~bzr-pqm/bzr/bzr.dev

« back to all changes in this revision

Viewing changes to bzrlib/repofmt/groupcompress_repo.py

  • Committer: John Arbash Meinel
  • Date: 2009-03-27 22:29:55 UTC
  • mto: (3735.39.2 clean)
  • mto: This revision was merged to the branch mainline in revision 4280.
  • Revision ID: john@arbash-meinel.com-20090327222955-utifmfm888zerixt
Implement apply_delta_to_source which doesn't have to malloc another string.

Show diffs side-by-side

added added

removed removed

Lines of Context:
1
 
# Copyright (C) 2008, 2009, 2010 Canonical Ltd
 
1
# Copyright (C) 2008, 2009 Canonical Ltd
2
2
#
3
3
# This program is free software; you can redistribute it and/or modify
4
4
# it under the terms of the GNU General Public License as published by
14
14
# along with this program; if not, write to the Free Software
15
15
# Foundation, Inc., 51 Franklin Street, Fifth Floor, Boston, MA 02110-1301 USA
16
16
 
17
 
"""Repository formats using CHK inventories and groupcompress compression."""
 
17
"""Repostory formats using B+Tree indices and groupcompress compression."""
18
18
 
19
19
import time
20
20
 
29
29
    knit,
30
30
    osutils,
31
31
    pack,
 
32
    repository,
32
33
    revision as _mod_revision,
33
34
    trace,
34
35
    ui,
35
36
    )
36
 
from bzrlib.btree_index import (
37
 
    BTreeGraphIndex,
38
 
    BTreeBuilder,
39
 
    )
 
37
from bzrlib.index import GraphIndex, GraphIndexBuilder
40
38
from bzrlib.groupcompress import (
41
39
    _GCGraphIndex,
42
40
    GroupCompressVersionedFiles,
45
43
    Pack,
46
44
    NewPack,
47
45
    KnitPackRepository,
48
 
    KnitPackStreamSource,
49
46
    PackRootCommitBuilder,
50
47
    RepositoryPackCollection,
51
 
    RepositoryFormatPack,
52
 
    ResumedPack,
 
48
    RepositoryFormatKnitPack6,
53
49
    Packer,
 
50
    CHKInventoryRepository,
 
51
    RepositoryFormatPackDevelopment5Hash16,
 
52
    RepositoryFormatPackDevelopment5Hash255,
54
53
    )
55
 
from bzrlib.static_tuple import StaticTuple
 
54
 
56
55
 
57
56
 
58
57
class GCPack(NewPack):
87
86
            # have a regular 2-list index giving parents and compression
88
87
            # source.
89
88
            index_builder_class(reference_lists=1),
90
 
            # Texts: per file graph, for all fileids - so one reference list
91
 
            # and two elements in the key tuple.
 
89
            # Texts: compression and per file graph, for all fileids - so two
 
90
            # reference lists and two elements in the key tuple.
92
91
            index_builder_class(reference_lists=1, key_elements=2),
93
92
            # Signatures: Just blobs to store, no compression, no parents
94
93
            # listing.
154
153
        self._writer.begin()
155
154
        # what state is the pack in? (open, finished, aborted)
156
155
        self._state = 'open'
157
 
        # no name until we finish writing the content
158
 
        self.name = None
159
156
 
160
157
    def _check_references(self):
161
158
        """Make sure our external references are present.
165
162
        have deltas based on a fallback repository.
166
163
        (See <https://bugs.launchpad.net/bzr/+bug/288751>)
167
164
        """
168
 
        # Groupcompress packs don't have any external references, arguably CHK
169
 
        # pages have external references, but we cannot 'cheaply' determine
170
 
        # them without actually walking all of the chk pages.
171
 
 
172
 
 
173
 
class ResumedGCPack(ResumedPack):
174
 
 
175
 
    def _check_references(self):
176
 
        """Make sure our external compression parents are present."""
177
 
        # See GCPack._check_references for why this is empty
178
 
 
179
 
    def _get_external_refs(self, index):
180
 
        # GC repositories don't have compression parents external to a given
181
 
        # pack file
182
 
        return set()
 
165
        # Groupcompress packs don't have any external references
183
166
 
184
167
 
185
168
class GCCHKPacker(Packer):
192
175
                                          reload_func=reload_func)
193
176
        self._pack_collection = pack_collection
194
177
        # ATM, We only support this for GCCHK repositories
195
 
        if pack_collection.chk_index is None:
196
 
            raise AssertionError('pack_collection.chk_index should not be None')
 
178
        assert pack_collection.chk_index is not None
197
179
        self._gather_text_refs = False
198
180
        self._chk_id_roots = []
199
181
        self._chk_p_id_roots = []
218
200
            p_id_roots_set = set()
219
201
            stream = source_vf.get_record_stream(keys, 'groupcompress', True)
220
202
            for idx, record in enumerate(stream):
221
 
                # Inventories should always be with revisions; assume success.
222
203
                bytes = record.get_bytes_as('fulltext')
223
204
                chk_inv = inventory.CHKInventory.deserialise(None, bytes,
224
205
                                                             record.key)
229
210
                    self._chk_id_roots.append(key)
230
211
                    id_roots_set.add(key)
231
212
                p_id_map = chk_inv.parent_id_basename_to_file_id
232
 
                if p_id_map is None:
233
 
                    raise AssertionError('Parent id -> file_id map not set')
 
213
                assert p_id_map is not None
234
214
                key = p_id_map.key()
235
215
                if key not in p_id_roots_set:
236
216
                    p_id_roots_set.add(key)
262
242
        remaining_keys = set(keys)
263
243
        counter = [0]
264
244
        if self._gather_text_refs:
 
245
            bytes_to_info = inventory.CHKInventory._bytes_to_utf8name_key
265
246
            self._text_refs = set()
266
247
        def _get_referenced_stream(root_keys, parse_leaf_nodes=False):
267
248
            cur_keys = root_keys
288
269
                    # Store is None, because we know we have a LeafNode, and we
289
270
                    # just want its entries
290
271
                    for file_id, bytes in node.iteritems(None):
291
 
                        self._text_refs.add(chk_map._bytes_to_text_key(bytes))
 
272
                        name_utf8, file_id, revision_id = bytes_to_info(bytes)
 
273
                        self._text_refs.add((file_id, revision))
292
274
                def next_stream():
293
275
                    stream = source_vf.get_record_stream(cur_keys,
294
276
                                                         'as-requested', True)
295
277
                    for record in stream:
296
 
                        if record.storage_kind == 'absent':
297
 
                            # An absent CHK record: we assume that the missing
298
 
                            # record is in a different pack - e.g. a page not
299
 
                            # altered by the commit we're packing.
300
 
                            continue
301
278
                        bytes = record.get_bytes_as('fulltext')
302
279
                        # We don't care about search_key_func for this code,
303
280
                        # because we only care about external references.
350
327
        """Build a VersionedFiles instance on top of this group of packs."""
351
328
        index_name = index_name + '_index'
352
329
        index_to_pack = {}
353
 
        access = knit._DirectPackAccess(index_to_pack,
354
 
                                        reload_func=self._reload_func)
 
330
        access = knit._DirectPackAccess(index_to_pack)
355
331
        if for_write:
356
332
            # Use new_pack
357
 
            if self.new_pack is None:
358
 
                raise AssertionError('No new pack has been set')
 
333
            assert self.new_pack is not None
359
334
            index = getattr(self.new_pack, index_name)
360
335
            index_to_pack[index] = self.new_pack.access_tuple()
361
336
            index.set_optimize(for_size=True)
411
386
 
412
387
    def _copy_inventory_texts(self):
413
388
        source_vf, target_vf = self._build_vfs('inventory', True, True)
414
 
        # It is not sufficient to just use self.revision_keys, as stacked
415
 
        # repositories can have more inventories than they have revisions.
416
 
        # One alternative would be to do something with
417
 
        # get_parent_map(self.revision_keys), but that shouldn't be any faster
418
 
        # than this.
419
 
        inventory_keys = source_vf.keys()
420
 
        missing_inventories = set(self.revision_keys).difference(inventory_keys)
421
 
        if missing_inventories:
422
 
            missing_inventories = sorted(missing_inventories)
423
 
            raise ValueError('We are missing inventories for revisions: %s'
424
 
                % (missing_inventories,))
425
 
        self._copy_stream(source_vf, target_vf, inventory_keys,
 
389
        self._copy_stream(source_vf, target_vf, self.revision_keys,
426
390
                          'inventories', self._get_filtered_inv_stream, 2)
427
391
 
428
392
    def _copy_chk_texts(self):
455
419
        #      is grabbing too many keys...
456
420
        text_keys = source_vf.keys()
457
421
        self._copy_stream(source_vf, target_vf, text_keys,
458
 
                          'texts', self._get_progress_stream, 4)
 
422
                          'text', self._get_progress_stream, 4)
459
423
 
460
424
    def _copy_signature_texts(self):
461
425
        source_vf, target_vf = self._build_vfs('signature', False, False)
478
442
        if not self._use_pack(self.new_pack):
479
443
            self.new_pack.abort()
480
444
            return None
481
 
        self.new_pack.finish_content()
482
 
        if len(self.packs) == 1:
483
 
            old_pack = self.packs[0]
484
 
            if old_pack.name == self.new_pack._hash.hexdigest():
485
 
                # The single old pack was already optimally packed.
486
 
                trace.mutter('single pack %s was already optimally packed',
487
 
                    old_pack.name)
488
 
                self.new_pack.abort()
489
 
                return None
490
445
        self.pb.update('finishing repack', 6, 7)
491
446
        self.new_pack.finish()
492
447
        self._pack_collection.allocate(self.new_pack)
581
536
class GCRepositoryPackCollection(RepositoryPackCollection):
582
537
 
583
538
    pack_factory = GCPack
584
 
    resumed_pack_factory = ResumedGCPack
585
 
 
586
 
    def _check_new_inventories(self):
587
 
        """Detect missing inventories or chk root entries for the new revisions
588
 
        in this write group.
589
 
 
590
 
        :returns: list of strs, summarising any problems found.  If the list is
591
 
            empty no problems were found.
592
 
        """
593
 
        # Ensure that all revisions added in this write group have:
594
 
        #   - corresponding inventories,
595
 
        #   - chk root entries for those inventories,
596
 
        #   - and any present parent inventories have their chk root
597
 
        #     entries too.
598
 
        # And all this should be independent of any fallback repository.
599
 
        problems = []
600
 
        key_deps = self.repo.revisions._index._key_dependencies
601
 
        new_revisions_keys = key_deps.get_new_keys()
602
 
        no_fallback_inv_index = self.repo.inventories._index
603
 
        no_fallback_chk_bytes_index = self.repo.chk_bytes._index
604
 
        no_fallback_texts_index = self.repo.texts._index
605
 
        inv_parent_map = no_fallback_inv_index.get_parent_map(
606
 
            new_revisions_keys)
607
 
        # Are any inventories for corresponding to the new revisions missing?
608
 
        corresponding_invs = set(inv_parent_map)
609
 
        missing_corresponding = set(new_revisions_keys)
610
 
        missing_corresponding.difference_update(corresponding_invs)
611
 
        if missing_corresponding:
612
 
            problems.append("inventories missing for revisions %s" %
613
 
                (sorted(missing_corresponding),))
614
 
            return problems
615
 
        # Are any chk root entries missing for any inventories?  This includes
616
 
        # any present parent inventories, which may be used when calculating
617
 
        # deltas for streaming.
618
 
        all_inv_keys = set(corresponding_invs)
619
 
        for parent_inv_keys in inv_parent_map.itervalues():
620
 
            all_inv_keys.update(parent_inv_keys)
621
 
        # Filter out ghost parents.
622
 
        all_inv_keys.intersection_update(
623
 
            no_fallback_inv_index.get_parent_map(all_inv_keys))
624
 
        parent_invs_only_keys = all_inv_keys.symmetric_difference(
625
 
            corresponding_invs)
626
 
        all_missing = set()
627
 
        inv_ids = [key[-1] for key in all_inv_keys]
628
 
        parent_invs_only_ids = [key[-1] for key in parent_invs_only_keys]
629
 
        root_key_info = _build_interesting_key_sets(
630
 
            self.repo, inv_ids, parent_invs_only_ids)
631
 
        expected_chk_roots = root_key_info.all_keys()
632
 
        present_chk_roots = no_fallback_chk_bytes_index.get_parent_map(
633
 
            expected_chk_roots)
634
 
        missing_chk_roots = expected_chk_roots.difference(present_chk_roots)
635
 
        if missing_chk_roots:
636
 
            problems.append("missing referenced chk root keys: %s"
637
 
                % (sorted(missing_chk_roots),))
638
 
            # Don't bother checking any further.
639
 
            return problems
640
 
        # Find all interesting chk_bytes records, and make sure they are
641
 
        # present, as well as the text keys they reference.
642
 
        chk_bytes_no_fallbacks = self.repo.chk_bytes.without_fallbacks()
643
 
        chk_bytes_no_fallbacks._search_key_func = \
644
 
            self.repo.chk_bytes._search_key_func
645
 
        chk_diff = chk_map.iter_interesting_nodes(
646
 
            chk_bytes_no_fallbacks, root_key_info.interesting_root_keys,
647
 
            root_key_info.uninteresting_root_keys)
648
 
        text_keys = set()
649
 
        try:
650
 
            for record in _filter_text_keys(chk_diff, text_keys,
651
 
                                            chk_map._bytes_to_text_key):
652
 
                pass
653
 
        except errors.NoSuchRevision, e:
654
 
            # XXX: It would be nice if we could give a more precise error here.
655
 
            problems.append("missing chk node(s) for id_to_entry maps")
656
 
        chk_diff = chk_map.iter_interesting_nodes(
657
 
            chk_bytes_no_fallbacks, root_key_info.interesting_pid_root_keys,
658
 
            root_key_info.uninteresting_pid_root_keys)
659
 
        try:
660
 
            for interesting_rec, interesting_map in chk_diff:
661
 
                pass
662
 
        except errors.NoSuchRevision, e:
663
 
            problems.append(
664
 
                "missing chk node(s) for parent_id_basename_to_file_id maps")
665
 
        present_text_keys = no_fallback_texts_index.get_parent_map(text_keys)
666
 
        missing_text_keys = text_keys.difference(present_text_keys)
667
 
        if missing_text_keys:
668
 
            problems.append("missing text keys: %r"
669
 
                % (sorted(missing_text_keys),))
670
 
        return problems
 
539
 
 
540
    def _already_packed(self):
 
541
        """Is the collection already packed?"""
 
542
        # Always repack GC repositories for now
 
543
        return False
671
544
 
672
545
    def _execute_pack_operations(self, pack_operations,
673
546
                                 _packer_class=GCCHKPacker,
687
560
            packer = GCCHKPacker(self, packs, '.autopack',
688
561
                                 reload_func=reload_func)
689
562
            try:
690
 
                result = packer.pack()
 
563
                packer.pack()
691
564
            except errors.RetryWithNewPacks:
692
565
                # An exception is propagating out of this context, make sure
693
566
                # this packer has cleaned up. Packer() doesn't set its new_pack
696
569
                if packer.new_pack is not None:
697
570
                    packer.new_pack.abort()
698
571
                raise
699
 
            if result is None:
700
 
                return
701
572
            for pack in packs:
702
573
                self._remove_pack_from_memory(pack)
703
574
        # record the newly available packs and stop advertising the old
704
575
        # packs
705
 
        to_be_obsoleted = []
706
 
        for _, packs in pack_operations:
707
 
            to_be_obsoleted.extend(packs)
708
 
        result = self._save_pack_names(clear_obsolete_packs=True,
709
 
                                       obsolete_packs=to_be_obsoleted)
710
 
        return result
711
 
 
712
 
 
713
 
class CHKInventoryRepository(KnitPackRepository):
714
 
    """subclass of KnitPackRepository that uses CHK based inventories."""
 
576
        self._save_pack_names(clear_obsolete_packs=True)
 
577
        # Move the old packs out of the way now they are no longer referenced.
 
578
        for revision_count, packs in pack_operations:
 
579
            self._obsolete_packs(packs)
 
580
 
 
581
 
 
582
# XXX: This format is scheduled for termination
 
583
#
 
584
# class GCPackRepository(KnitPackRepository):
 
585
#     """GC customisation of KnitPackRepository."""
 
586
#
 
587
#     def __init__(self, _format, a_bzrdir, control_files, _commit_builder_class,
 
588
#         _serializer):
 
589
#         """Overridden to change pack collection class."""
 
590
#         KnitPackRepository.__init__(self, _format, a_bzrdir, control_files,
 
591
#             _commit_builder_class, _serializer)
 
592
#         # and now replace everything it did :)
 
593
#         index_transport = self._transport.clone('indices')
 
594
#         self._pack_collection = GCRepositoryPackCollection(self,
 
595
#             self._transport, index_transport,
 
596
#             self._transport.clone('upload'),
 
597
#             self._transport.clone('packs'),
 
598
#             _format.index_builder_class,
 
599
#             _format.index_class,
 
600
#             use_chk_index=self._format.supports_chks,
 
601
#             )
 
602
#         self.inventories = GroupCompressVersionedFiles(
 
603
#             _GCGraphIndex(self._pack_collection.inventory_index.combined_index,
 
604
#                 add_callback=self._pack_collection.inventory_index.add_callback,
 
605
#                 parents=True, is_locked=self.is_locked),
 
606
#             access=self._pack_collection.inventory_index.data_access)
 
607
#         self.revisions = GroupCompressVersionedFiles(
 
608
#             _GCGraphIndex(self._pack_collection.revision_index.combined_index,
 
609
#                 add_callback=self._pack_collection.revision_index.add_callback,
 
610
#                 parents=True, is_locked=self.is_locked),
 
611
#             access=self._pack_collection.revision_index.data_access,
 
612
#             delta=False)
 
613
#         self.signatures = GroupCompressVersionedFiles(
 
614
#             _GCGraphIndex(self._pack_collection.signature_index.combined_index,
 
615
#                 add_callback=self._pack_collection.signature_index.add_callback,
 
616
#                 parents=False, is_locked=self.is_locked),
 
617
#             access=self._pack_collection.signature_index.data_access,
 
618
#             delta=False)
 
619
#         self.texts = GroupCompressVersionedFiles(
 
620
#             _GCGraphIndex(self._pack_collection.text_index.combined_index,
 
621
#                 add_callback=self._pack_collection.text_index.add_callback,
 
622
#                 parents=True, is_locked=self.is_locked),
 
623
#             access=self._pack_collection.text_index.data_access)
 
624
#         if _format.supports_chks:
 
625
#             # No graph, no compression:- references from chks are between
 
626
#             # different objects not temporal versions of the same; and without
 
627
#             # some sort of temporal structure knit compression will just fail.
 
628
#             self.chk_bytes = GroupCompressVersionedFiles(
 
629
#                 _GCGraphIndex(self._pack_collection.chk_index.combined_index,
 
630
#                     add_callback=self._pack_collection.chk_index.add_callback,
 
631
#                     parents=False, is_locked=self.is_locked),
 
632
#                 access=self._pack_collection.chk_index.data_access)
 
633
#         else:
 
634
#             self.chk_bytes = None
 
635
#         # True when the repository object is 'write locked' (as opposed to the
 
636
#         # physical lock only taken out around changes to the pack-names list.)
 
637
#         # Another way to represent this would be a decorator around the control
 
638
#         # files object that presents logical locks as physical ones - if this
 
639
#         # gets ugly consider that alternative design. RBC 20071011
 
640
#         self._write_lock_count = 0
 
641
#         self._transaction = None
 
642
#         # for tests
 
643
#         self._reconcile_does_inventory_gc = True
 
644
#         self._reconcile_fixes_text_parents = True
 
645
#         self._reconcile_backsup_inventory = False
 
646
#
 
647
#     def suspend_write_group(self):
 
648
#         raise errors.UnsuspendableWriteGroup(self)
 
649
#
 
650
#     def _resume_write_group(self, tokens):
 
651
#         raise errors.UnsuspendableWriteGroup(self)
 
652
#
 
653
#     def _reconcile_pack(self, collection, packs, extension, revs, pb):
 
654
#         bork
 
655
#         return packer.pack(pb)
 
656
 
 
657
 
 
658
class GCCHKPackRepository(CHKInventoryRepository):
 
659
    """GC customisation of CHKInventoryRepository."""
715
660
 
716
661
    def __init__(self, _format, a_bzrdir, control_files, _commit_builder_class,
717
662
        _serializer):
731
676
        self.inventories = GroupCompressVersionedFiles(
732
677
            _GCGraphIndex(self._pack_collection.inventory_index.combined_index,
733
678
                add_callback=self._pack_collection.inventory_index.add_callback,
734
 
                parents=True, is_locked=self.is_locked,
735
 
                inconsistency_fatal=False),
 
679
                parents=True, is_locked=self.is_locked),
736
680
            access=self._pack_collection.inventory_index.data_access)
737
681
        self.revisions = GroupCompressVersionedFiles(
738
682
            _GCGraphIndex(self._pack_collection.revision_index.combined_index,
739
683
                add_callback=self._pack_collection.revision_index.add_callback,
740
 
                parents=True, is_locked=self.is_locked,
741
 
                track_external_parent_refs=True, track_new_keys=True),
 
684
                parents=True, is_locked=self.is_locked),
742
685
            access=self._pack_collection.revision_index.data_access,
743
686
            delta=False)
744
687
        self.signatures = GroupCompressVersionedFiles(
745
688
            _GCGraphIndex(self._pack_collection.signature_index.combined_index,
746
689
                add_callback=self._pack_collection.signature_index.add_callback,
747
 
                parents=False, is_locked=self.is_locked,
748
 
                inconsistency_fatal=False),
 
690
                parents=False, is_locked=self.is_locked),
749
691
            access=self._pack_collection.signature_index.data_access,
750
692
            delta=False)
751
693
        self.texts = GroupCompressVersionedFiles(
752
694
            _GCGraphIndex(self._pack_collection.text_index.combined_index,
753
695
                add_callback=self._pack_collection.text_index.add_callback,
754
 
                parents=True, is_locked=self.is_locked,
755
 
                inconsistency_fatal=False),
 
696
                parents=True, is_locked=self.is_locked),
756
697
            access=self._pack_collection.text_index.data_access)
757
698
        # No parents, individual CHK pages don't have specific ancestry
758
699
        self.chk_bytes = GroupCompressVersionedFiles(
759
700
            _GCGraphIndex(self._pack_collection.chk_index.combined_index,
760
701
                add_callback=self._pack_collection.chk_index.add_callback,
761
 
                parents=False, is_locked=self.is_locked,
762
 
                inconsistency_fatal=False),
 
702
                parents=False, is_locked=self.is_locked),
763
703
            access=self._pack_collection.chk_index.data_access)
764
 
        search_key_name = self._format._serializer.search_key_name
765
 
        search_key_func = chk_map.search_key_registry.get(search_key_name)
766
 
        self.chk_bytes._search_key_func = search_key_func
767
704
        # True when the repository object is 'write locked' (as opposed to the
768
705
        # physical lock only taken out around changes to the pack-names list.)
769
706
        # Another way to represent this would be a decorator around the control
776
713
        self._reconcile_fixes_text_parents = True
777
714
        self._reconcile_backsup_inventory = False
778
715
 
779
 
    def _add_inventory_checked(self, revision_id, inv, parents):
780
 
        """Add inv to the repository after checking the inputs.
781
 
 
782
 
        This function can be overridden to allow different inventory styles.
783
 
 
784
 
        :seealso: add_inventory, for the contract.
785
 
        """
786
 
        # make inventory
787
 
        serializer = self._format._serializer
788
 
        result = inventory.CHKInventory.from_inventory(self.chk_bytes, inv,
789
 
            maximum_size=serializer.maximum_size,
790
 
            search_key_name=serializer.search_key_name)
791
 
        inv_lines = result.to_lines()
792
 
        return self._inventory_add_lines(revision_id, parents,
793
 
            inv_lines, check_content=False)
794
 
 
795
 
    def _create_inv_from_null(self, delta, revision_id):
796
 
        """This will mutate new_inv directly.
797
 
 
798
 
        This is a simplified form of create_by_apply_delta which knows that all
799
 
        the old values must be None, so everything is a create.
800
 
        """
801
 
        serializer = self._format._serializer
802
 
        new_inv = inventory.CHKInventory(serializer.search_key_name)
803
 
        new_inv.revision_id = revision_id
804
 
        entry_to_bytes = new_inv._entry_to_bytes
805
 
        id_to_entry_dict = {}
806
 
        parent_id_basename_dict = {}
807
 
        for old_path, new_path, file_id, entry in delta:
808
 
            if old_path is not None:
809
 
                raise ValueError('Invalid delta, somebody tried to delete %r'
810
 
                                 ' from the NULL_REVISION'
811
 
                                 % ((old_path, file_id),))
812
 
            if new_path is None:
813
 
                raise ValueError('Invalid delta, delta from NULL_REVISION has'
814
 
                                 ' no new_path %r' % (file_id,))
815
 
            if new_path == '':
816
 
                new_inv.root_id = file_id
817
 
                parent_id_basename_key = StaticTuple('', '').intern()
818
 
            else:
819
 
                utf8_entry_name = entry.name.encode('utf-8')
820
 
                parent_id_basename_key = StaticTuple(entry.parent_id,
821
 
                                                     utf8_entry_name).intern()
822
 
            new_value = entry_to_bytes(entry)
823
 
            # Populate Caches?
824
 
            # new_inv._path_to_fileid_cache[new_path] = file_id
825
 
            key = StaticTuple(file_id).intern()
826
 
            id_to_entry_dict[key] = new_value
827
 
            parent_id_basename_dict[parent_id_basename_key] = file_id
828
 
 
829
 
        new_inv._populate_from_dicts(self.chk_bytes, id_to_entry_dict,
830
 
            parent_id_basename_dict, maximum_size=serializer.maximum_size)
831
 
        return new_inv
832
 
 
833
 
    def add_inventory_by_delta(self, basis_revision_id, delta, new_revision_id,
834
 
                               parents, basis_inv=None, propagate_caches=False):
835
 
        """Add a new inventory expressed as a delta against another revision.
836
 
 
837
 
        :param basis_revision_id: The inventory id the delta was created
838
 
            against.
839
 
        :param delta: The inventory delta (see Inventory.apply_delta for
840
 
            details).
841
 
        :param new_revision_id: The revision id that the inventory is being
842
 
            added for.
843
 
        :param parents: The revision ids of the parents that revision_id is
844
 
            known to have and are in the repository already. These are supplied
845
 
            for repositories that depend on the inventory graph for revision
846
 
            graph access, as well as for those that pun ancestry with delta
847
 
            compression.
848
 
        :param basis_inv: The basis inventory if it is already known,
849
 
            otherwise None.
850
 
        :param propagate_caches: If True, the caches for this inventory are
851
 
          copied to and updated for the result if possible.
852
 
 
853
 
        :returns: (validator, new_inv)
854
 
            The validator(which is a sha1 digest, though what is sha'd is
855
 
            repository format specific) of the serialized inventory, and the
856
 
            resulting inventory.
857
 
        """
858
 
        if not self.is_in_write_group():
859
 
            raise AssertionError("%r not in write group" % (self,))
860
 
        _mod_revision.check_not_reserved_id(new_revision_id)
861
 
        basis_tree = None
862
 
        if basis_inv is None:
863
 
            if basis_revision_id == _mod_revision.NULL_REVISION:
864
 
                new_inv = self._create_inv_from_null(delta, new_revision_id)
865
 
                if new_inv.root_id is None:
866
 
                    raise errors.RootMissing()
867
 
                inv_lines = new_inv.to_lines()
868
 
                return self._inventory_add_lines(new_revision_id, parents,
869
 
                    inv_lines, check_content=False), new_inv
870
 
            else:
871
 
                basis_tree = self.revision_tree(basis_revision_id)
872
 
                basis_tree.lock_read()
873
 
                basis_inv = basis_tree.inventory
874
 
        try:
875
 
            result = basis_inv.create_by_apply_delta(delta, new_revision_id,
876
 
                propagate_caches=propagate_caches)
877
 
            inv_lines = result.to_lines()
878
 
            return self._inventory_add_lines(new_revision_id, parents,
879
 
                inv_lines, check_content=False), result
880
 
        finally:
881
 
            if basis_tree is not None:
882
 
                basis_tree.unlock()
883
 
 
884
 
    def _deserialise_inventory(self, revision_id, bytes):
885
 
        return inventory.CHKInventory.deserialise(self.chk_bytes, bytes,
886
 
            (revision_id,))
887
 
 
888
 
    def _iter_inventories(self, revision_ids, ordering):
889
 
        """Iterate over many inventory objects."""
890
 
        if ordering is None:
891
 
            ordering = 'unordered'
892
 
        keys = [(revision_id,) for revision_id in revision_ids]
893
 
        stream = self.inventories.get_record_stream(keys, ordering, True)
894
 
        texts = {}
895
 
        for record in stream:
896
 
            if record.storage_kind != 'absent':
897
 
                texts[record.key] = record.get_bytes_as('fulltext')
898
 
            else:
899
 
                raise errors.NoSuchRevision(self, record.key)
900
 
        for key in keys:
901
 
            yield inventory.CHKInventory.deserialise(self.chk_bytes, texts[key], key)
902
 
 
903
 
    def _iter_inventory_xmls(self, revision_ids, ordering):
904
 
        # Without a native 'xml' inventory, this method doesn't make sense.
905
 
        # However older working trees, and older bundles want it - so we supply
906
 
        # it allowing _get_inventory_xml to work. Bundles currently use the
907
 
        # serializer directly; this also isn't ideal, but there isn't an xml
908
 
        # iteration interface offered at all for repositories. We could make
909
 
        # _iter_inventory_xmls be part of the contract, even if kept private.
910
 
        inv_to_str = self._serializer.write_inventory_to_string
911
 
        for inv in self.iter_inventories(revision_ids, ordering=ordering):
912
 
            yield inv_to_str(inv), inv.revision_id
913
 
 
914
 
    def _find_present_inventory_keys(self, revision_keys):
915
 
        parent_map = self.inventories.get_parent_map(revision_keys)
916
 
        present_inventory_keys = set(k for k in parent_map)
917
 
        return present_inventory_keys
918
 
 
919
 
    def fileids_altered_by_revision_ids(self, revision_ids, _inv_weave=None):
920
 
        """Find the file ids and versions affected by revisions.
921
 
 
922
 
        :param revisions: an iterable containing revision ids.
923
 
        :param _inv_weave: The inventory weave from this repository or None.
924
 
            If None, the inventory weave will be opened automatically.
925
 
        :return: a dictionary mapping altered file-ids to an iterable of
926
 
            revision_ids. Each altered file-ids has the exact revision_ids that
927
 
            altered it listed explicitly.
928
 
        """
929
 
        rich_root = self.supports_rich_root()
930
 
        bytes_to_info = inventory.CHKInventory._bytes_to_utf8name_key
931
 
        file_id_revisions = {}
932
 
        pb = ui.ui_factory.nested_progress_bar()
933
 
        try:
934
 
            revision_keys = [(r,) for r in revision_ids]
935
 
            parent_keys = self._find_parent_keys_of_revisions(revision_keys)
936
 
            # TODO: instead of using _find_present_inventory_keys, change the
937
 
            #       code paths to allow missing inventories to be tolerated.
938
 
            #       However, we only want to tolerate missing parent
939
 
            #       inventories, not missing inventories for revision_ids
940
 
            present_parent_inv_keys = self._find_present_inventory_keys(
941
 
                                        parent_keys)
942
 
            present_parent_inv_ids = set(
943
 
                [k[-1] for k in present_parent_inv_keys])
944
 
            inventories_to_read = set(revision_ids)
945
 
            inventories_to_read.update(present_parent_inv_ids)
946
 
            root_key_info = _build_interesting_key_sets(
947
 
                self, inventories_to_read, present_parent_inv_ids)
948
 
            interesting_root_keys = root_key_info.interesting_root_keys
949
 
            uninteresting_root_keys = root_key_info.uninteresting_root_keys
950
 
            chk_bytes = self.chk_bytes
951
 
            for record, items in chk_map.iter_interesting_nodes(chk_bytes,
952
 
                        interesting_root_keys, uninteresting_root_keys,
953
 
                        pb=pb):
954
 
                for name, bytes in items:
955
 
                    (name_utf8, file_id, revision_id) = bytes_to_info(bytes)
956
 
                    # TODO: consider interning file_id, revision_id here, or
957
 
                    #       pushing that intern() into bytes_to_info()
958
 
                    # TODO: rich_root should always be True here, for all
959
 
                    #       repositories that support chk_bytes
960
 
                    if not rich_root and name_utf8 == '':
961
 
                        continue
962
 
                    try:
963
 
                        file_id_revisions[file_id].add(revision_id)
964
 
                    except KeyError:
965
 
                        file_id_revisions[file_id] = set([revision_id])
966
 
        finally:
967
 
            pb.finished()
968
 
        return file_id_revisions
969
 
 
970
 
    def find_text_key_references(self):
971
 
        """Find the text key references within the repository.
972
 
 
973
 
        :return: A dictionary mapping text keys ((fileid, revision_id) tuples)
974
 
            to whether they were referred to by the inventory of the
975
 
            revision_id that they contain. The inventory texts from all present
976
 
            revision ids are assessed to generate this report.
977
 
        """
978
 
        # XXX: Slow version but correct: rewrite as a series of delta
979
 
        # examinations/direct tree traversal. Note that that will require care
980
 
        # as a common node is reachable both from the inventory that added it,
981
 
        # and others afterwards.
982
 
        revision_keys = self.revisions.keys()
983
 
        result = {}
984
 
        rich_roots = self.supports_rich_root()
985
 
        pb = ui.ui_factory.nested_progress_bar()
986
 
        try:
987
 
            all_revs = self.all_revision_ids()
988
 
            total = len(all_revs)
989
 
            for pos, inv in enumerate(self.iter_inventories(all_revs)):
990
 
                pb.update("Finding text references", pos, total)
991
 
                for _, entry in inv.iter_entries():
992
 
                    if not rich_roots and entry.file_id == inv.root_id:
993
 
                        continue
994
 
                    key = (entry.file_id, entry.revision)
995
 
                    result.setdefault(key, False)
996
 
                    if entry.revision == inv.revision_id:
997
 
                        result[key] = True
998
 
            return result
999
 
        finally:
1000
 
            pb.finished()
 
716
    def suspend_write_group(self):
 
717
        raise errors.UnsuspendableWriteGroup(self)
 
718
 
 
719
    def _resume_write_group(self, tokens):
 
720
        raise errors.UnsuspendableWriteGroup(self)
1001
721
 
1002
722
    def _reconcile_pack(self, collection, packs, extension, revs, pb):
 
723
        # assert revs is None
1003
724
        packer = GCCHKReconcilePacker(collection, packs, extension)
1004
725
        return packer.pack(pb)
1005
726
 
1006
 
    def _get_source(self, to_format):
1007
 
        """Return a source for streaming from this repository."""
1008
 
        if self._format._serializer == to_format._serializer:
1009
 
            # We must be exactly the same format, otherwise stuff like the chk
1010
 
            # page layout might be different.
1011
 
            # Actually, this test is just slightly looser than exact so that
1012
 
            # CHK2 <-> 2a transfers will work.
1013
 
            return GroupCHKStreamSource(self, to_format)
1014
 
        return super(CHKInventoryRepository, self)._get_source(to_format)
1015
 
 
1016
 
 
1017
 
class GroupCHKStreamSource(KnitPackStreamSource):
1018
 
    """Used when both the source and target repo are GroupCHK repos."""
1019
 
 
1020
 
    def __init__(self, from_repository, to_format):
1021
 
        """Create a StreamSource streaming from from_repository."""
1022
 
        super(GroupCHKStreamSource, self).__init__(from_repository, to_format)
1023
 
        self._revision_keys = None
1024
 
        self._text_keys = None
1025
 
        self._text_fetch_order = 'groupcompress'
1026
 
        self._chk_id_roots = None
1027
 
        self._chk_p_id_roots = None
1028
 
 
1029
 
    def _get_inventory_stream(self, inventory_keys, allow_absent=False):
1030
 
        """Get a stream of inventory texts.
1031
 
 
1032
 
        When this function returns, self._chk_id_roots and self._chk_p_id_roots
1033
 
        should be populated.
1034
 
        """
1035
 
        self._chk_id_roots = []
1036
 
        self._chk_p_id_roots = []
1037
 
        def _filtered_inv_stream():
1038
 
            id_roots_set = set()
1039
 
            p_id_roots_set = set()
1040
 
            source_vf = self.from_repository.inventories
1041
 
            stream = source_vf.get_record_stream(inventory_keys,
1042
 
                                                 'groupcompress', True)
1043
 
            for record in stream:
1044
 
                if record.storage_kind == 'absent':
1045
 
                    if allow_absent:
1046
 
                        continue
1047
 
                    else:
1048
 
                        raise errors.NoSuchRevision(self, record.key)
1049
 
                bytes = record.get_bytes_as('fulltext')
1050
 
                chk_inv = inventory.CHKInventory.deserialise(None, bytes,
1051
 
                                                             record.key)
1052
 
                key = chk_inv.id_to_entry.key()
1053
 
                if key not in id_roots_set:
1054
 
                    self._chk_id_roots.append(key)
1055
 
                    id_roots_set.add(key)
1056
 
                p_id_map = chk_inv.parent_id_basename_to_file_id
1057
 
                if p_id_map is None:
1058
 
                    raise AssertionError('Parent id -> file_id map not set')
1059
 
                key = p_id_map.key()
1060
 
                if key not in p_id_roots_set:
1061
 
                    p_id_roots_set.add(key)
1062
 
                    self._chk_p_id_roots.append(key)
1063
 
                yield record
1064
 
            # We have finished processing all of the inventory records, we
1065
 
            # don't need these sets anymore
1066
 
            id_roots_set.clear()
1067
 
            p_id_roots_set.clear()
1068
 
        return ('inventories', _filtered_inv_stream())
1069
 
 
1070
 
    def _get_filtered_chk_streams(self, excluded_revision_keys):
1071
 
        self._text_keys = set()
1072
 
        excluded_revision_keys.discard(_mod_revision.NULL_REVISION)
1073
 
        if not excluded_revision_keys:
1074
 
            uninteresting_root_keys = set()
1075
 
            uninteresting_pid_root_keys = set()
1076
 
        else:
1077
 
            # filter out any excluded revisions whose inventories are not
1078
 
            # actually present
1079
 
            # TODO: Update Repository.iter_inventories() to add
1080
 
            #       ignore_missing=True
1081
 
            present_keys = self.from_repository._find_present_inventory_keys(
1082
 
                            excluded_revision_keys)
1083
 
            present_ids = [k[-1] for k in present_keys]
1084
 
            uninteresting_root_keys = set()
1085
 
            uninteresting_pid_root_keys = set()
1086
 
            for inv in self.from_repository.iter_inventories(present_ids):
1087
 
                uninteresting_root_keys.add(inv.id_to_entry.key())
1088
 
                uninteresting_pid_root_keys.add(
1089
 
                    inv.parent_id_basename_to_file_id.key())
1090
 
        chk_bytes = self.from_repository.chk_bytes
1091
 
        def _filter_id_to_entry():
1092
 
            interesting_nodes = chk_map.iter_interesting_nodes(chk_bytes,
1093
 
                        self._chk_id_roots, uninteresting_root_keys)
1094
 
            for record in _filter_text_keys(interesting_nodes, self._text_keys,
1095
 
                    chk_map._bytes_to_text_key):
1096
 
                if record is not None:
1097
 
                    yield record
1098
 
            # Consumed
1099
 
            self._chk_id_roots = None
1100
 
        yield 'chk_bytes', _filter_id_to_entry()
1101
 
        def _get_parent_id_basename_to_file_id_pages():
1102
 
            for record, items in chk_map.iter_interesting_nodes(chk_bytes,
1103
 
                        self._chk_p_id_roots, uninteresting_pid_root_keys):
1104
 
                if record is not None:
1105
 
                    yield record
1106
 
            # Consumed
1107
 
            self._chk_p_id_roots = None
1108
 
        yield 'chk_bytes', _get_parent_id_basename_to_file_id_pages()
1109
 
 
1110
 
    def get_stream(self, search):
1111
 
        def wrap_and_count(pb, rc, stream):
1112
 
            """Yield records from stream while showing progress."""
1113
 
            count = 0
1114
 
            for record in stream:
1115
 
                if count == rc.STEP:
1116
 
                    rc.increment(count)
1117
 
                    pb.update('Estimate', rc.current, rc.max)
1118
 
                    count = 0
1119
 
                count += 1
1120
 
                yield record
1121
 
 
1122
 
        revision_ids = search.get_keys()
1123
 
        pb = ui.ui_factory.nested_progress_bar()
1124
 
        rc = self._record_counter
1125
 
        self._record_counter.setup(len(revision_ids))
1126
 
        for stream_info in self._fetch_revision_texts(revision_ids):
1127
 
            yield (stream_info[0],
1128
 
                wrap_and_count(pb, rc, stream_info[1]))
1129
 
        self._revision_keys = [(rev_id,) for rev_id in revision_ids]
1130
 
        self.from_repository.revisions.clear_cache()
1131
 
        self.from_repository.signatures.clear_cache()
1132
 
        s = self._get_inventory_stream(self._revision_keys)
1133
 
        yield (s[0], wrap_and_count(pb, rc, s[1]))
1134
 
        self.from_repository.inventories.clear_cache()
1135
 
        # TODO: The keys to exclude might be part of the search recipe
1136
 
        # For now, exclude all parents that are at the edge of ancestry, for
1137
 
        # which we have inventories
1138
 
        from_repo = self.from_repository
1139
 
        parent_keys = from_repo._find_parent_keys_of_revisions(
1140
 
                        self._revision_keys)
1141
 
        for stream_info in self._get_filtered_chk_streams(parent_keys):
1142
 
            yield (stream_info[0], wrap_and_count(pb, rc, stream_info[1]))
1143
 
        self.from_repository.chk_bytes.clear_cache()
1144
 
        s = self._get_text_stream()
1145
 
        yield (s[0], wrap_and_count(pb, rc, s[1]))
1146
 
        self.from_repository.texts.clear_cache()
1147
 
        pb.update('Done', rc.max, rc.max)
1148
 
        pb.finished()
1149
 
 
1150
 
    def get_stream_for_missing_keys(self, missing_keys):
1151
 
        # missing keys can only occur when we are byte copying and not
1152
 
        # translating (because translation means we don't send
1153
 
        # unreconstructable deltas ever).
1154
 
        missing_inventory_keys = set()
1155
 
        for key in missing_keys:
1156
 
            if key[0] != 'inventories':
1157
 
                raise AssertionError('The only missing keys we should'
1158
 
                    ' be filling in are inventory keys, not %s'
1159
 
                    % (key[0],))
1160
 
            missing_inventory_keys.add(key[1:])
1161
 
        if self._chk_id_roots or self._chk_p_id_roots:
1162
 
            raise AssertionError('Cannot call get_stream_for_missing_keys'
1163
 
                ' until all of get_stream() has been consumed.')
1164
 
        # Yield the inventory stream, so we can find the chk stream
1165
 
        # Some of the missing_keys will be missing because they are ghosts.
1166
 
        # As such, we can ignore them. The Sink is required to verify there are
1167
 
        # no unavailable texts when the ghost inventories are not filled in.
1168
 
        yield self._get_inventory_stream(missing_inventory_keys,
1169
 
                                         allow_absent=True)
1170
 
        # We use the empty set for excluded_revision_keys, to make it clear
1171
 
        # that we want to transmit all referenced chk pages.
1172
 
        for stream_info in self._get_filtered_chk_streams(set()):
1173
 
            yield stream_info
1174
 
 
1175
 
 
1176
 
class _InterestingKeyInfo(object):
1177
 
    def __init__(self):
1178
 
        self.interesting_root_keys = set()
1179
 
        self.interesting_pid_root_keys = set()
1180
 
        self.uninteresting_root_keys = set()
1181
 
        self.uninteresting_pid_root_keys = set()
1182
 
 
1183
 
    def all_interesting(self):
1184
 
        return self.interesting_root_keys.union(self.interesting_pid_root_keys)
1185
 
 
1186
 
    def all_uninteresting(self):
1187
 
        return self.uninteresting_root_keys.union(
1188
 
            self.uninteresting_pid_root_keys)
1189
 
 
1190
 
    def all_keys(self):
1191
 
        return self.all_interesting().union(self.all_uninteresting())
1192
 
 
1193
 
 
1194
 
def _build_interesting_key_sets(repo, inventory_ids, parent_only_inv_ids):
1195
 
    result = _InterestingKeyInfo()
1196
 
    for inv in repo.iter_inventories(inventory_ids, 'unordered'):
1197
 
        root_key = inv.id_to_entry.key()
1198
 
        pid_root_key = inv.parent_id_basename_to_file_id.key()
1199
 
        if inv.revision_id in parent_only_inv_ids:
1200
 
            result.uninteresting_root_keys.add(root_key)
1201
 
            result.uninteresting_pid_root_keys.add(pid_root_key)
1202
 
        else:
1203
 
            result.interesting_root_keys.add(root_key)
1204
 
            result.interesting_pid_root_keys.add(pid_root_key)
1205
 
    return result
1206
 
 
1207
 
 
1208
 
def _filter_text_keys(interesting_nodes_iterable, text_keys, bytes_to_text_key):
1209
 
    """Iterate the result of iter_interesting_nodes, yielding the records
1210
 
    and adding to text_keys.
1211
 
    """
1212
 
    text_keys_update = text_keys.update
1213
 
    for record, items in interesting_nodes_iterable:
1214
 
        text_keys_update([bytes_to_text_key(b) for n,b in items])
1215
 
        yield record
1216
 
 
1217
 
 
1218
 
 
1219
 
 
1220
 
class RepositoryFormatCHK1(RepositoryFormatPack):
 
727
 
 
728
class RepositoryFormatPackGCCHK16(RepositoryFormatPackDevelopment5Hash16):
1221
729
    """A hashed CHK+group compress pack repository."""
1222
730
 
1223
 
    repository_class = CHKInventoryRepository
 
731
    repository_class = GCCHKPackRepository
 
732
    _commit_builder_class = PackRootCommitBuilder
 
733
    rich_root_data = True
1224
734
    supports_external_lookups = True
 
735
    supports_tree_reference = True
 
736
    supports_chks = True
 
737
    # Note: We cannot unpack a delta that references a text we haven't
 
738
    # seen yet. There are 2 options, work in fulltexts, or require
 
739
    # topological sorting. Using fulltexts is more optimal for local
 
740
    # operations, because the source can be smart about extracting
 
741
    # multiple in-a-row (and sharing strings). Topological is better
 
742
    # for remote, because we access less data.
 
743
    _fetch_order = 'unordered'
 
744
    _fetch_uses_deltas = False
 
745
 
 
746
    def _get_matching_bzrdir(self):
 
747
        return bzrdir.format_registry.make_bzrdir('gc-chk16')
 
748
 
 
749
    def _ignore_setting_bzrdir(self, format):
 
750
        pass
 
751
 
 
752
    _matchingbzrdir = property(_get_matching_bzrdir, _ignore_setting_bzrdir)
 
753
 
 
754
    def get_format_string(self):
 
755
        """See RepositoryFormat.get_format_string()."""
 
756
        return ('Bazaar development format - hash16chk+gc rich-root'
 
757
                ' (needs bzr.dev from 1.14)\n')
 
758
 
 
759
    def get_format_description(self):
 
760
        """See RepositoryFormat.get_format_description()."""
 
761
        return ("Development repository format - hash16chk+groupcompress")
 
762
 
 
763
    def check_conversion_target(self, target_format):
 
764
        if not target_format.rich_root_data:
 
765
            raise errors.BadConversionTarget(
 
766
                'Does not support rich root data.', target_format)
 
767
        if not getattr(target_format, 'supports_tree_reference', False):
 
768
            raise errors.BadConversionTarget(
 
769
                'Does not support nested trees', target_format)
 
770
 
 
771
 
 
772
class RepositoryFormatPackGCCHK255(RepositoryFormatPackDevelopment5Hash255):
 
773
    """A hashed CHK+group compress pack repository."""
 
774
 
 
775
    repository_class = GCCHKPackRepository
 
776
    supports_chks = True
 
777
    # Setting this to True causes us to use InterModel1And2, so for now set
 
778
    # it to False which uses InterDifferingSerializer. When IM1&2 is
 
779
    # removed (as it is in bzr.dev) we can set this back to True.
 
780
    _commit_builder_class = PackRootCommitBuilder
 
781
    rich_root_data = True
 
782
 
 
783
    def _get_matching_bzrdir(self):
 
784
        return bzrdir.format_registry.make_bzrdir('gc-chk255')
 
785
 
 
786
    def _ignore_setting_bzrdir(self, format):
 
787
        pass
 
788
 
 
789
    _matchingbzrdir = property(_get_matching_bzrdir, _ignore_setting_bzrdir)
 
790
 
 
791
    def get_format_string(self):
 
792
        """See RepositoryFormat.get_format_string()."""
 
793
        return ('Bazaar development format - hash255chk+gc rich-root'
 
794
                ' (needs bzr.dev from 1.14)\n')
 
795
 
 
796
    def get_format_description(self):
 
797
        """See RepositoryFormat.get_format_description()."""
 
798
        return ("Development repository format - hash255chk+groupcompress")
 
799
 
 
800
    def check_conversion_target(self, target_format):
 
801
        if not target_format.rich_root_data:
 
802
            raise errors.BadConversionTarget(
 
803
                'Does not support rich root data.', target_format)
 
804
        if not getattr(target_format, 'supports_tree_reference', False):
 
805
            raise errors.BadConversionTarget(
 
806
                'Does not support nested trees', target_format)
 
807
 
 
808
 
 
809
class RepositoryFormatPackGCCHK255Big(RepositoryFormatPackGCCHK255):
 
810
    """A hashed CHK+group compress pack repository."""
 
811
 
 
812
    repository_class = GCCHKPackRepository
1225
813
    supports_chks = True
1226
814
    # For right now, setting this to True gives us InterModel1And2 rather
1227
815
    # than InterDifferingSerializer
1228
816
    _commit_builder_class = PackRootCommitBuilder
1229
817
    rich_root_data = True
1230
818
    _serializer = chk_serializer.chk_serializer_255_bigpage
1231
 
    _commit_inv_deltas = True
1232
 
    # What index classes to use
1233
 
    index_builder_class = BTreeBuilder
1234
 
    index_class = BTreeGraphIndex
1235
819
    # Note: We cannot unpack a delta that references a text we haven't
1236
820
    # seen yet. There are 2 options, work in fulltexts, or require
1237
821
    # topological sorting. Using fulltexts is more optimal for local
1239
823
    # multiple in-a-row (and sharing strings). Topological is better
1240
824
    # for remote, because we access less data.
1241
825
    _fetch_order = 'unordered'
1242
 
    _fetch_uses_deltas = False # essentially ignored by the groupcompress code.
1243
 
    fast_deltas = True
1244
 
    pack_compresses = True
 
826
    _fetch_uses_deltas = False
1245
827
 
1246
828
    def _get_matching_bzrdir(self):
1247
 
        return bzrdir.format_registry.make_bzrdir('development6-rich-root')
 
829
        return bzrdir.format_registry.make_bzrdir('gc-chk255-big')
1248
830
 
1249
831
    def _ignore_setting_bzrdir(self, format):
1250
832
        pass
1253
835
 
1254
836
    def get_format_string(self):
1255
837
        """See RepositoryFormat.get_format_string()."""
1256
 
        return ('Bazaar development format - group compression and chk inventory'
 
838
        return ('Bazaar development format - hash255chk+gc rich-root bigpage'
1257
839
                ' (needs bzr.dev from 1.14)\n')
1258
840
 
1259
841
    def get_format_description(self):
1260
842
        """See RepositoryFormat.get_format_description()."""
1261
 
        return ("Development repository format - rich roots, group compression"
1262
 
            " and chk inventories")
1263
 
 
1264
 
 
1265
 
class RepositoryFormatCHK2(RepositoryFormatCHK1):
1266
 
    """A CHK repository that uses the bencode revision serializer."""
1267
 
 
1268
 
    _serializer = chk_serializer.chk_bencode_serializer
1269
 
 
1270
 
    def _get_matching_bzrdir(self):
1271
 
        return bzrdir.format_registry.make_bzrdir('development7-rich-root')
1272
 
 
1273
 
    def _ignore_setting_bzrdir(self, format):
1274
 
        pass
1275
 
 
1276
 
    _matchingbzrdir = property(_get_matching_bzrdir, _ignore_setting_bzrdir)
1277
 
 
1278
 
    def get_format_string(self):
1279
 
        """See RepositoryFormat.get_format_string()."""
1280
 
        return ('Bazaar development format - chk repository with bencode '
1281
 
                'revision serialization (needs bzr.dev from 1.16)\n')
1282
 
 
1283
 
 
1284
 
class RepositoryFormat2a(RepositoryFormatCHK2):
1285
 
    """A CHK repository that uses the bencode revision serializer.
1286
 
 
1287
 
    This is the same as RepositoryFormatCHK2 but with a public name.
1288
 
    """
1289
 
 
1290
 
    _serializer = chk_serializer.chk_bencode_serializer
1291
 
 
1292
 
    def _get_matching_bzrdir(self):
1293
 
        return bzrdir.format_registry.make_bzrdir('2a')
1294
 
 
1295
 
    def _ignore_setting_bzrdir(self, format):
1296
 
        pass
1297
 
 
1298
 
    _matchingbzrdir = property(_get_matching_bzrdir, _ignore_setting_bzrdir)
1299
 
 
1300
 
    def get_format_string(self):
1301
 
        return ('Bazaar repository format 2a (needs bzr 1.16 or later)\n')
1302
 
 
1303
 
    def get_format_description(self):
1304
 
        """See RepositoryFormat.get_format_description()."""
1305
 
        return ("Repository format 2a - rich roots, group compression"
1306
 
            " and chk inventories")
 
843
        return ("Development repository format - hash255chk+groupcompress + bigpage")
 
844
 
 
845
    def check_conversion_target(self, target_format):
 
846
        if not target_format.rich_root_data:
 
847
            raise errors.BadConversionTarget(
 
848
                'Does not support rich root data.', target_format)
 
849
        if not getattr(target_format, 'supports_tree_reference', False):
 
850
            raise errors.BadConversionTarget(
 
851
                'Does not support nested trees', target_format)