~bzr-pqm/bzr/bzr.dev

3735.31.2 by John Arbash Meinel
Cleanup trailing whitespace, get test_source to pass by removing asserts.
1
# Copyright (C) 2008, 2009 Canonical Ltd
2
#
0.17.9 by Robert Collins
Initial stab at repository format support.
3
# This program is free software; you can redistribute it and/or modify
3735.31.2 by John Arbash Meinel
Cleanup trailing whitespace, get test_source to pass by removing asserts.
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.
7
#
0.17.9 by Robert Collins
Initial stab at repository format support.
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.
3735.31.2 by John Arbash Meinel
Cleanup trailing whitespace, get test_source to pass by removing asserts.
12
#
0.17.9 by Robert Collins
Initial stab at repository format support.
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
3735.36.3 by John Arbash Meinel
Add the new address for FSF to the new files.
15
# Foundation, Inc., 51 Franklin Street, Fifth Floor, Boston, MA 02110-1301 USA
0.17.9 by Robert Collins
Initial stab at repository format support.
16
4277.2.1 by Vincent Ladeuil
Fix misleading signature and one typo.
17
"""Repository formats using CHK inventories and groupcompress compression."""
0.17.9 by Robert Collins
Initial stab at repository format support.
18
19
import time
20
0.20.19 by John Arbash Meinel
Groupcompress now supports 'autopack' and 'pack'.
21
from bzrlib import (
3735.31.16 by John Arbash Meinel
Add _get_matching_bzrdir methods to all of the GC Formats.
22
    bzrdir,
3735.31.3 by John Arbash Meinel
Rename gc_repo to groupcompress_repo, start trying to remove some of the special casing and get tests passing.
23
    chk_map,
3735.31.4 by John Arbash Meinel
Continue trying to reduce the friction for the groupcompress repo formats.
24
    chk_serializer,
0.20.19 by John Arbash Meinel
Groupcompress now supports 'autopack' and 'pack'.
25
    debug,
26
    errors,
3735.31.15 by John Arbash Meinel
A bit more refactoring, try to clean up the _execute_packs code.
27
    index as _mod_index,
3735.31.2 by John Arbash Meinel
Cleanup trailing whitespace, get test_source to pass by removing asserts.
28
    inventory,
0.20.19 by John Arbash Meinel
Groupcompress now supports 'autopack' and 'pack'.
29
    knit,
3735.31.2 by John Arbash Meinel
Cleanup trailing whitespace, get test_source to pass by removing asserts.
30
    osutils,
0.20.19 by John Arbash Meinel
Groupcompress now supports 'autopack' and 'pack'.
31
    pack,
4241.6.8 by Robert Collins, John Arbash Meinel, Ian Clatworthy, Vincent Ladeuil
Add --development6-rich-root, disabling the legacy and unneeded development2 format, and activating the tests for CHK features disabled pending this format. (Robert Collins, John Arbash Meinel, Ian Clatworthy, Vincent Ladeuil)
32
    remote,
0.20.19 by John Arbash Meinel
Groupcompress now supports 'autopack' and 'pack'.
33
    repository,
3735.31.27 by John Arbash Meinel
Implement GCCHKReconcilePacker.
34
    revision as _mod_revision,
0.23.28 by John Arbash Meinel
Gotta import 'trace' if you want to use trace.mutter()
35
    trace,
0.20.19 by John Arbash Meinel
Groupcompress now supports 'autopack' and 'pack'.
36
    ui,
37
    )
4241.6.8 by Robert Collins, John Arbash Meinel, Ian Clatworthy, Vincent Ladeuil
Add --development6-rich-root, disabling the legacy and unneeded development2 format, and activating the tests for CHK features disabled pending this format. (Robert Collins, John Arbash Meinel, Ian Clatworthy, Vincent Ladeuil)
38
from bzrlib.btree_index import (
39
    BTreeGraphIndex,
40
    BTreeBuilder,
41
    )
0.17.9 by Robert Collins
Initial stab at repository format support.
42
from bzrlib.index import GraphIndex, GraphIndexBuilder
3735.31.1 by John Arbash Meinel
Bring the groupcompress plugin into the brisbane-core branch.
43
from bzrlib.groupcompress import (
0.17.9 by Robert Collins
Initial stab at repository format support.
44
    _GCGraphIndex,
45
    GroupCompressVersionedFiles,
46
    )
47
from bzrlib.repofmt.pack_repo import (
48
    Pack,
49
    NewPack,
50
    KnitPackRepository,
3735.31.9 by John Arbash Meinel
Setting the _commit_builder_class makes commit with rich_root_data=True work.
51
    PackRootCommitBuilder,
0.17.9 by Robert Collins
Initial stab at repository format support.
52
    RepositoryPackCollection,
4241.6.8 by Robert Collins, John Arbash Meinel, Ian Clatworthy, Vincent Ladeuil
Add --development6-rich-root, disabling the legacy and unneeded development2 format, and activating the tests for CHK features disabled pending this format. (Robert Collins, John Arbash Meinel, Ian Clatworthy, Vincent Ladeuil)
53
    RepositoryFormatPack,
4343.3.2 by John Arbash Meinel
All stacking tests seem to be passing for dev6 repos
54
    ResumedPack,
0.17.9 by Robert Collins
Initial stab at repository format support.
55
    Packer,
0.17.25 by Robert Collins
Preliminary --gc-plain-chk support.
56
    )
0.17.9 by Robert Collins
Initial stab at repository format support.
57
58
59
class GCPack(NewPack):
60
0.17.22 by Robert Collins
really get gc working with 1.10
61
    def __init__(self, pack_collection, upload_suffix='', file_mode=None):
0.17.9 by Robert Collins
Initial stab at repository format support.
62
        """Create a NewPack instance.
63
0.20.30 by Ian Clatworthy
repofmt.py code cleanups
64
        :param pack_collection: A PackCollection into which this is being
65
            inserted.
0.17.9 by Robert Collins
Initial stab at repository format support.
66
        :param upload_suffix: An optional suffix to be given to any temporary
67
            files created during the pack creation. e.g '.autopack'
68
        :param file_mode: An optional file mode to create the new files with.
69
        """
3735.31.5 by John Arbash Meinel
Move some of the monkey patching into the correct locations.
70
        # replaced from NewPack to:
0.17.25 by Robert Collins
Preliminary --gc-plain-chk support.
71
        # - change inventory reference list length to 1
72
        # - change texts reference lists to 1
3735.31.5 by John Arbash Meinel
Move some of the monkey patching into the correct locations.
73
        # TODO: patch this to be parameterised
3735.31.2 by John Arbash Meinel
Cleanup trailing whitespace, get test_source to pass by removing asserts.
74
0.17.9 by Robert Collins
Initial stab at repository format support.
75
        # The relative locations of the packs are constrained, but all are
76
        # passed in because the caller has them, so as to avoid object churn.
0.17.22 by Robert Collins
really get gc working with 1.10
77
        index_builder_class = pack_collection._index_builder_class
3735.31.4 by John Arbash Meinel
Continue trying to reduce the friction for the groupcompress repo formats.
78
        # from brisbane-core
79
        if pack_collection.chk_index is not None:
80
            chk_index = index_builder_class(reference_lists=0)
0.17.25 by Robert Collins
Preliminary --gc-plain-chk support.
81
        else:
3735.31.4 by John Arbash Meinel
Continue trying to reduce the friction for the groupcompress repo formats.
82
            chk_index = None
83
        Pack.__init__(self,
84
            # Revisions: parents list, no text compression.
85
            index_builder_class(reference_lists=1),
86
            # Inventory: We want to map compression only, but currently the
87
            # knit code hasn't been updated enough to understand that, so we
88
            # have a regular 2-list index giving parents and compression
89
            # source.
90
            index_builder_class(reference_lists=1),
4371.2.2 by Vincent Ladeuil
Fix comment as per John's review.
91
            # Texts: per file graph, for all fileids - so one reference list
92
            # and two elements in the key tuple.
3735.31.4 by John Arbash Meinel
Continue trying to reduce the friction for the groupcompress repo formats.
93
            index_builder_class(reference_lists=1, key_elements=2),
94
            # Signatures: Just blobs to store, no compression, no parents
95
            # listing.
96
            index_builder_class(reference_lists=0),
97
            # CHK based storage - just blobs, no compression or parents.
98
            chk_index=chk_index
99
            )
0.17.22 by Robert Collins
really get gc working with 1.10
100
        self._pack_collection = pack_collection
101
        # When we make readonly indices, we need this.
102
        self.index_class = pack_collection._index_class
0.17.9 by Robert Collins
Initial stab at repository format support.
103
        # where should the new pack be opened
0.17.22 by Robert Collins
really get gc working with 1.10
104
        self.upload_transport = pack_collection._upload_transport
0.17.9 by Robert Collins
Initial stab at repository format support.
105
        # where are indices written out to
0.17.22 by Robert Collins
really get gc working with 1.10
106
        self.index_transport = pack_collection._index_transport
0.17.9 by Robert Collins
Initial stab at repository format support.
107
        # where is the pack renamed to when it is finished?
0.17.22 by Robert Collins
really get gc working with 1.10
108
        self.pack_transport = pack_collection._pack_transport
0.17.9 by Robert Collins
Initial stab at repository format support.
109
        # What file mode to upload the pack and indices with.
110
        self._file_mode = file_mode
111
        # tracks the content written to the .pack file.
3735.31.2 by John Arbash Meinel
Cleanup trailing whitespace, get test_source to pass by removing asserts.
112
        self._hash = osutils.md5()
0.17.9 by Robert Collins
Initial stab at repository format support.
113
        # a four-tuple with the length in bytes of the indices, once the pack
114
        # is finalised. (rev, inv, text, sigs)
115
        self.index_sizes = None
116
        # How much data to cache when writing packs. Note that this is not
117
        # synchronised with reads, because it's not in the transport layer, so
118
        # is not safe unless the client knows it won't be reading from the pack
119
        # under creation.
120
        self._cache_limit = 0
121
        # the temporary pack file name.
3735.31.2 by John Arbash Meinel
Cleanup trailing whitespace, get test_source to pass by removing asserts.
122
        self.random_name = osutils.rand_chars(20) + upload_suffix
0.17.9 by Robert Collins
Initial stab at repository format support.
123
        # when was this pack started ?
124
        self.start_time = time.time()
125
        # open an output stream for the data added to the pack.
126
        self.write_stream = self.upload_transport.open_write_stream(
127
            self.random_name, mode=self._file_mode)
128
        if 'pack' in debug.debug_flags:
0.23.30 by John Arbash Meinel
Merge in Ian's groupcompress trunk updates
129
            trace.mutter('%s: create_pack: pack stream open: %s%s t+%6.3fs',
0.17.9 by Robert Collins
Initial stab at repository format support.
130
                time.ctime(), self.upload_transport.base, self.random_name,
131
                time.time() - self.start_time)
3735.31.2 by John Arbash Meinel
Cleanup trailing whitespace, get test_source to pass by removing asserts.
132
        # A list of byte sequences to be written to the new pack, and the
133
        # aggregate size of them.  Stored as a list rather than separate
0.17.9 by Robert Collins
Initial stab at repository format support.
134
        # variables so that the _write_data closure below can update them.
135
        self._buffer = [[], 0]
3735.31.2 by John Arbash Meinel
Cleanup trailing whitespace, get test_source to pass by removing asserts.
136
        # create a callable for adding data
0.17.9 by Robert Collins
Initial stab at repository format support.
137
        #
138
        # robertc says- this is a closure rather than a method on the object
139
        # so that the variables are locals, and faster than accessing object
140
        # members.
141
        def _write_data(bytes, flush=False, _buffer=self._buffer,
142
            _write=self.write_stream.write, _update=self._hash.update):
143
            _buffer[0].append(bytes)
144
            _buffer[1] += len(bytes)
145
            # buffer cap
146
            if _buffer[1] > self._cache_limit or flush:
147
                bytes = ''.join(_buffer[0])
148
                _write(bytes)
149
                _update(bytes)
150
                _buffer[:] = [[], 0]
151
        # expose this on self, for the occasion when clients want to add data.
152
        self._write_data = _write_data
153
        # a pack writer object to serialise pack records.
154
        self._writer = pack.ContainerWriter(self._write_data)
155
        self._writer.begin()
156
        # what state is the pack in? (open, finished, aborted)
157
        self._state = 'open'
158
3735.31.17 by John Arbash Meinel
Properly support suspend/resume by disallowing them :).
159
    def _check_references(self):
160
        """Make sure our external references are present.
161
162
        Packs are allowed to have deltas whose base is not in the pack, but it
163
        must be present somewhere in this collection.  It is not allowed to
164
        have deltas based on a fallback repository.
165
        (See <https://bugs.launchpad.net/bzr/+bug/288751>)
166
        """
4343.3.1 by John Arbash Meinel
Set 'supports_external_lookups=True' for dev6 repositories.
167
        # Groupcompress packs don't have any external references, arguably CHK
168
        # pages have external references, but we cannot 'cheaply' determine
169
        # them without actually walking all of the chk pages.
3735.31.17 by John Arbash Meinel
Properly support suspend/resume by disallowing them :).
170
0.17.9 by Robert Collins
Initial stab at repository format support.
171
4343.3.2 by John Arbash Meinel
All stacking tests seem to be passing for dev6 repos
172
class ResumedGCPack(ResumedPack):
173
4343.3.8 by John Arbash Meinel
Some cleanup passes.
174
    def _check_references(self):
175
        """Make sure our external compression parents are present."""
176
        # See GCPack._check_references for why this is empty
177
4343.3.2 by John Arbash Meinel
All stacking tests seem to be passing for dev6 repos
178
    def _get_external_refs(self, index):
179
        # GC repositories don't have compression parents external to a given
180
        # pack file
181
        return set()
182
183
3735.31.27 by John Arbash Meinel
Implement GCCHKReconcilePacker.
184
class GCCHKPacker(Packer):
3735.31.22 by John Arbash Meinel
Factor out the copying code into GCPacker.
185
    """This class understand what it takes to collect a GCCHK repo."""
0.17.9 by Robert Collins
Initial stab at repository format support.
186
3735.31.24 by John Arbash Meinel
Start moving more things deeper inside the code.
187
    def __init__(self, pack_collection, packs, suffix, revision_ids=None,
188
                 reload_func=None):
3735.31.27 by John Arbash Meinel
Implement GCCHKReconcilePacker.
189
        super(GCCHKPacker, self).__init__(pack_collection, packs, suffix,
190
                                          revision_ids=revision_ids,
191
                                          reload_func=reload_func)
3735.31.22 by John Arbash Meinel
Factor out the copying code into GCPacker.
192
        self._pack_collection = pack_collection
193
        # ATM, We only support this for GCCHK repositories
4241.6.8 by Robert Collins, John Arbash Meinel, Ian Clatworthy, Vincent Ladeuil
Add --development6-rich-root, disabling the legacy and unneeded development2 format, and activating the tests for CHK features disabled pending this format. (Robert Collins, John Arbash Meinel, Ian Clatworthy, Vincent Ladeuil)
194
        if pack_collection.chk_index is None:
195
            raise AssertionError('pack_collection.chk_index should not be None')
3735.31.27 by John Arbash Meinel
Implement GCCHKReconcilePacker.
196
        self._gather_text_refs = False
3735.31.22 by John Arbash Meinel
Factor out the copying code into GCPacker.
197
        self._chk_id_roots = []
198
        self._chk_p_id_roots = []
3735.31.27 by John Arbash Meinel
Implement GCCHKReconcilePacker.
199
        self._text_refs = None
3735.31.25 by John Arbash Meinel
Finish refactoring everything into _copy_X_texts() style.
200
        # set by .pack() if self.revision_ids is not None
201
        self.revision_keys = None
0.17.9 by Robert Collins
Initial stab at repository format support.
202
3735.31.25 by John Arbash Meinel
Finish refactoring everything into _copy_X_texts() style.
203
    def _get_progress_stream(self, source_vf, keys, message, pb):
3735.31.15 by John Arbash Meinel
A bit more refactoring, try to clean up the _execute_packs code.
204
        def pb_stream():
205
            substream = source_vf.get_record_stream(keys, 'groupcompress', True)
206
            for idx, record in enumerate(substream):
207
                if pb is not None:
3735.31.25 by John Arbash Meinel
Finish refactoring everything into _copy_X_texts() style.
208
                    pb.update(message, idx + 1, len(keys))
3735.31.15 by John Arbash Meinel
A bit more refactoring, try to clean up the _execute_packs code.
209
                yield record
210
        return pb_stream()
211
3735.31.25 by John Arbash Meinel
Finish refactoring everything into _copy_X_texts() style.
212
    def _get_filtered_inv_stream(self, source_vf, keys, message, pb=None):
0.22.5 by John Arbash Meinel
Try a different method of streaming the chk pages.
213
        """Filter the texts of inventories, to find the chk pages."""
0.17.43 by John Arbash Meinel
Give better chk and inventory progress indication during pack/autopack
214
        total_keys = len(keys)
3735.31.22 by John Arbash Meinel
Factor out the copying code into GCPacker.
215
        def _filtered_inv_stream():
216
            id_roots_set = set()
217
            p_id_roots_set = set()
218
            stream = source_vf.get_record_stream(keys, 'groupcompress', True)
0.22.5 by John Arbash Meinel
Try a different method of streaming the chk pages.
219
            for idx, record in enumerate(stream):
220
                bytes = record.get_bytes_as('fulltext')
3735.31.22 by John Arbash Meinel
Factor out the copying code into GCPacker.
221
                chk_inv = inventory.CHKInventory.deserialise(None, bytes,
222
                                                             record.key)
0.17.43 by John Arbash Meinel
Give better chk and inventory progress indication during pack/autopack
223
                if pb is not None:
224
                    pb.update('inv', idx, total_keys)
0.22.6 by John Arbash Meinel
Clustering chk pages properly makes a big difference.
225
                key = chk_inv.id_to_entry.key()
226
                if key not in id_roots_set:
3735.31.22 by John Arbash Meinel
Factor out the copying code into GCPacker.
227
                    self._chk_id_roots.append(key)
0.22.6 by John Arbash Meinel
Clustering chk pages properly makes a big difference.
228
                    id_roots_set.add(key)
0.22.5 by John Arbash Meinel
Try a different method of streaming the chk pages.
229
                p_id_map = chk_inv.parent_id_basename_to_file_id
4241.6.8 by Robert Collins, John Arbash Meinel, Ian Clatworthy, Vincent Ladeuil
Add --development6-rich-root, disabling the legacy and unneeded development2 format, and activating the tests for CHK features disabled pending this format. (Robert Collins, John Arbash Meinel, Ian Clatworthy, Vincent Ladeuil)
230
                if p_id_map is None:
231
                    raise AssertionError('Parent id -> file_id map not set')
3735.31.22 by John Arbash Meinel
Factor out the copying code into GCPacker.
232
                key = p_id_map.key()
233
                if key not in p_id_roots_set:
234
                    p_id_roots_set.add(key)
235
                    self._chk_p_id_roots.append(key)
0.22.5 by John Arbash Meinel
Try a different method of streaming the chk pages.
236
                yield record
3735.31.22 by John Arbash Meinel
Factor out the copying code into GCPacker.
237
            # We have finished processing all of the inventory records, we
238
            # don't need these sets anymore
239
            id_roots_set.clear()
240
            p_id_roots_set.clear()
3735.42.3 by John Arbash Meinel
Fix a bug introduced into GCPacker.
241
        return _filtered_inv_stream()
0.22.5 by John Arbash Meinel
Try a different method of streaming the chk pages.
242
3735.31.22 by John Arbash Meinel
Factor out the copying code into GCPacker.
243
    def _get_chk_streams(self, source_vf, keys, pb=None):
0.22.5 by John Arbash Meinel
Try a different method of streaming the chk pages.
244
        # We want to stream the keys from 'id_roots', and things they
245
        # reference, and then stream things from p_id_roots and things they
246
        # reference, and then any remaining keys that we didn't get to.
247
0.22.6 by John Arbash Meinel
Clustering chk pages properly makes a big difference.
248
        # We also group referenced texts together, so if one root references a
249
        # text with prefix 'a', and another root references a node with prefix
250
        # 'a', we want to yield those nodes before we yield the nodes for 'b'
0.20.30 by Ian Clatworthy
repofmt.py code cleanups
251
        # This keeps 'similar' nodes together.
0.22.6 by John Arbash Meinel
Clustering chk pages properly makes a big difference.
252
0.22.5 by John Arbash Meinel
Try a different method of streaming the chk pages.
253
        # Note: We probably actually want multiple streams here, to help the
254
        #       client understand that the different levels won't compress well
0.20.30 by Ian Clatworthy
repofmt.py code cleanups
255
        #       against each other.
0.20.27 by John Arbash Meinel
Update a Note/Todo
256
        #       Test the difference between using one Group per level, and
257
        #       using 1 Group per prefix. (so '' (root) would get a group, then
258
        #       all the references to search-key 'a' would get a group, etc.)
0.17.43 by John Arbash Meinel
Give better chk and inventory progress indication during pack/autopack
259
        total_keys = len(keys)
0.22.5 by John Arbash Meinel
Try a different method of streaming the chk pages.
260
        remaining_keys = set(keys)
0.20.25 by John Arbash Meinel
As expected, splitting things up into streams of streams
261
        counter = [0]
3735.31.27 by John Arbash Meinel
Implement GCCHKReconcilePacker.
262
        if self._gather_text_refs:
3735.36.12 by John Arbash Meinel
Add some direct tests for CHKInventory._entry_to_bytes
263
            bytes_to_info = inventory.CHKInventory._bytes_to_utf8name_key
3735.31.27 by John Arbash Meinel
Implement GCCHKReconcilePacker.
264
            self._text_refs = set()
265
        def _get_referenced_stream(root_keys, parse_leaf_nodes=False):
0.22.5 by John Arbash Meinel
Try a different method of streaming the chk pages.
266
            cur_keys = root_keys
267
            while cur_keys:
0.22.6 by John Arbash Meinel
Clustering chk pages properly makes a big difference.
268
                keys_by_search_prefix = {}
0.22.5 by John Arbash Meinel
Try a different method of streaming the chk pages.
269
                remaining_keys.difference_update(cur_keys)
270
                next_keys = set()
3735.31.26 by John Arbash Meinel
A bit more internal refactoring.
271
                def handle_internal_node(node):
272
                    for prefix, value in node._items.iteritems():
3735.2.147 by John Arbash Meinel
Track down and fix an unexpected edge case.
273
                        # We don't want to request the same key twice, and we
274
                        # want to order it by the first time it is seen.
275
                        # Even further, we don't want to request a key which is
276
                        # not in this group of pack files (it should be in the
277
                        # repo, but it doesn't have to be in the group being
278
                        # packed.)
279
                        # TODO: consider how to treat externally referenced chk
280
                        #       pages as 'external_references' so that we
281
                        #       always fill them in for stacked branches
282
                        if value not in next_keys and value in remaining_keys:
3735.31.26 by John Arbash Meinel
A bit more internal refactoring.
283
                            keys_by_search_prefix.setdefault(prefix,
284
                                []).append(value)
285
                            next_keys.add(value)
3735.31.27 by John Arbash Meinel
Implement GCCHKReconcilePacker.
286
                def handle_leaf_node(node):
287
                    # Store is None, because we know we have a LeafNode, and we
288
                    # just want its entries
289
                    for file_id, bytes in node.iteritems(None):
3735.36.12 by John Arbash Meinel
Add some direct tests for CHKInventory._entry_to_bytes
290
                        name_utf8, file_id, revision_id = bytes_to_info(bytes)
3735.36.16 by John Arbash Meinel
Handle a typo in 'handle_leaf_node' during Reconcile tests.
291
                        self._text_refs.add((file_id, revision_id))
0.20.26 by John Arbash Meinel
Try even harder, now with even *more* streams.
292
                def next_stream():
3735.2.147 by John Arbash Meinel
Track down and fix an unexpected edge case.
293
                    stream = source_vf.get_record_stream(cur_keys,
294
                                                         'as-requested', True)
0.20.26 by John Arbash Meinel
Try even harder, now with even *more* streams.
295
                    for record in stream:
296
                        bytes = record.get_bytes_as('fulltext')
297
                        # We don't care about search_key_func for this code,
298
                        # because we only care about external references.
299
                        node = chk_map._deserialise(bytes, record.key,
300
                                                    search_key_func=None)
301
                        common_base = node._search_prefix
302
                        if isinstance(node, chk_map.InternalNode):
3735.31.26 by John Arbash Meinel
A bit more internal refactoring.
303
                            handle_internal_node(node)
3735.31.27 by John Arbash Meinel
Implement GCCHKReconcilePacker.
304
                        elif parse_leaf_nodes:
305
                            handle_leaf_node(node)
0.20.26 by John Arbash Meinel
Try even harder, now with even *more* streams.
306
                        counter[0] += 1
307
                        if pb is not None:
0.17.43 by John Arbash Meinel
Give better chk and inventory progress indication during pack/autopack
308
                            pb.update('chk node', counter[0], total_keys)
0.20.26 by John Arbash Meinel
Try even harder, now with even *more* streams.
309
                        yield record
310
                yield next_stream()
0.22.6 by John Arbash Meinel
Clustering chk pages properly makes a big difference.
311
                # Double check that we won't be emitting any keys twice
3735.31.26 by John Arbash Meinel
A bit more internal refactoring.
312
                # If we get rid of the pre-calculation of all keys, we could
313
                # turn this around and do
314
                # next_keys.difference_update(seen_keys)
3735.2.147 by John Arbash Meinel
Track down and fix an unexpected edge case.
315
                # However, we also may have references to chk pages in another
316
                # pack file during autopack. We filter earlier, so we should no
317
                # longer need to do this
318
                # next_keys = next_keys.intersection(remaining_keys)
0.22.6 by John Arbash Meinel
Clustering chk pages properly makes a big difference.
319
                cur_keys = []
320
                for prefix in sorted(keys_by_search_prefix):
3735.34.1 by John Arbash Meinel
Some testing to see if we can decrease the peak memory consumption a bit.
321
                    cur_keys.extend(keys_by_search_prefix.pop(prefix))
3735.31.27 by John Arbash Meinel
Implement GCCHKReconcilePacker.
322
        for stream in _get_referenced_stream(self._chk_id_roots,
323
                                             self._gather_text_refs):
0.20.26 by John Arbash Meinel
Try even harder, now with even *more* streams.
324
            yield stream
3735.31.25 by John Arbash Meinel
Finish refactoring everything into _copy_X_texts() style.
325
        del self._chk_id_roots
3735.2.147 by John Arbash Meinel
Track down and fix an unexpected edge case.
326
        # while it isn't really possible for chk_id_roots to not be in the
327
        # local group of packs, it is possible that the tree shape has not
328
        # changed recently, so we need to filter _chk_p_id_roots by the
329
        # available keys
330
        chk_p_id_roots = [key for key in self._chk_p_id_roots
331
                          if key in remaining_keys]
3735.31.25 by John Arbash Meinel
Finish refactoring everything into _copy_X_texts() style.
332
        del self._chk_p_id_roots
3735.2.147 by John Arbash Meinel
Track down and fix an unexpected edge case.
333
        for stream in _get_referenced_stream(chk_p_id_roots, False):
334
            yield stream
0.22.5 by John Arbash Meinel
Try a different method of streaming the chk pages.
335
        if remaining_keys:
3735.31.25 by John Arbash Meinel
Finish refactoring everything into _copy_X_texts() style.
336
            trace.mutter('There were %d keys in the chk index, %d of which'
337
                         ' were not referenced', total_keys,
338
                         len(remaining_keys))
339
            if self.revision_ids is None:
340
                stream = source_vf.get_record_stream(remaining_keys,
341
                                                     'unordered', True)
342
                yield stream
0.22.5 by John Arbash Meinel
Try a different method of streaming the chk pages.
343
3735.31.22 by John Arbash Meinel
Factor out the copying code into GCPacker.
344
    def _build_vf(self, index_name, parents, delta, for_write=False):
3735.31.15 by John Arbash Meinel
A bit more refactoring, try to clean up the _execute_packs code.
345
        """Build a VersionedFiles instance on top of this group of packs."""
346
        index_name = index_name + '_index'
347
        index_to_pack = {}
348
        access = knit._DirectPackAccess(index_to_pack)
349
        if for_write:
3735.31.22 by John Arbash Meinel
Factor out the copying code into GCPacker.
350
            # Use new_pack
4241.6.8 by Robert Collins, John Arbash Meinel, Ian Clatworthy, Vincent Ladeuil
Add --development6-rich-root, disabling the legacy and unneeded development2 format, and activating the tests for CHK features disabled pending this format. (Robert Collins, John Arbash Meinel, Ian Clatworthy, Vincent Ladeuil)
351
            if self.new_pack is None:
352
                raise AssertionError('No new pack has been set')
3735.31.23 by John Arbash Meinel
Change the GCPacker to inherit from Packer now that there isn't as much difference.
353
            index = getattr(self.new_pack, index_name)
354
            index_to_pack[index] = self.new_pack.access_tuple()
3735.31.15 by John Arbash Meinel
A bit more refactoring, try to clean up the _execute_packs code.
355
            index.set_optimize(for_size=True)
3735.31.23 by John Arbash Meinel
Change the GCPacker to inherit from Packer now that there isn't as much difference.
356
            access.set_writer(self.new_pack._writer, index,
357
                              self.new_pack.access_tuple())
3735.31.15 by John Arbash Meinel
A bit more refactoring, try to clean up the _execute_packs code.
358
            add_callback = index.add_nodes
359
        else:
360
            indices = []
3735.31.23 by John Arbash Meinel
Change the GCPacker to inherit from Packer now that there isn't as much difference.
361
            for pack in self.packs:
3735.31.15 by John Arbash Meinel
A bit more refactoring, try to clean up the _execute_packs code.
362
                sub_index = getattr(pack, index_name)
363
                index_to_pack[sub_index] = pack.access_tuple()
364
                indices.append(sub_index)
365
            index = _mod_index.CombinedGraphIndex(indices)
366
            add_callback = None
367
        vf = GroupCompressVersionedFiles(
368
            _GCGraphIndex(index,
369
                          add_callback=add_callback,
370
                          parents=parents,
3735.31.22 by John Arbash Meinel
Factor out the copying code into GCPacker.
371
                          is_locked=self._pack_collection.repo.is_locked),
3735.31.15 by John Arbash Meinel
A bit more refactoring, try to clean up the _execute_packs code.
372
            access=access,
373
            delta=delta)
374
        return vf
375
3735.31.24 by John Arbash Meinel
Start moving more things deeper inside the code.
376
    def _build_vfs(self, index_name, parents, delta):
377
        """Build the source and target VersionedFiles."""
378
        source_vf = self._build_vf(index_name, parents,
379
                                   delta, for_write=False)
380
        target_vf = self._build_vf(index_name, parents,
381
                                   delta, for_write=True)
382
        return source_vf, target_vf
383
3735.31.25 by John Arbash Meinel
Finish refactoring everything into _copy_X_texts() style.
384
    def _copy_stream(self, source_vf, target_vf, keys, message, vf_to_stream,
385
                     pb_offset):
386
        trace.mutter('repacking %d %s', len(keys), message)
3735.2.146 by John Arbash Meinel
Give the right progress message.
387
        self.pb.update('repacking %s' % (message,), pb_offset)
3735.31.25 by John Arbash Meinel
Finish refactoring everything into _copy_X_texts() style.
388
        child_pb = ui.ui_factory.nested_progress_bar()
389
        try:
390
            stream = vf_to_stream(source_vf, keys, message, child_pb)
3735.32.22 by John Arbash Meinel
Change the code so that the GCPacker knows to set reuse_blocks=False.
391
            for _ in target_vf._insert_record_stream(stream,
3735.32.28 by John Arbash Meinel
We can use 'random_id=True' when copying the streams.
392
                                                     random_id=True,
3735.32.22 by John Arbash Meinel
Change the code so that the GCPacker knows to set reuse_blocks=False.
393
                                                     reuse_blocks=False):
394
                pass
3735.31.25 by John Arbash Meinel
Finish refactoring everything into _copy_X_texts() style.
395
        finally:
396
            child_pb.finished()
397
398
    def _copy_revision_texts(self):
399
        source_vf, target_vf = self._build_vfs('revision', True, False)
400
        if not self.revision_keys:
401
            # We are doing a full fetch, aka 'pack'
402
            self.revision_keys = source_vf.keys()
403
        self._copy_stream(source_vf, target_vf, self.revision_keys,
404
                          'revisions', self._get_progress_stream, 1)
405
406
    def _copy_inventory_texts(self):
407
        source_vf, target_vf = self._build_vfs('inventory', True, True)
408
        self._copy_stream(source_vf, target_vf, self.revision_keys,
3735.31.27 by John Arbash Meinel
Implement GCCHKReconcilePacker.
409
                          'inventories', self._get_filtered_inv_stream, 2)
3735.31.25 by John Arbash Meinel
Finish refactoring everything into _copy_X_texts() style.
410
411
    def _copy_chk_texts(self):
412
        source_vf, target_vf = self._build_vfs('chk', False, False)
413
        # TODO: This is technically spurious... if it is a performance issue,
414
        #       remove it
415
        total_keys = source_vf.keys()
416
        trace.mutter('repacking chk: %d id_to_entry roots,'
417
                     ' %d p_id_map roots, %d total keys',
418
                     len(self._chk_id_roots), len(self._chk_p_id_roots),
419
                     len(total_keys))
420
        self.pb.update('repacking chk', 3)
421
        child_pb = ui.ui_factory.nested_progress_bar()
422
        try:
423
            for stream in self._get_chk_streams(source_vf, total_keys,
424
                                                pb=child_pb):
3735.32.22 by John Arbash Meinel
Change the code so that the GCPacker knows to set reuse_blocks=False.
425
                for _ in target_vf._insert_record_stream(stream,
3735.32.28 by John Arbash Meinel
We can use 'random_id=True' when copying the streams.
426
                                                         random_id=True,
3735.32.22 by John Arbash Meinel
Change the code so that the GCPacker knows to set reuse_blocks=False.
427
                                                         reuse_blocks=False):
428
                    pass
3735.31.25 by John Arbash Meinel
Finish refactoring everything into _copy_X_texts() style.
429
        finally:
430
            child_pb.finished()
431
432
    def _copy_text_texts(self):
433
        source_vf, target_vf = self._build_vfs('text', True, True)
434
        # XXX: We don't walk the chk map to determine referenced (file_id,
435
        #      revision_id) keys.  We don't do it yet because you really need
436
        #      to filter out the ones that are present in the parents of the
437
        #      rev just before the ones you are copying, otherwise the filter
438
        #      is grabbing too many keys...
439
        text_keys = source_vf.keys()
440
        self._copy_stream(source_vf, target_vf, text_keys,
3735.31.27 by John Arbash Meinel
Implement GCCHKReconcilePacker.
441
                          'text', self._get_progress_stream, 4)
3735.31.25 by John Arbash Meinel
Finish refactoring everything into _copy_X_texts() style.
442
443
    def _copy_signature_texts(self):
444
        source_vf, target_vf = self._build_vfs('signature', False, False)
445
        signature_keys = source_vf.keys()
446
        signature_keys.intersection(self.revision_keys)
447
        self._copy_stream(source_vf, target_vf, signature_keys,
448
                          'signatures', self._get_progress_stream, 5)
449
3735.31.23 by John Arbash Meinel
Change the GCPacker to inherit from Packer now that there isn't as much difference.
450
    def _create_pack_from_packs(self):
3735.31.25 by John Arbash Meinel
Finish refactoring everything into _copy_X_texts() style.
451
        self.pb.update('repacking', 0, 7)
3735.31.23 by John Arbash Meinel
Change the GCPacker to inherit from Packer now that there isn't as much difference.
452
        self.new_pack = self.open_pack()
453
        # Is this necessary for GC ?
454
        self.new_pack.set_write_cache_size(1024*1024)
3735.31.25 by John Arbash Meinel
Finish refactoring everything into _copy_X_texts() style.
455
        self._copy_revision_texts()
456
        self._copy_inventory_texts()
457
        self._copy_chk_texts()
458
        self._copy_text_texts()
459
        self._copy_signature_texts()
3735.31.23 by John Arbash Meinel
Change the GCPacker to inherit from Packer now that there isn't as much difference.
460
        self.new_pack._check_references()
461
        if not self._use_pack(self.new_pack):
462
            self.new_pack.abort()
463
            return None
3735.31.25 by John Arbash Meinel
Finish refactoring everything into _copy_X_texts() style.
464
        self.pb.update('finishing repack', 6, 7)
3735.31.23 by John Arbash Meinel
Change the GCPacker to inherit from Packer now that there isn't as much difference.
465
        self.new_pack.finish()
466
        self._pack_collection.allocate(self.new_pack)
467
        return self.new_pack
3735.31.15 by John Arbash Meinel
A bit more refactoring, try to clean up the _execute_packs code.
468
3735.31.22 by John Arbash Meinel
Factor out the copying code into GCPacker.
469
3735.31.27 by John Arbash Meinel
Implement GCCHKReconcilePacker.
470
class GCCHKReconcilePacker(GCCHKPacker):
471
    """A packer which regenerates indices etc as it copies.
472
473
    This is used by ``bzr reconcile`` to cause parent text pointers to be
474
    regenerated.
475
    """
476
477
    def __init__(self, *args, **kwargs):
478
        super(GCCHKReconcilePacker, self).__init__(*args, **kwargs)
479
        self._data_changed = False
480
        self._gather_text_refs = True
481
482
    def _copy_inventory_texts(self):
483
        source_vf, target_vf = self._build_vfs('inventory', True, True)
484
        self._copy_stream(source_vf, target_vf, self.revision_keys,
485
                          'inventories', self._get_filtered_inv_stream, 2)
486
        if source_vf.keys() != self.revision_keys:
487
            self._data_changed = True
488
489
    def _copy_text_texts(self):
490
        """generate what texts we should have and then copy."""
491
        source_vf, target_vf = self._build_vfs('text', True, True)
492
        trace.mutter('repacking %d texts', len(self._text_refs))
493
        self.pb.update("repacking texts", 4)
494
        # we have three major tasks here:
495
        # 1) generate the ideal index
496
        repo = self._pack_collection.repo
497
        # We want the one we just wrote, so base it on self.new_pack
498
        revision_vf = self._build_vf('revision', True, False, for_write=True)
499
        ancestor_keys = revision_vf.get_parent_map(revision_vf.keys())
500
        # Strip keys back into revision_ids.
501
        ancestors = dict((k[0], tuple([p[0] for p in parents]))
502
                         for k, parents in ancestor_keys.iteritems())
503
        del ancestor_keys
504
        # TODO: _generate_text_key_index should be much cheaper to generate from
505
        #       a chk repository, rather than the current implementation
506
        ideal_index = repo._generate_text_key_index(None, ancestors)
507
        file_id_parent_map = source_vf.get_parent_map(self._text_refs)
508
        # 2) generate a keys list that contains all the entries that can
509
        #    be used as-is, with corrected parents.
510
        ok_keys = []
511
        new_parent_keys = {} # (key, parent_keys)
512
        discarded_keys = []
513
        NULL_REVISION = _mod_revision.NULL_REVISION
514
        for key in self._text_refs:
515
            # 0 - index
516
            # 1 - key
517
            # 2 - value
518
            # 3 - refs
519
            try:
520
                ideal_parents = tuple(ideal_index[key])
521
            except KeyError:
522
                discarded_keys.append(key)
523
                self._data_changed = True
524
            else:
525
                if ideal_parents == (NULL_REVISION,):
526
                    ideal_parents = ()
527
                source_parents = file_id_parent_map[key]
528
                if ideal_parents == source_parents:
529
                    # no change needed.
530
                    ok_keys.append(key)
531
                else:
532
                    # We need to change the parent graph, but we don't need to
533
                    # re-insert the text (since we don't pun the compression
534
                    # parent with the parents list)
535
                    self._data_changed = True
536
                    new_parent_keys[key] = ideal_parents
537
        # we're finished with some data.
538
        del ideal_index
539
        del file_id_parent_map
540
        # 3) bulk copy the data, updating records than need it
541
        def _update_parents_for_texts():
542
            stream = source_vf.get_record_stream(self._text_refs,
543
                'groupcompress', False)
544
            for record in stream:
545
                if record.key in new_parent_keys:
546
                    record.parents = new_parent_keys[record.key]
547
                yield record
548
        target_vf.insert_record_stream(_update_parents_for_texts())
549
550
    def _use_pack(self, new_pack):
551
        """Override _use_pack to check for reconcile having changed content."""
552
        return new_pack.data_inserted() and self._data_changed
553
554
3735.31.22 by John Arbash Meinel
Factor out the copying code into GCPacker.
555
class GCRepositoryPackCollection(RepositoryPackCollection):
556
557
    pack_factory = GCPack
4343.3.2 by John Arbash Meinel
All stacking tests seem to be passing for dev6 repos
558
    resumed_pack_factory = ResumedGCPack
3735.31.22 by John Arbash Meinel
Factor out the copying code into GCPacker.
559
3735.2.150 by Ian Clatworthy
always repack gc repositories for now, even if only one pack there
560
    def _already_packed(self):
561
        """Is the collection already packed?"""
562
        # Always repack GC repositories for now
563
        return False
564
3735.31.27 by John Arbash Meinel
Implement GCCHKReconcilePacker.
565
    def _execute_pack_operations(self, pack_operations,
566
                                 _packer_class=GCCHKPacker,
0.22.1 by John Arbash Meinel
A first-cut at implementing an auto-pack by copying everything.
567
                                 reload_func=None):
568
        """Execute a series of pack operations.
569
570
        :param pack_operations: A list of [revision_count, packs_to_combine].
571
        :param _packer_class: The class of packer to use (default: Packer).
572
        :return: None.
573
        """
3735.31.23 by John Arbash Meinel
Change the GCPacker to inherit from Packer now that there isn't as much difference.
574
        # XXX: Copied across from RepositoryPackCollection simply because we
575
        #      want to override the _packer_class ... :(
0.22.1 by John Arbash Meinel
A first-cut at implementing an auto-pack by copying everything.
576
        for revision_count, packs in pack_operations:
577
            # we may have no-ops from the setup logic
578
            if len(packs) == 0:
579
                continue
3735.31.27 by John Arbash Meinel
Implement GCCHKReconcilePacker.
580
            packer = GCCHKPacker(self, packs, '.autopack',
581
                                 reload_func=reload_func)
0.22.1 by John Arbash Meinel
A first-cut at implementing an auto-pack by copying everything.
582
            try:
3735.31.23 by John Arbash Meinel
Change the GCPacker to inherit from Packer now that there isn't as much difference.
583
                packer.pack()
584
            except errors.RetryWithNewPacks:
585
                # An exception is propagating out of this context, make sure
586
                # this packer has cleaned up. Packer() doesn't set its new_pack
587
                # state into the RepositoryPackCollection object, so we only
588
                # have access to it directly here.
589
                if packer.new_pack is not None:
590
                    packer.new_pack.abort()
0.20.19 by John Arbash Meinel
Groupcompress now supports 'autopack' and 'pack'.
591
                raise
0.22.1 by John Arbash Meinel
A first-cut at implementing an auto-pack by copying everything.
592
            for pack in packs:
593
                self._remove_pack_from_memory(pack)
594
        # record the newly available packs and stop advertising the old
595
        # packs
596
        self._save_pack_names(clear_obsolete_packs=True)
597
        # Move the old packs out of the way now they are no longer referenced.
598
        for revision_count, packs in pack_operations:
599
            self._obsolete_packs(packs)
0.20.7 by John Arbash Meinel
(ugly hack) autopacking doesn't work, so don't do it.
600
0.17.9 by Robert Collins
Initial stab at repository format support.
601
4241.6.8 by Robert Collins, John Arbash Meinel, Ian Clatworthy, Vincent Ladeuil
Add --development6-rich-root, disabling the legacy and unneeded development2 format, and activating the tests for CHK features disabled pending this format. (Robert Collins, John Arbash Meinel, Ian Clatworthy, Vincent Ladeuil)
602
class CHKInventoryRepository(KnitPackRepository):
603
    """subclass of KnitPackRepository that uses CHK based inventories."""
0.17.26 by Robert Collins
Working better --gc-plain-chk.
604
3735.31.4 by John Arbash Meinel
Continue trying to reduce the friction for the groupcompress repo formats.
605
    def __init__(self, _format, a_bzrdir, control_files, _commit_builder_class,
606
        _serializer):
607
        """Overridden to change pack collection class."""
608
        KnitPackRepository.__init__(self, _format, a_bzrdir, control_files,
609
            _commit_builder_class, _serializer)
610
        # and now replace everything it did :)
611
        index_transport = self._transport.clone('indices')
612
        self._pack_collection = GCRepositoryPackCollection(self,
613
            self._transport, index_transport,
614
            self._transport.clone('upload'),
615
            self._transport.clone('packs'),
616
            _format.index_builder_class,
617
            _format.index_class,
618
            use_chk_index=self._format.supports_chks,
619
            )
620
        self.inventories = GroupCompressVersionedFiles(
621
            _GCGraphIndex(self._pack_collection.inventory_index.combined_index,
622
                add_callback=self._pack_collection.inventory_index.add_callback,
623
                parents=True, is_locked=self.is_locked),
624
            access=self._pack_collection.inventory_index.data_access)
625
        self.revisions = GroupCompressVersionedFiles(
626
            _GCGraphIndex(self._pack_collection.revision_index.combined_index,
627
                add_callback=self._pack_collection.revision_index.add_callback,
4343.3.21 by John Arbash Meinel
Implement get_missing_parents in terms of _KeyRefs.
628
                parents=True, is_locked=self.is_locked,
629
                track_external_parent_refs=True),
3735.31.4 by John Arbash Meinel
Continue trying to reduce the friction for the groupcompress repo formats.
630
            access=self._pack_collection.revision_index.data_access,
631
            delta=False)
632
        self.signatures = GroupCompressVersionedFiles(
633
            _GCGraphIndex(self._pack_collection.signature_index.combined_index,
634
                add_callback=self._pack_collection.signature_index.add_callback,
635
                parents=False, is_locked=self.is_locked),
636
            access=self._pack_collection.signature_index.data_access,
637
            delta=False)
638
        self.texts = GroupCompressVersionedFiles(
639
            _GCGraphIndex(self._pack_collection.text_index.combined_index,
640
                add_callback=self._pack_collection.text_index.add_callback,
641
                parents=True, is_locked=self.is_locked),
642
            access=self._pack_collection.text_index.data_access)
643
        # No parents, individual CHK pages don't have specific ancestry
644
        self.chk_bytes = GroupCompressVersionedFiles(
645
            _GCGraphIndex(self._pack_collection.chk_index.combined_index,
646
                add_callback=self._pack_collection.chk_index.add_callback,
647
                parents=False, is_locked=self.is_locked),
648
            access=self._pack_collection.chk_index.data_access)
649
        # True when the repository object is 'write locked' (as opposed to the
650
        # physical lock only taken out around changes to the pack-names list.)
651
        # Another way to represent this would be a decorator around the control
652
        # files object that presents logical locks as physical ones - if this
653
        # gets ugly consider that alternative design. RBC 20071011
654
        self._write_lock_count = 0
655
        self._transaction = None
656
        # for tests
657
        self._reconcile_does_inventory_gc = True
658
        self._reconcile_fixes_text_parents = True
659
        self._reconcile_backsup_inventory = False
0.17.26 by Robert Collins
Working better --gc-plain-chk.
660
4241.6.8 by Robert Collins, John Arbash Meinel, Ian Clatworthy, Vincent Ladeuil
Add --development6-rich-root, disabling the legacy and unneeded development2 format, and activating the tests for CHK features disabled pending this format. (Robert Collins, John Arbash Meinel, Ian Clatworthy, Vincent Ladeuil)
661
    def _add_inventory_checked(self, revision_id, inv, parents):
662
        """Add inv to the repository after checking the inputs.
663
664
        This function can be overridden to allow different inventory styles.
665
666
        :seealso: add_inventory, for the contract.
667
        """
668
        # make inventory
669
        serializer = self._format._serializer
670
        result = inventory.CHKInventory.from_inventory(self.chk_bytes, inv,
671
            maximum_size=serializer.maximum_size,
672
            search_key_name=serializer.search_key_name)
673
        inv_lines = result.to_lines()
674
        return self._inventory_add_lines(revision_id, parents,
675
            inv_lines, check_content=False)
676
677
    def add_inventory_by_delta(self, basis_revision_id, delta, new_revision_id,
678
                               parents, basis_inv=None, propagate_caches=False):
679
        """Add a new inventory expressed as a delta against another revision.
680
681
        :param basis_revision_id: The inventory id the delta was created
682
            against.
683
        :param delta: The inventory delta (see Inventory.apply_delta for
684
            details).
685
        :param new_revision_id: The revision id that the inventory is being
686
            added for.
687
        :param parents: The revision ids of the parents that revision_id is
688
            known to have and are in the repository already. These are supplied
689
            for repositories that depend on the inventory graph for revision
690
            graph access, as well as for those that pun ancestry with delta
691
            compression.
692
        :param basis_inv: The basis inventory if it is already known,
693
            otherwise None.
694
        :param propagate_caches: If True, the caches for this inventory are
695
          copied to and updated for the result if possible.
696
697
        :returns: (validator, new_inv)
698
            The validator(which is a sha1 digest, though what is sha'd is
699
            repository format specific) of the serialized inventory, and the
700
            resulting inventory.
701
        """
702
        if basis_revision_id == _mod_revision.NULL_REVISION:
703
            return KnitPackRepository.add_inventory_by_delta(self,
704
                basis_revision_id, delta, new_revision_id, parents)
705
        if not self.is_in_write_group():
706
            raise AssertionError("%r not in write group" % (self,))
707
        _mod_revision.check_not_reserved_id(new_revision_id)
708
        basis_tree = self.revision_tree(basis_revision_id)
709
        basis_tree.lock_read()
710
        try:
711
            if basis_inv is None:
712
                basis_inv = basis_tree.inventory
713
            result = basis_inv.create_by_apply_delta(delta, new_revision_id,
714
                propagate_caches=propagate_caches)
715
            inv_lines = result.to_lines()
716
            return self._inventory_add_lines(new_revision_id, parents,
717
                inv_lines, check_content=False), result
718
        finally:
719
            basis_tree.unlock()
720
721
    def _iter_inventories(self, revision_ids):
722
        """Iterate over many inventory objects."""
723
        keys = [(revision_id,) for revision_id in revision_ids]
724
        stream = self.inventories.get_record_stream(keys, 'unordered', True)
725
        texts = {}
726
        for record in stream:
727
            if record.storage_kind != 'absent':
728
                texts[record.key] = record.get_bytes_as('fulltext')
729
            else:
730
                raise errors.NoSuchRevision(self, record.key)
731
        for key in keys:
732
            yield inventory.CHKInventory.deserialise(self.chk_bytes, texts[key], key)
733
734
    def _iter_inventory_xmls(self, revision_ids):
735
        # Without a native 'xml' inventory, this method doesn't make sense, so
736
        # make it raise to trap naughty direct users.
737
        raise NotImplementedError(self._iter_inventory_xmls)
738
4343.3.23 by John Arbash Meinel
Change RepositoryCHK1.fileids_altered_by to handle ghosts the way XML does.
739
    def _find_parent_ids_of_revisions(self, revision_ids):
740
        # TODO: we probably want to make this a helper that other code can get
741
        #       at
742
        parent_map = self.get_parent_map(revision_ids)
743
        parents = set()
744
        map(parents.update, parent_map.itervalues())
745
        parents.difference_update(revision_ids)
746
        parents.discard(_mod_revision.NULL_REVISION)
747
        return parents
4241.6.8 by Robert Collins, John Arbash Meinel, Ian Clatworthy, Vincent Ladeuil
Add --development6-rich-root, disabling the legacy and unneeded development2 format, and activating the tests for CHK features disabled pending this format. (Robert Collins, John Arbash Meinel, Ian Clatworthy, Vincent Ladeuil)
748
4343.3.23 by John Arbash Meinel
Change RepositoryCHK1.fileids_altered_by to handle ghosts the way XML does.
749
    def _find_present_inventory_ids(self, revision_ids):
750
        keys = [(r,) for r in revision_ids]
751
        parent_map = self.inventories.get_parent_map(keys)
752
        present_inventory_ids = set(k[-1] for k in parent_map)
753
        return present_inventory_ids
4241.6.8 by Robert Collins, John Arbash Meinel, Ian Clatworthy, Vincent Ladeuil
Add --development6-rich-root, disabling the legacy and unneeded development2 format, and activating the tests for CHK features disabled pending this format. (Robert Collins, John Arbash Meinel, Ian Clatworthy, Vincent Ladeuil)
754
755
    def fileids_altered_by_revision_ids(self, revision_ids, _inv_weave=None):
756
        """Find the file ids and versions affected by revisions.
757
758
        :param revisions: an iterable containing revision ids.
759
        :param _inv_weave: The inventory weave from this repository or None.
760
            If None, the inventory weave will be opened automatically.
761
        :return: a dictionary mapping altered file-ids to an iterable of
762
            revision_ids. Each altered file-ids has the exact revision_ids that
763
            altered it listed explicitly.
764
        """
4343.3.23 by John Arbash Meinel
Change RepositoryCHK1.fileids_altered_by to handle ghosts the way XML does.
765
        rich_root = self.supports_rich_root()
766
        bytes_to_info = inventory.CHKInventory._bytes_to_utf8name_key
767
        file_id_revisions = {}
4241.6.8 by Robert Collins, John Arbash Meinel, Ian Clatworthy, Vincent Ladeuil
Add --development6-rich-root, disabling the legacy and unneeded development2 format, and activating the tests for CHK features disabled pending this format. (Robert Collins, John Arbash Meinel, Ian Clatworthy, Vincent Ladeuil)
768
        pb = ui.ui_factory.nested_progress_bar()
769
        try:
4343.3.23 by John Arbash Meinel
Change RepositoryCHK1.fileids_altered_by to handle ghosts the way XML does.
770
            parent_ids = self._find_parent_ids_of_revisions(revision_ids)
771
            present_parent_inv_ids = self._find_present_inventory_ids(parent_ids)
772
            uninteresting_root_keys = set()
773
            interesting_root_keys = set()
774
            inventories_to_read = set(present_parent_inv_ids)
775
            inventories_to_read.update(revision_ids)
776
            for inv in self.iter_inventories(inventories_to_read):
777
                entry_chk_root_key = inv.id_to_entry.key()
778
                if inv.revision_id in present_parent_inv_ids:
779
                    uninteresting_root_keys.add(entry_chk_root_key)
780
                else:
781
                    interesting_root_keys.add(entry_chk_root_key)
782
783
            chk_bytes = self.chk_bytes
784
            for record, items in chk_map.iter_interesting_nodes(chk_bytes,
785
                        interesting_root_keys, uninteresting_root_keys,
786
                        pb=pb):
787
                for name, bytes in items:
788
                    (name_utf8, file_id, revision_id) = bytes_to_info(bytes)
789
                    if not rich_root and name_utf8 == '':
790
                        continue
791
                    try:
792
                        file_id_revisions[file_id].add(revision_id)
793
                    except KeyError:
794
                        file_id_revisions[file_id] = set([revision_id])
4241.6.8 by Robert Collins, John Arbash Meinel, Ian Clatworthy, Vincent Ladeuil
Add --development6-rich-root, disabling the legacy and unneeded development2 format, and activating the tests for CHK features disabled pending this format. (Robert Collins, John Arbash Meinel, Ian Clatworthy, Vincent Ladeuil)
795
        finally:
796
            pb.finished()
4343.3.23 by John Arbash Meinel
Change RepositoryCHK1.fileids_altered_by to handle ghosts the way XML does.
797
        return file_id_revisions
4241.6.8 by Robert Collins, John Arbash Meinel, Ian Clatworthy, Vincent Ladeuil
Add --development6-rich-root, disabling the legacy and unneeded development2 format, and activating the tests for CHK features disabled pending this format. (Robert Collins, John Arbash Meinel, Ian Clatworthy, Vincent Ladeuil)
798
799
    def find_text_key_references(self):
800
        """Find the text key references within the repository.
801
802
        :return: A dictionary mapping text keys ((fileid, revision_id) tuples)
803
            to whether they were referred to by the inventory of the
804
            revision_id that they contain. The inventory texts from all present
805
            revision ids are assessed to generate this report.
806
        """
807
        # XXX: Slow version but correct: rewrite as a series of delta
808
        # examinations/direct tree traversal. Note that that will require care
809
        # as a common node is reachable both from the inventory that added it,
810
        # and others afterwards.
811
        revision_keys = self.revisions.keys()
812
        result = {}
813
        rich_roots = self.supports_rich_root()
814
        pb = ui.ui_factory.nested_progress_bar()
815
        try:
816
            all_revs = self.all_revision_ids()
817
            total = len(all_revs)
818
            for pos, inv in enumerate(self.iter_inventories(all_revs)):
819
                pb.update("Finding text references", pos, total)
820
                for _, entry in inv.iter_entries():
821
                    if not rich_roots and entry.file_id == inv.root_id:
822
                        continue
823
                    key = (entry.file_id, entry.revision)
824
                    result.setdefault(key, False)
825
                    if entry.revision == inv.revision_id:
826
                        result[key] = True
827
            return result
828
        finally:
829
            pb.finished()
830
831
    def _reconcile_pack(self, collection, packs, extension, revs, pb):
832
        packer = GCCHKReconcilePacker(collection, packs, extension)
833
        return packer.pack(pb)
834
3735.42.1 by John Arbash Meinel
Create a GC => GC fetch code path.
835
    def _get_source(self, to_format):
836
        """Return a source for streaming from this repository."""
4241.6.8 by Robert Collins, John Arbash Meinel, Ian Clatworthy, Vincent Ladeuil
Add --development6-rich-root, disabling the legacy and unneeded development2 format, and activating the tests for CHK features disabled pending this format. (Robert Collins, John Arbash Meinel, Ian Clatworthy, Vincent Ladeuil)
837
        if isinstance(to_format, remote.RemoteRepositoryFormat):
838
            # Can't just check attributes on to_format with the current code,
839
            # work around this:
840
            to_format._ensure_real()
841
            to_format = to_format._custom_format
3735.42.1 by John Arbash Meinel
Create a GC => GC fetch code path.
842
        if to_format.__class__ is self._format.__class__:
843
            # We must be exactly the same format, otherwise stuff like the chk
844
            # page layout might be different
845
            return GroupCHKStreamSource(self, to_format)
4241.6.8 by Robert Collins, John Arbash Meinel, Ian Clatworthy, Vincent Ladeuil
Add --development6-rich-root, disabling the legacy and unneeded development2 format, and activating the tests for CHK features disabled pending this format. (Robert Collins, John Arbash Meinel, Ian Clatworthy, Vincent Ladeuil)
846
        return super(CHKInventoryRepository, self)._get_source(to_format)
3735.42.1 by John Arbash Meinel
Create a GC => GC fetch code path.
847
848
849
class GroupCHKStreamSource(repository.StreamSource):
850
    """Used when both the source and target repo are GroupCHK repos."""
851
852
    def __init__(self, from_repository, to_format):
853
        """Create a StreamSource streaming from from_repository."""
854
        super(GroupCHKStreamSource, self).__init__(from_repository, to_format)
855
        self._revision_keys = None
856
        self._text_keys = None
857
        self._chk_id_roots = None
858
        self._chk_p_id_roots = None
859
4392.2.2 by John Arbash Meinel
Add tests that ensure we can fetch branches with ghosts in their ancestry.
860
    def _get_inventory_stream(self, inventory_keys, allow_absent=False):
3735.42.1 by John Arbash Meinel
Create a GC => GC fetch code path.
861
        """Get a stream of inventory texts.
862
863
        When this function returns, self._chk_id_roots and self._chk_p_id_roots
864
        should be populated.
865
        """
866
        self._chk_id_roots = []
867
        self._chk_p_id_roots = []
868
        def _filtered_inv_stream():
869
            id_roots_set = set()
870
            p_id_roots_set = set()
871
            source_vf = self.from_repository.inventories
4343.3.4 by John Arbash Meinel
Custom implementation for GroupCHKStreamSource.get_stream_for_missing_keys.
872
            stream = source_vf.get_record_stream(inventory_keys,
3735.42.1 by John Arbash Meinel
Create a GC => GC fetch code path.
873
                                                 'groupcompress', True)
874
            for record in stream:
4392.2.2 by John Arbash Meinel
Add tests that ensure we can fetch branches with ghosts in their ancestry.
875
                if record.storage_kind == 'absent':
876
                    if allow_absent:
877
                        continue
878
                    else:
879
                        raise errors.NoSuchRevision(self, record.key)
3735.42.1 by John Arbash Meinel
Create a GC => GC fetch code path.
880
                bytes = record.get_bytes_as('fulltext')
881
                chk_inv = inventory.CHKInventory.deserialise(None, bytes,
882
                                                             record.key)
883
                key = chk_inv.id_to_entry.key()
884
                if key not in id_roots_set:
885
                    self._chk_id_roots.append(key)
886
                    id_roots_set.add(key)
887
                p_id_map = chk_inv.parent_id_basename_to_file_id
888
                if p_id_map is None:
889
                    raise AssertionError('Parent id -> file_id map not set')
890
                key = p_id_map.key()
891
                if key not in p_id_roots_set:
892
                    p_id_roots_set.add(key)
893
                    self._chk_p_id_roots.append(key)
894
                yield record
895
            # We have finished processing all of the inventory records, we
896
            # don't need these sets anymore
897
            id_roots_set.clear()
898
            p_id_roots_set.clear()
899
        return ('inventories', _filtered_inv_stream())
900
4343.3.26 by John Arbash Meinel
Handle get_stream when a parent is actually a ghost.
901
    def _find_present_inventories(self, revision_ids):
902
        revision_keys = [(r,) for r in revision_ids]
903
        inventories = self.from_repository.inventories
904
        present_inventories = inventories.get_parent_map(revision_keys)
905
        return [p[-1] for p in present_inventories]
906
907
    def _get_filtered_chk_streams(self, excluded_revision_ids):
3735.42.1 by John Arbash Meinel
Create a GC => GC fetch code path.
908
        self._text_keys = set()
4343.3.26 by John Arbash Meinel
Handle get_stream when a parent is actually a ghost.
909
        excluded_revision_ids.discard(_mod_revision.NULL_REVISION)
910
        if not excluded_revision_ids:
3735.42.1 by John Arbash Meinel
Create a GC => GC fetch code path.
911
            uninteresting_root_keys = set()
912
            uninteresting_pid_root_keys = set()
913
        else:
4343.3.26 by John Arbash Meinel
Handle get_stream when a parent is actually a ghost.
914
            # filter out any excluded revisions whose inventories are not
915
            # actually present
916
            # TODO: Update Repository.iter_inventories() to add
917
            #       ignore_missing=True
4343.3.36 by John Arbash Meinel
Some review feedback from Andrew.
918
            present_ids = self.from_repository._find_present_inventory_ids(
919
                            excluded_revision_ids)
4343.3.26 by John Arbash Meinel
Handle get_stream when a parent is actually a ghost.
920
            present_ids = self._find_present_inventories(excluded_revision_ids)
3735.42.1 by John Arbash Meinel
Create a GC => GC fetch code path.
921
            uninteresting_root_keys = set()
922
            uninteresting_pid_root_keys = set()
4343.3.26 by John Arbash Meinel
Handle get_stream when a parent is actually a ghost.
923
            for inv in self.from_repository.iter_inventories(present_ids):
3735.42.1 by John Arbash Meinel
Create a GC => GC fetch code path.
924
                uninteresting_root_keys.add(inv.id_to_entry.key())
925
                uninteresting_pid_root_keys.add(
926
                    inv.parent_id_basename_to_file_id.key())
927
        bytes_to_info = inventory.CHKInventory._bytes_to_utf8name_key
928
        chk_bytes = self.from_repository.chk_bytes
929
        def _filter_id_to_entry():
930
            for record, items in chk_map.iter_interesting_nodes(chk_bytes,
931
                        self._chk_id_roots, uninteresting_root_keys):
932
                for name, bytes in items:
933
                    # Note: we don't care about name_utf8, because we are always
934
                    # rich-root = True
935
                    _, file_id, revision_id = bytes_to_info(bytes)
936
                    self._text_keys.add((file_id, revision_id))
937
                if record is not None:
938
                    yield record
4343.3.8 by John Arbash Meinel
Some cleanup passes.
939
            # Consumed
940
            self._chk_id_roots = None
3735.42.1 by John Arbash Meinel
Create a GC => GC fetch code path.
941
        yield 'chk_bytes', _filter_id_to_entry()
942
        def _get_parent_id_basename_to_file_id_pages():
943
            for record, items in chk_map.iter_interesting_nodes(chk_bytes,
944
                        self._chk_p_id_roots, uninteresting_pid_root_keys):
945
                if record is not None:
946
                    yield record
4343.3.8 by John Arbash Meinel
Some cleanup passes.
947
            # Consumed
948
            self._chk_p_id_roots = None
3735.42.1 by John Arbash Meinel
Create a GC => GC fetch code path.
949
        yield 'chk_bytes', _get_parent_id_basename_to_file_id_pages()
950
951
    def _get_text_stream(self):
952
        # Note: We know we don't have to handle adding root keys, because both
953
        # the source and target are GCCHK, and those always support rich-roots
3735.42.3 by John Arbash Meinel
Fix a bug introduced into GCPacker.
954
        # We may want to request as 'unordered', in case the source has done a
955
        # 'split' packing
3735.42.1 by John Arbash Meinel
Create a GC => GC fetch code path.
956
        return ('texts', self.from_repository.texts.get_record_stream(
957
                            self._text_keys, 'groupcompress', False))
958
959
    def get_stream(self, search):
960
        revision_ids = search.get_keys()
961
        for stream_info in self._fetch_revision_texts(revision_ids):
962
            yield stream_info
963
        self._revision_keys = [(rev_id,) for rev_id in revision_ids]
4343.3.4 by John Arbash Meinel
Custom implementation for GroupCHKStreamSource.get_stream_for_missing_keys.
964
        yield self._get_inventory_stream(self._revision_keys)
4343.3.12 by John Arbash Meinel
instead of trusting the Search, use a get_parent_map call.
965
        # TODO: The keys to exclude might be part of the search recipe
966
        # For now, exclude all parents that are at the edge of ancestry, for
967
        # which we have inventories
4343.3.36 by John Arbash Meinel
Some review feedback from Andrew.
968
        from_repo = self.from_repository
969
        parent_ids = from_repo._find_parent_ids_of_revisions(revision_ids)
4343.3.26 by John Arbash Meinel
Handle get_stream when a parent is actually a ghost.
970
        for stream_info in self._get_filtered_chk_streams(parent_ids):
3735.42.1 by John Arbash Meinel
Create a GC => GC fetch code path.
971
            yield stream_info
972
        yield self._get_text_stream()
4241.6.8 by Robert Collins, John Arbash Meinel, Ian Clatworthy, Vincent Ladeuil
Add --development6-rich-root, disabling the legacy and unneeded development2 format, and activating the tests for CHK features disabled pending this format. (Robert Collins, John Arbash Meinel, Ian Clatworthy, Vincent Ladeuil)
973
4343.3.4 by John Arbash Meinel
Custom implementation for GroupCHKStreamSource.get_stream_for_missing_keys.
974
    def get_stream_for_missing_keys(self, missing_keys):
975
        # missing keys can only occur when we are byte copying and not
976
        # translating (because translation means we don't send
977
        # unreconstructable deltas ever).
4343.3.8 by John Arbash Meinel
Some cleanup passes.
978
        missing_inventory_keys = set()
4343.3.4 by John Arbash Meinel
Custom implementation for GroupCHKStreamSource.get_stream_for_missing_keys.
979
        for key in missing_keys:
4343.3.8 by John Arbash Meinel
Some cleanup passes.
980
            if key[0] != 'inventories':
4343.3.4 by John Arbash Meinel
Custom implementation for GroupCHKStreamSource.get_stream_for_missing_keys.
981
                raise AssertionError('The only missing keys we should'
982
                    ' be filling in are inventory keys, not %s'
4343.3.8 by John Arbash Meinel
Some cleanup passes.
983
                    % (key[0],))
984
            missing_inventory_keys.add(key[1:])
985
        if self._chk_id_roots or self._chk_p_id_roots:
986
            raise AssertionError('Cannot call get_stream_for_missing_keys'
987
                ' untill all of get_stream() has been consumed.')
4343.3.4 by John Arbash Meinel
Custom implementation for GroupCHKStreamSource.get_stream_for_missing_keys.
988
        # Yield the inventory stream, so we can find the chk stream
4392.2.2 by John Arbash Meinel
Add tests that ensure we can fetch branches with ghosts in their ancestry.
989
        # Some of the missing_keys will be missing because they are ghosts.
990
        # As such, we can ignore them. The Sink is required to verify there are
991
        # no unavailable texts when the ghost inventories are not filled in.
992
        yield self._get_inventory_stream(missing_inventory_keys,
993
                                         allow_absent=True)
4343.3.26 by John Arbash Meinel
Handle get_stream when a parent is actually a ghost.
994
        # We use the empty set for excluded_revision_ids, to make it clear that
995
        # we want to transmit all referenced chk pages.
4343.3.4 by John Arbash Meinel
Custom implementation for GroupCHKStreamSource.get_stream_for_missing_keys.
996
        for stream_info in self._get_filtered_chk_streams(set()):
997
            yield stream_info
998
4241.6.8 by Robert Collins, John Arbash Meinel, Ian Clatworthy, Vincent Ladeuil
Add --development6-rich-root, disabling the legacy and unneeded development2 format, and activating the tests for CHK features disabled pending this format. (Robert Collins, John Arbash Meinel, Ian Clatworthy, Vincent Ladeuil)
999
1000
class RepositoryFormatCHK1(RepositoryFormatPack):
1001
    """A hashed CHK+group compress pack repository."""
1002
1003
    repository_class = CHKInventoryRepository
4343.3.1 by John Arbash Meinel
Set 'supports_external_lookups=True' for dev6 repositories.
1004
    supports_external_lookups = True
4241.6.8 by Robert Collins, John Arbash Meinel, Ian Clatworthy, Vincent Ladeuil
Add --development6-rich-root, disabling the legacy and unneeded development2 format, and activating the tests for CHK features disabled pending this format. (Robert Collins, John Arbash Meinel, Ian Clatworthy, Vincent Ladeuil)
1005
    supports_chks = True
1006
    # For right now, setting this to True gives us InterModel1And2 rather
1007
    # than InterDifferingSerializer
1008
    _commit_builder_class = PackRootCommitBuilder
1009
    rich_root_data = True
1010
    _serializer = chk_serializer.chk_serializer_255_bigpage
1011
    _commit_inv_deltas = True
1012
    # What index classes to use
1013
    index_builder_class = BTreeBuilder
1014
    index_class = BTreeGraphIndex
1015
    # Note: We cannot unpack a delta that references a text we haven't
1016
    # seen yet. There are 2 options, work in fulltexts, or require
1017
    # topological sorting. Using fulltexts is more optimal for local
1018
    # operations, because the source can be smart about extracting
1019
    # multiple in-a-row (and sharing strings). Topological is better
1020
    # for remote, because we access less data.
1021
    _fetch_order = 'unordered'
4265.1.2 by John Arbash Meinel
Set _fetch_uses_deltas back to False, so that conversions from arbitrary sources can still fetch 'unordered'
1022
    _fetch_uses_deltas = False # essentially ignored by the groupcompress code.
4241.6.8 by Robert Collins, John Arbash Meinel, Ian Clatworthy, Vincent Ladeuil
Add --development6-rich-root, disabling the legacy and unneeded development2 format, and activating the tests for CHK features disabled pending this format. (Robert Collins, John Arbash Meinel, Ian Clatworthy, Vincent Ladeuil)
1023
    fast_deltas = True
1024
1025
    def _get_matching_bzrdir(self):
1026
        return bzrdir.format_registry.make_bzrdir('development6-rich-root')
1027
1028
    def _ignore_setting_bzrdir(self, format):
1029
        pass
1030
1031
    _matchingbzrdir = property(_get_matching_bzrdir, _ignore_setting_bzrdir)
1032
1033
    def get_format_string(self):
1034
        """See RepositoryFormat.get_format_string()."""
1035
        return ('Bazaar development format - group compression and chk inventory'
1036
                ' (needs bzr.dev from 1.14)\n')
1037
1038
    def get_format_description(self):
1039
        """See RepositoryFormat.get_format_description()."""
1040
        return ("Development repository format - rich roots, group compression"
1041
            " and chk inventories")
1042
1043
    def check_conversion_target(self, target_format):
1044
        if not target_format.rich_root_data:
1045
            raise errors.BadConversionTarget(
1046
                'Does not support rich root data.', target_format)
1047
        if not getattr(target_format, 'supports_tree_reference', False):
1048
            raise errors.BadConversionTarget(
1049
                'Does not support nested trees', target_format)
1050
1051