/brz/remove-bazaar

To get this branch, use:
bzr branch http://gegoxaren.bato24.eu/bzr/brz/remove-bazaar
5374.2.5 by John Arbash Meinel
Rework things a bit so the logic can be shared.
1
# Copyright (C) 2007-2010 Canonical Ltd
2520.4.85 by Aaron Bentley
Get all test passing (which just proves there aren't enough tests!)
2
#
3
# This program is free software; you can redistribute it and/or modify
4
# it under the terms of the GNU General Public License as published by
5
# the Free Software Foundation; either version 2 of the License, or
6
# (at your option) any later version.
7
#
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.
12
#
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
4183.7.1 by Sabin Iacob
update FSF mailing address
15
# Foundation, Inc., 51 Franklin Street, Fifth Floor, Boston, MA 02110-1301 USA
2520.4.85 by Aaron Bentley
Get all test passing (which just proves there aren't enough tests!)
16
6379.6.3 by Jelmer Vernooij
Use absolute_import.
17
from __future__ import absolute_import
18
2520.4.26 by Aaron Bentley
Make decompression reasonably memory-efficient
19
import bz2
2520.4.130 by Aaron Bentley
Finish tweaking decode_name
20
import re
2520.4.20 by Aaron Bentley
Compress and base64-encode bundle contents
21
6624 by Jelmer Vernooij
Merge Python3 porting work ('py3 pokes')
22
from ... import (
6621.22.2 by Martin
Use BytesIO or StringIO from bzrlib.sixish
23
    bencode,
2520.4.34 by Aaron Bentley
Add signature support
24
    errors,
2520.4.26 by Aaron Bentley
Make decompression reasonably memory-efficient
25
    iterablefile,
4543.2.16 by John Arbash Meinel
Adding an inventory text cache.
26
    lru_cache,
2520.4.13 by Aaron Bentley
Use real container implementation
27
    multiparent,
2520.4.97 by Aaron Bentley
Hack in support for inventory conversion
28
    osutils,
2520.4.40 by Aaron Bentley
Add human-readable diff to bundles
29
    revision as _mod_revision,
2520.4.45 by Aaron Bentley
Handle inconsistencies in last-modified-revision between vf and inventory
30
    trace,
4543.2.16 by John Arbash Meinel
Adding an inventory text cache.
31
    ui,
6670.4.1 by Jelmer Vernooij
Update imports.
32
    )
33
from ...bzr import (
34
    pack,
6670.4.10 by Jelmer Vernooij
Move serializer to bzr.
35
    serializer,
5374.2.5 by John Arbash Meinel
Rework things a bit so the logic can be shared.
36
    versionedfile as _mod_versionedfile,
2520.4.13 by Aaron Bentley
Use real container implementation
37
    )
6624 by Jelmer Vernooij
Merge Python3 porting work ('py3 pokes')
38
from ...bundle import bundle_data, serializer as bundle_serializer
39
from ...i18n import ngettext
40
from ...sixish import (
6621.22.2 by Martin
Use BytesIO or StringIO from bzrlib.sixish
41
    BytesIO,
6656.1.1 by Martin
Apply 2to3 dict fixer and clean up resulting mess using view helpers
42
    viewitems,
6621.22.2 by Martin
Use BytesIO or StringIO from bzrlib.sixish
43
    )
2520.4.14 by Aaron Bentley
Get most tests passing, use format header
44
2520.4.4 by Aaron Bentley
Get basis support for a new bundle format in place
45
5374.2.5 by John Arbash Meinel
Rework things a bit so the logic can be shared.
46
class _MPDiffInventoryGenerator(_mod_versionedfile._MPDiffGenerator):
47
    """Generate Inventory diffs serialized inventories."""
48
49
    def __init__(self, repo, inventory_keys):
50
        super(_MPDiffInventoryGenerator, self).__init__(repo.inventories,
51
            inventory_keys)
52
        self.repo = repo
53
        self.sha1s = {}
54
55
    def iter_diffs(self):
56
        """Compute the diffs one at a time."""
57
        # This is instead of compute_diffs() since we guarantee our ordering of
58
        # inventories, we don't have to do any buffering
59
        self._find_needed_keys()
60
        # We actually use a slightly different ordering. We grab all of the
61
        # parents first, and then grab the ordered requests.
62
        needed_ids = [k[-1] for k in self.present_parents]
63
        needed_ids.extend([k[-1] for k in self.ordered_keys])
64
        inv_to_str = self.repo._serializer.write_inventory_to_string
65
        for inv in self.repo.iter_inventories(needed_ids):
66
            revision_id = inv.revision_id
67
            key = (revision_id,)
68
            if key in self.present_parents:
69
                # Not a key we will transmit, which is a shame, since because
70
                # of that bundles don't work with stacked branches
71
                parent_ids = None
72
            else:
73
                parent_ids = [k[-1] for k in self.parent_map[key]]
74
            as_bytes = inv_to_str(inv)
75
            self._process_one_record(key, (as_bytes,))
76
            if parent_ids is None:
77
                continue
78
            diff = self.diffs.pop(key)
79
            sha1 = osutils.sha_string(as_bytes)
80
            yield revision_id, parent_ids, sha1, diff
81
82
2520.4.25 by Aaron Bentley
Rename ContainerWriter/ContainerReader to BundleWriter/BundleReader
83
class BundleWriter(object):
2520.4.118 by Aaron Bentley
Add docs
84
    """Writer for bundle-format files.
85
86
    This serves roughly the same purpose as ContainerReader, but acts as a
87
    layer on top of it.
88
2520.4.123 by Aaron Bentley
Cleanup of bundle code
89
    Provides ways of writing the specific record types supported this bundle
2520.4.118 by Aaron Bentley
Add docs
90
    format.
91
    """
2520.4.123 by Aaron Bentley
Cleanup of bundle code
92
2520.4.23 by Aaron Bentley
Move responsability for encoding into container objects
93
    def __init__(self, fileobj):
2520.4.27 by Aaron Bentley
Use less memory when writing bzip-encoded files
94
        self._container = pack.ContainerWriter(self._write_encoded)
2520.4.23 by Aaron Bentley
Move responsability for encoding into container objects
95
        self._fileobj = fileobj
2520.4.27 by Aaron Bentley
Use less memory when writing bzip-encoded files
96
        self._compressor = bz2.BZ2Compressor()
2520.4.21 by Aaron Bentley
Finish turning ContainerWriter into a new layer
97
2520.4.118 by Aaron Bentley
Add docs
98
    def _write_encoded(self, bytes):
99
        """Write bzip2-encoded bytes to the file"""
100
        self._fileobj.write(self._compressor.compress(bytes))
101
2520.4.21 by Aaron Bentley
Finish turning ContainerWriter into a new layer
102
    def begin(self):
2520.4.118 by Aaron Bentley
Add docs
103
        """Start writing the bundle"""
6989.1.1 by Jelmer Vernooij
Use format registry for bundles.
104
        self._fileobj.write(bundle_serializer._get_bundle_header('4'))
6973.6.1 by Jelmer Vernooij
More bees.
105
        self._fileobj.write(b'#\n')
2520.4.21 by Aaron Bentley
Finish turning ContainerWriter into a new layer
106
        self._container.begin()
107
108
    def end(self):
2520.4.118 by Aaron Bentley
Add docs
109
        """Finish writing the bundle"""
2520.4.21 by Aaron Bentley
Finish turning ContainerWriter into a new layer
110
        self._container.end()
2520.4.76 by Aaron Bentley
Move base64-encoding into merge directives
111
        self._fileobj.write(self._compressor.flush())
2520.4.21 by Aaron Bentley
Finish turning ContainerWriter into a new layer
112
2520.4.60 by Aaron Bentley
Add sha1 verification for mpdiffs
113
    def add_multiparent_record(self, mp_bytes, sha1, parents, repo_kind,
2520.4.21 by Aaron Bentley
Finish turning ContainerWriter into a new layer
114
                               revision_id, file_id):
2520.4.118 by Aaron Bentley
Add docs
115
        """Add a record for a multi-parent diff
116
117
        :mp_bytes: A multi-parent diff, as a bytestring
2520.4.123 by Aaron Bentley
Cleanup of bundle code
118
        :sha1: The sha1 hash of the fulltext
2520.4.118 by Aaron Bentley
Add docs
119
        :parents: a list of revision-ids of the parents
120
        :repo_kind: The kind of object in the repository.  May be 'file' or
121
            'inventory'
122
        :revision_id: The revision id of the mpdiff being added.
123
        :file_id: The file-id of the file, or None for inventories.
124
        """
2520.4.60 by Aaron Bentley
Add sha1 verification for mpdiffs
125
        metadata = {'parents': parents,
126
                    'storage_kind': 'mpdiff',
127
                    'sha1': sha1}
128
        self._add_record(mp_bytes, metadata, repo_kind, revision_id, file_id)
2520.4.21 by Aaron Bentley
Finish turning ContainerWriter into a new layer
129
2520.4.123 by Aaron Bentley
Cleanup of bundle code
130
    def add_fulltext_record(self, bytes, parents, repo_kind, revision_id):
2520.4.118 by Aaron Bentley
Add docs
131
        """Add a record for a fulltext
132
133
        :bytes: The fulltext, as a bytestring
134
        :parents: a list of revision-ids of the parents
135
        :repo_kind: The kind of object in the repository.  May be 'revision' or
136
            'signature'
137
        :revision_id: The revision id of the fulltext being added.
138
        """
139
        metadata = {'parents': parents,
2520.5.3 by Aaron Bentley
fix sha1 in bundle format 4
140
                    'storage_kind': 'mpdiff'}
2520.4.60 by Aaron Bentley
Add sha1 verification for mpdiffs
141
        self._add_record(bytes, {'parents': parents,
2520.4.123 by Aaron Bentley
Cleanup of bundle code
142
            'storage_kind': 'fulltext'}, repo_kind, revision_id, None)
2520.4.21 by Aaron Bentley
Finish turning ContainerWriter into a new layer
143
2520.4.95 by Aaron Bentley
Add support for header/info records
144
    def add_info_record(self, **kwargs):
2520.4.118 by Aaron Bentley
Add docs
145
        """Add an info record to the bundle
146
147
        Any parameters may be supplied, except 'self' and 'storage_kind'.
148
        Values must be lists, strings, integers, dicts, or a combination.
149
        """
2520.4.95 by Aaron Bentley
Add support for header/info records
150
        kwargs['storage_kind'] = 'header'
151
        self._add_record(None, kwargs, 'info', None, None)
152
2520.4.21 by Aaron Bentley
Finish turning ContainerWriter into a new layer
153
    @staticmethod
2520.4.68 by Aaron Bentley
Change name separators to all-slash
154
    def encode_name(content_kind, revision_id, file_id=None):
2520.4.118 by Aaron Bentley
Add docs
155
        """Encode semantic ids as a container name"""
3376.2.4 by Martin Pool
Remove every assert statement from bzrlib!
156
        if content_kind not in ('revision', 'file', 'inventory', 'signature',
157
                'info'):
158
            raise ValueError(content_kind)
2520.4.118 by Aaron Bentley
Add docs
159
        if content_kind == 'file':
3376.2.4 by Martin Pool
Remove every assert statement from bzrlib!
160
            if file_id is None:
161
                raise AssertionError()
2520.4.118 by Aaron Bentley
Add docs
162
        else:
3376.2.4 by Martin Pool
Remove every assert statement from bzrlib!
163
            if file_id is not None:
164
                raise AssertionError()
2520.4.95 by Aaron Bentley
Add support for header/info records
165
        if content_kind == 'info':
3376.2.4 by Martin Pool
Remove every assert statement from bzrlib!
166
            if revision_id is not None:
167
                raise AssertionError()
168
        elif revision_id is None:
169
            raise AssertionError()
2520.4.127 by Aaron Bentley
Fix up name encoding to handle revision-ids with slashes
170
        names = [n.replace('/', '//') for n in
171
                 (content_kind, revision_id, file_id) if n is not None]
2520.4.68 by Aaron Bentley
Change name separators to all-slash
172
        return '/'.join(names)
2520.4.21 by Aaron Bentley
Finish turning ContainerWriter into a new layer
173
2520.4.56 by Aaron Bentley
Begin adding support for arbitrary metadata
174
    def _add_record(self, bytes, metadata, repo_kind, revision_id, file_id):
2520.4.118 by Aaron Bentley
Add docs
175
        """Add a bundle record to the container.
176
177
        Most bundle records are recorded as header/body pairs, with the
178
        body being nameless.  Records with storage_kind 'header' have no
179
        body.
180
        """
2520.4.21 by Aaron Bentley
Finish turning ContainerWriter into a new layer
181
        name = self.encode_name(repo_kind, revision_id, file_id)
2520.4.95 by Aaron Bentley
Add support for header/info records
182
        encoded_metadata = bencode.bencode(metadata)
2682.1.1 by Robert Collins
* The ``bzrlib.pack`` interface has changed to use tuples of bytestrings
183
        self._container.add_bytes_record(encoded_metadata, [(name, )])
2520.4.95 by Aaron Bentley
Add support for header/info records
184
        if metadata['storage_kind'] != 'header':
185
            self._container.add_bytes_record(bytes, [])
2520.4.13 by Aaron Bentley
Use real container implementation
186
2520.4.7 by Aaron Bentley
Fix patch deserialization
187
2520.4.25 by Aaron Bentley
Rename ContainerWriter/ContainerReader to BundleWriter/BundleReader
188
class BundleReader(object):
2520.4.118 by Aaron Bentley
Add docs
189
    """Reader for bundle-format files.
190
191
    This serves roughly the same purpose as ContainerReader, but acts as a
192
    layer on top of it, providing metadata, a semantic name, and a record
193
    body
194
    """
2520.4.123 by Aaron Bentley
Cleanup of bundle code
195
4543.2.14 by John Arbash Meinel
Clarify some comments, fix up a debugging change.
196
    def __init__(self, fileobj, stream_input=True):
2520.4.145 by Aaron Bentley
Add memory_friendly toggle, be memory-unfriendly for merge directives
197
        """Constructor
198
199
        :param fileobj: a file containing a bzip-encoded container
2520.4.148 by Aaron Bentley
Updates from review
200
        :param stream_input: If True, the BundleReader stream input rather than
201
            reading it all into memory at once.  Reading it into memory all at
202
            once is (currently) faster.
2520.4.145 by Aaron Bentley
Add memory_friendly toggle, be memory-unfriendly for merge directives
203
        """
2520.4.23 by Aaron Bentley
Move responsability for encoding into container objects
204
        line = fileobj.readline()
205
        if line != '\n':
206
            fileobj.readline()
2520.4.40 by Aaron Bentley
Add human-readable diff to bundles
207
        self.patch_lines = []
2520.4.148 by Aaron Bentley
Updates from review
208
        if stream_input:
2520.4.145 by Aaron Bentley
Add memory_friendly toggle, be memory-unfriendly for merge directives
209
            source_file = iterablefile.IterableFile(self.iter_decode(fileobj))
210
        else:
6621.22.2 by Martin
Use BytesIO or StringIO from bzrlib.sixish
211
            source_file = BytesIO(bz2.decompress(fileobj.read()))
2916.2.18 by Andrew Bennetts
Use iter_records_from_file rather than ContainerReader.
212
        self._container_file = source_file
2520.4.26 by Aaron Bentley
Make decompression reasonably memory-efficient
213
214
    @staticmethod
215
    def iter_decode(fileobj):
2520.4.118 by Aaron Bentley
Add docs
216
        """Iterate through decoded fragments of the file"""
2520.4.26 by Aaron Bentley
Make decompression reasonably memory-efficient
217
        decompressor = bz2.BZ2Decompressor()
218
        for line in fileobj:
2916.2.18 by Andrew Bennetts
Use iter_records_from_file rather than ContainerReader.
219
            try:
220
                yield decompressor.decompress(line)
221
            except EOFError:
222
                return
2520.4.22 by Aaron Bentley
Create ContainerReader
223
224
    @staticmethod
225
    def decode_name(name):
2520.4.118 by Aaron Bentley
Add docs
226
        """Decode a name from its container form into a semantic form
227
228
        :retval: content_kind, revision_id, file_id
229
        """
2520.4.130 by Aaron Bentley
Finish tweaking decode_name
230
        segments = re.split('(//?)', name)
231
        names = ['']
2520.4.127 by Aaron Bentley
Fix up name encoding to handle revision-ids with slashes
232
        for segment in segments:
233
            if segment == '//':
234
                names[-1] += '/'
2520.4.130 by Aaron Bentley
Finish tweaking decode_name
235
            elif segment == '/':
2520.4.127 by Aaron Bentley
Fix up name encoding to handle revision-ids with slashes
236
                names.append('')
237
            else:
238
                names[-1] += segment
2520.4.130 by Aaron Bentley
Finish tweaking decode_name
239
        content_kind = names[0]
2520.4.95 by Aaron Bentley
Add support for header/info records
240
        revision_id = None
241
        file_id = None
242
        if len(names) > 1:
243
            revision_id = names[1]
2520.4.68 by Aaron Bentley
Change name separators to all-slash
244
        if len(names) > 2:
245
            file_id = names[2]
246
        return content_kind, revision_id, file_id
2520.4.22 by Aaron Bentley
Create ContainerReader
247
248
    def iter_records(self):
2520.4.118 by Aaron Bentley
Add docs
249
        """Iterate through bundle records
250
251
        :return: a generator of (bytes, metadata, content_kind, revision_id,
252
            file_id)
253
        """
2916.2.18 by Andrew Bennetts
Use iter_records_from_file rather than ContainerReader.
254
        iterator = pack.iter_records_from_file(self._container_file)
255
        for names, bytes in iterator:
2520.4.131 by Aaron Bentley
Raise BadBundle for records with wrong number of names
256
            if len(names) != 1:
257
                raise errors.BadBundle('Record has %d names instead of 1'
258
                                       % len(names))
2916.2.18 by Andrew Bennetts
Use iter_records_from_file rather than ContainerReader.
259
            metadata = bencode.bdecode(bytes)
2520.4.95 by Aaron Bentley
Add support for header/info records
260
            if metadata['storage_kind'] == 'header':
261
                bytes = None
262
            else:
6634.2.1 by Martin
Apply 2to3 next fixer and make compatible
263
                _unused, bytes = next(iterator)
2682.1.1 by Robert Collins
* The ``bzrlib.pack`` interface has changed to use tuples of bytestrings
264
            yield (bytes, metadata) + self.decode_name(names[0][0])
2520.4.22 by Aaron Bentley
Create ContainerReader
265
266
4237.3.1 by Jelmer Vernooij
Add new module with generic serializer information; keep XML-specific bits in
267
class BundleSerializerV4(bundle_serializer.BundleSerializer):
2520.4.118 by Aaron Bentley
Add docs
268
    """Implement the high-level bundle interface"""
2520.4.123 by Aaron Bentley
Cleanup of bundle code
269
2520.4.4 by Aaron Bentley
Get basis support for a new bundle format in place
270
    def write(self, repository, revision_ids, forced_bases, fileobj):
2520.4.118 by Aaron Bentley
Add docs
271
        """Write a bundle to a file-like object
272
273
        For backwards-compatibility only
274
        """
2520.4.50 by Aaron Bentley
Split write functionality out into a separate object
275
        write_op = BundleWriteOperation.from_old_args(repository, revision_ids,
276
                                                      forced_bases, fileobj)
2520.4.53 by Aaron Bentley
refactor bundle serialization to make write_bundle primary
277
        return write_op.do_write()
278
279
    def write_bundle(self, repository, target, base, fileobj):
2520.4.118 by Aaron Bentley
Add docs
280
        """Write a bundle to a file object
281
282
        :param repository: The repository to retrieve revision data from
283
        :param target: The head revision to include ancestors of
284
        :param base: The ancestor of the target to stop including acestors
285
            at.
286
        :param fileobj: The file-like object to write to
287
        """
2520.4.53 by Aaron Bentley
refactor bundle serialization to make write_bundle primary
288
        write_op =  BundleWriteOperation(base, target, repository, fileobj)
289
        return write_op.do_write()
2520.4.50 by Aaron Bentley
Split write functionality out into a separate object
290
291
    def read(self, file):
2520.4.118 by Aaron Bentley
Add docs
292
        """return a reader object for a given file"""
2520.4.72 by Aaron Bentley
Rename format to 4alpha
293
        bundle = BundleInfoV4(file, self)
2520.4.50 by Aaron Bentley
Split write functionality out into a separate object
294
        return bundle
295
2520.4.101 by Aaron Bentley
Use a registry to look up xml serializers by format
296
    @staticmethod
297
    def get_source_serializer(info):
2520.4.118 by Aaron Bentley
Add docs
298
        """Retrieve the serializer for a given info object"""
4237.3.1 by Jelmer Vernooij
Add new module with generic serializer information; keep XML-specific bits in
299
        return serializer.format_registry.get(info['serializer'])
2520.4.101 by Aaron Bentley
Use a registry to look up xml serializers by format
300
2520.4.50 by Aaron Bentley
Split write functionality out into a separate object
301
302
class BundleWriteOperation(object):
2520.4.118 by Aaron Bentley
Add docs
303
    """Perform the operation of writing revisions to a bundle"""
2520.4.123 by Aaron Bentley
Cleanup of bundle code
304
2520.4.50 by Aaron Bentley
Split write functionality out into a separate object
305
    @classmethod
306
    def from_old_args(cls, repository, revision_ids, forced_bases, fileobj):
2520.4.123 by Aaron Bentley
Cleanup of bundle code
307
        """Create a BundleWriteOperation from old-style arguments"""
2520.4.50 by Aaron Bentley
Split write functionality out into a separate object
308
        base, target = cls.get_base_target(revision_ids, forced_bases,
309
                                           repository)
310
        return BundleWriteOperation(base, target, repository, fileobj,
311
                                    revision_ids)
312
2520.4.53 by Aaron Bentley
refactor bundle serialization to make write_bundle primary
313
    def __init__(self, base, target, repository, fileobj, revision_ids=None):
2520.4.50 by Aaron Bentley
Split write functionality out into a separate object
314
        self.base = base
315
        self.target = target
316
        self.repository = repository
2520.4.39 by Aaron Bentley
Rename container => bundle(reader) where appropriate
317
        bundle = BundleWriter(fileobj)
2520.4.50 by Aaron Bentley
Split write functionality out into a separate object
318
        self.bundle = bundle
2520.4.53 by Aaron Bentley
refactor bundle serialization to make write_bundle primary
319
        if revision_ids is not None:
320
            self.revision_ids = revision_ids
321
        else:
4154.1.1 by Ian Clatworthy
make send use graph.find_difference() instead of walking all of history twice
322
            graph = repository.get_graph()
4154.1.3 by Ian Clatworthy
strip ghosts so test_bundle_with_ghosts works again
323
            revision_ids = graph.find_unique_ancestors(target, [base])
324
            # Strip ghosts
325
            parents = graph.get_parent_map(revision_ids)
326
            self.revision_ids = [r for r in revision_ids if r in parents]
6619.3.12 by Jelmer Vernooij
Use 2to3 set_literal fixer.
327
        self.revision_keys = {(revid,) for revid in self.revision_ids}
2520.4.50 by Aaron Bentley
Split write functionality out into a separate object
328
329
    def do_write(self):
2520.4.118 by Aaron Bentley
Add docs
330
        """Write all data to the bundle"""
6143.1.3 by Jonathan Riddell
more plurals
331
        trace.note(ngettext('Bundling %d revision.', 'Bundling %d revisions.',
332
                            len(self.revision_ids)), len(self.revision_ids))
6754.8.4 by Jelmer Vernooij
Use new context stuff.
333
        with self.repository.lock_read():
3350.6.4 by Robert Collins
First cut at pluralised VersionedFiles. Some rather massive API incompatabilities, primarily because of the difficulty of coherence among competing stores.
334
            self.bundle.begin()
335
            self.write_info()
336
            self.write_files()
337
            self.write_revisions()
338
            self.bundle.end()
2520.4.53 by Aaron Bentley
refactor bundle serialization to make write_bundle primary
339
        return self.revision_ids
2520.4.50 by Aaron Bentley
Split write functionality out into a separate object
340
2520.4.97 by Aaron Bentley
Hack in support for inventory conversion
341
    def write_info(self):
2520.4.118 by Aaron Bentley
Add docs
342
        """Write format info"""
2520.4.113 by Aaron Bentley
Avoid peeking at Repository._serializer
343
        serializer_format = self.repository.get_serializer_format()
2520.4.99 by Aaron Bentley
Test conversion across models
344
        supports_rich_root = {True: 1, False: 0}[
345
            self.repository.supports_rich_root()]
2520.4.113 by Aaron Bentley
Avoid peeking at Repository._serializer
346
        self.bundle.add_info_record(serializer=serializer_format,
2520.4.99 by Aaron Bentley
Test conversion across models
347
                                    supports_rich_root=supports_rich_root)
2520.4.97 by Aaron Bentley
Hack in support for inventory conversion
348
2520.4.51 by Aaron Bentley
Split iteration through file revisions into a method, so we can vary it
349
    def write_files(self):
2520.4.118 by Aaron Bentley
Add docs
350
        """Write bundle records for all revisions of all files"""
3350.6.4 by Robert Collins
First cut at pluralised VersionedFiles. Some rather massive API incompatabilities, primarily because of the difficulty of coherence among competing stores.
351
        text_keys = []
3350.6.7 by Robert Collins
Review feedback, making things more clear, adding documentation on what is used where.
352
        altered_fileids = self.repository.fileids_altered_by_revision_ids(
353
                self.revision_ids)
6656.1.1 by Martin
Apply 2to3 dict fixer and clean up resulting mess using view helpers
354
        for file_id, revision_ids in viewitems(altered_fileids):
3350.6.4 by Robert Collins
First cut at pluralised VersionedFiles. Some rather massive API incompatabilities, primarily because of the difficulty of coherence among competing stores.
355
            for revision_id in revision_ids:
356
                text_keys.append((file_id, revision_id))
3350.6.10 by Martin Pool
VersionedFiles review cleanups
357
        self._add_mp_records_keys('file', self.repository.texts, text_keys)
2520.4.50 by Aaron Bentley
Split write functionality out into a separate object
358
359
    def write_revisions(self):
2520.4.118 by Aaron Bentley
Add docs
360
        """Write bundle records for all revisions and signatures"""
3350.6.4 by Robert Collins
First cut at pluralised VersionedFiles. Some rather massive API incompatabilities, primarily because of the difficulty of coherence among competing stores.
361
        inv_vf = self.repository.inventories
4543.2.12 by John Arbash Meinel
Always sort the inventories in pure topological order
362
        topological_order = [key[-1] for key in multiparent.topo_iter_keys(
363
                                inv_vf, self.revision_keys)]
364
        revision_order = topological_order
2520.4.75 by Aaron Bentley
Fix traceback on empty bundles.
365
        if self.target is not None and self.target in self.revision_ids:
4543.2.20 by John Arbash Meinel
Update from Martin's review feedback.
366
            # Make sure the target revision is always the last entry
4543.2.12 by John Arbash Meinel
Always sort the inventories in pure topological order
367
            revision_order = list(topological_order)
2520.4.50 by Aaron Bentley
Split write functionality out into a separate object
368
            revision_order.remove(self.target)
369
            revision_order.append(self.target)
4543.2.3 by John Arbash Meinel
Change the name to test_merge_directive
370
        if self.repository._serializer.support_altered_by_hack:
4543.2.20 by John Arbash Meinel
Update from Martin's review feedback.
371
            # Repositories that support_altered_by_hack means that
372
            # inventories.make_mpdiffs() contains all the data about the tree
373
            # shape. Formats without support_altered_by_hack require
374
            # chk_bytes/etc, so we use a different code path.
4543.2.3 by John Arbash Meinel
Change the name to test_merge_directive
375
            self._add_mp_records_keys('inventory', inv_vf,
4543.2.12 by John Arbash Meinel
Always sort the inventories in pure topological order
376
                                      [(revid,) for revid in topological_order])
4543.2.3 by John Arbash Meinel
Change the name to test_merge_directive
377
        else:
4543.2.20 by John Arbash Meinel
Update from Martin's review feedback.
378
            # Inventories should always be added in pure-topological order, so
379
            # that we can apply the mpdiff for the child to the parent texts.
4543.2.12 by John Arbash Meinel
Always sort the inventories in pure topological order
380
            self._add_inventory_mpdiffs_from_serializer(topological_order)
4543.2.3 by John Arbash Meinel
Change the name to test_merge_directive
381
        self._add_revision_texts(revision_order)
382
4543.2.4 by John Arbash Meinel
Start working on code that will use Repository._serializer.write_inventory_to_strig.
383
    def _add_inventory_mpdiffs_from_serializer(self, revision_order):
4543.2.20 by John Arbash Meinel
Update from Martin's review feedback.
384
        """Generate mpdiffs by serializing inventories.
385
386
        The current repository only has part of the tree shape information in
387
        the 'inventories' vf. So we use serializer.write_inventory_to_string to
388
        get a 'full' representation of the tree shape, and then generate
389
        mpdiffs on that data stream. This stream can then be reconstructed on
390
        the other side.
391
        """
4543.2.4 by John Arbash Meinel
Start working on code that will use Repository._serializer.write_inventory_to_strig.
392
        inventory_key_order = [(r,) for r in revision_order]
5374.2.5 by John Arbash Meinel
Rework things a bit so the logic can be shared.
393
        generator = _MPDiffInventoryGenerator(self.repository,
394
                                              inventory_key_order)
395
        for revision_id, parent_ids, sha1, diff in generator.iter_diffs():
4543.2.4 by John Arbash Meinel
Start working on code that will use Repository._serializer.write_inventory_to_strig.
396
            text = ''.join(diff.to_patch())
397
            self.bundle.add_multiparent_record(text, sha1, parent_ids,
398
                                               'inventory', revision_id, None)
399
4543.2.3 by John Arbash Meinel
Change the name to test_merge_directive
400
    def _add_revision_texts(self, revision_order):
3099.3.5 by John Arbash Meinel
Update the last couple of places that referred to Provider.get_parents() directly.
401
        parent_map = self.repository.get_parent_map(revision_order)
4202.3.1 by Andrew Bennetts
Don't use get_revision_xml when writing a bundle, instead get all the revisions together.
402
        revision_to_str = self.repository._serializer.write_revision_to_string
403
        revisions = self.repository.get_revisions(revision_order)
404
        for revision in revisions:
405
            revision_id = revision.revision_id
3099.3.5 by John Arbash Meinel
Update the last couple of places that referred to Provider.get_parents() directly.
406
            parents = parent_map.get(revision_id, None)
4202.3.1 by Andrew Bennetts
Don't use get_revision_xml when writing a bundle, instead get all the revisions together.
407
            revision_text = revision_to_str(revision)
2520.4.50 by Aaron Bentley
Split write functionality out into a separate object
408
            self.bundle.add_fulltext_record(revision_text, parents,
2520.4.123 by Aaron Bentley
Cleanup of bundle code
409
                                       'revision', revision_id)
2520.4.34 by Aaron Bentley
Add signature support
410
            try:
2520.4.50 by Aaron Bentley
Split write functionality out into a separate object
411
                self.bundle.add_fulltext_record(
412
                    self.repository.get_signature_text(
2520.4.123 by Aaron Bentley
Cleanup of bundle code
413
                    revision_id), parents, 'signature', revision_id)
2520.4.34 by Aaron Bentley
Add signature support
414
            except errors.NoSuchRevision:
415
                pass
416
2520.4.50 by Aaron Bentley
Split write functionality out into a separate object
417
    @staticmethod
418
    def get_base_target(revision_ids, forced_bases, repository):
2520.4.123 by Aaron Bentley
Cleanup of bundle code
419
        """Determine the base and target from old-style revision ids"""
2520.4.50 by Aaron Bentley
Split write functionality out into a separate object
420
        if len(revision_ids) == 0:
421
            return None, None
422
        target = revision_ids[0]
423
        base = forced_bases.get(target)
424
        if base is None:
425
            parents = repository.get_revision(target).parent_ids
426
            if len(parents) == 0:
427
                base = _mod_revision.NULL_REVISION
428
            else:
429
                base = parents[0]
430
        return base, target
431
3350.6.10 by Martin Pool
VersionedFiles review cleanups
432
    def _add_mp_records_keys(self, repo_kind, vf, keys):
2520.4.118 by Aaron Bentley
Add docs
433
        """Add multi-parent diff records to a bundle"""
3350.6.4 by Robert Collins
First cut at pluralised VersionedFiles. Some rather massive API incompatabilities, primarily because of the difficulty of coherence among competing stores.
434
        ordered_keys = list(multiparent.topo_iter_keys(vf, keys))
435
        mpdiffs = vf.make_mpdiffs(ordered_keys)
436
        sha1s = vf.get_sha1s(ordered_keys)
437
        parent_map = vf.get_parent_map(ordered_keys)
3350.8.3 by Robert Collins
VF.get_sha1s needed changing to be stackable.
438
        for mpdiff, item_key, in zip(mpdiffs, ordered_keys):
439
            sha1 = sha1s[item_key]
3350.6.4 by Robert Collins
First cut at pluralised VersionedFiles. Some rather massive API incompatabilities, primarily because of the difficulty of coherence among competing stores.
440
            parents = [key[-1] for key in parent_map[item_key]]
2520.4.41 by Aaron Bentley
Accelerate mpdiff generation
441
            text = ''.join(mpdiff.to_patch())
3350.6.4 by Robert Collins
First cut at pluralised VersionedFiles. Some rather massive API incompatabilities, primarily because of the difficulty of coherence among competing stores.
442
            # Infer file id records as appropriate.
443
            if len(item_key) == 2:
444
                file_id = item_key[0]
445
            else:
446
                file_id = None
2520.4.60 by Aaron Bentley
Add sha1 verification for mpdiffs
447
            self.bundle.add_multiparent_record(text, sha1, parents, repo_kind,
3350.6.4 by Robert Collins
First cut at pluralised VersionedFiles. Some rather massive API incompatabilities, primarily because of the difficulty of coherence among competing stores.
448
                                               item_key[-1], file_id)
2520.4.6 by Aaron Bentley
Get installation started
449
450
2520.4.72 by Aaron Bentley
Rename format to 4alpha
451
class BundleInfoV4(object):
2520.4.6 by Aaron Bentley
Get installation started
452
2520.4.118 by Aaron Bentley
Add docs
453
    """Provide (most of) the BundleInfo interface"""
2520.4.6 by Aaron Bentley
Get installation started
454
    def __init__(self, fileobj, serializer):
2520.4.14 by Aaron Bentley
Get most tests passing, use format header
455
        self._fileobj = fileobj
456
        self._serializer = serializer
457
        self.__real_revisions = None
458
        self.__revisions = None
459
460
    def install(self, repository):
461
        return self.install_revisions(repository)
462
2520.4.148 by Aaron Bentley
Updates from review
463
    def install_revisions(self, repository, stream_input=True):
464
        """Install this bundle's revisions into the specified repository
465
466
        :param target_repo: The repository to install into
467
        :param stream_input: If True, will stream input rather than reading it
468
            all into memory at once.  Reading it into memory all at once is
469
            (currently) faster.
470
        """
2520.4.18 by Aaron Bentley
Generate mpdiffs for inventory
471
        repository.lock_write()
472
        try:
2520.4.148 by Aaron Bentley
Updates from review
473
            ri = RevisionInstaller(self.get_bundle_reader(stream_input),
2520.4.21 by Aaron Bentley
Finish turning ContainerWriter into a new layer
474
                                   self._serializer, repository)
2520.4.18 by Aaron Bentley
Generate mpdiffs for inventory
475
            return ri.install()
476
        finally:
477
            repository.unlock()
2520.4.14 by Aaron Bentley
Get most tests passing, use format header
478
2520.4.109 by Aaron Bentley
start work on directive cherry-picking
479
    def get_merge_request(self, target_repo):
480
        """Provide data for performing a merge
481
482
        Returns suggested base, suggested target, and patch verification status
483
        """
484
        return None, self.target, 'inapplicable'
485
2520.4.148 by Aaron Bentley
Updates from review
486
    def get_bundle_reader(self, stream_input=True):
487
        """Return a new BundleReader for the associated bundle
488
489
        :param stream_input: If True, the BundleReader stream input rather than
490
            reading it all into memory at once.  Reading it into memory all at
491
            once is (currently) faster.
492
        """
2520.4.21 by Aaron Bentley
Finish turning ContainerWriter into a new layer
493
        self._fileobj.seek(0)
2520.4.148 by Aaron Bentley
Updates from review
494
        return BundleReader(self._fileobj, stream_input)
2520.4.21 by Aaron Bentley
Finish turning ContainerWriter into a new layer
495
2520.4.14 by Aaron Bentley
Get most tests passing, use format header
496
    def _get_real_revisions(self):
497
        if self.__real_revisions is None:
498
            self.__real_revisions = []
2520.4.39 by Aaron Bentley
Rename container => bundle(reader) where appropriate
499
            bundle_reader = self.get_bundle_reader()
2520.4.102 by Aaron Bentley
rename parents to metadata
500
            for bytes, metadata, repo_kind, revision_id, file_id in \
2520.4.39 by Aaron Bentley
Rename container => bundle(reader) where appropriate
501
                bundle_reader.iter_records():
2520.4.101 by Aaron Bentley
Use a registry to look up xml serializers by format
502
                if repo_kind == 'info':
503
                    serializer =\
2520.4.102 by Aaron Bentley
rename parents to metadata
504
                        self._serializer.get_source_serializer(metadata)
2520.4.22 by Aaron Bentley
Create ContainerReader
505
                if repo_kind == 'revision':
2520.4.101 by Aaron Bentley
Use a registry to look up xml serializers by format
506
                    rev = serializer.read_revision_from_string(bytes)
2520.4.14 by Aaron Bentley
Get most tests passing, use format header
507
                    self.__real_revisions.append(rev)
508
        return self.__real_revisions
509
    real_revisions = property(_get_real_revisions)
510
511
    def _get_revisions(self):
512
        if self.__revisions is None:
513
            self.__revisions = []
514
            for revision in self.real_revisions:
2520.4.33 by Aaron Bentley
remove test dependencies on serialization minutia
515
                self.__revisions.append(
516
                    bundle_data.RevisionInfo.from_revision(revision))
2520.4.14 by Aaron Bentley
Get most tests passing, use format header
517
        return self.__revisions
518
519
    revisions = property(_get_revisions)
520
2520.4.29 by Aaron Bentley
Reactivate some testing, fix topo_iter
521
    def _get_target(self):
522
        return self.revisions[-1].revision_id
523
524
    target = property(_get_target)
525
2520.4.14 by Aaron Bentley
Get most tests passing, use format header
526
527
class RevisionInstaller(object):
2520.4.123 by Aaron Bentley
Cleanup of bundle code
528
    """Installs revisions into a repository"""
2520.4.14 by Aaron Bentley
Get most tests passing, use format header
529
2520.4.21 by Aaron Bentley
Finish turning ContainerWriter into a new layer
530
    def __init__(self, container, serializer, repository):
531
        self._container = container
2520.4.6 by Aaron Bentley
Get installation started
532
        self._serializer = serializer
2520.4.14 by Aaron Bentley
Get most tests passing, use format header
533
        self._repository = repository
2520.4.97 by Aaron Bentley
Hack in support for inventory conversion
534
        self._info = None
2520.4.99 by Aaron Bentley
Test conversion across models
535
2520.4.14 by Aaron Bentley
Get most tests passing, use format header
536
    def install(self):
2592.4.1 by Martin Pool
RevisionInstaller now creates a write group for its work
537
        """Perform the installation.
3943.8.1 by Marius Kruger
remove all trailing whitespace from bzr source
538
2592.4.1 by Martin Pool
RevisionInstaller now creates a write group for its work
539
        Must be called with the Repository locked.
540
        """
541
        self._repository.start_write_group()
542
        try:
2856.1.2 by Robert Collins
Review feedback.
543
            result = self._install_in_write_group()
2592.4.1 by Martin Pool
RevisionInstaller now creates a write group for its work
544
        except:
545
            self._repository.abort_write_group()
546
            raise
547
        self._repository.commit_write_group()
548
        return result
549
2856.1.2 by Robert Collins
Review feedback.
550
    def _install_in_write_group(self):
2520.4.6 by Aaron Bentley
Get installation started
551
        current_file = None
552
        current_versionedfile = None
553
        pending_file_records = []
2520.4.142 by Aaron Bentley
Clean up installation of inventory records
554
        inventory_vf = None
555
        pending_inventory_records = []
2520.4.8 by Aaron Bentley
Serialize inventory
556
        added_inv = set()
2520.4.29 by Aaron Bentley
Reactivate some testing, fix topo_iter
557
        target_revision = None
2520.4.58 by Aaron Bentley
Propogate support for metadata to iter_revisions, add storage kind
558
        for bytes, metadata, repo_kind, revision_id, file_id in\
2520.4.22 by Aaron Bentley
Create ContainerReader
559
            self._container.iter_records():
2520.4.97 by Aaron Bentley
Hack in support for inventory conversion
560
            if repo_kind == 'info':
3376.2.4 by Martin Pool
Remove every assert statement from bzrlib!
561
                if self._info is not None:
562
                    raise AssertionError()
2520.4.123 by Aaron Bentley
Cleanup of bundle code
563
                self._handle_info(metadata)
3350.6.4 by Robert Collins
First cut at pluralised VersionedFiles. Some rather massive API incompatabilities, primarily because of the difficulty of coherence among competing stores.
564
            if (pending_file_records and
565
                (repo_kind, file_id) != ('file', current_file)):
566
                # Flush the data for a single file - prevents memory
567
                # spiking due to buffering all files in memory.
568
                self._install_mp_records_keys(self._repository.texts,
569
                    pending_file_records)
2520.4.8 by Aaron Bentley
Serialize inventory
570
                current_file = None
3350.6.4 by Robert Collins
First cut at pluralised VersionedFiles. Some rather massive API incompatabilities, primarily because of the difficulty of coherence among competing stores.
571
                del pending_file_records[:]
2520.4.142 by Aaron Bentley
Clean up installation of inventory records
572
            if len(pending_inventory_records) > 0 and repo_kind != 'inventory':
3350.6.4 by Robert Collins
First cut at pluralised VersionedFiles. Some rather massive API incompatabilities, primarily because of the difficulty of coherence among competing stores.
573
                self._install_inventory_records(pending_inventory_records)
2520.4.142 by Aaron Bentley
Clean up installation of inventory records
574
                pending_inventory_records = []
575
            if repo_kind == 'inventory':
3350.6.4 by Robert Collins
First cut at pluralised VersionedFiles. Some rather massive API incompatabilities, primarily because of the difficulty of coherence among competing stores.
576
                pending_inventory_records.append(((revision_id,), metadata, bytes))
2520.4.142 by Aaron Bentley
Clean up installation of inventory records
577
            if repo_kind == 'revision':
578
                target_revision = revision_id
579
                self._install_revision(revision_id, metadata, bytes)
580
            if repo_kind == 'signature':
581
                self._install_signature(revision_id, metadata, bytes)
2520.4.22 by Aaron Bentley
Create ContainerReader
582
            if repo_kind == 'file':
2520.4.142 by Aaron Bentley
Clean up installation of inventory records
583
                current_file = file_id
3350.6.4 by Robert Collins
First cut at pluralised VersionedFiles. Some rather massive API incompatabilities, primarily because of the difficulty of coherence among competing stores.
584
                pending_file_records.append(((file_id, revision_id), metadata, bytes))
585
        self._install_mp_records_keys(self._repository.texts, pending_file_records)
2520.4.14 by Aaron Bentley
Get most tests passing, use format header
586
        return target_revision
2520.4.6 by Aaron Bentley
Get installation started
587
2520.4.123 by Aaron Bentley
Cleanup of bundle code
588
    def _handle_info(self, info):
589
        """Extract data from an info record"""
590
        self._info = info
591
        self._source_serializer = self._serializer.get_source_serializer(info)
592
        if (info['supports_rich_root'] == 0 and
593
            self._repository.supports_rich_root()):
594
            self.update_root = True
595
        else:
596
            self.update_root = False
597
2520.4.60 by Aaron Bentley
Add sha1 verification for mpdiffs
598
    def _install_mp_records(self, versionedfile, records):
2520.4.61 by Aaron Bentley
Do bulk insertion of records
599
        if len(records) == 0:
600
            return
601
        d_func = multiparent.MultiParent.from_patch
602
        vf_records = [(r, m['parents'], m['sha1'], d_func(t)) for r, m, t in
603
                      records if r not in versionedfile]
604
        versionedfile.add_mpdiffs(vf_records)
2520.4.8 by Aaron Bentley
Serialize inventory
605
3350.6.4 by Robert Collins
First cut at pluralised VersionedFiles. Some rather massive API incompatabilities, primarily because of the difficulty of coherence among competing stores.
606
    def _install_mp_records_keys(self, versionedfile, records):
607
        d_func = multiparent.MultiParent.from_patch
608
        vf_records = []
609
        for key, meta, text in records:
3350.6.7 by Robert Collins
Review feedback, making things more clear, adding documentation on what is used where.
610
            # Adapt to tuple interface: A length two key is a file_id,
611
            # revision_id pair, a length 1 key is a
612
            # revision/signature/inventory. We need to do this because
613
            # the metadata extraction from the bundle has not yet been updated
614
            # to use the consistent tuple interface itself.
3350.6.4 by Robert Collins
First cut at pluralised VersionedFiles. Some rather massive API incompatabilities, primarily because of the difficulty of coherence among competing stores.
615
            if len(key) == 2:
616
                prefix = key[:1]
617
            else:
618
                prefix = ()
619
            parents = [prefix + (parent,) for parent in meta['parents']]
620
            vf_records.append((key, parents, meta['sha1'], d_func(text)))
621
        versionedfile.add_mpdiffs(vf_records)
622
4543.2.17 by John Arbash Meinel
Adding a parent inventory cache, and then using add_inventory_by_delta.
623
    def _get_parent_inventory_texts(self, inventory_text_cache,
624
                                    inventory_cache, parent_ids):
4543.2.16 by John Arbash Meinel
Adding an inventory text cache.
625
        cached_parent_texts = {}
626
        remaining_parent_ids = []
627
        for parent_id in parent_ids:
628
            p_text = inventory_text_cache.get(parent_id, None)
629
            if p_text is None:
630
                remaining_parent_ids.append(parent_id)
631
            else:
632
                cached_parent_texts[parent_id] = p_text
633
        ghosts = ()
4543.2.17 by John Arbash Meinel
Adding a parent inventory cache, and then using add_inventory_by_delta.
634
        # TODO: Use inventory_cache to grab inventories we already have in
635
        #       memory
4543.2.16 by John Arbash Meinel
Adding an inventory text cache.
636
        if remaining_parent_ids:
637
            # first determine what keys are actually present in the local
638
            # inventories object (don't use revisions as they haven't been
639
            # installed yet.)
640
            parent_keys = [(r,) for r in remaining_parent_ids]
641
            present_parent_map = self._repository.inventories.get_parent_map(
642
                                        parent_keys)
643
            present_parent_ids = []
644
            ghosts = set()
645
            for p_id in remaining_parent_ids:
646
                if (p_id,) in present_parent_map:
647
                    present_parent_ids.append(p_id)
648
                else:
649
                    ghosts.add(p_id)
650
            to_string = self._source_serializer.write_inventory_to_string
651
            for parent_inv in self._repository.iter_inventories(
652
                                    present_parent_ids):
653
                p_text = to_string(parent_inv)
4543.2.17 by John Arbash Meinel
Adding a parent inventory cache, and then using add_inventory_by_delta.
654
                inventory_cache[parent_inv.revision_id] = parent_inv
4543.2.16 by John Arbash Meinel
Adding an inventory text cache.
655
                cached_parent_texts[parent_inv.revision_id] = p_text
656
                inventory_text_cache[parent_inv.revision_id] = p_text
657
658
        parent_texts = [cached_parent_texts[parent_id]
659
                        for parent_id in parent_ids
660
                         if parent_id not in ghosts]
661
        return parent_texts
662
3350.6.4 by Robert Collins
First cut at pluralised VersionedFiles. Some rather massive API incompatabilities, primarily because of the difficulty of coherence among competing stores.
663
    def _install_inventory_records(self, records):
4543.2.3 by John Arbash Meinel
Change the name to test_merge_directive
664
        if (self._info['serializer'] == self._repository._serializer.format_num
665
            and self._repository._serializer.support_altered_by_hack):
3350.6.4 by Robert Collins
First cut at pluralised VersionedFiles. Some rather massive API incompatabilities, primarily because of the difficulty of coherence among competing stores.
666
            return self._install_mp_records_keys(self._repository.inventories,
667
                records)
4543.2.16 by John Arbash Meinel
Adding an inventory text cache.
668
        # Use a 10MB text cache, since these are string xml inventories. Note
669
        # that 10MB is fairly small for large projects (a single inventory can
670
        # be >5MB). Another possibility is to cache 10-20 inventory texts
671
        # instead
672
        inventory_text_cache = lru_cache.LRUSizeCache(10*1024*1024)
4543.2.21 by John Arbash Meinel
A few more tiny tweaks to comments, etc.
673
        # Also cache the in-memory representation. This allows us to create
674
        # inventory deltas to apply rather than calling add_inventory from
675
        # scratch each time.
4543.2.17 by John Arbash Meinel
Adding a parent inventory cache, and then using add_inventory_by_delta.
676
        inventory_cache = lru_cache.LRUCache(10)
6861.4.1 by Jelmer Vernooij
Make progress bars context managers.
677
        with ui.ui_factory.nested_progress_bar() as pb:
4543.2.16 by John Arbash Meinel
Adding an inventory text cache.
678
            num_records = len(records)
679
            for idx, (key, metadata, bytes) in enumerate(records):
680
                pb.update('installing inventory', idx, num_records)
681
                revision_id = key[-1]
682
                parent_ids = metadata['parents']
683
                # Note: This assumes the local ghosts are identical to the
684
                #       ghosts in the source, as the Bundle serialization
685
                #       format doesn't record ghosts.
686
                p_texts = self._get_parent_inventory_texts(inventory_text_cache,
4543.2.17 by John Arbash Meinel
Adding a parent inventory cache, and then using add_inventory_by_delta.
687
                                                           inventory_cache,
4543.2.16 by John Arbash Meinel
Adding an inventory text cache.
688
                                                           parent_ids)
689
                # Why does to_lines() take strings as the source, it seems that
690
                # it would have to cast to a list of lines, which we get back
691
                # as lines and then cast back to a string.
692
                target_lines = multiparent.MultiParent.from_patch(bytes
693
                            ).to_lines(p_texts)
694
                inv_text = ''.join(target_lines)
695
                del target_lines
696
                sha1 = osutils.sha_string(inv_text)
697
                if sha1 != metadata['sha1']:
698
                    raise errors.BadBundle("Can't convert to target format")
699
                # Add this to the cache so we don't have to extract it again.
700
                inventory_text_cache[revision_id] = inv_text
701
                target_inv = self._source_serializer.read_inventory_from_string(
702
                    inv_text)
703
                self._handle_root(target_inv, parent_ids)
4543.2.17 by John Arbash Meinel
Adding a parent inventory cache, and then using add_inventory_by_delta.
704
                parent_inv = None
705
                if parent_ids:
706
                    parent_inv = inventory_cache.get(parent_ids[0], None)
4543.2.16 by John Arbash Meinel
Adding an inventory text cache.
707
                try:
4543.2.17 by John Arbash Meinel
Adding a parent inventory cache, and then using add_inventory_by_delta.
708
                    if parent_inv is None:
709
                        self._repository.add_inventory(revision_id, target_inv,
710
                                                       parent_ids)
711
                    else:
712
                        delta = target_inv._make_delta(parent_inv)
713
                        self._repository.add_inventory_by_delta(parent_ids[0],
714
                            delta, revision_id, parent_ids)
4543.2.16 by John Arbash Meinel
Adding an inventory text cache.
715
                except errors.UnsupportedInventoryKind:
716
                    raise errors.IncompatibleRevision(repr(self._repository))
4543.2.17 by John Arbash Meinel
Adding a parent inventory cache, and then using add_inventory_by_delta.
717
                inventory_cache[revision_id] = target_inv
2520.4.99 by Aaron Bentley
Test conversion across models
718
719
    def _handle_root(self, target_inv, parent_ids):
720
        revision_id = target_inv.revision_id
721
        if self.update_root:
3350.6.4 by Robert Collins
First cut at pluralised VersionedFiles. Some rather massive API incompatabilities, primarily because of the difficulty of coherence among competing stores.
722
            text_key = (target_inv.root.file_id, revision_id)
723
            parent_keys = [(target_inv.root.file_id, parent) for
724
                parent in parent_ids]
725
            self._repository.texts.add_lines(text_key, parent_keys, [])
2520.4.99 by Aaron Bentley
Test conversion across models
726
        elif not self._repository.supports_rich_root():
727
            if target_inv.root.revision != revision_id:
728
                raise errors.IncompatibleRevision(repr(self._repository))
729
2520.4.59 by Aaron Bentley
Push metadata down the stack
730
    def _install_revision(self, revision_id, metadata, text):
2520.4.14 by Aaron Bentley
Get most tests passing, use format header
731
        if self._repository.has_revision(revision_id):
732
            return
3350.6.4 by Robert Collins
First cut at pluralised VersionedFiles. Some rather massive API incompatabilities, primarily because of the difficulty of coherence among competing stores.
733
        revision = self._source_serializer.read_revision_from_string(text)
734
        self._repository.add_revision(revision.revision_id, revision)
2520.4.34 by Aaron Bentley
Add signature support
735
2520.4.59 by Aaron Bentley
Push metadata down the stack
736
    def _install_signature(self, revision_id, metadata, text):
2520.4.100 by Aaron Bentley
Fix repeat signature installs
737
        transaction = self._repository.get_transaction()
3350.6.4 by Robert Collins
First cut at pluralised VersionedFiles. Some rather massive API incompatabilities, primarily because of the difficulty of coherence among competing stores.
738
        if self._repository.has_signature_for_revision_id(revision_id):
2520.4.100 by Aaron Bentley
Fix repeat signature installs
739
            return
3350.6.4 by Robert Collins
First cut at pluralised VersionedFiles. Some rather massive API incompatabilities, primarily because of the difficulty of coherence among competing stores.
740
        self._repository.add_signature_text(revision_id, text)