a_bzrdir, _found=False, _override_transport=None):
2040
"""See RepositoryFormat.open().
2042
:param _override_transport: INTERNAL USE ONLY. Allows opening the
2043
repository at a slightly different url
2044
than normal. I.e. during 'upgrade'.
2047
format = RepositoryFormat.find_format(a_bzrdir)
2048
if _override_transport is not None:
2049
repo_transport = _override_transport
2051
repo_transport = a_bzrdir.get_repository_transport(None)
2052
control_files = lockable_files.LockableFiles(repo_transport,
2053
'lock', lockdir.LockDir)
2054
return self.repository_class(_format=self,
2056
control_files=control_files,
2057
_commit_builder_class=self._commit_builder_class,
2058
_serializer=self._serializer)
2061
class RepositoryFormatKnitPack1(RepositoryFormatPack):
2062
"""A no-subtrees parameterized Pack repository.
2064
This format was introduced in 0.92.
2067
repository_class = KnitPackRepository
2068
_commit_builder_class = PackCommitBuilder
2070
def _serializer(self):
2071
return xml5.serializer_v5
2072
# What index classes to use
2073
index_builder_class = InMemoryGraphIndex
2074
index_class = GraphIndex
2076
def _get_matching_bzrdir(self):
2077
return bzrdir.format_registry.make_bzrdir('pack-0.92')
2079
def _ignore_setting_bzrdir(self, format):
2082
_matchingbzrdir = property(_get_matching_bzrdir, _ignore_setting_bzrdir)
2084
def get_format_string(self):
2085
"""See RepositoryFormat.get_format_string()."""
2086
return "Bazaar pack repository format 1 (needs bzr 0.92)\n"
2088
def get_format_description(self):
2089
"""See RepositoryFormat.get_format_description()."""
2090
return "Packs containing knits without subtree support"
2092
def check_conversion_target(self, target_format):
2096
class RepositoryFormatKnitPack3(RepositoryFormatPack):
2097
"""A subtrees parameterized Pack repository.
2099
This repository format uses the xml7 serializer to get:
2100
- support for recording full info about the tree root
2101
- support for recording tree-references
2103
This format was introduced in 0.92.
2106
repository_class = KnitPackRepository
2107
_commit_builder_class = PackRootCommitBuilder
2108
rich_root_data = True
2109
supports_tree_reference = True
2111
def _serializer(self):
2112
return xml7.serializer_v7
2113
# What index classes to use
2114
index_builder_class = InMemoryGraphIndex
2115
index_class = GraphIndex
2117
def _get_matching_bzrdir(self):
2118
return bzrdir.format_registry.make_bzrdir(
2119
'pack-0.92-subtree')
2121
def _ignore_setting_bzrdir(self, format):
2124
_matchingbzrdir = property(_get_matching_bzrdir, _ignore_setting_bzrdir)
2126
def check_conversion_target(self, target_format):
2127
if not target_format.rich_root_data:
2128
raise errors.BadConversionTarget(
2129
'Does not support rich root data.', target_format)
2130
if not getattr(target_format, 'supports_tree_reference', False):
2131
raise errors.BadConversionTarget(
2132
'Does not support nested trees', target_format)
2134
def get_format_string(self):
2135
"""See RepositoryFormat.get_format_string()."""
2136
return "Bazaar pack repository format 1 with subtree support (needs bzr 0.92)\n"
2138
def get_format_description(self):
2139
"""See RepositoryFormat.get_format_description()."""
2140
return "Packs containing knits with subtree support\n"
2143
class RepositoryFormatKnitPack4(RepositoryFormatPack):
2144
"""A rich-root, no subtrees parameterized Pack repository.
2146
This repository format uses the xml6 serializer to get:
2147
- support for recording full info about the tree root
2149
This format was introduced in 1.0.
2152
repository_class = KnitPackRepository
2153
_commit_builder_class = PackRootCommitBuilder
2154
rich_root_data = True
2155
supports_tree_reference = False
2157
def _serializer(self):
2158
return xml6.serializer_v6
2159
# What index classes to use
2160
index_builder_class = InMemoryGraphIndex
2161
index_class = GraphIndex
2163
def _get_matching_bzrdir(self):
2164
return bzrdir.format_registry.make_bzrdir(
2167
def _ignore_setting_bzrdir(self, format):
2170
_matchingbzrdir = property(_get_matching_bzrdir, _ignore_setting_bzrdir)
2172
def check_conversion_target(self, target_format):
2173
if not target_format.rich_root_data:
2174
raise errors.BadConversionTarget(
2175
'Does not support rich root data.', target_format)
2177
def get_format_string(self):
2178
"""See RepositoryFormat.get_format_string()."""
2179
return ("Bazaar pack repository format 1 with rich root"
2180
" (needs bzr 1.0)\n")
2182
def get_format_description(self):
2183
"""See RepositoryFormat.get_format_description()."""
2184
return "Packs containing knits with rich root support\n"
2187
class RepositoryFormatKnitPack5(RepositoryFormatPack):
2188
"""Repository that supports external references to allow stacking.
2192
Supports external lookups, which results in non-truncated ghosts after
2193
reconcile compared to pack-0.92 formats.
2196
repository_class = KnitPackRepository
2197
_commit_builder_class = PackCommitBuilder
2198
supports_external_lookups = True
2199
# What index classes to use
2200
index_builder_class = InMemoryGraphIndex
2201
index_class = GraphIndex
2204
def _serializer(self):
2205
return xml5.serializer_v5
2207
def _get_matching_bzrdir(self):
2208
return bzrdir.format_registry.make_bzrdir('1.6')
2210
def _ignore_setting_bzrdir(self, format):
2213
_matchingbzrdir = property(_get_matching_bzrdir, _ignore_setting_bzrdir)
2215
def get_format_string(self):
2216
"""See RepositoryFormat.get_format_string()."""
2217
return "Bazaar RepositoryFormatKnitPack5 (bzr 1.6)\n"
2219
def get_format_description(self):
2220
"""See RepositoryFormat.get_format_description()."""
2221
return "Packs 5 (adds stacking support, requires bzr 1.6)"
2223
def check_conversion_target(self, target_format):
2227
class RepositoryFormatKnitPack5RichRoot(RepositoryFormatPack):
2228
"""A repository with rich roots and stacking.
2230
New in release 1.6.1.
2232
Supports stacking on other repositories, allowing data to be accessed
2233
without being stored locally.
2236
repository_class = KnitPackRepository
2237
_commit_builder_class = PackRootCommitBuilder
2238
rich_root_data = True
2239
supports_tree_reference = False # no subtrees
2240
supports_external_lookups = True
2241
# What index classes to use
2242
index_builder_class = InMemoryGraphIndex
2243
index_class = GraphIndex
2246
def _serializer(self):
2247
return xml6.serializer_v6
2249
def _get_matching_bzrdir(self):
2250
return bzrdir.format_registry.make_bzrdir(
2253
def _ignore_setting_bzrdir(self, format):
2256
_matchingbzrdir = property(_get_matching_bzrdir, _ignore_setting_bzrdir)
2258
def check_conversion_target(self, target_format):
2259
if not target_format.rich_root_data:
2260
raise errors.BadConversionTarget(
2261
'Does not support rich root data.', target_format)
2263
def get_format_string(self):
2264
"""See RepositoryFormat.get_format_string()."""
2265
return "Bazaar RepositoryFormatKnitPack5RichRoot (bzr 1.6.1)\n"
2267
def get_format_description(self):
2268
return "Packs 5 rich-root (adds stacking support, requires bzr 1.6.1)"
2271
class RepositoryFormatKnitPack5RichRootBroken(RepositoryFormatPack):
2272
"""A repository with rich roots and external references.
2276
Supports external lookups, which results in non-truncated ghosts after
2277
reconcile compared to pack-0.92 formats.
2279
This format was deprecated because the serializer it uses accidentally
2280
supported subtrees, when the format was not intended to. This meant that
2281
someone could accidentally fetch from an incorrect repository.
2284
repository_class = KnitPackRepository
2285
_commit_builder_class = PackRootCommitBuilder
2286
rich_root_data = True
2287
supports_tree_reference = False # no subtrees
2289
supports_external_lookups = True
2290
# What index classes to use
2291
index_builder_class = InMemoryGraphIndex
2292
index_class = GraphIndex
2295
def _serializer(self):
2296
return xml7.serializer_v7
2298
def _get_matching_bzrdir(self):
2299
return bzrdir.format_registry.make_bzrdir(
2302
def _ignore_setting_bzrdir(self, format):
2305
_matchingbzrdir = property(_get_matching_bzrdir, _ignore_setting_bzrdir)
2307
def check_conversion_target(self, target_format):
2308
if not target_format.rich_root_data:
2309
raise errors.BadConversionTarget(
2310
'Does not support rich root data.', target_format)
2312
def get_format_string(self):
2313
"""See RepositoryFormat.get_format_string()."""
2314
return "Bazaar RepositoryFormatKnitPack5RichRoot (bzr 1.6)\n"
2316
def get_format_description(self):
2317
return ("Packs 5 rich-root (adds stacking support, requires bzr 1.6)"
2321
class RepositoryFormatKnitPack6(RepositoryFormatPack):
2322
"""A repository with stacking and btree indexes,
2323
without rich roots or subtrees.
2325
This is equivalent to pack-1.6 with B+Tree indices.
2328
repository_class = KnitPackRepository
2329
_commit_builder_class = PackCommitBuilder
2330
supports_external_lookups = True
2331
# What index classes to use
2332
index_builder_class = BTreeBuilder
2333
index_class = BTreeGraphIndex
2336
def _serializer(self):
2337
return xml5.serializer_v5
2339
def _get_matching_bzrdir(self):
2340
return bzrdir.format_registry.make_bzrdir('1.9')
2342
def _ignore_setting_bzrdir(self, format):
2345
_matchingbzrdir = property(_get_matching_bzrdir, _ignore_setting_bzrdir)
2347
def get_format_string(self):
2348
"""See RepositoryFormat.get_format_string()."""
2349
return "Bazaar RepositoryFormatKnitPack6 (bzr 1.9)\n"
2351
def get_format_description(self):
2352
"""See RepositoryFormat.get_format_description()."""
2353
return "Packs 6 (uses btree indexes, requires bzr 1.9)"
2355
def check_conversion_target(self, target_format):
2359
class RepositoryFormatKnitPack6RichRoot(RepositoryFormatPack):
2360
"""A repository with rich roots, no subtrees, stacking and btree indexes.
2362
1.6-rich-root with B+Tree indices.
2365
repository_class = KnitPackRepository
2366
_commit_builder_class = PackRootCommitBuilder
2367
rich_root_data = True
2368
supports_tree_reference = False # no subtrees
2369
supports_external_lookups = True
2370
# What index classes to use
2371
index_builder_class = BTreeBuilder
2372
index_class = BTreeGraphIndex
2375
def _serializer(self):
2376
return xml6.serializer_v6
2378
def _get_matching_bzrdir(self):
2379
return bzrdir.format_registry.make_bzrdir(
2382
def _ignore_setting_bzrdir(self, format):
2385
_matchingbzrdir = property(_get_matching_bzrdir, _ignore_setting_bzrdir)
2387
def check_conversion_target(self, target_format):
2388
if not target_format.rich_root_data:
2389
raise errors.BadConversionTarget(
2390
'Does not support rich root data.', target_format)
2392
def get_format_string(self):
2393
"""See RepositoryFormat.get_format_string()."""
2394
return "Bazaar RepositoryFormatKnitPack6RichRoot (bzr 1.9)\n"
2396
def get_format_description(self):
2397
return "Packs 6 rich-root (uses btree indexes, requires bzr 1.9)"
2400
class RepositoryFormatPackDevelopment2(RepositoryFormatPack):
2401
"""A no-subtrees development repository.
2403
This format should be retained until the second release after bzr 1.7.
2405
This is pack-1.6.1 with B+Tree indices.
2408
repository_class = KnitPackRepository
2409
_commit_builder_class = PackCommitBuilder
2410
supports_external_lookups = True
2411
# What index classes to use
2412
index_builder_class = BTreeBuilder
2413
index_class = BTreeGraphIndex
2416
def _serializer(self):
2417
return xml5.serializer_v5
2419
def _get_matching_bzrdir(self):
2420
return bzrdir.format_registry.make_bzrdir('development2')
2422
def _ignore_setting_bzrdir(self, format):
2425
_matchingbzrdir = property(_get_matching_bzrdir, _ignore_setting_bzrdir)
2427
def get_format_string(self):
2428
"""See RepositoryFormat.get_format_string()."""
2429
return "Bazaar development format 2 (needs bzr.dev from before 1.8)\n"
2431
def get_format_description(self):
2432
"""See RepositoryFormat.get_format_description()."""
2433
return ("Development repository format, currently the same as "
2434
"1.6.1 with B+Trees.\n")
2436
def check_conversion_target(self, target_format):
2440
class RepositoryFormatPackDevelopment2Subtree(RepositoryFormatPack):
2441
"""A subtrees development repository.
2443
This format should be retained until the second release after bzr 1.7.
2445
1.6.1-subtree[as it might have been] with B+Tree indices.
2448
repository_class = KnitPackRepository
2449
_commit_builder_class = PackRootCommitBuilder
2450
rich_root_data = True
2451
supports_tree_reference = True
2452
supports_external_lookups = True
2453
# What index classes to use
2454
index_builder_class = BTreeBuilder
2455
index_class = BTreeGraphIndex
2458
def _serializer(self):
2459
return xml7.serializer_v7
2461
def _get_matching_bzrdir(self):
2462
return bzrdir.format_registry.make_bzrdir(
2463
'development2-subtree')
2465
def _ignore_setting_bzrdir(self, format):
2468
_matchingbzrdir = property(_get_matching_bzrdir, _ignore_setting_bzrdir)
2470
def check_conversion_target(self, target_format):
2471
if not target_format.rich_root_data:
2472
raise errors.BadConversionTarget(
2473
'Does not support rich root data.', target_format)
2474
if not getattr(target_format, 'supports_tree_reference', False):
2475
raise errors.BadConversionTarget(
2476
'Does not support nested trees', target_format)
2478
def get_format_string(self):
2479
"""See RepositoryFormat.get_format_string()."""
2480
return ("Bazaar development format 2 with subtree support "
2481
"(needs bzr.dev from before 1.8)\n")
2483
def get_format_description(self):
2484
"""See RepositoryFormat.get_format_description()."""
2485
return ("Development repository format, currently the same as "
2486
"1.6.1-subtree with B+Tree indices.\n")
added
removed
bzrlib/repofmt/pack_repo.py
