/brz/remove-bazaar

To get this branch, use:
bzr branch http://gegoxaren.bato24.eu/bzr/brz/remove-bazaar
2220.2.1 by Martin Pool
Start adding space for tags stored in the repository
1
# Copyright (C) 2005, 2006, 2007 Canonical Ltd
1887.1.1 by Adeodato Simó
Do not separate paragraphs in the copyright statement with blank lines,
2
#
1185.65.1 by Aaron Bentley
Refactored out ControlFiles and RevisionStore from _Branch
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.
1887.1.1 by Adeodato Simó
Do not separate paragraphs in the copyright statement with blank lines,
7
#
1185.65.1 by Aaron Bentley
Refactored out ControlFiles and RevisionStore from _Branch
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.
1887.1.1 by Adeodato Simó
Do not separate paragraphs in the copyright statement with blank lines,
12
#
1185.65.1 by Aaron Bentley
Refactored out ControlFiles and RevisionStore from _Branch
13
# You should have received a copy of the GNU General Public License
14
# along with this program; if not, write to the Free Software
15
# Foundation, Inc., 59 Temple Place, Suite 330, Boston, MA  02111-1307  USA
1185.65.10 by Robert Collins
Rename Controlfiles to LockableFiles.
16
1996.3.4 by John Arbash Meinel
lazy_import bzrlib/repository.py
17
from cStringIO import StringIO
18
19
from bzrlib.lazy_import import lazy_import
20
lazy_import(globals(), """
1740.3.7 by Jelmer Vernooij
Move committer, log, revprops, timestamp and timezone to CommitBuilder.
21
import re
22
import time
1996.3.4 by John Arbash Meinel
lazy_import bzrlib/repository.py
23
1910.2.22 by Aaron Bentley
Make commits preserve root entry data
24
from bzrlib import (
1996.3.4 by John Arbash Meinel
lazy_import bzrlib/repository.py
25
    bzrdir,
26
    check,
2745.1.1 by Robert Collins
Add a number of -Devil checkpoints.
27
    debug,
2490.2.21 by Aaron Bentley
Rename graph to deprecated_graph
28
    deprecated_graph,
1996.3.4 by John Arbash Meinel
lazy_import bzrlib/repository.py
29
    errors,
2116.4.1 by John Arbash Meinel
Update file and revision id generators.
30
    generate_ids,
1996.3.4 by John Arbash Meinel
lazy_import bzrlib/repository.py
31
    gpg,
32
    graph,
2163.2.1 by John Arbash Meinel
Speed up the fileids_altered_by_revision_ids processing
33
    lazy_regex,
1996.3.4 by John Arbash Meinel
lazy_import bzrlib/repository.py
34
    lockable_files,
35
    lockdir,
2988.1.5 by Robert Collins
Use a LRU cache when generating the text index to reduce inventory deserialisations.
36
    lru_cache,
1910.2.22 by Aaron Bentley
Make commits preserve root entry data
37
    osutils,
2241.1.1 by Martin Pool
Change RepositoryFormat to use a Registry rather than ad-hoc dictionary
38
    registry,
2018.5.163 by Andrew Bennetts
Deal with various review comments from Robert.
39
    remote,
1996.3.4 by John Arbash Meinel
lazy_import bzrlib/repository.py
40
    revision as _mod_revision,
41
    symbol_versioning,
1910.2.22 by Aaron Bentley
Make commits preserve root entry data
42
    transactions,
2988.1.3 by Robert Collins
Add a new repositoy method _generate_text_key_index for use by reconcile/check.
43
    tsort,
1996.3.4 by John Arbash Meinel
lazy_import bzrlib/repository.py
44
    ui,
45
    )
2520.4.54 by Aaron Bentley
Hang a create_bundle method off repository
46
from bzrlib.bundle import serializer
1996.3.4 by John Arbash Meinel
lazy_import bzrlib/repository.py
47
from bzrlib.revisiontree import RevisionTree
48
from bzrlib.store.versioned import VersionedFileStore
49
from bzrlib.store.text import TextStore
50
from bzrlib.testament import Testament
2535.3.40 by Andrew Bennetts
Tidy up more XXXs.
51
from bzrlib.util import bencode
1996.3.4 by John Arbash Meinel
lazy_import bzrlib/repository.py
52
""")
53
1534.4.28 by Robert Collins
first cut at merge from integration.
54
from bzrlib.decorators import needs_read_lock, needs_write_lock
1563.2.12 by Robert Collins
Checkpointing: created InterObject to factor out common inter object worker code, added InterVersionedFile and tests to allow making join work between any versionedfile.
55
from bzrlib.inter import InterObject
1910.2.3 by Aaron Bentley
All tests pass
56
from bzrlib.inventory import Inventory, InventoryDirectory, ROOT_ID
1996.3.4 by John Arbash Meinel
lazy_import bzrlib/repository.py
57
from bzrlib.symbol_versioning import (
58
        deprecated_method,
1773.4.1 by Martin Pool
Add pyflakes makefile target; fix many warnings
59
        )
2745.1.1 by Robert Collins
Add a number of -Devil checkpoints.
60
from bzrlib.trace import mutter, mutter_callsite, note, warning
1185.70.3 by Martin Pool
Various updates to make storage branch mergeable:
61
1185.65.1 by Aaron Bentley
Refactored out ControlFiles and RevisionStore from _Branch
62
1904.2.5 by Martin Pool
Fix format warning inside test suite and add test
63
# Old formats display a warning, but only once
64
_deprecation_warning_done = False
65
66
2592.3.135 by Robert Collins
Do not create many transient knit objects, saving 4% on commit.
67
class CommitBuilder(object):
68
    """Provides an interface to build up a commit.
69
70
    This allows describing a tree to be committed without needing to 
71
    know the internals of the format of the repository.
72
    """
73
    
74
    # all clients should supply tree roots.
75
    record_root_entry = True
2825.5.2 by Robert Collins
Review feedback, and fix pointless commits with nested trees to raise PointlessCommit appropriately.
76
    # the default CommitBuilder does not manage trees whose root is versioned.
77
    _versioned_root = False
2592.3.135 by Robert Collins
Do not create many transient knit objects, saving 4% on commit.
78
2979.2.2 by Robert Collins
Per-file graph heads detection during commit for pack repositories.
79
    def __init__(self, repository, parents, config, timestamp=None,
80
                 timezone=None, committer=None, revprops=None,
2592.3.135 by Robert Collins
Do not create many transient knit objects, saving 4% on commit.
81
                 revision_id=None):
82
        """Initiate a CommitBuilder.
83
84
        :param repository: Repository to commit to.
85
        :param parents: Revision ids of the parents of the new revision.
86
        :param config: Configuration to use.
87
        :param timestamp: Optional timestamp recorded for commit.
88
        :param timezone: Optional timezone for timestamp.
89
        :param committer: Optional committer to set for commit.
90
        :param revprops: Optional dictionary of revision properties.
91
        :param revision_id: Optional revision id.
92
        """
93
        self._config = config
94
95
        if committer is None:
96
            self._committer = self._config.username()
97
        else:
98
            assert isinstance(committer, basestring), type(committer)
99
            self._committer = committer
100
101
        self.new_inventory = Inventory(None)
2858.2.1 by Martin Pool
Remove most calls to safe_file_id and safe_revision_id.
102
        self._new_revision_id = revision_id
2592.3.135 by Robert Collins
Do not create many transient knit objects, saving 4% on commit.
103
        self.parents = parents
104
        self.repository = repository
105
106
        self._revprops = {}
107
        if revprops is not None:
108
            self._revprops.update(revprops)
109
110
        if timestamp is None:
111
            timestamp = time.time()
112
        # Restrict resolution to 1ms
113
        self._timestamp = round(timestamp, 3)
114
115
        if timezone is None:
116
            self._timezone = osutils.local_time_offset()
117
        else:
118
            self._timezone = int(timezone)
119
120
        self._generate_revision_if_needed()
2979.2.5 by Robert Collins
Make CommitBuilder.heads be _heads as its internal to CommitBuilder only.
121
        self.__heads = graph.HeadsCache(repository.get_graph()).heads
2592.3.135 by Robert Collins
Do not create many transient knit objects, saving 4% on commit.
122
123
    def commit(self, message):
124
        """Make the actual commit.
125
126
        :return: The revision id of the recorded revision.
127
        """
128
        rev = _mod_revision.Revision(
129
                       timestamp=self._timestamp,
130
                       timezone=self._timezone,
131
                       committer=self._committer,
132
                       message=message,
133
                       inventory_sha1=self.inv_sha1,
134
                       revision_id=self._new_revision_id,
135
                       properties=self._revprops)
136
        rev.parent_ids = self.parents
137
        self.repository.add_revision(self._new_revision_id, rev,
138
            self.new_inventory, self._config)
139
        self.repository.commit_write_group()
140
        return self._new_revision_id
141
142
    def abort(self):
143
        """Abort the commit that is being built.
144
        """
145
        self.repository.abort_write_group()
146
147
    def revision_tree(self):
148
        """Return the tree that was just committed.
149
150
        After calling commit() this can be called to get a RevisionTree
151
        representing the newly committed tree. This is preferred to
152
        calling Repository.revision_tree() because that may require
153
        deserializing the inventory, while we already have a copy in
154
        memory.
155
        """
156
        return RevisionTree(self.repository, self.new_inventory,
157
                            self._new_revision_id)
158
159
    def finish_inventory(self):
160
        """Tell the builder that the inventory is finished."""
161
        if self.new_inventory.root is None:
2903.2.9 by Martin Pool
Review cleanups, mostly documentation
162
            raise AssertionError('Root entry should be supplied to'
2592.3.135 by Robert Collins
Do not create many transient knit objects, saving 4% on commit.
163
                ' record_entry_contents, as of bzr 0.10.',
164
                 DeprecationWarning, stacklevel=2)
165
            self.new_inventory.add(InventoryDirectory(ROOT_ID, '', None))
166
        self.new_inventory.revision_id = self._new_revision_id
167
        self.inv_sha1 = self.repository.add_inventory(
168
            self._new_revision_id,
169
            self.new_inventory,
170
            self.parents
171
            )
172
173
    def _gen_revision_id(self):
174
        """Return new revision-id."""
175
        return generate_ids.gen_revision_id(self._config.username(),
176
                                            self._timestamp)
177
178
    def _generate_revision_if_needed(self):
179
        """Create a revision id if None was supplied.
180
        
181
        If the repository can not support user-specified revision ids
182
        they should override this function and raise CannotSetRevisionId
183
        if _new_revision_id is not None.
184
185
        :raises: CannotSetRevisionId
186
        """
187
        if self._new_revision_id is None:
188
            self._new_revision_id = self._gen_revision_id()
189
            self.random_revid = True
190
        else:
191
            self.random_revid = False
192
2979.2.5 by Robert Collins
Make CommitBuilder.heads be _heads as its internal to CommitBuilder only.
193
    def _heads(self, file_id, revision_ids):
2979.2.1 by Robert Collins
Make it possible for different commit builders to override heads().
194
        """Calculate the graph heads for revision_ids in the graph of file_id.
195
196
        This can use either a per-file graph or a global revision graph as we
197
        have an identity relationship between the two graphs.
198
        """
2979.2.5 by Robert Collins
Make CommitBuilder.heads be _heads as its internal to CommitBuilder only.
199
        return self.__heads(revision_ids)
2979.2.1 by Robert Collins
Make it possible for different commit builders to override heads().
200
2592.3.135 by Robert Collins
Do not create many transient knit objects, saving 4% on commit.
201
    def _check_root(self, ie, parent_invs, tree):
202
        """Helper for record_entry_contents.
203
204
        :param ie: An entry being added.
205
        :param parent_invs: The inventories of the parent revisions of the
206
            commit.
207
        :param tree: The tree that is being committed.
208
        """
2871.1.2 by Robert Collins
* ``CommitBuilder.record_entry_contents`` now requires the root entry of a
209
        # In this revision format, root entries have no knit or weave When
210
        # serializing out to disk and back in root.revision is always
211
        # _new_revision_id
212
        ie.revision = self._new_revision_id
2818.3.1 by Robert Collins
Change CommitBuilder factory delegation to allow simple declaration.
213
2871.1.4 by Robert Collins
Merge bzr.dev.
214
    def _get_delta(self, ie, basis_inv, path):
215
        """Get a delta against the basis inventory for ie."""
216
        if ie.file_id not in basis_inv:
217
            # add
218
            return (None, path, ie.file_id, ie)
219
        elif ie != basis_inv[ie.file_id]:
220
            # common but altered
221
            # TODO: avoid tis id2path call.
222
            return (basis_inv.id2path(ie.file_id), path, ie.file_id, ie)
2592.3.135 by Robert Collins
Do not create many transient knit objects, saving 4% on commit.
223
        else:
2871.1.4 by Robert Collins
Merge bzr.dev.
224
            # common, unaltered
225
            return None
2592.3.135 by Robert Collins
Do not create many transient knit objects, saving 4% on commit.
226
227
    def record_entry_contents(self, ie, parent_invs, path, tree,
228
        content_summary):
229
        """Record the content of ie from tree into the commit if needed.
230
231
        Side effect: sets ie.revision when unchanged
232
233
        :param ie: An inventory entry present in the commit.
234
        :param parent_invs: The inventories of the parent revisions of the
235
            commit.
236
        :param path: The path the entry is at in the tree.
237
        :param tree: The tree which contains this entry and should be used to 
238
            obtain content.
239
        :param content_summary: Summary data from the tree about the paths
240
            content - stat, length, exec, sha/link target. This is only
241
            accessed when the entry has a revision of None - that is when it is
242
            a candidate to commit.
2871.1.3 by Robert Collins
* The CommitBuilder method ``record_entry_contents`` now returns summary
243
        :return: A tuple (change_delta, version_recorded). change_delta is 
244
            an inventory_delta change for this entry against the basis tree of
245
            the commit, or None if no change occured against the basis tree.
246
            version_recorded is True if a new version of the entry has been
247
            recorded. For instance, committing a merge where a file was only
248
            changed on the other side will return (delta, False).
2592.3.135 by Robert Collins
Do not create many transient knit objects, saving 4% on commit.
249
        """
250
        if self.new_inventory.root is None:
2871.1.2 by Robert Collins
* ``CommitBuilder.record_entry_contents`` now requires the root entry of a
251
            if ie.parent_id is not None:
252
                raise errors.RootMissing()
2592.3.135 by Robert Collins
Do not create many transient knit objects, saving 4% on commit.
253
            self._check_root(ie, parent_invs, tree)
254
        if ie.revision is None:
255
            kind = content_summary[0]
256
        else:
257
            # ie is carried over from a prior commit
258
            kind = ie.kind
259
        # XXX: repository specific check for nested tree support goes here - if
260
        # the repo doesn't want nested trees we skip it ?
261
        if (kind == 'tree-reference' and
262
            not self.repository._format.supports_tree_reference):
263
            # mismatch between commit builder logic and repository:
264
            # this needs the entry creation pushed down into the builder.
2776.4.18 by Robert Collins
Review feedback.
265
            raise NotImplementedError('Missing repository subtree support.')
2592.3.135 by Robert Collins
Do not create many transient knit objects, saving 4% on commit.
266
        self.new_inventory.add(ie)
267
2871.1.3 by Robert Collins
* The CommitBuilder method ``record_entry_contents`` now returns summary
268
        # TODO: slow, take it out of the inner loop.
269
        try:
270
            basis_inv = parent_invs[0]
271
        except IndexError:
272
            basis_inv = Inventory(root_id=None)
273
2592.3.135 by Robert Collins
Do not create many transient knit objects, saving 4% on commit.
274
        # ie.revision is always None if the InventoryEntry is considered
2776.4.13 by Robert Collins
Merge bzr.dev.
275
        # for committing. We may record the previous parents revision if the
276
        # content is actually unchanged against a sole head.
2592.3.135 by Robert Collins
Do not create many transient knit objects, saving 4% on commit.
277
        if ie.revision is not None:
2903.2.5 by Martin Pool
record_entry_contents should give back deltas for changed roots; clean it up a bit
278
            if not self._versioned_root and path == '':
2871.1.3 by Robert Collins
* The CommitBuilder method ``record_entry_contents`` now returns summary
279
                # repositories that do not version the root set the root's
280
                # revision to the new commit even when no change occurs, and
281
                # this masks when a change may have occurred against the basis,
282
                # so calculate if one happened.
2903.2.5 by Martin Pool
record_entry_contents should give back deltas for changed roots; clean it up a bit
283
                if ie.file_id in basis_inv:
284
                    delta = (basis_inv.id2path(ie.file_id), path,
285
                        ie.file_id, ie)
286
                else:
2871.1.3 by Robert Collins
* The CommitBuilder method ``record_entry_contents`` now returns summary
287
                    # add
288
                    delta = (None, path, ie.file_id, ie)
2903.2.5 by Martin Pool
record_entry_contents should give back deltas for changed roots; clean it up a bit
289
                return delta, False
290
            else:
291
                # we don't need to commit this, because the caller already
292
                # determined that an existing revision of this file is
293
                # appropriate.
2903.2.9 by Martin Pool
Review cleanups, mostly documentation
294
                return None, (ie.revision == self._new_revision_id)
2592.3.135 by Robert Collins
Do not create many transient knit objects, saving 4% on commit.
295
        # XXX: Friction: parent_candidates should return a list not a dict
296
        #      so that we don't have to walk the inventories again.
297
        parent_candiate_entries = ie.parent_candidates(parent_invs)
2979.2.5 by Robert Collins
Make CommitBuilder.heads be _heads as its internal to CommitBuilder only.
298
        head_set = self._heads(ie.file_id, parent_candiate_entries.keys())
2592.3.135 by Robert Collins
Do not create many transient knit objects, saving 4% on commit.
299
        heads = []
300
        for inv in parent_invs:
301
            if ie.file_id in inv:
302
                old_rev = inv[ie.file_id].revision
303
                if old_rev in head_set:
304
                    heads.append(inv[ie.file_id].revision)
305
                    head_set.remove(inv[ie.file_id].revision)
306
307
        store = False
308
        # now we check to see if we need to write a new record to the
309
        # file-graph.
310
        # We write a new entry unless there is one head to the ancestors, and
311
        # the kind-derived content is unchanged.
312
313
        # Cheapest check first: no ancestors, or more the one head in the
314
        # ancestors, we write a new node.
315
        if len(heads) != 1:
316
            store = True
317
        if not store:
318
            # There is a single head, look it up for comparison
319
            parent_entry = parent_candiate_entries[heads[0]]
320
            # if the non-content specific data has changed, we'll be writing a
321
            # node:
322
            if (parent_entry.parent_id != ie.parent_id or
323
                parent_entry.name != ie.name):
324
                store = True
325
        # now we need to do content specific checks:
326
        if not store:
327
            # if the kind changed the content obviously has
328
            if kind != parent_entry.kind:
329
                store = True
330
        if kind == 'file':
2911.5.3 by John Arbash Meinel
Revert the changes to CommitBuilder.record_entry_contents.
331
            assert content_summary[2] is not None, \
332
                "Files must not have executable = None"
2592.3.135 by Robert Collins
Do not create many transient knit objects, saving 4% on commit.
333
            if not store:
334
                if (# if the file length changed we have to store:
335
                    parent_entry.text_size != content_summary[1] or
336
                    # if the exec bit has changed we have to store:
337
                    parent_entry.executable != content_summary[2]):
338
                    store = True
339
                elif parent_entry.text_sha1 == content_summary[3]:
340
                    # all meta and content is unchanged (using a hash cache
341
                    # hit to check the sha)
342
                    ie.revision = parent_entry.revision
343
                    ie.text_size = parent_entry.text_size
344
                    ie.text_sha1 = parent_entry.text_sha1
345
                    ie.executable = parent_entry.executable
2871.1.4 by Robert Collins
Merge bzr.dev.
346
                    return self._get_delta(ie, basis_inv, path), False
2592.3.135 by Robert Collins
Do not create many transient knit objects, saving 4% on commit.
347
                else:
348
                    # Either there is only a hash change(no hash cache entry,
349
                    # or same size content change), or there is no change on
350
                    # this file at all.
2776.4.19 by Robert Collins
Final review tweaks.
351
                    # Provide the parent's hash to the store layer, so that the
352
                    # content is unchanged we will not store a new node.
2592.3.135 by Robert Collins
Do not create many transient knit objects, saving 4% on commit.
353
                    nostore_sha = parent_entry.text_sha1
354
            if store:
2776.4.18 by Robert Collins
Review feedback.
355
                # We want to record a new node regardless of the presence or
356
                # absence of a content change in the file.
2592.3.135 by Robert Collins
Do not create many transient knit objects, saving 4% on commit.
357
                nostore_sha = None
2776.4.18 by Robert Collins
Review feedback.
358
            ie.executable = content_summary[2]
359
            lines = tree.get_file(ie.file_id, path).readlines()
2592.3.135 by Robert Collins
Do not create many transient knit objects, saving 4% on commit.
360
            try:
361
                ie.text_sha1, ie.text_size = self._add_text_to_weave(
362
                    ie.file_id, lines, heads, nostore_sha)
363
            except errors.ExistingContent:
2776.4.18 by Robert Collins
Review feedback.
364
                # Turns out that the file content was unchanged, and we were
365
                # only going to store a new node if it was changed. Carry over
366
                # the entry.
2592.3.135 by Robert Collins
Do not create many transient knit objects, saving 4% on commit.
367
                ie.revision = parent_entry.revision
368
                ie.text_size = parent_entry.text_size
369
                ie.text_sha1 = parent_entry.text_sha1
370
                ie.executable = parent_entry.executable
2871.1.4 by Robert Collins
Merge bzr.dev.
371
                return self._get_delta(ie, basis_inv, path), False
2592.3.135 by Robert Collins
Do not create many transient knit objects, saving 4% on commit.
372
        elif kind == 'directory':
373
            if not store:
374
                # all data is meta here, nothing specific to directory, so
375
                # carry over:
376
                ie.revision = parent_entry.revision
2871.1.4 by Robert Collins
Merge bzr.dev.
377
                return self._get_delta(ie, basis_inv, path), False
2592.3.135 by Robert Collins
Do not create many transient knit objects, saving 4% on commit.
378
            lines = []
379
            self._add_text_to_weave(ie.file_id, lines, heads, None)
380
        elif kind == 'symlink':
381
            current_link_target = content_summary[3]
382
            if not store:
2776.4.18 by Robert Collins
Review feedback.
383
                # symlink target is not generic metadata, check if it has
2592.3.135 by Robert Collins
Do not create many transient knit objects, saving 4% on commit.
384
                # changed.
385
                if current_link_target != parent_entry.symlink_target:
386
                    store = True
387
            if not store:
388
                # unchanged, carry over.
389
                ie.revision = parent_entry.revision
390
                ie.symlink_target = parent_entry.symlink_target
2871.1.4 by Robert Collins
Merge bzr.dev.
391
                return self._get_delta(ie, basis_inv, path), False
2592.3.135 by Robert Collins
Do not create many transient knit objects, saving 4% on commit.
392
            ie.symlink_target = current_link_target
393
            lines = []
394
            self._add_text_to_weave(ie.file_id, lines, heads, None)
395
        elif kind == 'tree-reference':
396
            if not store:
397
                if content_summary[3] != parent_entry.reference_revision:
398
                    store = True
399
            if not store:
400
                # unchanged, carry over.
401
                ie.reference_revision = parent_entry.reference_revision
402
                ie.revision = parent_entry.revision
2871.1.4 by Robert Collins
Merge bzr.dev.
403
                return self._get_delta(ie, basis_inv, path), False
2592.3.135 by Robert Collins
Do not create many transient knit objects, saving 4% on commit.
404
            ie.reference_revision = content_summary[3]
405
            lines = []
406
            self._add_text_to_weave(ie.file_id, lines, heads, None)
407
        else:
408
            raise NotImplementedError('unknown kind')
409
        ie.revision = self._new_revision_id
2871.1.4 by Robert Collins
Merge bzr.dev.
410
        return self._get_delta(ie, basis_inv, path), True
2592.3.135 by Robert Collins
Do not create many transient knit objects, saving 4% on commit.
411
412
    def _add_text_to_weave(self, file_id, new_lines, parents, nostore_sha):
413
        versionedfile = self.repository.weave_store.get_weave_or_empty(
414
            file_id, self.repository.get_transaction())
415
        # Don't change this to add_lines - add_lines_with_ghosts is cheaper
416
        # than add_lines, and allows committing when a parent is ghosted for
417
        # some reason.
2592.3.136 by Robert Collins
Merge bzr.dev.
418
        # Note: as we read the content directly from the tree, we know its not
419
        # been turned into unicode or badly split - but a broken tree
420
        # implementation could give us bad output from readlines() so this is
421
        # not a guarantee of safety. What would be better is always checking
422
        # the content during test suite execution. RBC 20070912
2592.3.135 by Robert Collins
Do not create many transient knit objects, saving 4% on commit.
423
        try:
424
            return versionedfile.add_lines_with_ghosts(
425
                self._new_revision_id, parents, new_lines,
2592.3.136 by Robert Collins
Merge bzr.dev.
426
                nostore_sha=nostore_sha, random_id=self.random_revid,
427
                check_content=False)[0:2]
2592.3.135 by Robert Collins
Do not create many transient knit objects, saving 4% on commit.
428
        finally:
429
            versionedfile.clear_cache()
430
431
432
class RootCommitBuilder(CommitBuilder):
433
    """This commitbuilder actually records the root id"""
434
    
2825.5.2 by Robert Collins
Review feedback, and fix pointless commits with nested trees to raise PointlessCommit appropriately.
435
    # the root entry gets versioned properly by this builder.
2840.1.1 by Ian Clatworthy
faster pointless commit detection (Robert Collins)
436
    _versioned_root = True
2825.5.2 by Robert Collins
Review feedback, and fix pointless commits with nested trees to raise PointlessCommit appropriately.
437
2592.3.135 by Robert Collins
Do not create many transient knit objects, saving 4% on commit.
438
    def _check_root(self, ie, parent_invs, tree):
439
        """Helper for record_entry_contents.
440
441
        :param ie: An entry being added.
442
        :param parent_invs: The inventories of the parent revisions of the
443
            commit.
444
        :param tree: The tree that is being committed.
445
        """
446
447
2220.2.3 by Martin Pool
Add tag: revision namespace.
448
######################################################################
449
# Repositories
450
1185.66.5 by Aaron Bentley
Renamed RevisionStorage to Repository
451
class Repository(object):
1185.70.3 by Martin Pool
Various updates to make storage branch mergeable:
452
    """Repository holding history for one or more branches.
453
454
    The repository holds and retrieves historical information including
455
    revisions and file history.  It's normally accessed only by the Branch,
456
    which views a particular line of development through that history.
457
458
    The Repository builds on top of Stores and a Transport, which respectively 
459
    describe the disk data format and the way of accessing the (possibly 
460
    remote) disk.
461
    """
1185.65.17 by Robert Collins
Merge from integration, mode-changes are broken.
462
2592.3.135 by Robert Collins
Do not create many transient knit objects, saving 4% on commit.
463
    # What class to use for a CommitBuilder. Often its simpler to change this
464
    # in a Repository class subclass rather than to override
465
    # get_commit_builder.
466
    _commit_builder_class = CommitBuilder
467
    # The search regex used by xml based repositories to determine what things
468
    # where changed in a single commit.
2163.2.1 by John Arbash Meinel
Speed up the fileids_altered_by_revision_ids processing
469
    _file_ids_altered_regex = lazy_regex.lazy_compile(
470
        r'file_id="(?P<file_id>[^"]+)"'
2776.4.6 by Robert Collins
Fixup various commit test failures falling out from the other commit changes.
471
        r'.* revision="(?P<revision_id>[^"]+)"'
2163.2.1 by John Arbash Meinel
Speed up the fileids_altered_by_revision_ids processing
472
        )
473
2617.6.2 by Robert Collins
Add abort_write_group and wire write_groups into fetch and commit.
474
    def abort_write_group(self):
475
        """Commit the contents accrued within the current write group.
476
477
        :seealso: start_write_group.
478
        """
479
        if self._write_group is not self.get_transaction():
480
            # has an unlock or relock occured ?
481
            raise errors.BzrError('mismatched lock context and write group.')
482
        self._abort_write_group()
483
        self._write_group = None
484
485
    def _abort_write_group(self):
486
        """Template method for per-repository write group cleanup.
487
        
488
        This is called during abort before the write group is considered to be 
489
        finished and should cleanup any internal state accrued during the write
490
        group. There is no requirement that data handed to the repository be
491
        *not* made available - this is not a rollback - but neither should any
492
        attempt be made to ensure that data added is fully commited. Abort is
493
        invoked when an error has occured so futher disk or network operations
494
        may not be possible or may error and if possible should not be
495
        attempted.
496
        """
497
1570.1.2 by Robert Collins
Import bzrtools' 'fix' command as 'bzr reconcile.'
498
    @needs_write_lock
2249.5.12 by John Arbash Meinel
Change the APIs for VersionedFile, Store, and some of Repository into utf-8
499
    def add_inventory(self, revision_id, inv, parents):
500
        """Add the inventory inv to the repository as revision_id.
1570.1.2 by Robert Collins
Import bzrtools' 'fix' command as 'bzr reconcile.'
501
        
2249.5.12 by John Arbash Meinel
Change the APIs for VersionedFile, Store, and some of Repository into utf-8
502
        :param parents: The revision ids of the parents that revision_id
1570.1.2 by Robert Collins
Import bzrtools' 'fix' command as 'bzr reconcile.'
503
                        is known to have and are in the repository already.
504
505
        returns the sha1 of the serialized inventory.
506
        """
2592.3.61 by Robert Collins
Remove inventory.kndx.
507
        assert self.is_in_write_group()
2249.5.12 by John Arbash Meinel
Change the APIs for VersionedFile, Store, and some of Repository into utf-8
508
        _mod_revision.check_not_reserved_id(revision_id)
509
        assert inv.revision_id is None or inv.revision_id == revision_id, \
1740.2.2 by Aaron Bentley
Add test for the basis inventory automatically adding the revision id.
510
            "Mismatch between inventory revision" \
2249.5.12 by John Arbash Meinel
Change the APIs for VersionedFile, Store, and some of Repository into utf-8
511
            " id and insertion revid (%r, %r)" % (inv.revision_id, revision_id)
1910.2.6 by Aaron Bentley
Update for merge review, handle deprecations
512
        assert inv.root is not None
2817.2.1 by Robert Collins
* Inventory serialisation no longer double-sha's the content.
513
        inv_lines = self._serialise_inventory_to_lines(inv)
2592.3.61 by Robert Collins
Remove inventory.kndx.
514
        inv_vf = self.get_inventory_weave()
2817.2.1 by Robert Collins
* Inventory serialisation no longer double-sha's the content.
515
        return self._inventory_add_lines(inv_vf, revision_id, parents,
516
            inv_lines, check_content=False)
1570.1.2 by Robert Collins
Import bzrtools' 'fix' command as 'bzr reconcile.'
517
2805.6.7 by Robert Collins
Review feedback.
518
    def _inventory_add_lines(self, inv_vf, revision_id, parents, lines,
519
        check_content=True):
2817.2.1 by Robert Collins
* Inventory serialisation no longer double-sha's the content.
520
        """Store lines in inv_vf and return the sha1 of the inventory."""
1740.3.6 by Jelmer Vernooij
Move inventory writing to the commit builder.
521
        final_parents = []
522
        for parent in parents:
523
            if parent in inv_vf:
524
                final_parents.append(parent)
2817.2.1 by Robert Collins
* Inventory serialisation no longer double-sha's the content.
525
        return inv_vf.add_lines(revision_id, final_parents, lines,
526
            check_content=check_content)[0]
1740.3.6 by Jelmer Vernooij
Move inventory writing to the commit builder.
527
1570.1.2 by Robert Collins
Import bzrtools' 'fix' command as 'bzr reconcile.'
528
    @needs_write_lock
2249.5.12 by John Arbash Meinel
Change the APIs for VersionedFile, Store, and some of Repository into utf-8
529
    def add_revision(self, revision_id, rev, inv=None, config=None):
530
        """Add rev to the revision store as revision_id.
1570.1.2 by Robert Collins
Import bzrtools' 'fix' command as 'bzr reconcile.'
531
2249.5.12 by John Arbash Meinel
Change the APIs for VersionedFile, Store, and some of Repository into utf-8
532
        :param revision_id: the revision id to use.
1570.1.2 by Robert Collins
Import bzrtools' 'fix' command as 'bzr reconcile.'
533
        :param rev: The revision object.
534
        :param inv: The inventory for the revision. if None, it will be looked
535
                    up in the inventory storer
536
        :param config: If None no digital signature will be created.
537
                       If supplied its signature_needed method will be used
538
                       to determine if a signature should be made.
539
        """
2249.5.13 by John Arbash Meinel
Finish auditing Repository, and fix generate_ids to always generate utf8 ids.
540
        # TODO: jam 20070210 Shouldn't we check rev.revision_id and
541
        #       rev.parent_ids?
2249.5.12 by John Arbash Meinel
Change the APIs for VersionedFile, Store, and some of Repository into utf-8
542
        _mod_revision.check_not_reserved_id(revision_id)
1570.1.2 by Robert Collins
Import bzrtools' 'fix' command as 'bzr reconcile.'
543
        if config is not None and config.signature_needed():
544
            if inv is None:
2249.5.12 by John Arbash Meinel
Change the APIs for VersionedFile, Store, and some of Repository into utf-8
545
                inv = self.get_inventory(revision_id)
1570.1.2 by Robert Collins
Import bzrtools' 'fix' command as 'bzr reconcile.'
546
            plaintext = Testament(rev, inv).as_short_text()
547
            self.store_revision_signature(
2249.5.12 by John Arbash Meinel
Change the APIs for VersionedFile, Store, and some of Repository into utf-8
548
                gpg.GPGStrategy(config), plaintext, revision_id)
549
        if not revision_id in self.get_inventory_weave():
1570.1.2 by Robert Collins
Import bzrtools' 'fix' command as 'bzr reconcile.'
550
            if inv is None:
2249.5.12 by John Arbash Meinel
Change the APIs for VersionedFile, Store, and some of Repository into utf-8
551
                raise errors.WeaveRevisionNotPresent(revision_id,
1570.1.2 by Robert Collins
Import bzrtools' 'fix' command as 'bzr reconcile.'
552
                                                     self.get_inventory_weave())
553
            else:
554
                # yes, this is not suitable for adding with ghosts.
2249.5.12 by John Arbash Meinel
Change the APIs for VersionedFile, Store, and some of Repository into utf-8
555
                self.add_inventory(revision_id, inv, rev.parent_ids)
1608.2.1 by Martin Pool
[merge] Storage filename escaping
556
        self._revision_store.add_revision(rev, self.get_transaction())
1570.1.2 by Robert Collins
Import bzrtools' 'fix' command as 'bzr reconcile.'
557
2520.4.10 by Aaron Bentley
Enable installation of revisions
558
    def _add_revision_text(self, revision_id, text):
559
        revision = self._revision_store._serializer.read_revision_from_string(
560
            text)
561
        self._revision_store._add_revision(revision, StringIO(text),
562
                                           self.get_transaction())
563
1732.2.4 by Martin Pool
Split check into Branch.check and Repository.check
564
    def all_revision_ids(self):
565
        """Returns a list of all the revision ids in the repository. 
566
567
        This is deprecated because code should generally work on the graph
568
        reachable from a particular revision, and ignore any other revisions
569
        that might be present.  There is no direct replacement method.
570
        """
2592.3.114 by Robert Collins
More evil mutterings.
571
        if 'evil' in debug.debug_flags:
2850.3.1 by Robert Collins
Move various weave specific code out of the base Repository class to weaverepo.py.
572
            mutter_callsite(2, "all_revision_ids is linear with history.")
1732.2.4 by Martin Pool
Split check into Branch.check and Repository.check
573
        return self._all_revision_ids()
574
575
    def _all_revision_ids(self):
1534.4.50 by Robert Collins
Got the bzrdir api straightened out, plenty of refactoring to use it pending, but the api is up and running.
576
        """Returns a list of all the revision ids in the repository. 
577
578
        These are in as much topological order as the underlying store can 
2850.3.1 by Robert Collins
Move various weave specific code out of the base Repository class to weaverepo.py.
579
        present.
1534.4.50 by Robert Collins
Got the bzrdir api straightened out, plenty of refactoring to use it pending, but the api is up and running.
580
        """
2850.3.1 by Robert Collins
Move various weave specific code out of the base Repository class to weaverepo.py.
581
        raise NotImplementedError(self._all_revision_ids)
1534.4.50 by Robert Collins
Got the bzrdir api straightened out, plenty of refactoring to use it pending, but the api is up and running.
582
1687.1.7 by Robert Collins
Teach Repository about break_lock.
583
    def break_lock(self):
584
        """Break a lock if one is present from another instance.
585
586
        Uses the ui factory to ask for confirmation if the lock may be from
587
        an active process.
588
        """
589
        self.control_files.break_lock()
590
1534.4.50 by Robert Collins
Got the bzrdir api straightened out, plenty of refactoring to use it pending, but the api is up and running.
591
    @needs_read_lock
592
    def _eliminate_revisions_not_present(self, revision_ids):
593
        """Check every revision id in revision_ids to see if we have it.
594
595
        Returns a set of the present revisions.
596
        """
1534.4.41 by Robert Collins
Branch now uses BzrDir reasonably sanely.
597
        result = []
1534.4.50 by Robert Collins
Got the bzrdir api straightened out, plenty of refactoring to use it pending, but the api is up and running.
598
        for id in revision_ids:
1534.4.41 by Robert Collins
Branch now uses BzrDir reasonably sanely.
599
            if self.has_revision(id):
600
               result.append(id)
601
        return result
602
1534.4.40 by Robert Collins
Add RepositoryFormats and allow bzrdir.open or create _repository to be used.
603
    @staticmethod
604
    def create(a_bzrdir):
605
        """Construct the current default format repository in a_bzrdir."""
606
        return RepositoryFormat.get_default_format().initialize(a_bzrdir)
607
1563.2.29 by Robert Collins
Remove all but fetch references to repository.revision_store.
608
    def __init__(self, _format, a_bzrdir, control_files, _revision_store, control_store, text_store):
1556.1.3 by Robert Collins
Rearrangment of Repository logic to be less type code driven, and bugfix InterRepository.missing_revision_ids
609
        """instantiate a Repository.
610
611
        :param _format: The format of the repository on disk.
612
        :param a_bzrdir: The BzrDir of the repository.
613
614
        In the future we will have a single api for all stores for
615
        getting file texts, inventories and revisions, then
616
        this construct will accept instances of those things.
617
        """
1608.2.1 by Martin Pool
[merge] Storage filename escaping
618
        super(Repository, self).__init__()
1534.4.40 by Robert Collins
Add RepositoryFormats and allow bzrdir.open or create _repository to be used.
619
        self._format = _format
1556.1.3 by Robert Collins
Rearrangment of Repository logic to be less type code driven, and bugfix InterRepository.missing_revision_ids
620
        # the following are part of the public API for Repository:
1534.4.40 by Robert Collins
Add RepositoryFormats and allow bzrdir.open or create _repository to be used.
621
        self.bzrdir = a_bzrdir
1556.1.3 by Robert Collins
Rearrangment of Repository logic to be less type code driven, and bugfix InterRepository.missing_revision_ids
622
        self.control_files = control_files
1563.2.29 by Robert Collins
Remove all but fetch references to repository.revision_store.
623
        self._revision_store = _revision_store
1759.2.1 by Jelmer Vernooij
Fix some types (found using aspell).
624
        # backwards compatibility
1563.2.17 by Robert Collins
Change knits repositories to use a knit versioned file store for file texts.
625
        self.weave_store = text_store
2671.4.2 by Robert Collins
Review feedback.
626
        # for tests
627
        self._reconcile_does_inventory_gc = True
2745.6.16 by Aaron Bentley
Update from review
628
        self._reconcile_fixes_text_parents = False
2951.1.3 by Robert Collins
Partial support for native reconcile with packs.
629
        self._reconcile_backsup_inventory = True
1563.2.23 by Robert Collins
Add add_revision and get_revision methods to RevisionStore
630
        # not right yet - should be more semantically clear ? 
631
        # 
632
        self.control_store = control_store
633
        self.control_weaves = control_store
1608.2.1 by Martin Pool
[merge] Storage filename escaping
634
        # TODO: make sure to construct the right store classes, etc, depending
635
        # on whether escaping is required.
1904.2.3 by Martin Pool
Give a warning on access to old repository formats
636
        self._warn_if_deprecated()
2617.6.1 by Robert Collins
* New method on Repository - ``start_write_group``, ``end_write_group``
637
        self._write_group = None
2592.4.5 by Martin Pool
Add Repository.base on all repositories.
638
        self.base = control_files._transport.base
1185.65.1 by Aaron Bentley
Refactored out ControlFiles and RevisionStore from _Branch
639
1668.1.3 by Martin Pool
[patch] use the correct transaction when committing snapshot (Malone: #43959)
640
    def __repr__(self):
2592.4.5 by Martin Pool
Add Repository.base on all repositories.
641
        return '%s(%r)' % (self.__class__.__name__,
642
                           self.base)
1668.1.3 by Martin Pool
[patch] use the correct transaction when committing snapshot (Malone: #43959)
643
2671.1.4 by Andrew Bennetts
Rename is_same_repository to has_same_location, thanks Aaron!
644
    def has_same_location(self, other):
2671.1.3 by Andrew Bennetts
Remove Repository.__eq__/__ne__ methods, replace with is_same_repository method.
645
        """Returns a boolean indicating if this repository is at the same
646
        location as another repository.
647
648
        This might return False even when two repository objects are accessing
649
        the same physical repository via different URLs.
650
        """
2592.3.162 by Robert Collins
Remove some arbitrary differences from bzr.dev.
651
        if self.__class__ is not other.__class__:
652
            return False
653
        return (self.control_files._transport.base ==
654
                other.control_files._transport.base)
2671.1.1 by Andrew Bennetts
Add support for comparing Repositories with == and != operators.
655
2617.6.1 by Robert Collins
* New method on Repository - ``start_write_group``, ``end_write_group``
656
    def is_in_write_group(self):
657
        """Return True if there is an open write group.
658
659
        :seealso: start_write_group.
660
        """
661
        return self._write_group is not None
662
1694.2.6 by Martin Pool
[merge] bzr.dev
663
    def is_locked(self):
664
        return self.control_files.is_locked()
665
2592.3.188 by Robert Collins
Allow pack repositories to have multiple writers active at one time, for greater concurrency.
666
    def is_write_locked(self):
667
        """Return True if this object is write locked."""
668
        return self.is_locked() and self.control_files._lock_mode == 'w'
669
2018.5.75 by Andrew Bennetts
Add Repository.{dont_,}leave_lock_in_place.
670
    def lock_write(self, token=None):
671
        """Lock this repository for writing.
2617.6.8 by Robert Collins
Review feedback and documentation.
672
673
        This causes caching within the repository obejct to start accumlating
674
        data during reads, and allows a 'write_group' to be obtained. Write
675
        groups must be used for actual data insertion.
2018.5.75 by Andrew Bennetts
Add Repository.{dont_,}leave_lock_in_place.
676
        
677
        :param token: if this is already locked, then lock_write will fail
678
            unless the token matches the existing lock.
679
        :returns: a token if this instance supports tokens, otherwise None.
680
        :raises TokenLockingNotSupported: when a token is given but this
681
            instance doesn't support using token locks.
682
        :raises MismatchedToken: if the specified token doesn't match the token
683
            of the existing lock.
2617.6.8 by Robert Collins
Review feedback and documentation.
684
        :seealso: start_write_group.
2018.5.75 by Andrew Bennetts
Add Repository.{dont_,}leave_lock_in_place.
685
2018.5.145 by Andrew Bennetts
Add a brief explanation of what tokens are used for to lock_write docstrings.
686
        A token should be passed in if you know that you have locked the object
687
        some other way, and need to synchronise this object's state with that
688
        fact.
689
2018.5.75 by Andrew Bennetts
Add Repository.{dont_,}leave_lock_in_place.
690
        XXX: this docstring is duplicated in many places, e.g. lockable_files.py
691
        """
2617.6.2 by Robert Collins
Add abort_write_group and wire write_groups into fetch and commit.
692
        result = self.control_files.lock_write(token=token)
693
        self._refresh_data()
694
        return result
1185.65.1 by Aaron Bentley
Refactored out ControlFiles and RevisionStore from _Branch
695
696
    def lock_read(self):
1553.5.55 by Martin Pool
[revert] broken changes
697
        self.control_files.lock_read()
2617.6.2 by Robert Collins
Add abort_write_group and wire write_groups into fetch and commit.
698
        self._refresh_data()
1185.65.1 by Aaron Bentley
Refactored out ControlFiles and RevisionStore from _Branch
699
1694.2.6 by Martin Pool
[merge] bzr.dev
700
    def get_physical_lock_status(self):
701
        return self.control_files.get_physical_lock_status()
1624.3.36 by Olaf Conradi
Rename is_transport_locked() to get_physical_lock_status() as the
702
2018.5.75 by Andrew Bennetts
Add Repository.{dont_,}leave_lock_in_place.
703
    def leave_lock_in_place(self):
704
        """Tell this repository not to release the physical lock when this
705
        object is unlocked.
2018.5.76 by Andrew Bennetts
Testing that repository.{dont_,}leave_lock_in_place raises NotImplementedError if lock_write returns None.
706
        
707
        If lock_write doesn't return a token, then this method is not supported.
2018.5.75 by Andrew Bennetts
Add Repository.{dont_,}leave_lock_in_place.
708
        """
709
        self.control_files.leave_in_place()
710
711
    def dont_leave_lock_in_place(self):
712
        """Tell this repository to release the physical lock when this
713
        object is unlocked, even if it didn't originally acquire it.
2018.5.76 by Andrew Bennetts
Testing that repository.{dont_,}leave_lock_in_place raises NotImplementedError if lock_write returns None.
714
715
        If lock_write doesn't return a token, then this method is not supported.
2018.5.75 by Andrew Bennetts
Add Repository.{dont_,}leave_lock_in_place.
716
        """
717
        self.control_files.dont_leave_in_place()
718
1534.4.50 by Robert Collins
Got the bzrdir api straightened out, plenty of refactoring to use it pending, but the api is up and running.
719
    @needs_read_lock
2258.1.2 by Robert Collins
New version of gather_stats which gathers aggregate data too.
720
    def gather_stats(self, revid=None, committers=None):
2258.1.1 by Robert Collins
Move info branch statistics gathering into the repository to allow smart server optimisation (Robert Collins).
721
        """Gather statistics from a revision id.
722
2258.1.2 by Robert Collins
New version of gather_stats which gathers aggregate data too.
723
        :param revid: The revision id to gather statistics from, if None, then
724
            no revision specific statistics are gathered.
2258.1.1 by Robert Collins
Move info branch statistics gathering into the repository to allow smart server optimisation (Robert Collins).
725
        :param committers: Optional parameter controlling whether to grab
2258.1.2 by Robert Collins
New version of gather_stats which gathers aggregate data too.
726
            a count of committers from the revision specific statistics.
2258.1.1 by Robert Collins
Move info branch statistics gathering into the repository to allow smart server optimisation (Robert Collins).
727
        :return: A dictionary of statistics. Currently this contains:
728
            committers: The number of committers if requested.
729
            firstrev: A tuple with timestamp, timezone for the penultimate left
730
                most ancestor of revid, if revid is not the NULL_REVISION.
731
            latestrev: A tuple with timestamp, timezone for revid, if revid is
732
                not the NULL_REVISION.
2258.1.2 by Robert Collins
New version of gather_stats which gathers aggregate data too.
733
            revisions: The total revision count in the repository.
734
            size: An estimate disk size of the repository in bytes.
2258.1.1 by Robert Collins
Move info branch statistics gathering into the repository to allow smart server optimisation (Robert Collins).
735
        """
736
        result = {}
2258.1.2 by Robert Collins
New version of gather_stats which gathers aggregate data too.
737
        if revid and committers:
2258.1.1 by Robert Collins
Move info branch statistics gathering into the repository to allow smart server optimisation (Robert Collins).
738
            result['committers'] = 0
2258.1.2 by Robert Collins
New version of gather_stats which gathers aggregate data too.
739
        if revid and revid != _mod_revision.NULL_REVISION:
740
            if committers:
741
                all_committers = set()
742
            revisions = self.get_ancestry(revid)
743
            # pop the leading None
744
            revisions.pop(0)
745
            first_revision = None
746
            if not committers:
747
                # ignore the revisions in the middle - just grab first and last
748
                revisions = revisions[0], revisions[-1]
749
            for revision in self.get_revisions(revisions):
750
                if not first_revision:
751
                    first_revision = revision
752
                if committers:
753
                    all_committers.add(revision.committer)
754
            last_revision = revision
755
            if committers:
756
                result['committers'] = len(all_committers)
757
            result['firstrev'] = (first_revision.timestamp,
758
                first_revision.timezone)
759
            result['latestrev'] = (last_revision.timestamp,
760
                last_revision.timezone)
761
762
        # now gather global repository information
763
        if self.bzrdir.root_transport.listable():
764
            c, t = self._revision_store.total_size(self.get_transaction())
765
            result['revisions'] = c
766
            result['size'] = t
2258.1.1 by Robert Collins
Move info branch statistics gathering into the repository to allow smart server optimisation (Robert Collins).
767
        return result
768
2535.3.12 by Andrew Bennetts
Add a first cut of a get_data_stream method to Repository.
769
    def get_data_stream(self, revision_ids):
770
        raise NotImplementedError(self.get_data_stream)
771
2535.3.17 by Andrew Bennetts
[broken] Closer to a working Repository.fetch_revisions smart request.
772
    def insert_data_stream(self, stream):
2592.3.214 by Robert Collins
Merge bzr.dev.
773
        """XXX What does this really do? 
774
        
775
        Is it a substitute for fetch? 
776
        Should it manage its own write group ?
777
        """
2535.3.50 by Andrew Bennetts
Use tuple names in data streams rather than concatenated strings.
778
        for item_key, bytes in stream:
779
            if item_key[0] == 'file':
780
                (file_id,) = item_key[1:]
2535.3.27 by Andrew Bennetts
Merge from bzr.dev.
781
                knit = self.weave_store.get_weave_or_empty(
2535.3.17 by Andrew Bennetts
[broken] Closer to a working Repository.fetch_revisions smart request.
782
                    file_id, self.get_transaction())
2535.3.50 by Andrew Bennetts
Use tuple names in data streams rather than concatenated strings.
783
            elif item_key == ('inventory',):
2535.3.17 by Andrew Bennetts
[broken] Closer to a working Repository.fetch_revisions smart request.
784
                knit = self.get_inventory_weave()
2535.3.50 by Andrew Bennetts
Use tuple names in data streams rather than concatenated strings.
785
            elif item_key == ('revisions',):
2535.3.28 by Andrew Bennetts
Use self._revision_store.get_revision_file rather that self.control_store.get_weave('revisions', ...)
786
                knit = self._revision_store.get_revision_file(
787
                    self.get_transaction())
2535.3.50 by Andrew Bennetts
Use tuple names in data streams rather than concatenated strings.
788
            elif item_key == ('signatures',):
2535.3.25 by Andrew Bennetts
Fetch signatures too.
789
                knit = self._revision_store.get_signature_file(
790
                    self.get_transaction())
2535.3.17 by Andrew Bennetts
[broken] Closer to a working Repository.fetch_revisions smart request.
791
            else:
2535.3.40 by Andrew Bennetts
Tidy up more XXXs.
792
                raise RepositoryDataStreamError(
2535.3.50 by Andrew Bennetts
Use tuple names in data streams rather than concatenated strings.
793
                    "Unrecognised data stream key '%s'" % (item_key,))
2535.3.17 by Andrew Bennetts
[broken] Closer to a working Repository.fetch_revisions smart request.
794
            decoded_list = bencode.bdecode(bytes)
795
            format = decoded_list.pop(0)
796
            data_list = []
797
            knit_bytes = ''
798
            for version, options, parents, some_bytes in decoded_list:
799
                data_list.append((version, options, len(some_bytes), parents))
800
                knit_bytes += some_bytes
2535.3.40 by Andrew Bennetts
Tidy up more XXXs.
801
            knit.insert_data_stream(
802
                (format, data_list, StringIO(knit_bytes).read))
2535.3.17 by Andrew Bennetts
[broken] Closer to a working Repository.fetch_revisions smart request.
803
2258.1.1 by Robert Collins
Move info branch statistics gathering into the repository to allow smart server optimisation (Robert Collins).
804
    @needs_read_lock
3010.1.5 by Robert Collins
Test that missing_revision_ids handles the case of the source not having the requested revision correctly with and without find_ghosts.
805
    def missing_revision_ids(self, other, revision_id=None, find_ghosts=True):
1534.4.50 by Robert Collins
Got the bzrdir api straightened out, plenty of refactoring to use it pending, but the api is up and running.
806
        """Return the revision ids that other has that this does not.
807
        
808
        These are returned in topological order.
809
810
        revision_id: only return revision ids included by revision_id.
811
        """
3010.1.5 by Robert Collins
Test that missing_revision_ids handles the case of the source not having the requested revision correctly with and without find_ghosts.
812
        return InterRepository.get(other, self).missing_revision_ids(
813
            revision_id, find_ghosts)
1534.4.50 by Robert Collins
Got the bzrdir api straightened out, plenty of refactoring to use it pending, but the api is up and running.
814
1534.4.40 by Robert Collins
Add RepositoryFormats and allow bzrdir.open or create _repository to be used.
815
    @staticmethod
816
    def open(base):
817
        """Open the repository rooted at base.
818
819
        For instance, if the repository is at URL/.bzr/repository,
820
        Repository.open(URL) -> a Repository instance.
821
        """
1773.4.1 by Martin Pool
Add pyflakes makefile target; fix many warnings
822
        control = bzrdir.BzrDir.open(base)
1534.4.40 by Robert Collins
Add RepositoryFormats and allow bzrdir.open or create _repository to be used.
823
        return control.open_repository()
824
2387.1.1 by Robert Collins
Remove the --basis parameter to clone etc. (Robert Collins)
825
    def copy_content_into(self, destination, revision_id=None):
1534.6.6 by Robert Collins
Move find_repository to bzrdir, its not quite ideal there but its simpler and until someone chooses to vary the search by branch type its completely sufficient.
826
        """Make a complete copy of the content in self into destination.
827
        
828
        This is a destructive operation! Do not use it on existing 
829
        repositories.
830
        """
2387.1.1 by Robert Collins
Remove the --basis parameter to clone etc. (Robert Collins)
831
        return InterRepository.get(self, destination).copy_content(revision_id)
1534.4.50 by Robert Collins
Got the bzrdir api straightened out, plenty of refactoring to use it pending, but the api is up and running.
832
2617.6.2 by Robert Collins
Add abort_write_group and wire write_groups into fetch and commit.
833
    def commit_write_group(self):
834
        """Commit the contents accrued within the current write group.
2617.6.1 by Robert Collins
* New method on Repository - ``start_write_group``, ``end_write_group``
835
836
        :seealso: start_write_group.
837
        """
838
        if self._write_group is not self.get_transaction():
839
            # has an unlock or relock occured ?
2592.3.38 by Robert Collins
All experimental format tests passing again.
840
            raise errors.BzrError('mismatched lock context %r and '
841
                'write group %r.' %
842
                (self.get_transaction(), self._write_group))
2617.6.2 by Robert Collins
Add abort_write_group and wire write_groups into fetch and commit.
843
        self._commit_write_group()
2617.6.1 by Robert Collins
* New method on Repository - ``start_write_group``, ``end_write_group``
844
        self._write_group = None
845
2617.6.2 by Robert Collins
Add abort_write_group and wire write_groups into fetch and commit.
846
    def _commit_write_group(self):
847
        """Template method for per-repository write group cleanup.
848
        
849
        This is called before the write group is considered to be 
850
        finished and should ensure that all data handed to the repository
851
        for writing during the write group is safely committed (to the 
852
        extent possible considering file system caching etc).
853
        """
854
2949.1.1 by Robert Collins
Change Repository.fetch to provide a find_ghosts parameter which triggers ghost filling.
855
    def fetch(self, source, revision_id=None, pb=None, find_ghosts=False):
1534.4.50 by Robert Collins
Got the bzrdir api straightened out, plenty of refactoring to use it pending, but the api is up and running.
856
        """Fetch the content required to construct revision_id from source.
857
858
        If revision_id is None all content is copied.
2949.1.1 by Robert Collins
Change Repository.fetch to provide a find_ghosts parameter which triggers ghost filling.
859
        :param find_ghosts: Find and copy revisions in the source that are
860
            ghosts in the target (and not reachable directly by walking out to
861
            the first-present revision in target from revision_id).
1534.4.50 by Robert Collins
Got the bzrdir api straightened out, plenty of refactoring to use it pending, but the api is up and running.
862
        """
2592.3.115 by Robert Collins
Move same repository check up to Repository.fetch to allow all fetch implementations to benefit.
863
        # fast path same-url fetch operations
864
        if self.has_same_location(source):
865
            # check that last_revision is in 'from' and then return a
866
            # no-operation.
867
            if (revision_id is not None and
868
                not _mod_revision.is_null(revision_id)):
869
                self.get_revision(revision_id)
870
            return 0, []
2323.8.3 by Aaron Bentley
Reduce scope of try/except, update NEWS
871
        inter = InterRepository.get(source, self)
2323.8.2 by Aaron Bentley
Give a nicer error on fetch when repos are in incompatible formats
872
        try:
2949.1.2 by Robert Collins
* Fetch with pack repositories will no longer read the entire history graph.
873
            return inter.fetch(revision_id=revision_id, pb=pb, find_ghosts=find_ghosts)
2323.8.2 by Aaron Bentley
Give a nicer error on fetch when repos are in incompatible formats
874
        except NotImplementedError:
875
            raise errors.IncompatibleRepositories(source, self)
1534.4.41 by Robert Collins
Branch now uses BzrDir reasonably sanely.
876
2520.4.54 by Aaron Bentley
Hang a create_bundle method off repository
877
    def create_bundle(self, target, base, fileobj, format=None):
878
        return serializer.write_bundle(self, target, base, fileobj, format)
879
2803.2.1 by Robert Collins
* CommitBuilder now advertises itself as requiring the root entry to be
880
    def get_commit_builder(self, branch, parents, config, timestamp=None,
881
                           timezone=None, committer=None, revprops=None,
1740.3.7 by Jelmer Vernooij
Move committer, log, revprops, timestamp and timezone to CommitBuilder.
882
                           revision_id=None):
883
        """Obtain a CommitBuilder for this repository.
884
        
885
        :param branch: Branch to commit to.
886
        :param parents: Revision ids of the parents of the new revision.
887
        :param config: Configuration to use.
888
        :param timestamp: Optional timestamp recorded for commit.
889
        :param timezone: Optional timezone for timestamp.
890
        :param committer: Optional committer to set for commit.
891
        :param revprops: Optional dictionary of revision properties.
892
        :param revision_id: Optional revision id.
893
        """
2818.3.2 by Robert Collins
Review feedback.
894
        result = self._commit_builder_class(self, parents, config,
2592.3.135 by Robert Collins
Do not create many transient knit objects, saving 4% on commit.
895
            timestamp, timezone, committer, revprops, revision_id)
2617.6.2 by Robert Collins
Add abort_write_group and wire write_groups into fetch and commit.
896
        self.start_write_group()
897
        return result
1740.3.1 by Jelmer Vernooij
Introduce and use CommitBuilder objects.
898
1185.65.1 by Aaron Bentley
Refactored out ControlFiles and RevisionStore from _Branch
899
    def unlock(self):
2617.6.1 by Robert Collins
* New method on Repository - ``start_write_group``, ``end_write_group``
900
        if (self.control_files._lock_count == 1 and
901
            self.control_files._lock_mode == 'w'):
902
            if self._write_group is not None:
2592.3.244 by Martin Pool
unlock while in a write group now aborts the write group, unlocks, and errors.
903
                self.abort_write_group()
904
                self.control_files.unlock()
2617.6.1 by Robert Collins
* New method on Repository - ``start_write_group``, ``end_write_group``
905
                raise errors.BzrError(
906
                    'Must end write groups before releasing write locks.')
1185.65.1 by Aaron Bentley
Refactored out ControlFiles and RevisionStore from _Branch
907
        self.control_files.unlock()
908
1185.65.27 by Robert Collins
Tweak storage towards mergability.
909
    @needs_read_lock
2387.1.1 by Robert Collins
Remove the --basis parameter to clone etc. (Robert Collins)
910
    def clone(self, a_bzrdir, revision_id=None):
1534.4.41 by Robert Collins
Branch now uses BzrDir reasonably sanely.
911
        """Clone this repository into a_bzrdir using the current format.
912
913
        Currently no check is made that the format of this repository and
914
        the bzrdir format are compatible. FIXME RBC 20060201.
2241.1.4 by Martin Pool
Moved old weave-based repository formats into bzrlib.repofmt.weaverepo.
915
916
        :return: The newly created destination repository.
1534.4.41 by Robert Collins
Branch now uses BzrDir reasonably sanely.
917
        """
2440.1.1 by Martin Pool
Add new Repository.sprout,
918
        # TODO: deprecate after 0.16; cloning this with all its settings is
919
        # probably not very useful -- mbp 20070423
920
        dest_repo = self._create_sprouting_repo(a_bzrdir, shared=self.is_shared())
921
        self.copy_content_into(dest_repo, revision_id)
922
        return dest_repo
923
2617.6.1 by Robert Collins
* New method on Repository - ``start_write_group``, ``end_write_group``
924
    def start_write_group(self):
925
        """Start a write group in the repository.
926
927
        Write groups are used by repositories which do not have a 1:1 mapping
928
        between file ids and backend store to manage the insertion of data from
929
        both fetch and commit operations.
930
2617.6.2 by Robert Collins
Add abort_write_group and wire write_groups into fetch and commit.
931
        A write lock is required around the start_write_group/commit_write_group
2617.6.1 by Robert Collins
* New method on Repository - ``start_write_group``, ``end_write_group``
932
        for the support of lock-requiring repository formats.
2617.6.8 by Robert Collins
Review feedback and documentation.
933
934
        One can only insert data into a repository inside a write group.
935
2617.6.6 by Robert Collins
Some review feedback.
936
        :return: None.
2617.6.1 by Robert Collins
* New method on Repository - ``start_write_group``, ``end_write_group``
937
        """
2592.3.188 by Robert Collins
Allow pack repositories to have multiple writers active at one time, for greater concurrency.
938
        if not self.is_write_locked():
2617.6.1 by Robert Collins
* New method on Repository - ``start_write_group``, ``end_write_group``
939
            raise errors.NotWriteLocked(self)
940
        if self._write_group:
941
            raise errors.BzrError('already in a write group')
2617.6.2 by Robert Collins
Add abort_write_group and wire write_groups into fetch and commit.
942
        self._start_write_group()
943
        # so we can detect unlock/relock - the write group is now entered.
2617.6.1 by Robert Collins
* New method on Repository - ``start_write_group``, ``end_write_group``
944
        self._write_group = self.get_transaction()
945
2617.6.2 by Robert Collins
Add abort_write_group and wire write_groups into fetch and commit.
946
    def _start_write_group(self):
947
        """Template method for per-repository write group startup.
948
        
949
        This is called before the write group is considered to be 
950
        entered.
951
        """
952
2440.1.1 by Martin Pool
Add new Repository.sprout,
953
    @needs_read_lock
954
    def sprout(self, to_bzrdir, revision_id=None):
955
        """Create a descendent repository for new development.
956
957
        Unlike clone, this does not copy the settings of the repository.
958
        """
959
        dest_repo = self._create_sprouting_repo(to_bzrdir, shared=False)
960
        dest_repo.fetch(self, revision_id=revision_id)
961
        return dest_repo
962
963
    def _create_sprouting_repo(self, a_bzrdir, shared):
1534.4.50 by Robert Collins
Got the bzrdir api straightened out, plenty of refactoring to use it pending, but the api is up and running.
964
        if not isinstance(a_bzrdir._format, self.bzrdir._format.__class__):
965
            # use target default format.
2241.1.4 by Martin Pool
Moved old weave-based repository formats into bzrlib.repofmt.weaverepo.
966
            dest_repo = a_bzrdir.create_repository()
1534.4.50 by Robert Collins
Got the bzrdir api straightened out, plenty of refactoring to use it pending, but the api is up and running.
967
        else:
2241.1.4 by Martin Pool
Moved old weave-based repository formats into bzrlib.repofmt.weaverepo.
968
            # Most control formats need the repository to be specifically
969
            # created, but on some old all-in-one formats it's not needed
970
            try:
2440.1.1 by Martin Pool
Add new Repository.sprout,
971
                dest_repo = self._format.initialize(a_bzrdir, shared=shared)
2241.1.4 by Martin Pool
Moved old weave-based repository formats into bzrlib.repofmt.weaverepo.
972
            except errors.UninitializableFormat:
973
                dest_repo = a_bzrdir.open_repository()
974
        return dest_repo
1534.4.41 by Robert Collins
Branch now uses BzrDir reasonably sanely.
975
1563.2.22 by Robert Collins
Move responsibility for repository.has_revision into RevisionStore
976
    @needs_read_lock
1185.65.1 by Aaron Bentley
Refactored out ControlFiles and RevisionStore from _Branch
977
    def has_revision(self, revision_id):
1563.2.22 by Robert Collins
Move responsibility for repository.has_revision into RevisionStore
978
        """True if this repository has a copy of the revision."""
2745.1.1 by Robert Collins
Add a number of -Devil checkpoints.
979
        if 'evil' in debug.debug_flags:
2592.3.112 by Robert Collins
Various fixups found dogfooding.
980
            mutter_callsite(3, "has_revision is a LBYL symptom.")
1563.2.22 by Robert Collins
Move responsibility for repository.has_revision into RevisionStore
981
        return self._revision_store.has_revision_id(revision_id,
982
                                                    self.get_transaction())
1185.65.1 by Aaron Bentley
Refactored out ControlFiles and RevisionStore from _Branch
983
1185.65.27 by Robert Collins
Tweak storage towards mergability.
984
    @needs_read_lock
2850.3.1 by Robert Collins
Move various weave specific code out of the base Repository class to weaverepo.py.
985
    def get_revision(self, revision_id):
986
        """Return the Revision object for a named revision."""
987
        return self.get_revisions([revision_id])[0]
988
989
    @needs_read_lock
1570.1.13 by Robert Collins
Check for incorrect revision parentage in the weave during revision access.
990
    def get_revision_reconcile(self, revision_id):
991
        """'reconcile' helper routine that allows access to a revision always.
992
        
993
        This variant of get_revision does not cross check the weave graph
994
        against the revision one as get_revision does: but it should only
995
        be used by reconcile, or reconcile-alike commands that are correcting
996
        or testing the revision graph.
997
        """
2850.3.1 by Robert Collins
Move various weave specific code out of the base Repository class to weaverepo.py.
998
        return self._get_revisions([revision_id])[0]
2249.5.13 by John Arbash Meinel
Finish auditing Repository, and fix generate_ids to always generate utf8 ids.
999
1756.1.2 by Aaron Bentley
Show logs using get_revisions
1000
    @needs_read_lock
1001
    def get_revisions(self, revision_ids):
2850.3.1 by Robert Collins
Move various weave specific code out of the base Repository class to weaverepo.py.
1002
        """Get many revisions at once."""
1003
        return self._get_revisions(revision_ids)
1004
1005
    @needs_read_lock
1006
    def _get_revisions(self, revision_ids):
1007
        """Core work logic to get many revisions without sanity checks."""
1008
        for rev_id in revision_ids:
1009
            if not rev_id or not isinstance(rev_id, basestring):
1010
                raise errors.InvalidRevisionId(revision_id=rev_id, branch=self)
2249.5.13 by John Arbash Meinel
Finish auditing Repository, and fix generate_ids to always generate utf8 ids.
1011
        revs = self._revision_store.get_revisions(revision_ids,
1756.1.2 by Aaron Bentley
Show logs using get_revisions
1012
                                                  self.get_transaction())
2249.5.13 by John Arbash Meinel
Finish auditing Repository, and fix generate_ids to always generate utf8 ids.
1013
        for rev in revs:
1014
            assert not isinstance(rev.revision_id, unicode)
1015
            for parent_id in rev.parent_ids:
1016
                assert not isinstance(parent_id, unicode)
1017
        return revs
1185.65.1 by Aaron Bentley
Refactored out ControlFiles and RevisionStore from _Branch
1018
1185.65.27 by Robert Collins
Tweak storage towards mergability.
1019
    @needs_read_lock
1563.2.29 by Robert Collins
Remove all but fetch references to repository.revision_store.
1020
    def get_revision_xml(self, revision_id):
2249.5.13 by John Arbash Meinel
Finish auditing Repository, and fix generate_ids to always generate utf8 ids.
1021
        # TODO: jam 20070210 This shouldn't be necessary since get_revision
1022
        #       would have already do it.
1023
        # TODO: jam 20070210 Just use _serializer.write_revision_to_string()
1024
        rev = self.get_revision(revision_id)
1563.2.29 by Robert Collins
Remove all but fetch references to repository.revision_store.
1025
        rev_tmp = StringIO()
1026
        # the current serializer..
1027
        self._revision_store._serializer.write_revision(rev, rev_tmp)
1028
        rev_tmp.seek(0)
1029
        return rev_tmp.getvalue()
1030
1031
    @needs_read_lock
1756.3.22 by Aaron Bentley
Tweaks from review
1032
    def get_deltas_for_revisions(self, revisions):
1756.3.19 by Aaron Bentley
Documentation and cleanups
1033
        """Produce a generator of revision deltas.
1034
        
1035
        Note that the input is a sequence of REVISIONS, not revision_ids.
1036
        Trees will be held in memory until the generator exits.
1037
        Each delta is relative to the revision's lefthand predecessor.
1038
        """
1756.3.3 by Aaron Bentley
More refactoring, introduce revision_trees.
1039
        required_trees = set()
1040
        for revision in revisions:
1041
            required_trees.add(revision.revision_id)
1042
            required_trees.update(revision.parent_ids[:1])
1043
        trees = dict((t.get_revision_id(), t) for 
1044
                     t in self.revision_trees(required_trees))
1045
        for revision in revisions:
1046
            if not revision.parent_ids:
1852.5.1 by Robert Collins
Deprecate EmptyTree in favour of using Repository.revision_tree.
1047
                old_tree = self.revision_tree(None)
1756.3.3 by Aaron Bentley
More refactoring, introduce revision_trees.
1048
            else:
1049
                old_tree = trees[revision.parent_ids[0]]
1852.10.3 by Robert Collins
Remove all uses of compare_trees and replace with Tree.changes_from throughout bzrlib.
1050
            yield trees[revision.revision_id].changes_from(old_tree)
1756.3.3 by Aaron Bentley
More refactoring, introduce revision_trees.
1051
1756.3.19 by Aaron Bentley
Documentation and cleanups
1052
    @needs_read_lock
1744.2.2 by Johan Rydberg
Add get_revision_delta to Repository; and make Branch.get_revision_delta use it.
1053
    def get_revision_delta(self, revision_id):
1054
        """Return the delta for one revision.
1055
1056
        The delta is relative to the left-hand predecessor of the
1057
        revision.
1058
        """
1756.3.3 by Aaron Bentley
More refactoring, introduce revision_trees.
1059
        r = self.get_revision(revision_id)
1756.3.22 by Aaron Bentley
Tweaks from review
1060
        return list(self.get_deltas_for_revisions([r]))[0]
1744.2.2 by Johan Rydberg
Add get_revision_delta to Repository; and make Branch.get_revision_delta use it.
1061
1185.65.1 by Aaron Bentley
Refactored out ControlFiles and RevisionStore from _Branch
1062
    @needs_write_lock
1063
    def store_revision_signature(self, gpg_strategy, plaintext, revision_id):
1563.2.29 by Robert Collins
Remove all but fetch references to repository.revision_store.
1064
        signature = gpg_strategy.sign(plaintext)
2996.2.4 by Aaron Bentley
Rename function to add_signature_text
1065
        self.add_signature_text(revision_id, signature)
2996.2.3 by Aaron Bentley
Add tests for install_revisions and add_signature
1066
1067
    @needs_write_lock
2996.2.4 by Aaron Bentley
Rename function to add_signature_text
1068
    def add_signature_text(self, revision_id, signature):
1563.2.29 by Robert Collins
Remove all but fetch references to repository.revision_store.
1069
        self._revision_store.add_revision_signature_text(revision_id,
1070
                                                         signature,
1071
                                                         self.get_transaction())
1185.65.1 by Aaron Bentley
Refactored out ControlFiles and RevisionStore from _Branch
1072
2988.1.2 by Robert Collins
New Repository API find_text_key_references for use by reconcile and check.
1073
    def find_text_key_references(self):
1074
        """Find the text key references within the repository.
1075
1076
        :return: a dictionary mapping (file_id, revision_id) tuples to altered file-ids to an iterable of
1077
        revision_ids. Each altered file-ids has the exact revision_ids that
1078
        altered it listed explicitly.
1079
        :return: A dictionary mapping text keys ((fileid, revision_id) tuples)
1080
            to whether they were referred to by the inventory of the
1081
            revision_id that they contain. The inventory texts from all present
1082
            revision ids are assessed to generate this report.
1083
        """
1084
        revision_ids = self.all_revision_ids()
1085
        w = self.get_inventory_weave()
1086
        pb = ui.ui_factory.nested_progress_bar()
1087
        try:
1088
            return self._find_text_key_references_from_xml_inventory_lines(
1089
                w.iter_lines_added_or_present_in_versions(revision_ids, pb=pb))
1090
        finally:
1091
            pb.finished()
1092
2988.1.1 by Robert Collins
Refactor fetch's xml inventory parsing into a core routine that extracts the data and a separate one that filters for fetch.
1093
    def _find_text_key_references_from_xml_inventory_lines(self,
1094
        line_iterator):
1095
        """Core routine for extracting references to texts from inventories.
2592.3.110 by Robert Collins
Filter out texts and signatures not referenced by the revisions being copied during pack to pack fetching.
1096
1097
        This performs the translation of xml lines to revision ids.
1098
2975.3.1 by Robert Collins
Change (without backwards compatibility) the
1099
        :param line_iterator: An iterator of lines, origin_version_id
2988.1.1 by Robert Collins
Refactor fetch's xml inventory parsing into a core routine that extracts the data and a separate one that filters for fetch.
1100
        :return: A dictionary mapping text keys ((fileid, revision_id) tuples)
1101
            to whether they were referred to by the inventory of the
1102
            revision_id that they contain. Note that if that revision_id was
1103
            not part of the line_iterator's output then False will be given -
1104
            even though it may actually refer to that key.
1534.4.41 by Robert Collins
Branch now uses BzrDir reasonably sanely.
1105
        """
2988.2.2 by Robert Collins
Review feedback.
1106
        if not self._serializer.support_altered_by_hack:
1107
            raise AssertionError(
1108
                "_find_text_key_references_from_xml_inventory_lines only "
1109
                "supported for branches which store inventory as unnested xml"
1110
                ", not on %r" % self)
1694.2.6 by Martin Pool
[merge] bzr.dev
1111
        result = {}
1563.2.35 by Robert Collins
cleanup deprecation warnings and finish conversion so the inventory is knit based too.
1112
1694.2.6 by Martin Pool
[merge] bzr.dev
1113
        # this code needs to read every new line in every inventory for the
1114
        # inventories [revision_ids]. Seeing a line twice is ok. Seeing a line
1759.2.1 by Jelmer Vernooij
Fix some types (found using aspell).
1115
        # not present in one of those inventories is unnecessary but not 
1594.2.6 by Robert Collins
Introduce a api specifically for looking at lines in some versions of the inventory, for fileid_involved.
1116
        # harmful because we are filtering by the revision id marker in the
1694.2.6 by Martin Pool
[merge] bzr.dev
1117
        # inventory lines : we only select file ids altered in one of those  
1759.2.2 by Jelmer Vernooij
Revert some of my spelling fixes and fix some typos after review by Aaron.
1118
        # revisions. We don't need to see all lines in the inventory because
1594.2.6 by Robert Collins
Introduce a api specifically for looking at lines in some versions of the inventory, for fileid_involved.
1119
        # only those added in an inventory in rev X can contain a revision=X
1120
        # line.
2163.2.3 by John Arbash Meinel
Change to local variables to save another 300ms
1121
        unescape_revid_cache = {}
1122
        unescape_fileid_cache = {}
1123
2163.2.5 by John Arbash Meinel
Inline the cache lookup, and explain why
1124
        # jam 20061218 In a big fetch, this handles hundreds of thousands
1125
        # of lines, so it has had a lot of inlining and optimizing done.
1126
        # Sorry that it is a little bit messy.
2163.2.3 by John Arbash Meinel
Change to local variables to save another 300ms
1127
        # Move several functions to be local variables, since this is a long
1128
        # running loop.
1129
        search = self._file_ids_altered_regex.search
2163.2.5 by John Arbash Meinel
Inline the cache lookup, and explain why
1130
        unescape = _unescape_xml
2163.2.3 by John Arbash Meinel
Change to local variables to save another 300ms
1131
        setdefault = result.setdefault
2975.3.1 by Robert Collins
Change (without backwards compatibility) the
1132
        for line, version_id in line_iterator:
2592.3.110 by Robert Collins
Filter out texts and signatures not referenced by the revisions being copied during pack to pack fetching.
1133
            match = search(line)
1134
            if match is None:
1135
                continue
1136
            # One call to match.group() returning multiple items is quite a
1137
            # bit faster than 2 calls to match.group() each returning 1
1138
            file_id, revision_id = match.group('file_id', 'revision_id')
1139
1140
            # Inlining the cache lookups helps a lot when you make 170,000
1141
            # lines and 350k ids, versus 8.4 unique ids.
1142
            # Using a cache helps in 2 ways:
1143
            #   1) Avoids unnecessary decoding calls
1144
            #   2) Re-uses cached strings, which helps in future set and
1145
            #      equality checks.
1146
            # (2) is enough that removing encoding entirely along with
1147
            # the cache (so we are using plain strings) results in no
1148
            # performance improvement.
1149
            try:
1150
                revision_id = unescape_revid_cache[revision_id]
1151
            except KeyError:
1152
                unescaped = unescape(revision_id)
1153
                unescape_revid_cache[revision_id] = unescaped
1154
                revision_id = unescaped
1155
2988.2.2 by Robert Collins
Review feedback.
1156
            # Note that unconditionally unescaping means that we deserialise
1157
            # every fileid, which for general 'pull' is not great, but we don't
1158
            # really want to have some many fulltexts that this matters anyway.
1159
            # RBC 20071114.
2988.1.1 by Robert Collins
Refactor fetch's xml inventory parsing into a core routine that extracts the data and a separate one that filters for fetch.
1160
            try:
1161
                file_id = unescape_fileid_cache[file_id]
1162
            except KeyError:
1163
                unescaped = unescape(file_id)
1164
                unescape_fileid_cache[file_id] = unescaped
1165
                file_id = unescaped
1166
1167
            key = (file_id, revision_id)
1168
            setdefault(key, False)
1169
            if revision_id == version_id:
1170
                result[key] = True
1171
        return result
1172
1173
    def _find_file_ids_from_xml_inventory_lines(self, line_iterator,
1174
        revision_ids):
1175
        """Helper routine for fileids_altered_by_revision_ids.
1176
1177
        This performs the translation of xml lines to revision ids.
1178
1179
        :param line_iterator: An iterator of lines, origin_version_id
1180
        :param revision_ids: The revision ids to filter for. This should be a
1181
            set or other type which supports efficient __contains__ lookups, as
1182
            the revision id from each parsed line will be looked up in the
1183
            revision_ids filter.
1184
        :return: a dictionary mapping altered file-ids to an iterable of
1185
        revision_ids. Each altered file-ids has the exact revision_ids that
1186
        altered it listed explicitly.
1187
        """
1188
        result = {}
1189
        setdefault = result.setdefault
1190
        for file_id, revision_id in \
1191
            self._find_text_key_references_from_xml_inventory_lines(
1192
                line_iterator).iterkeys():
2975.3.1 by Robert Collins
Change (without backwards compatibility) the
1193
            # once data is all ensured-consistent; then this is
1194
            # if revision_id == version_id
2592.3.110 by Robert Collins
Filter out texts and signatures not referenced by the revisions being copied during pack to pack fetching.
1195
            if revision_id in revision_ids:
1196
                setdefault(file_id, set()).add(revision_id)
1197
        return result
1198
1199
    def fileids_altered_by_revision_ids(self, revision_ids):
1200
        """Find the file ids and versions affected by revisions.
1201
1202
        :param revisions: an iterable containing revision ids.
1203
        :return: a dictionary mapping altered file-ids to an iterable of
1204
        revision_ids. Each altered file-ids has the exact revision_ids that
1205
        altered it listed explicitly.
1206
        """
2592.3.185 by Robert Collins
Merge bzr.dev.
1207
        selected_revision_ids = set(revision_ids)
2592.3.110 by Robert Collins
Filter out texts and signatures not referenced by the revisions being copied during pack to pack fetching.
1208
        w = self.get_inventory_weave()
2039.1.1 by Aaron Bentley
Clean up progress properly when interrupted during fetch (#54000)
1209
        pb = ui.ui_factory.nested_progress_bar()
1210
        try:
2592.3.110 by Robert Collins
Filter out texts and signatures not referenced by the revisions being copied during pack to pack fetching.
1211
            return self._find_file_ids_from_xml_inventory_lines(
1212
                w.iter_lines_added_or_present_in_versions(
1213
                    selected_revision_ids, pb=pb),
1214
                selected_revision_ids)
2039.1.1 by Aaron Bentley
Clean up progress properly when interrupted during fetch (#54000)
1215
        finally:
1216
            pb.finished()
1534.4.41 by Robert Collins
Branch now uses BzrDir reasonably sanely.
1217
2708.1.7 by Aaron Bentley
Rename extract_files_bytes to iter_files_bytes
1218
    def iter_files_bytes(self, desired_files):
2708.1.9 by Aaron Bentley
Clean-up docs and imports
1219
        """Iterate through file versions.
1220
2708.1.10 by Aaron Bentley
Update docstrings
1221
        Files will not necessarily be returned in the order they occur in
1222
        desired_files.  No specific order is guaranteed.
1223
2708.1.9 by Aaron Bentley
Clean-up docs and imports
1224
        Yields pairs of identifier, bytes_iterator.  identifier is an opaque
2708.1.10 by Aaron Bentley
Update docstrings
1225
        value supplied by the caller as part of desired_files.  It should
1226
        uniquely identify the file version in the caller's context.  (Examples:
1227
        an index number or a TreeTransform trans_id.)
1228
1229
        bytes_iterator is an iterable of bytestrings for the file.  The
1230
        kind of iterable and length of the bytestrings are unspecified, but for
1231
        this implementation, it is a list of lines produced by
1232
        VersionedFile.get_lines().
1233
2708.1.9 by Aaron Bentley
Clean-up docs and imports
1234
        :param desired_files: a list of (file_id, revision_id, identifier)
2708.1.10 by Aaron Bentley
Update docstrings
1235
            triples
2708.1.9 by Aaron Bentley
Clean-up docs and imports
1236
        """
2708.1.3 by Aaron Bentley
Implement extract_files_bytes on Repository
1237
        transaction = self.get_transaction()
1238
        for file_id, revision_id, callable_data in desired_files:
2708.1.11 by Aaron Bentley
Test and tweak error handling
1239
            try:
1240
                weave = self.weave_store.get_weave(file_id, transaction)
1241
            except errors.NoSuchFile:
1242
                raise errors.NoSuchIdInRepository(self, file_id)
2708.1.6 by Aaron Bentley
Turn extract_files_bytes into an iterator
1243
            yield callable_data, weave.get_lines(revision_id)
2708.1.3 by Aaron Bentley
Implement extract_files_bytes on Repository
1244
3063.2.1 by Robert Collins
Solve reconciling erroring when multiple portions of a single delta chain are being reinserted.
1245
    def _generate_text_key_index(self, text_key_references=None,
1246
        ancestors=None):
2988.1.3 by Robert Collins
Add a new repositoy method _generate_text_key_index for use by reconcile/check.
1247
        """Generate a new text key index for the repository.
1248
1249
        This is an expensive function that will take considerable time to run.
1250
1251
        :return: A dict mapping text keys ((file_id, revision_id) tuples) to a
1252
            list of parents, also text keys. When a given key has no parents,
1253
            the parents list will be [NULL_REVISION].
1254
        """
1255
        # All revisions, to find inventory parents.
3063.2.1 by Robert Collins
Solve reconciling erroring when multiple portions of a single delta chain are being reinserted.
1256
        if ancestors is None:
1257
            revision_graph = self.get_revision_graph_with_ghosts()
1258
            ancestors = revision_graph.get_ancestors()
2951.2.9 by Robert Collins
* ``pack-0.92`` repositories can now be reconciled.
1259
        if text_key_references is None:
1260
            text_key_references = self.find_text_key_references()
2988.3.1 by Robert Collins
Handle the progress bar in _generate_text_key_index correctly.
1261
        pb = ui.ui_factory.nested_progress_bar()
1262
        try:
1263
            return self._do_generate_text_key_index(ancestors,
1264
                text_key_references, pb)
1265
        finally:
1266
            pb.finished()
1267
1268
    def _do_generate_text_key_index(self, ancestors, text_key_references, pb):
1269
        """Helper for _generate_text_key_index to avoid deep nesting."""
2988.1.3 by Robert Collins
Add a new repositoy method _generate_text_key_index for use by reconcile/check.
1270
        revision_order = tsort.topo_sort(ancestors)
1271
        invalid_keys = set()
1272
        revision_keys = {}
1273
        for revision_id in revision_order:
1274
            revision_keys[revision_id] = set()
1275
        text_count = len(text_key_references)
1276
        # a cache of the text keys to allow reuse; costs a dict of all the
1277
        # keys, but saves a 2-tuple for every child of a given key.
1278
        text_key_cache = {}
1279
        for text_key, valid in text_key_references.iteritems():
1280
            if not valid:
1281
                invalid_keys.add(text_key)
1282
            else:
1283
                revision_keys[text_key[1]].add(text_key)
1284
            text_key_cache[text_key] = text_key
1285
        del text_key_references
1286
        text_index = {}
1287
        text_graph = graph.Graph(graph.DictParentsProvider(text_index))
1288
        NULL_REVISION = _mod_revision.NULL_REVISION
2988.1.5 by Robert Collins
Use a LRU cache when generating the text index to reduce inventory deserialisations.
1289
        # Set a cache with a size of 10 - this suffices for bzr.dev but may be
1290
        # too small for large or very branchy trees. However, for 55K path
1291
        # trees, it would be easy to use too much memory trivially. Ideally we
1292
        # could gauge this by looking at available real memory etc, but this is
1293
        # always a tricky proposition.
1294
        inventory_cache = lru_cache.LRUCache(10)
2988.1.3 by Robert Collins
Add a new repositoy method _generate_text_key_index for use by reconcile/check.
1295
        batch_size = 10 # should be ~150MB on a 55K path tree
1296
        batch_count = len(revision_order) / batch_size + 1
1297
        processed_texts = 0
1298
        pb.update("Calculating text parents.", processed_texts, text_count)
1299
        for offset in xrange(batch_count):
1300
            to_query = revision_order[offset * batch_size:(offset + 1) *
1301
                batch_size]
1302
            if not to_query:
1303
                break
1304
            for rev_tree in self.revision_trees(to_query):
1305
                revision_id = rev_tree.get_revision_id()
1306
                parent_ids = ancestors[revision_id]
1307
                for text_key in revision_keys[revision_id]:
1308
                    pb.update("Calculating text parents.", processed_texts)
1309
                    processed_texts += 1
1310
                    candidate_parents = []
1311
                    for parent_id in parent_ids:
1312
                        parent_text_key = (text_key[0], parent_id)
1313
                        try:
1314
                            check_parent = parent_text_key not in \
1315
                                revision_keys[parent_id]
1316
                        except KeyError:
1317
                            # the parent parent_id is a ghost:
1318
                            check_parent = False
1319
                            # truncate the derived graph against this ghost.
1320
                            parent_text_key = None
1321
                        if check_parent:
1322
                            # look at the parent commit details inventories to
1323
                            # determine possible candidates in the per file graph.
1324
                            # TODO: cache here.
2988.1.5 by Robert Collins
Use a LRU cache when generating the text index to reduce inventory deserialisations.
1325
                            try:
1326
                                inv = inventory_cache[parent_id]
1327
                            except KeyError:
1328
                                inv = self.revision_tree(parent_id).inventory
1329
                                inventory_cache[parent_id] = inv
1330
                            parent_entry = inv._byid.get(text_key[0], None)
2988.1.3 by Robert Collins
Add a new repositoy method _generate_text_key_index for use by reconcile/check.
1331
                            if parent_entry is not None:
1332
                                parent_text_key = (
1333
                                    text_key[0], parent_entry.revision)
1334
                            else:
1335
                                parent_text_key = None
1336
                        if parent_text_key is not None:
1337
                            candidate_parents.append(
1338
                                text_key_cache[parent_text_key])
1339
                    parent_heads = text_graph.heads(candidate_parents)
1340
                    new_parents = list(parent_heads)
1341
                    new_parents.sort(key=lambda x:candidate_parents.index(x))
1342
                    if new_parents == []:
1343
                        new_parents = [NULL_REVISION]
1344
                    text_index[text_key] = new_parents
1345
1346
        for text_key in invalid_keys:
1347
            text_index[text_key] = [NULL_REVISION]
1348
        return text_index
1349
2668.2.8 by Andrew Bennetts
Rename get_data_to_fetch_for_revision_ids as item_keys_introduced_by.
1350
    def item_keys_introduced_by(self, revision_ids, _files_pb=None):
1351
        """Get an iterable listing the keys of all the data introduced by a set
1352
        of revision IDs.
1353
1354
        The keys will be ordered so that the corresponding items can be safely
1355
        fetched and inserted in that order.
1356
1357
        :returns: An iterable producing tuples of (knit-kind, file-id,
1358
            versions).  knit-kind is one of 'file', 'inventory', 'signatures',
1359
            'revisions'.  file-id is None unless knit-kind is 'file'.
2535.3.6 by Andrew Bennetts
Move some "what repo data to fetch logic" from RepoFetcher to Repository.
1360
        """
1361
        # XXX: it's a bit weird to control the inventory weave caching in this
2535.3.7 by Andrew Bennetts
Remove now unused _fetch_weave_texts, make progress reporting closer to how it was before I refactored __fetch.
1362
        # generator.  Ideally the caching would be done in fetch.py I think.  Or
1363
        # maybe this generator should explicitly have the contract that it
1364
        # should not be iterated until the previously yielded item has been
1365
        # processed?
2535.3.47 by Andrew Bennetts
Massively improve item_keys_introduced_by performance by calling self.lock_read().
1366
        self.lock_read()
2535.3.6 by Andrew Bennetts
Move some "what repo data to fetch logic" from RepoFetcher to Repository.
1367
        inv_w = self.get_inventory_weave()
1368
        inv_w.enable_cache()
1369
1370
        # file ids that changed
1371
        file_ids = self.fileids_altered_by_revision_ids(revision_ids)
2535.3.8 by Andrew Bennetts
Unbreak progress reporting.
1372
        count = 0
1373
        num_file_ids = len(file_ids)
2535.3.6 by Andrew Bennetts
Move some "what repo data to fetch logic" from RepoFetcher to Repository.
1374
        for file_id, altered_versions in file_ids.iteritems():
2668.2.8 by Andrew Bennetts
Rename get_data_to_fetch_for_revision_ids as item_keys_introduced_by.
1375
            if _files_pb is not None:
1376
                _files_pb.update("fetch texts", count, num_file_ids)
2535.3.8 by Andrew Bennetts
Unbreak progress reporting.
1377
            count += 1
2535.3.6 by Andrew Bennetts
Move some "what repo data to fetch logic" from RepoFetcher to Repository.
1378
            yield ("file", file_id, altered_versions)
2535.3.9 by Andrew Bennetts
More comments.
1379
        # We're done with the files_pb.  Note that it finished by the caller,
1380
        # just as it was created by the caller.
2668.2.8 by Andrew Bennetts
Rename get_data_to_fetch_for_revision_ids as item_keys_introduced_by.
1381
        del _files_pb
2535.3.6 by Andrew Bennetts
Move some "what repo data to fetch logic" from RepoFetcher to Repository.
1382
1383
        # inventory
1384
        yield ("inventory", None, revision_ids)
1385
        inv_w.clear_cache()
1386
1387
        # signatures
2535.3.25 by Andrew Bennetts
Fetch signatures too.
1388
        revisions_with_signatures = set()
1389
        for rev_id in revision_ids:
1390
            try:
1391
                self.get_signature_text(rev_id)
1392
            except errors.NoSuchRevision:
1393
                # not signed.
1394
                pass
1395
            else:
1396
                revisions_with_signatures.add(rev_id)
2535.3.47 by Andrew Bennetts
Massively improve item_keys_introduced_by performance by calling self.lock_read().
1397
        self.unlock()
2535.3.25 by Andrew Bennetts
Fetch signatures too.
1398
        yield ("signatures", None, revisions_with_signatures)
2535.3.6 by Andrew Bennetts
Move some "what repo data to fetch logic" from RepoFetcher to Repository.
1399
1400
        # revisions
1401
        yield ("revisions", None, revision_ids)
1402
1185.65.27 by Robert Collins
Tweak storage towards mergability.
1403
    @needs_read_lock
1185.65.1 by Aaron Bentley
Refactored out ControlFiles and RevisionStore from _Branch
1404
    def get_inventory_weave(self):
1405
        return self.control_weaves.get_weave('inventory',
1406
            self.get_transaction())
1407
1185.65.27 by Robert Collins
Tweak storage towards mergability.
1408
    @needs_read_lock
1185.65.1 by Aaron Bentley
Refactored out ControlFiles and RevisionStore from _Branch
1409
    def get_inventory(self, revision_id):
1410
        """Get Inventory object by hash."""
1740.2.3 by Aaron Bentley
Only reserialize the working tree basis inventory when needed.
1411
        return self.deserialise_inventory(
1412
            revision_id, self.get_inventory_xml(revision_id))
1413
1414
    def deserialise_inventory(self, revision_id, xml):
1415
        """Transform the xml into an inventory object. 
1416
1417
        :param revision_id: The expected revision id of the inventory.
1418
        :param xml: A serialised inventory.
1419
        """
2889.1.3 by Robert Collins
Merge bzr.dev.
1420
        return self._serializer.read_inventory_from_string(xml, revision_id)
1185.65.1 by Aaron Bentley
Refactored out ControlFiles and RevisionStore from _Branch
1421
1910.2.22 by Aaron Bentley
Make commits preserve root entry data
1422
    def serialise_inventory(self, inv):
1910.2.48 by Aaron Bentley
Update from review comments
1423
        return self._serializer.write_inventory_to_string(inv)
1910.2.22 by Aaron Bentley
Make commits preserve root entry data
1424
2817.2.1 by Robert Collins
* Inventory serialisation no longer double-sha's the content.
1425
    def _serialise_inventory_to_lines(self, inv):
1426
        return self._serializer.write_inventory_to_lines(inv)
1427
2520.4.113 by Aaron Bentley
Avoid peeking at Repository._serializer
1428
    def get_serializer_format(self):
1429
        return self._serializer.format_num
1430
1185.65.27 by Robert Collins
Tweak storage towards mergability.
1431
    @needs_read_lock
1185.65.1 by Aaron Bentley
Refactored out ControlFiles and RevisionStore from _Branch
1432
    def get_inventory_xml(self, revision_id):
1433
        """Get inventory XML as a file object."""
1434
        try:
2249.5.13 by John Arbash Meinel
Finish auditing Repository, and fix generate_ids to always generate utf8 ids.
1435
            assert isinstance(revision_id, str), type(revision_id)
1185.65.1 by Aaron Bentley
Refactored out ControlFiles and RevisionStore from _Branch
1436
            iw = self.get_inventory_weave()
1563.2.18 by Robert Collins
get knit repositories really using knits for text storage.
1437
            return iw.get_text(revision_id)
1185.65.1 by Aaron Bentley
Refactored out ControlFiles and RevisionStore from _Branch
1438
        except IndexError:
1773.4.1 by Martin Pool
Add pyflakes makefile target; fix many warnings
1439
            raise errors.HistoryMissing(self, 'inventory', revision_id)
1185.65.1 by Aaron Bentley
Refactored out ControlFiles and RevisionStore from _Branch
1440
1185.65.27 by Robert Collins
Tweak storage towards mergability.
1441
    @needs_read_lock
1185.65.1 by Aaron Bentley
Refactored out ControlFiles and RevisionStore from _Branch
1442
    def get_inventory_sha1(self, revision_id):
1443
        """Return the sha1 hash of the inventory entry
1444
        """
1445
        return self.get_revision(revision_id).inventory_sha1
1446
1185.65.27 by Robert Collins
Tweak storage towards mergability.
1447
    @needs_read_lock
2850.3.2 by Robert Collins
Review feedback.
1448
    def get_revision_graph(self, revision_id=None):
1449
        """Return a dictionary containing the revision graph.
1450
1451
        NB: This method should not be used as it accesses the entire graph all
1452
        at once, which is much more data than most operations should require.
1453
1454
        :param revision_id: The revision_id to get a graph from. If None, then
1455
        the entire revision graph is returned. This is a deprecated mode of
1456
        operation and will be removed in the future.
1457
        :return: a dictionary of revision_id->revision_parents_list.
1458
        """
1459
        raise NotImplementedError(self.get_revision_graph)
1460
1461
    @needs_read_lock
1594.2.3 by Robert Collins
bugfix revision.MultipleRevisionSources.get_revision_graph to integrate ghosts between sources. [slow on weaves, fast on knits.
1462
    def get_revision_graph_with_ghosts(self, revision_ids=None):
1463
        """Return a graph of the revisions with ghosts marked as applicable.
1464
1465
        :param revision_ids: an iterable of revisions to graph or None for all.
1466
        :return: a Graph object with the graph reachable from revision_ids.
1467
        """
2745.1.1 by Robert Collins
Add a number of -Devil checkpoints.
1468
        if 'evil' in debug.debug_flags:
2592.3.112 by Robert Collins
Various fixups found dogfooding.
1469
            mutter_callsite(3,
2745.1.1 by Robert Collins
Add a number of -Devil checkpoints.
1470
                "get_revision_graph_with_ghosts scales with size of history.")
2490.2.21 by Aaron Bentley
Rename graph to deprecated_graph
1471
        result = deprecated_graph.Graph()
1594.2.3 by Robert Collins
bugfix revision.MultipleRevisionSources.get_revision_graph to integrate ghosts between sources. [slow on weaves, fast on knits.
1472
        if not revision_ids:
1773.4.2 by Martin Pool
Cleanup of imports; undeprecate all_revision_ids()
1473
            pending = set(self.all_revision_ids())
1594.2.3 by Robert Collins
bugfix revision.MultipleRevisionSources.get_revision_graph to integrate ghosts between sources. [slow on weaves, fast on knits.
1474
            required = set([])
1475
        else:
2858.2.1 by Martin Pool
Remove most calls to safe_file_id and safe_revision_id.
1476
            pending = set(revision_ids)
1836.3.1 by Robert Collins
(robertc) Teach repository.get_revision_graph, and revision.common_ancestor, about NULL_REVISION.
1477
            # special case NULL_REVISION
1996.3.4 by John Arbash Meinel
lazy_import bzrlib/repository.py
1478
            if _mod_revision.NULL_REVISION in pending:
1479
                pending.remove(_mod_revision.NULL_REVISION)
1836.3.1 by Robert Collins
(robertc) Teach repository.get_revision_graph, and revision.common_ancestor, about NULL_REVISION.
1480
            required = set(pending)
1594.2.3 by Robert Collins
bugfix revision.MultipleRevisionSources.get_revision_graph to integrate ghosts between sources. [slow on weaves, fast on knits.
1481
        done = set([])
1482
        while len(pending):
1483
            revision_id = pending.pop()
1484
            try:
1485
                rev = self.get_revision(revision_id)
1486
            except errors.NoSuchRevision:
1487
                if revision_id in required:
1488
                    raise
1489
                # a ghost
1490
                result.add_ghost(revision_id)
1491
                continue
1492
            for parent_id in rev.parent_ids:
1493
                # is this queued or done ?
1494
                if (parent_id not in pending and
1495
                    parent_id not in done):
1496
                    # no, queue it.
1497
                    pending.add(parent_id)
1498
            result.add_node(revision_id, rev.parent_ids)
1594.2.15 by Robert Collins
Unfuck performance.
1499
            done.add(revision_id)
1594.2.3 by Robert Collins
bugfix revision.MultipleRevisionSources.get_revision_graph to integrate ghosts between sources. [slow on weaves, fast on knits.
1500
        return result
1501
2230.3.54 by Aaron Bentley
Move reverse history iteration to repository
1502
    def _get_history_vf(self):
1503
        """Get a versionedfile whose history graph reflects all revisions.
1504
1505
        For weave repositories, this is the inventory weave.
1506
        """
1507
        return self.get_inventory_weave()
1508
1509
    def iter_reverse_revision_history(self, revision_id):
1510
        """Iterate backwards through revision ids in the lefthand history
1511
1512
        :param revision_id: The revision id to start with.  All its lefthand
1513
            ancestors will be traversed.
1514
        """
1515
        if revision_id in (None, _mod_revision.NULL_REVISION):
1516
            return
1517
        next_id = revision_id
1518
        versionedfile = self._get_history_vf()
1519
        while True:
1520
            yield next_id
1521
            parents = versionedfile.get_parents(next_id)
1522
            if len(parents) == 0:
1523
                return
1524
            else:
1525
                next_id = parents[0]
1526
1594.2.3 by Robert Collins
bugfix revision.MultipleRevisionSources.get_revision_graph to integrate ghosts between sources. [slow on weaves, fast on knits.
1527
    @needs_read_lock
1185.65.1 by Aaron Bentley
Refactored out ControlFiles and RevisionStore from _Branch
1528
    def get_revision_inventory(self, revision_id):
1529
        """Return inventory of a past revision."""
1530
        # TODO: Unify this with get_inventory()
1531
        # bzr 0.0.6 and later imposes the constraint that the inventory_id
1532
        # must be the same as its revision, so this is trivial.
1534.4.28 by Robert Collins
first cut at merge from integration.
1533
        if revision_id is None:
1185.65.1 by Aaron Bentley
Refactored out ControlFiles and RevisionStore from _Branch
1534
            # This does not make sense: if there is no revision,
1535
            # then it is the current tree inventory surely ?!
1536
            # and thus get_root_id() is something that looks at the last
1537
            # commit on the branch, and the get_root_id is an inventory check.
1538
            raise NotImplementedError
1539
            # return Inventory(self.get_root_id())
1540
        else:
1541
            return self.get_inventory(revision_id)
1542
1185.65.27 by Robert Collins
Tweak storage towards mergability.
1543
    @needs_read_lock
1534.6.3 by Robert Collins
find_repository sufficiently robust.
1544
    def is_shared(self):
1545
        """Return True if this repository is flagged as a shared repository."""
1596.2.12 by Robert Collins
Merge and make Knit Repository use the revision store for all possible queries.
1546
        raise NotImplementedError(self.is_shared)
1534.6.3 by Robert Collins
find_repository sufficiently robust.
1547
1594.2.7 by Robert Collins
Add versionedfile.fix_parents api for correcting data post hoc.
1548
    @needs_write_lock
1692.1.1 by Robert Collins
* Repository.reconcile now takes a thorough keyword parameter to allow
1549
    def reconcile(self, other=None, thorough=False):
1594.2.7 by Robert Collins
Add versionedfile.fix_parents api for correcting data post hoc.
1550
        """Reconcile this repository."""
1551
        from bzrlib.reconcile import RepoReconciler
1692.1.1 by Robert Collins
* Repository.reconcile now takes a thorough keyword parameter to allow
1552
        reconciler = RepoReconciler(self, thorough=thorough)
1594.2.7 by Robert Collins
Add versionedfile.fix_parents api for correcting data post hoc.
1553
        reconciler.reconcile()
1554
        return reconciler
2440.1.1 by Martin Pool
Add new Repository.sprout,
1555
2617.6.2 by Robert Collins
Add abort_write_group and wire write_groups into fetch and commit.
1556
    def _refresh_data(self):
1557
        """Helper called from lock_* to ensure coherency with disk.
1558
1559
        The default implementation does nothing; it is however possible
1560
        for repositories to maintain loaded indices across multiple locks
1561
        by checking inside their implementation of this method to see
1562
        whether their indices are still valid. This depends of course on
1563
        the disk format being validatable in this manner.
1564
        """
1565
1534.6.3 by Robert Collins
find_repository sufficiently robust.
1566
    @needs_read_lock
1185.65.1 by Aaron Bentley
Refactored out ControlFiles and RevisionStore from _Branch
1567
    def revision_tree(self, revision_id):
1568
        """Return Tree for a revision on this branch.
1569
1852.5.1 by Robert Collins
Deprecate EmptyTree in favour of using Repository.revision_tree.
1570
        `revision_id` may be None for the empty tree revision.
1571
        """
1185.65.1 by Aaron Bentley
Refactored out ControlFiles and RevisionStore from _Branch
1572
        # TODO: refactor this to use an existing revision object
1573
        # so we don't need to read it in twice.
1996.3.4 by John Arbash Meinel
lazy_import bzrlib/repository.py
1574
        if revision_id is None or revision_id == _mod_revision.NULL_REVISION:
1731.1.61 by Aaron Bentley
Merge bzr.dev
1575
            return RevisionTree(self, Inventory(root_id=None), 
1576
                                _mod_revision.NULL_REVISION)
1185.65.1 by Aaron Bentley
Refactored out ControlFiles and RevisionStore from _Branch
1577
        else:
1578
            inv = self.get_revision_inventory(revision_id)
1185.65.17 by Robert Collins
Merge from integration, mode-changes are broken.
1579
            return RevisionTree(self, inv, revision_id)
1185.65.1 by Aaron Bentley
Refactored out ControlFiles and RevisionStore from _Branch
1580
1185.65.27 by Robert Collins
Tweak storage towards mergability.
1581
    @needs_read_lock
1756.3.3 by Aaron Bentley
More refactoring, introduce revision_trees.
1582
    def revision_trees(self, revision_ids):
1583
        """Return Tree for a revision on this branch.
1584
1756.3.19 by Aaron Bentley
Documentation and cleanups
1585
        `revision_id` may not be None or 'null:'"""
1756.3.3 by Aaron Bentley
More refactoring, introduce revision_trees.
1586
        assert None not in revision_ids
1996.3.4 by John Arbash Meinel
lazy_import bzrlib/repository.py
1587
        assert _mod_revision.NULL_REVISION not in revision_ids
1756.3.5 by Aaron Bentley
Switch to get_texts, optimize get_texts
1588
        texts = self.get_inventory_weave().get_texts(revision_ids)
1756.3.3 by Aaron Bentley
More refactoring, introduce revision_trees.
1589
        for text, revision_id in zip(texts, revision_ids):
1590
            inv = self.deserialise_inventory(revision_id, text)
1591
            yield RevisionTree(self, inv, revision_id)
1592
1593
    @needs_read_lock
2530.1.1 by Aaron Bentley
Make topological sorting optional for get_ancestry
1594
    def get_ancestry(self, revision_id, topo_sorted=True):
1185.66.2 by Aaron Bentley
Moved get_ancestry to RevisionStorage
1595
        """Return a list of revision-ids integrated by a revision.
1732.2.4 by Martin Pool
Split check into Branch.check and Repository.check
1596
1597
        The first element of the list is always None, indicating the origin 
1598
        revision.  This might change when we have history horizons, or 
1599
        perhaps we should have a new API.
1185.66.2 by Aaron Bentley
Moved get_ancestry to RevisionStorage
1600
        
1601
        This is topologically sorted.
1602
        """
2598.5.1 by Aaron Bentley
Start eliminating the use of None to indicate null revision
1603
        if _mod_revision.is_null(revision_id):
1185.66.2 by Aaron Bentley
Moved get_ancestry to RevisionStorage
1604
            return [None]
1534.4.41 by Robert Collins
Branch now uses BzrDir reasonably sanely.
1605
        if not self.has_revision(revision_id):
1606
            raise errors.NoSuchRevision(self, revision_id)
1185.66.2 by Aaron Bentley
Moved get_ancestry to RevisionStorage
1607
        w = self.get_inventory_weave()
2490.2.33 by Aaron Bentley
Disable topological sorting of get_ancestry where sensible
1608
        candidates = w.get_ancestry(revision_id, topo_sorted)
1594.2.9 by Robert Collins
Teach Knit repositories how to handle ghosts without corrupting at all.
1609
        return [None] + candidates # self._eliminate_revisions_not_present(candidates)
1185.66.2 by Aaron Bentley
Moved get_ancestry to RevisionStorage
1610
2604.2.1 by Robert Collins
(robertc) Introduce a pack command.
1611
    def pack(self):
1612
        """Compress the data within the repository.
1613
1614
        This operation only makes sense for some repository types. For other
1615
        types it should be a no-op that just returns.
1616
1617
        This stub method does not require a lock, but subclasses should use
1618
        @needs_write_lock as this is a long running call its reasonable to 
1619
        implicitly lock for the user.
1620
        """
1621
1185.65.4 by Aaron Bentley
Fixed cat command
1622
    @needs_read_lock
1623
    def print_file(self, file, revision_id):
1185.65.29 by Robert Collins
Implement final review suggestions.
1624
        """Print `file` to stdout.
1625
        
1626
        FIXME RBC 20060125 as John Meinel points out this is a bad api
1627
        - it writes to stdout, it assumes that that is valid etc. Fix
1628
        by creating a new more flexible convenience function.
1629
        """
1185.65.4 by Aaron Bentley
Fixed cat command
1630
        tree = self.revision_tree(revision_id)
1631
        # use inventory as it was in that revision
1632
        file_id = tree.inventory.path2id(file)
1633
        if not file_id:
1685.1.26 by John Arbash Meinel
Repository had a bug with what exception was raised when a file was missing
1634
            # TODO: jam 20060427 Write a test for this code path
1635
            #       it had a bug in it, and was raising the wrong
1636
            #       exception.
1637
            raise errors.BzrError("%r is not present in revision %s" % (file, revision_id))
1185.65.4 by Aaron Bentley
Fixed cat command
1638
        tree.print_file(file_id)
1639
1185.65.1 by Aaron Bentley
Refactored out ControlFiles and RevisionStore from _Branch
1640
    def get_transaction(self):
1641
        return self.control_files.get_transaction()
1642
2249.5.13 by John Arbash Meinel
Finish auditing Repository, and fix generate_ids to always generate utf8 ids.
1643
    def revision_parents(self, revision_id):
1644
        return self.get_inventory_weave().parent_names(revision_id)
1590.1.1 by Robert Collins
Improve common_ancestor performance.
1645
2490.2.13 by Aaron Bentley
Update distinct -> lowest, refactor, add ParentsProvider concept
1646
    def get_parents(self, revision_ids):
2490.2.22 by Aaron Bentley
Rename GraphWalker -> Graph, _AncestryWalker -> _BreadthFirstSearcher
1647
        """See StackedParentsProvider.get_parents"""
2490.2.13 by Aaron Bentley
Update distinct -> lowest, refactor, add ParentsProvider concept
1648
        parents_list = []
1649
        for revision_id in revision_ids:
1650
            if revision_id == _mod_revision.NULL_REVISION:
1651
                parents = []
1652
            else:
1653
                try:
1654
                    parents = self.get_revision(revision_id).parent_ids
1655
                except errors.NoSuchRevision:
1656
                    parents = None
1657
                else:
1658
                    if len(parents) == 0:
1659
                        parents = [_mod_revision.NULL_REVISION]
1660
            parents_list.append(parents)
1661
        return parents_list
1662
1663
    def _make_parents_provider(self):
1664
        return self
1665
2490.2.21 by Aaron Bentley
Rename graph to deprecated_graph
1666
    def get_graph(self, other_repository=None):
2490.2.13 by Aaron Bentley
Update distinct -> lowest, refactor, add ParentsProvider concept
1667
        """Return the graph walker for this repository format"""
1668
        parents_provider = self._make_parents_provider()
2490.2.14 by Aaron Bentley
Avoid StackedParentsProvider when underlying repos match
1669
        if (other_repository is not None and
1670
            other_repository.bzrdir.transport.base !=
1671
            self.bzrdir.transport.base):
2490.2.21 by Aaron Bentley
Rename graph to deprecated_graph
1672
            parents_provider = graph._StackedParentsProvider(
2490.2.13 by Aaron Bentley
Update distinct -> lowest, refactor, add ParentsProvider concept
1673
                [parents_provider, other_repository._make_parents_provider()])
2490.2.22 by Aaron Bentley
Rename GraphWalker -> Graph, _AncestryWalker -> _BreadthFirstSearcher
1674
        return graph.Graph(parents_provider)
2490.2.13 by Aaron Bentley
Update distinct -> lowest, refactor, add ParentsProvider concept
1675
3036.1.3 by Robert Collins
Privatise VersionedFileChecker.
1676
    def _get_versioned_file_checker(self):
2988.1.6 by Robert Collins
Change the contract for VersionedFileChecker to consolidate related parameters rather than splitting them across two api calls. This allows better reuse of a single checker object.
1677
        """Return an object suitable for checking versioned files."""
3036.1.3 by Robert Collins
Privatise VersionedFileChecker.
1678
        return _VersionedFileChecker(self)
2745.6.47 by Andrew Bennetts
Move check_parents out of VersionedFile.
1679
1185.65.27 by Robert Collins
Tweak storage towards mergability.
1680
    @needs_write_lock
1534.6.5 by Robert Collins
Cloning of repos preserves shared and make-working-tree attributes.
1681
    def set_make_working_trees(self, new_value):
1682
        """Set the policy flag for making working trees when creating branches.
1683
1684
        This only applies to branches that use this repository.
1685
1686
        The default is 'True'.
1687
        :param new_value: True to restore the default, False to disable making
1688
                          working trees.
1689
        """
1596.2.12 by Robert Collins
Merge and make Knit Repository use the revision store for all possible queries.
1690
        raise NotImplementedError(self.set_make_working_trees)
1534.6.5 by Robert Collins
Cloning of repos preserves shared and make-working-tree attributes.
1691
    
1692
    def make_working_trees(self):
1693
        """Returns the policy for making working trees on new branches."""
1596.2.12 by Robert Collins
Merge and make Knit Repository use the revision store for all possible queries.
1694
        raise NotImplementedError(self.make_working_trees)
1534.6.5 by Robert Collins
Cloning of repos preserves shared and make-working-tree attributes.
1695
1696
    @needs_write_lock
1185.65.1 by Aaron Bentley
Refactored out ControlFiles and RevisionStore from _Branch
1697
    def sign_revision(self, revision_id, gpg_strategy):
1698
        plaintext = Testament.from_revision(self, revision_id).as_short_text()
1699
        self.store_revision_signature(gpg_strategy, plaintext, revision_id)
1534.4.40 by Robert Collins
Add RepositoryFormats and allow bzrdir.open or create _repository to be used.
1700
1563.2.29 by Robert Collins
Remove all but fetch references to repository.revision_store.
1701
    @needs_read_lock
1702
    def has_signature_for_revision_id(self, revision_id):
1703
        """Query for a revision signature for revision_id in the repository."""
1704
        return self._revision_store.has_signature(revision_id,
1705
                                                  self.get_transaction())
1706
1563.2.31 by Robert Collins
Convert Knit repositories to use knits.
1707
    @needs_read_lock
1708
    def get_signature_text(self, revision_id):
1709
        """Return the text for a signature."""
1710
        return self._revision_store.get_signature_text(revision_id,
1711
                                                       self.get_transaction())
1712
1732.2.4 by Martin Pool
Split check into Branch.check and Repository.check
1713
    @needs_read_lock
2745.6.36 by Andrew Bennetts
Deprecate revision_ids arg to Repository.check and other tweaks.
1714
    def check(self, revision_ids=None):
1732.2.4 by Martin Pool
Split check into Branch.check and Repository.check
1715
        """Check consistency of all history of given revision_ids.
1716
1717
        Different repository implementations should override _check().
1718
1719
        :param revision_ids: A non-empty list of revision_ids whose ancestry
1720
             will be checked.  Typically the last revision_id of a branch.
1721
        """
1722
        return self._check(revision_ids)
1723
1724
    def _check(self, revision_ids):
1773.4.1 by Martin Pool
Add pyflakes makefile target; fix many warnings
1725
        result = check.Check(self)
1732.2.4 by Martin Pool
Split check into Branch.check and Repository.check
1726
        result.check()
1727
        return result
1728
1904.2.3 by Martin Pool
Give a warning on access to old repository formats
1729
    def _warn_if_deprecated(self):
1904.2.5 by Martin Pool
Fix format warning inside test suite and add test
1730
        global _deprecation_warning_done
1731
        if _deprecation_warning_done:
1732
            return
1733
        _deprecation_warning_done = True
1904.2.3 by Martin Pool
Give a warning on access to old repository formats
1734
        warning("Format %s for %s is deprecated - please use 'bzr upgrade' to get better performance"
1735
                % (self._format, self.bzrdir.transport.base))
1736
1910.2.63 by Aaron Bentley
Add supports_rich_root member to repository
1737
    def supports_rich_root(self):
1738
        return self._format.rich_root_data
1739
2150.2.2 by Robert Collins
Change the commit builder selected-revision-id test to use a unicode revision id where possible, leading to stricter testing of the hypothetical unicode revision id support in bzr.
1740
    def _check_ascii_revisionid(self, revision_id, method):
1741
        """Private helper for ascii-only repositories."""
1742
        # weave repositories refuse to store revisionids that are non-ascii.
1743
        if revision_id is not None:
1744
            # weaves require ascii revision ids.
1745
            if isinstance(revision_id, unicode):
1746
                try:
1747
                    revision_id.encode('ascii')
1748
                except UnicodeEncodeError:
1749
                    raise errors.NonAsciiRevisionId(method, self)
2249.5.12 by John Arbash Meinel
Change the APIs for VersionedFile, Store, and some of Repository into utf-8
1750
            else:
1751
                try:
1752
                    revision_id.decode('ascii')
1753
                except UnicodeDecodeError:
1754
                    raise errors.NonAsciiRevisionId(method, self)
2819.2.4 by Andrew Bennetts
Add a 'revision_graph_can_have_wrong_parents' method to repository.
1755
    
1756
    def revision_graph_can_have_wrong_parents(self):
1757
        """Is it possible for this repository to have a revision graph with
1758
        incorrect parents?
2150.2.2 by Robert Collins
Change the commit builder selected-revision-id test to use a unicode revision id where possible, leading to stricter testing of the hypothetical unicode revision id support in bzr.
1759
2819.2.4 by Andrew Bennetts
Add a 'revision_graph_can_have_wrong_parents' method to repository.
1760
        If True, then this repository must also implement
1761
        _find_inconsistent_revision_parents so that check and reconcile can
1762
        check for inconsistencies before proceeding with other checks that may
1763
        depend on the revision index being consistent.
1764
        """
1765
        raise NotImplementedError(self.revision_graph_can_have_wrong_parents)
1766
        
2241.1.18 by mbp at sourcefrog
Restore use of deprecating delegator for old formats in bzrlib.repository.
1767
# remove these delegates a while after bzr 0.15
1768
def __make_delegated(name, from_module):
1769
    def _deprecated_repository_forwarder():
1770
        symbol_versioning.warn('%s moved to %s in bzr 0.15'
1771
            % (name, from_module),
2241.1.20 by mbp at sourcefrog
update tests for new locations of weave repos
1772
            DeprecationWarning,
1773
            stacklevel=2)
2241.1.18 by mbp at sourcefrog
Restore use of deprecating delegator for old formats in bzrlib.repository.
1774
        m = __import__(from_module, globals(), locals(), [name])
1775
        try:
1776
            return getattr(m, name)
1777
        except AttributeError:
1778
            raise AttributeError('module %s has no name %s'
1779
                    % (m, name))
1780
    globals()[name] = _deprecated_repository_forwarder
1781
1782
for _name in [
1783
        'AllInOneRepository',
1784
        'WeaveMetaDirRepository',
1785
        'PreSplitOutRepositoryFormat',
1786
        'RepositoryFormat4',
1787
        'RepositoryFormat5',
1788
        'RepositoryFormat6',
1789
        'RepositoryFormat7',
1790
        ]:
1791
    __make_delegated(_name, 'bzrlib.repofmt.weaverepo')
1792
1793
for _name in [
1794
        'KnitRepository',
1795
        'RepositoryFormatKnit',
1796
        'RepositoryFormatKnit1',
1797
        ]:
1798
    __make_delegated(_name, 'bzrlib.repofmt.knitrepo')
1799
1800
2996.2.2 by Aaron Bentley
Create install_revisions function
1801
def install_revision(repository, rev, revision_tree):
1802
    """Install all revision data into a repository."""
1803
    install_revisions(repository, [(rev, revision_tree, None)])
1804
1805
1806
def install_revisions(repository, iterable):
2996.2.4 by Aaron Bentley
Rename function to add_signature_text
1807
    """Install all revision data into a repository.
1808
1809
    Accepts an iterable of revision, tree, signature tuples.  The signature
1810
    may be None.
1811
    """
2592.3.96 by Robert Collins
Merge index improvements (includes bzr.dev).
1812
    repository.start_write_group()
1813
    try:
2996.2.2 by Aaron Bentley
Create install_revisions function
1814
        for revision, revision_tree, signature in iterable:
1815
            _install_revision(repository, revision, revision_tree, signature)
2592.3.96 by Robert Collins
Merge index improvements (includes bzr.dev).
1816
    except:
1817
        repository.abort_write_group()
2592.3.101 by Robert Collins
Correctly propogate exceptions from repository.install_revisions.
1818
        raise
2592.3.96 by Robert Collins
Merge index improvements (includes bzr.dev).
1819
    else:
1820
        repository.commit_write_group()
1821
1822
2996.2.1 by Aaron Bentley
Add KnitRepositoryFormat4
1823
def _install_revision(repository, rev, revision_tree, signature):
2592.3.96 by Robert Collins
Merge index improvements (includes bzr.dev).
1824
    """Install all revision data into a repository."""
1185.82.84 by Aaron Bentley
Moved stuff around
1825
    present_parents = []
1826
    parent_trees = {}
1827
    for p_id in rev.parent_ids:
1828
        if repository.has_revision(p_id):
1829
            present_parents.append(p_id)
1830
            parent_trees[p_id] = repository.revision_tree(p_id)
1831
        else:
1852.5.1 by Robert Collins
Deprecate EmptyTree in favour of using Repository.revision_tree.
1832
            parent_trees[p_id] = repository.revision_tree(None)
1185.82.84 by Aaron Bentley
Moved stuff around
1833
1834
    inv = revision_tree.inventory
1910.2.51 by Aaron Bentley
Bundles now corrupt repositories
1835
    entries = inv.iter_entries()
2617.6.6 by Robert Collins
Some review feedback.
1836
    # backwards compatibility hack: skip the root id.
1910.2.63 by Aaron Bentley
Add supports_rich_root member to repository
1837
    if not repository.supports_rich_root():
1910.2.60 by Aaron Bentley
Ensure that new-model revisions aren't installed into old-model repos
1838
        path, root = entries.next()
1839
        if root.revision != rev.revision_id:
1910.2.63 by Aaron Bentley
Add supports_rich_root member to repository
1840
            raise errors.IncompatibleRevision(repr(repository))
1185.82.84 by Aaron Bentley
Moved stuff around
1841
    # Add the texts that are not already present
1852.6.3 by Robert Collins
Make iter(Tree) consistent for all tree types.
1842
    for path, ie in entries:
1185.82.84 by Aaron Bentley
Moved stuff around
1843
        w = repository.weave_store.get_weave_or_empty(ie.file_id,
1844
                repository.get_transaction())
1845
        if ie.revision not in w:
1846
            text_parents = []
1740.2.2 by Aaron Bentley
Add test for the basis inventory automatically adding the revision id.
1847
            # FIXME: TODO: The following loop *may* be overlapping/duplicate
1759.2.1 by Jelmer Vernooij
Fix some types (found using aspell).
1848
            # with InventoryEntry.find_previous_heads(). if it is, then there
1740.2.2 by Aaron Bentley
Add test for the basis inventory automatically adding the revision id.
1849
            # is a latent bug here where the parents may have ancestors of each
1850
            # other. RBC, AB
1185.82.84 by Aaron Bentley
Moved stuff around
1851
            for revision, tree in parent_trees.iteritems():
1852
                if ie.file_id not in tree:
1853
                    continue
1854
                parent_id = tree.inventory[ie.file_id].revision
1855
                if parent_id in text_parents:
1856
                    continue
1857
                text_parents.append(parent_id)
1858
                    
1859
            vfile = repository.weave_store.get_weave_or_empty(ie.file_id, 
1860
                repository.get_transaction())
1861
            lines = revision_tree.get_file(ie.file_id).readlines()
1862
            vfile.add_lines(rev.revision_id, text_parents, lines)
1863
    try:
1864
        # install the inventory
1865
        repository.add_inventory(rev.revision_id, inv, present_parents)
1866
    except errors.RevisionAlreadyPresent:
1867
        pass
2996.2.1 by Aaron Bentley
Add KnitRepositoryFormat4
1868
    if signature is not None:
2996.2.8 by Aaron Bentley
Fix add_signature discrepancies
1869
        repository.add_signature_text(rev.revision_id, signature)
1185.82.84 by Aaron Bentley
Moved stuff around
1870
    repository.add_revision(rev.revision_id, rev, inv)
1871
1872
1556.1.3 by Robert Collins
Rearrangment of Repository logic to be less type code driven, and bugfix InterRepository.missing_revision_ids
1873
class MetaDirRepository(Repository):
1874
    """Repositories in the new meta-dir layout."""
1875
1563.2.29 by Robert Collins
Remove all but fetch references to repository.revision_store.
1876
    def __init__(self, _format, a_bzrdir, control_files, _revision_store, control_store, text_store):
1556.1.3 by Robert Collins
Rearrangment of Repository logic to be less type code driven, and bugfix InterRepository.missing_revision_ids
1877
        super(MetaDirRepository, self).__init__(_format,
1878
                                                a_bzrdir,
1879
                                                control_files,
1563.2.29 by Robert Collins
Remove all but fetch references to repository.revision_store.
1880
                                                _revision_store,
1563.2.23 by Robert Collins
Add add_revision and get_revision methods to RevisionStore
1881
                                                control_store,
1563.2.17 by Robert Collins
Change knits repositories to use a knit versioned file store for file texts.
1882
                                                text_store)
1556.1.3 by Robert Collins
Rearrangment of Repository logic to be less type code driven, and bugfix InterRepository.missing_revision_ids
1883
        dir_mode = self.control_files._dir_mode
1884
        file_mode = self.control_files._file_mode
1885
1596.2.12 by Robert Collins
Merge and make Knit Repository use the revision store for all possible queries.
1886
    @needs_read_lock
1887
    def is_shared(self):
1888
        """Return True if this repository is flagged as a shared repository."""
1889
        return self.control_files._transport.has('shared-storage')
1890
1891
    @needs_write_lock
1892
    def set_make_working_trees(self, new_value):
1893
        """Set the policy flag for making working trees when creating branches.
1894
1895
        This only applies to branches that use this repository.
1896
1897
        The default is 'True'.
1898
        :param new_value: True to restore the default, False to disable making
1899
                          working trees.
1900
        """
1901
        if new_value:
1902
            try:
1903
                self.control_files._transport.delete('no-working-trees')
1904
            except errors.NoSuchFile:
1905
                pass
1906
        else:
1907
            self.control_files.put_utf8('no-working-trees', '')
1908
    
1909
    def make_working_trees(self):
1910
        """Returns the policy for making working trees on new branches."""
1911
        return not self.control_files._transport.has('no-working-trees')
1912
1556.1.3 by Robert Collins
Rearrangment of Repository logic to be less type code driven, and bugfix InterRepository.missing_revision_ids
1913
2241.1.2 by Martin Pool
change to using external Repository format registry
1914
class RepositoryFormatRegistry(registry.Registry):
2889.1.1 by Robert Collins
* The class ``bzrlib.repofmt.knitrepo.KnitRepository3`` has been folded into
1915
    """Registry of RepositoryFormats."""
2241.1.11 by Martin Pool
Get rid of RepositoryFormat*_instance objects. Instead the format
1916
1917
    def get(self, format_string):
1918
        r = registry.Registry.get(self, format_string)
1919
        if callable(r):
1920
            r = r()
1921
        return r
2241.1.2 by Martin Pool
change to using external Repository format registry
1922
    
1923
1924
format_registry = RepositoryFormatRegistry()
2241.1.11 by Martin Pool
Get rid of RepositoryFormat*_instance objects. Instead the format
1925
"""Registry of formats, indexed by their identifying format string.
1926
1927
This can contain either format instances themselves, or classes/factories that
1928
can be called to obtain one.
1929
"""
2241.1.2 by Martin Pool
change to using external Repository format registry
1930
2220.2.3 by Martin Pool
Add tag: revision namespace.
1931
1932
#####################################################################
1933
# Repository Formats
1910.2.46 by Aaron Bentley
Whitespace fix
1934
1534.4.40 by Robert Collins
Add RepositoryFormats and allow bzrdir.open or create _repository to be used.
1935
class RepositoryFormat(object):
1936
    """A repository format.
1937
1938
    Formats provide three things:
1939
     * An initialization routine to construct repository data on disk.
1940
     * a format string which is used when the BzrDir supports versioned
1941
       children.
1942
     * an open routine which returns a Repository instance.
1943
2889.1.2 by Robert Collins
Review feedback.
1944
    There is one and only one Format subclass for each on-disk format. But
1945
    there can be one Repository subclass that is used for several different
1946
    formats. The _format attribute on a Repository instance can be used to
1947
    determine the disk format.
2889.1.1 by Robert Collins
* The class ``bzrlib.repofmt.knitrepo.KnitRepository3`` has been folded into
1948
1534.4.40 by Robert Collins
Add RepositoryFormats and allow bzrdir.open or create _repository to be used.
1949
    Formats are placed in an dict by their format string for reference 
1950
    during opening. These should be subclasses of RepositoryFormat
1951
    for consistency.
1952
1953
    Once a format is deprecated, just deprecate the initialize and open
1954
    methods on the format class. Do not deprecate the object, as the 
1955
    object will be created every system load.
1956
1957
    Common instance attributes:
1958
    _matchingbzrdir - the bzrdir format that the repository format was
1959
    originally written to work with. This can be used if manually
1960
    constructing a bzrdir and repository, or more commonly for test suite
1961
    parameterisation.
1962
    """
1963
2949.1.2 by Robert Collins
* Fetch with pack repositories will no longer read the entire history graph.
1964
    # Set to True or False in derived classes. True indicates that the format
1965
    # supports ghosts gracefully.
1966
    supports_ghosts = None
1967
1904.2.3 by Martin Pool
Give a warning on access to old repository formats
1968
    def __str__(self):
1969
        return "<%s>" % self.__class__.__name__
1970
2241.1.11 by Martin Pool
Get rid of RepositoryFormat*_instance objects. Instead the format
1971
    def __eq__(self, other):
1972
        # format objects are generally stateless
1973
        return isinstance(other, self.__class__)
1974
2100.3.35 by Aaron Bentley
equality operations on bzrdir
1975
    def __ne__(self, other):
2100.3.31 by Aaron Bentley
Merged bzr.dev (17 tests failing)
1976
        return not self == other
1977
1534.4.40 by Robert Collins
Add RepositoryFormats and allow bzrdir.open or create _repository to be used.
1978
    @classmethod
1534.4.47 by Robert Collins
Split out repository into .bzr/repository
1979
    def find_format(klass, a_bzrdir):
2241.1.1 by Martin Pool
Change RepositoryFormat to use a Registry rather than ad-hoc dictionary
1980
        """Return the format for the repository object in a_bzrdir.
1981
        
1982
        This is used by bzr native formats that have a "format" file in
1983
        the repository.  Other methods may be used by different types of 
1984
        control directory.
1985
        """
1534.4.47 by Robert Collins
Split out repository into .bzr/repository
1986
        try:
1987
            transport = a_bzrdir.get_repository_transport(None)
1988
            format_string = transport.get("format").read()
2241.1.2 by Martin Pool
change to using external Repository format registry
1989
            return format_registry.get(format_string)
1534.4.47 by Robert Collins
Split out repository into .bzr/repository
1990
        except errors.NoSuchFile:
1991
            raise errors.NoRepositoryPresent(a_bzrdir)
1992
        except KeyError:
1740.5.6 by Martin Pool
Clean up many exception classes.
1993
            raise errors.UnknownFormatError(format=format_string)
1534.4.47 by Robert Collins
Split out repository into .bzr/repository
1994
2241.1.1 by Martin Pool
Change RepositoryFormat to use a Registry rather than ad-hoc dictionary
1995
    @classmethod
2241.1.2 by Martin Pool
change to using external Repository format registry
1996
    def register_format(klass, format):
1997
        format_registry.register(format.get_format_string(), format)
2241.1.1 by Martin Pool
Change RepositoryFormat to use a Registry rather than ad-hoc dictionary
1998
1999
    @classmethod
2000
    def unregister_format(klass, format):
2241.1.2 by Martin Pool
change to using external Repository format registry
2001
        format_registry.remove(format.get_format_string())
1563.2.23 by Robert Collins
Add add_revision and get_revision methods to RevisionStore
2002
    
1534.4.47 by Robert Collins
Split out repository into .bzr/repository
2003
    @classmethod
1534.4.40 by Robert Collins
Add RepositoryFormats and allow bzrdir.open or create _repository to be used.
2004
    def get_default_format(klass):
2005
        """Return the current default format."""
2204.5.3 by Aaron Bentley
zap old repository default handling
2006
        from bzrlib import bzrdir
2007
        return bzrdir.format_registry.make_bzrdir('default').repository_format
2241.1.1 by Martin Pool
Change RepositoryFormat to use a Registry rather than ad-hoc dictionary
2008
2009
    def _get_control_store(self, repo_transport, control_files):
2010
        """Return the control store for this repository."""
2011
        raise NotImplementedError(self._get_control_store)
1534.4.40 by Robert Collins
Add RepositoryFormats and allow bzrdir.open or create _repository to be used.
2012
2013
    def get_format_string(self):
2014
        """Return the ASCII format string that identifies this format.
2015
        
2016
        Note that in pre format ?? repositories the format string is 
2017
        not permitted nor written to disk.
2018
        """
2019
        raise NotImplementedError(self.get_format_string)
2020
1624.3.19 by Olaf Conradi
New call get_format_description to give a user-friendly description of a
2021
    def get_format_description(self):
1759.2.1 by Jelmer Vernooij
Fix some types (found using aspell).
2022
        """Return the short description for this format."""
1624.3.19 by Olaf Conradi
New call get_format_description to give a user-friendly description of a
2023
        raise NotImplementedError(self.get_format_description)
2024
1556.1.3 by Robert Collins
Rearrangment of Repository logic to be less type code driven, and bugfix InterRepository.missing_revision_ids
2025
    def _get_revision_store(self, repo_transport, control_files):
2026
        """Return the revision store object for this a_bzrdir."""
1556.1.5 by Robert Collins
Review feedback.
2027
        raise NotImplementedError(self._get_revision_store)
1556.1.3 by Robert Collins
Rearrangment of Repository logic to be less type code driven, and bugfix InterRepository.missing_revision_ids
2028
1563.2.22 by Robert Collins
Move responsibility for repository.has_revision into RevisionStore
2029
    def _get_text_rev_store(self,
2030
                            transport,
2031
                            control_files,
2032
                            name,
2033
                            compressed=True,
1563.2.28 by Robert Collins
Add total_size to the revision_store api.
2034
                            prefixed=False,
2035
                            serializer=None):
1556.1.3 by Robert Collins
Rearrangment of Repository logic to be less type code driven, and bugfix InterRepository.missing_revision_ids
2036
        """Common logic for getting a revision store for a repository.
2037
        
1563.2.17 by Robert Collins
Change knits repositories to use a knit versioned file store for file texts.
2038
        see self._get_revision_store for the subclass-overridable method to 
1556.1.3 by Robert Collins
Rearrangment of Repository logic to be less type code driven, and bugfix InterRepository.missing_revision_ids
2039
        get the store for a repository.
2040
        """
1563.2.22 by Robert Collins
Move responsibility for repository.has_revision into RevisionStore
2041
        from bzrlib.store.revision.text import TextRevisionStore
1556.1.3 by Robert Collins
Rearrangment of Repository logic to be less type code driven, and bugfix InterRepository.missing_revision_ids
2042
        dir_mode = control_files._dir_mode
2043
        file_mode = control_files._file_mode
2220.2.2 by Martin Pool
Add tag command and basic implementation
2044
        text_store = TextStore(transport.clone(name),
1563.2.22 by Robert Collins
Move responsibility for repository.has_revision into RevisionStore
2045
                              prefixed=prefixed,
2046
                              compressed=compressed,
2047
                              dir_mode=dir_mode,
2048
                              file_mode=file_mode)
1563.2.29 by Robert Collins
Remove all but fetch references to repository.revision_store.
2049
        _revision_store = TextRevisionStore(text_store, serializer)
2050
        return _revision_store
1556.1.3 by Robert Collins
Rearrangment of Repository logic to be less type code driven, and bugfix InterRepository.missing_revision_ids
2051
2241.1.6 by Martin Pool
Move Knit repositories into the submodule bzrlib.repofmt.knitrepo and
2052
    # TODO: this shouldn't be in the base class, it's specific to things that
2053
    # use weaves or knits -- mbp 20070207
1563.2.17 by Robert Collins
Change knits repositories to use a knit versioned file store for file texts.
2054
    def _get_versioned_file_store(self,
2055
                                  name,
2056
                                  transport,
2057
                                  control_files,
2058
                                  prefixed=True,
2241.1.10 by Martin Pool
Remove more references to weaves from the repository.py file
2059
                                  versionedfile_class=None,
1946.2.5 by John Arbash Meinel
Make knit stores delay creation, but not control stores
2060
                                  versionedfile_kwargs={},
1608.2.12 by Martin Pool
Store-escaping must quote uppercase characters too, so that they're safely
2061
                                  escaped=False):
2241.1.10 by Martin Pool
Remove more references to weaves from the repository.py file
2062
        if versionedfile_class is None:
2063
            versionedfile_class = self._versionedfile_class
1563.2.17 by Robert Collins
Change knits repositories to use a knit versioned file store for file texts.
2064
        weave_transport = control_files._transport.clone(name)
2065
        dir_mode = control_files._dir_mode
2066
        file_mode = control_files._file_mode
2067
        return VersionedFileStore(weave_transport, prefixed=prefixed,
1608.2.12 by Martin Pool
Store-escaping must quote uppercase characters too, so that they're safely
2068
                                  dir_mode=dir_mode,
2069
                                  file_mode=file_mode,
2070
                                  versionedfile_class=versionedfile_class,
1946.2.5 by John Arbash Meinel
Make knit stores delay creation, but not control stores
2071
                                  versionedfile_kwargs=versionedfile_kwargs,
1608.2.12 by Martin Pool
Store-escaping must quote uppercase characters too, so that they're safely
2072
                                  escaped=escaped)
1563.2.17 by Robert Collins
Change knits repositories to use a knit versioned file store for file texts.
2073
1534.6.1 by Robert Collins
allow API creation of shared repositories
2074
    def initialize(self, a_bzrdir, shared=False):
2075
        """Initialize a repository of this format in a_bzrdir.
2076
2077
        :param a_bzrdir: The bzrdir to put the new repository in it.
2078
        :param shared: The repository should be initialized as a sharable one.
1752.2.52 by Andrew Bennetts
Flesh out more Remote* methods needed to open and initialise remote branches/trees/repositories.
2079
        :returns: The new repository object.
2080
        
1534.6.1 by Robert Collins
allow API creation of shared repositories
2081
        This may raise UninitializableFormat if shared repository are not
2082
        compatible the a_bzrdir.
1534.4.40 by Robert Collins
Add RepositoryFormats and allow bzrdir.open or create _repository to be used.
2083
        """
1752.2.52 by Andrew Bennetts
Flesh out more Remote* methods needed to open and initialise remote branches/trees/repositories.
2084
        raise NotImplementedError(self.initialize)
1534.4.40 by Robert Collins
Add RepositoryFormats and allow bzrdir.open or create _repository to be used.
2085
2086
    def is_supported(self):
2087
        """Is this format supported?
2088
2089
        Supported formats must be initializable and openable.
2090
        Unsupported formats may not support initialization or committing or 
2091
        some other features depending on the reason for not being supported.
2092
        """
2093
        return True
2094
1910.2.12 by Aaron Bentley
Implement knit repo format 2
2095
    def check_conversion_target(self, target_format):
2096
        raise NotImplementedError(self.check_conversion_target)
2097
1534.4.40 by Robert Collins
Add RepositoryFormats and allow bzrdir.open or create _repository to be used.
2098
    def open(self, a_bzrdir, _found=False):
2099
        """Return an instance of this format for the bzrdir a_bzrdir.
2100
        
2101
        _found is a private parameter, do not use it.
2102
        """
1556.1.3 by Robert Collins
Rearrangment of Repository logic to be less type code driven, and bugfix InterRepository.missing_revision_ids
2103
        raise NotImplementedError(self.open)
1534.4.40 by Robert Collins
Add RepositoryFormats and allow bzrdir.open or create _repository to be used.
2104
1556.1.3 by Robert Collins
Rearrangment of Repository logic to be less type code driven, and bugfix InterRepository.missing_revision_ids
2105
2106
class MetaDirRepositoryFormat(RepositoryFormat):
1759.2.1 by Jelmer Vernooij
Fix some types (found using aspell).
2107
    """Common base class for the new repositories using the metadir layout."""
1556.1.3 by Robert Collins
Rearrangment of Repository logic to be less type code driven, and bugfix InterRepository.missing_revision_ids
2108
1910.2.14 by Aaron Bentley
Fail when trying to use interrepository on Knit2 and Knit1
2109
    rich_root_data = False
2323.5.17 by Martin Pool
Add supports_tree_reference to all repo formats (robert)
2110
    supports_tree_reference = False
2241.1.11 by Martin Pool
Get rid of RepositoryFormat*_instance objects. Instead the format
2111
    _matchingbzrdir = bzrdir.BzrDirMetaFormat1()
1910.2.14 by Aaron Bentley
Fail when trying to use interrepository on Knit2 and Knit1
2112
1556.1.4 by Robert Collins
Add a new format for what will become knit, and the surrounding logic to upgrade repositories within metadirs, and tests for the same.
2113
    def __init__(self):
2114
        super(MetaDirRepositoryFormat, self).__init__()
2115
1556.1.3 by Robert Collins
Rearrangment of Repository logic to be less type code driven, and bugfix InterRepository.missing_revision_ids
2116
    def _create_control_files(self, a_bzrdir):
2117
        """Create the required files and the initial control_files object."""
1759.2.2 by Jelmer Vernooij
Revert some of my spelling fixes and fix some typos after review by Aaron.
2118
        # FIXME: RBC 20060125 don't peek under the covers
1534.4.47 by Robert Collins
Split out repository into .bzr/repository
2119
        # NB: no need to escape relative paths that are url safe.
2120
        repository_transport = a_bzrdir.get_repository_transport(self)
1996.3.4 by John Arbash Meinel
lazy_import bzrlib/repository.py
2121
        control_files = lockable_files.LockableFiles(repository_transport,
2122
                                'lock', lockdir.LockDir)
1553.5.61 by Martin Pool
Locks protecting LockableFiles must now be explicitly created before use.
2123
        control_files.create_lock()
1556.1.3 by Robert Collins
Rearrangment of Repository logic to be less type code driven, and bugfix InterRepository.missing_revision_ids
2124
        return control_files
2125
2126
    def _upload_blank_content(self, a_bzrdir, dirs, files, utf8_files, shared):
2127
        """Upload the initial blank content."""
2128
        control_files = self._create_control_files(a_bzrdir)
1534.4.47 by Robert Collins
Split out repository into .bzr/repository
2129
        control_files.lock_write()
2130
        try:
1553.5.49 by Martin Pool
Use LockDirs for repo format 7
2131
            control_files._transport.mkdir_multi(dirs,
2132
                    mode=control_files._dir_mode)
1534.4.47 by Robert Collins
Split out repository into .bzr/repository
2133
            for file, content in files:
2134
                control_files.put(file, content)
2135
            for file, content in utf8_files:
2136
                control_files.put_utf8(file, content)
1534.6.1 by Robert Collins
allow API creation of shared repositories
2137
            if shared == True:
2138
                control_files.put_utf8('shared-storage', '')
1534.4.47 by Robert Collins
Split out repository into .bzr/repository
2139
        finally:
2140
            control_files.unlock()
1556.1.3 by Robert Collins
Rearrangment of Repository logic to be less type code driven, and bugfix InterRepository.missing_revision_ids
2141
2142
1534.4.40 by Robert Collins
Add RepositoryFormats and allow bzrdir.open or create _repository to be used.
2143
# formats which have no format string are not discoverable
2241.1.4 by Martin Pool
Moved old weave-based repository formats into bzrlib.repofmt.weaverepo.
2144
# and not independently creatable, so are not registered.  They're 
2241.1.11 by Martin Pool
Get rid of RepositoryFormat*_instance objects. Instead the format
2145
# all in bzrlib.repofmt.weaverepo now.  When an instance of one of these is
2146
# needed, it's constructed directly by the BzrDir.  Non-native formats where
2147
# the repository is not separately opened are similar.
2148
2241.1.4 by Martin Pool
Moved old weave-based repository formats into bzrlib.repofmt.weaverepo.
2149
format_registry.register_lazy(
2150
    'Bazaar-NG Repository format 7',
2151
    'bzrlib.repofmt.weaverepo',
2241.1.11 by Martin Pool
Get rid of RepositoryFormat*_instance objects. Instead the format
2152
    'RepositoryFormat7'
2241.1.4 by Martin Pool
Moved old weave-based repository formats into bzrlib.repofmt.weaverepo.
2153
    )
2592.3.22 by Robert Collins
Add new experimental repository formats.
2154
2241.1.6 by Martin Pool
Move Knit repositories into the submodule bzrlib.repofmt.knitrepo and
2155
format_registry.register_lazy(
2156
    'Bazaar-NG Knit Repository Format 1',
2157
    'bzrlib.repofmt.knitrepo',
2241.1.11 by Martin Pool
Get rid of RepositoryFormat*_instance objects. Instead the format
2158
    'RepositoryFormatKnit1',
2241.1.6 by Martin Pool
Move Knit repositories into the submodule bzrlib.repofmt.knitrepo and
2159
    )
2160
2241.1.5 by Martin Pool
Move KnitFormat2 into repofmt
2161
format_registry.register_lazy(
2255.2.230 by Robert Collins
Update tree format signatures to mention introducing bzr version.
2162
    'Bazaar Knit Repository Format 3 (bzr 0.15)\n',
2100.3.31 by Aaron Bentley
Merged bzr.dev (17 tests failing)
2163
    'bzrlib.repofmt.knitrepo',
2164
    'RepositoryFormatKnit3',
2165
    )
1534.4.40 by Robert Collins
Add RepositoryFormats and allow bzrdir.open or create _repository to be used.
2166
2996.2.1 by Aaron Bentley
Add KnitRepositoryFormat4
2167
format_registry.register_lazy(
2168
    'Bazaar Knit Repository Format 4 (bzr 1.0)\n',
2169
    'bzrlib.repofmt.knitrepo',
2170
    'RepositoryFormatKnit4',
2171
    )
2172
2939.2.1 by Ian Clatworthy
use 'knitpack' naming instead of 'experimental' for pack formats
2173
# Pack-based formats. There is one format for pre-subtrees, and one for
2174
# post-subtrees to allow ease of testing.
2939.2.5 by Ian Clatworthy
review feedback from lifeless
2175
# NOTE: These are experimental in 0.92.
2592.3.22 by Robert Collins
Add new experimental repository formats.
2176
format_registry.register_lazy(
2939.2.6 by Ian Clatworthy
more review feedback from lifeless and poolie
2177
    'Bazaar pack repository format 1 (needs bzr 0.92)\n',
2592.3.88 by Robert Collins
Move Pack repository logic to bzrlib.repofmt.pack_repo.
2178
    'bzrlib.repofmt.pack_repo',
2592.3.224 by Martin Pool
Rename GraphKnitRepository etc to KnitPackRepository
2179
    'RepositoryFormatKnitPack1',
2592.3.22 by Robert Collins
Add new experimental repository formats.
2180
    )
2181
format_registry.register_lazy(
2939.2.6 by Ian Clatworthy
more review feedback from lifeless and poolie
2182
    'Bazaar pack repository format 1 with subtree support (needs bzr 0.92)\n',
2592.3.88 by Robert Collins
Move Pack repository logic to bzrlib.repofmt.pack_repo.
2183
    'bzrlib.repofmt.pack_repo',
2592.3.224 by Martin Pool
Rename GraphKnitRepository etc to KnitPackRepository
2184
    'RepositoryFormatKnitPack3',
2592.3.22 by Robert Collins
Add new experimental repository formats.
2185
    )
2996.2.11 by Aaron Bentley
Implement rich-root-pack format ( #164639)
2186
format_registry.register_lazy(
2187
    'Bazaar pack repository format 1 with rich root (needs bzr 1.0)\n',
2188
    'bzrlib.repofmt.pack_repo',
2189
    'RepositoryFormatKnitPack4',
2190
    )
2592.3.22 by Robert Collins
Add new experimental repository formats.
2191
1534.4.40 by Robert Collins
Add RepositoryFormats and allow bzrdir.open or create _repository to be used.
2192
1563.2.12 by Robert Collins
Checkpointing: created InterObject to factor out common inter object worker code, added InterVersionedFile and tests to allow making join work between any versionedfile.
2193
class InterRepository(InterObject):
1534.1.27 by Robert Collins
Start InterRepository with InterRepository.get.
2194
    """This class represents operations taking place between two repositories.
2195
1534.1.33 by Robert Collins
Move copy_content_into into InterRepository and InterWeaveRepo, and disable the default codepath test as we have optimised paths for all current combinations.
2196
    Its instances have methods like copy_content and fetch, and contain
1534.1.27 by Robert Collins
Start InterRepository with InterRepository.get.
2197
    references to the source and target repositories these operations can be 
2198
    carried out on.
2199
2200
    Often we will provide convenience methods on 'repository' which carry out
2201
    operations with another repository - they will always forward to
2202
    InterRepository.get(other).method_name(parameters).
2203
    """
2204
1910.2.15 by Aaron Bentley
Back out inter.get changes, make optimizers an ordered list
2205
    _optimisers = []
1534.1.28 by Robert Collins
Allow for optimised InterRepository selection.
2206
    """The available optimised InterRepository types."""
2207
2387.1.1 by Robert Collins
Remove the --basis parameter to clone etc. (Robert Collins)
2208
    def copy_content(self, revision_id=None):
1910.2.15 by Aaron Bentley
Back out inter.get changes, make optimizers an ordered list
2209
        raise NotImplementedError(self.copy_content)
2210
2949.1.2 by Robert Collins
* Fetch with pack repositories will no longer read the entire history graph.
2211
    def fetch(self, revision_id=None, pb=None, find_ghosts=False):
1534.1.31 by Robert Collins
Deprecated fetch.fetch and fetch.greedy_fetch for branch.fetch, and move the Repository.fetch internals to InterRepo and InterWeaveRepo.
2212
        """Fetch the content required to construct revision_id.
2213
1910.7.17 by Andrew Bennetts
Various cosmetic changes.
2214
        The content is copied from self.source to self.target.
1534.1.31 by Robert Collins
Deprecated fetch.fetch and fetch.greedy_fetch for branch.fetch, and move the Repository.fetch internals to InterRepo and InterWeaveRepo.
2215
2216
        :param revision_id: if None all content is copied, if NULL_REVISION no
2217
                            content is copied.
2218
        :param pb: optional progress bar to use for progress reports. If not
2219
                   provided a default one will be created.
2220
2221
        Returns the copied revision count and the failed revisions in a tuple:
2222
        (copied, failures).
2223
        """
1910.2.15 by Aaron Bentley
Back out inter.get changes, make optimizers an ordered list
2224
        raise NotImplementedError(self.fetch)
2225
   
2226
    @needs_read_lock
3010.1.5 by Robert Collins
Test that missing_revision_ids handles the case of the source not having the requested revision correctly with and without find_ghosts.
2227
    def missing_revision_ids(self, revision_id=None, find_ghosts=True):
1910.2.15 by Aaron Bentley
Back out inter.get changes, make optimizers an ordered list
2228
        """Return the revision ids that source has that target does not.
2229
        
2230
        These are returned in topological order.
2231
2232
        :param revision_id: only return revision ids included by this
2233
                            revision_id.
2234
        """
2235
        # generic, possibly worst case, slow code path.
2236
        target_ids = set(self.target.all_revision_ids())
2237
        if revision_id is not None:
2238
            source_ids = self.source.get_ancestry(revision_id)
1963.2.6 by Robey Pointer
pychecker is on crack; go back to using 'is None'.
2239
            assert source_ids[0] is None
1910.2.15 by Aaron Bentley
Back out inter.get changes, make optimizers an ordered list
2240
            source_ids.pop(0)
2241
        else:
2242
            source_ids = self.source.all_revision_ids()
2243
        result_set = set(source_ids).difference(target_ids)
2244
        # this may look like a no-op: its not. It preserves the ordering
2245
        # other_ids had while only returning the members from other_ids
2246
        # that we've decided we need.
2247
        return [rev_id for rev_id in source_ids if rev_id in result_set]
2248
2592.3.28 by Robert Collins
Make InterKnitOptimiser be used between any same-model knit repository.
2249
    @staticmethod
2250
    def _same_model(source, target):
2251
        """True if source and target have the same data representation."""
2252
        if source.supports_rich_root() != target.supports_rich_root():
2253
            return False
2254
        if source._serializer != target._serializer:
2255
            return False
2256
        return True
2257
1910.2.15 by Aaron Bentley
Back out inter.get changes, make optimizers an ordered list
2258
2259
class InterSameDataRepository(InterRepository):
2260
    """Code for converting between repositories that represent the same data.
2261
    
2262
    Data format and model must match for this to work.
2263
    """
2264
2241.1.6 by Martin Pool
Move Knit repositories into the submodule bzrlib.repofmt.knitrepo and
2265
    @classmethod
2241.1.7 by Martin Pool
rename method
2266
    def _get_repo_format_to_test(self):
2814.1.1 by Robert Collins
* Pushing, pulling and branching branches with subtree references was not
2267
        """Repository format for testing with.
2268
        
2269
        InterSameData can pull from subtree to subtree and from non-subtree to
2270
        non-subtree, so we test this with the richest repository format.
2271
        """
2272
        from bzrlib.repofmt import knitrepo
2273
        return knitrepo.RepositoryFormatKnit3()
1910.2.15 by Aaron Bentley
Back out inter.get changes, make optimizers an ordered list
2274
1910.2.14 by Aaron Bentley
Fail when trying to use interrepository on Knit2 and Knit1
2275
    @staticmethod
2276
    def is_compatible(source, target):
2592.3.28 by Robert Collins
Make InterKnitOptimiser be used between any same-model knit repository.
2277
        return InterRepository._same_model(source, target)
1910.2.14 by Aaron Bentley
Fail when trying to use interrepository on Knit2 and Knit1
2278
1534.1.33 by Robert Collins
Move copy_content_into into InterRepository and InterWeaveRepo, and disable the default codepath test as we have optimised paths for all current combinations.
2279
    @needs_write_lock
2387.1.1 by Robert Collins
Remove the --basis parameter to clone etc. (Robert Collins)
2280
    def copy_content(self, revision_id=None):
1534.1.33 by Robert Collins
Move copy_content_into into InterRepository and InterWeaveRepo, and disable the default codepath test as we have optimised paths for all current combinations.
2281
        """Make a complete copy of the content in self into destination.
2440.1.1 by Martin Pool
Add new Repository.sprout,
2282
2283
        This copies both the repository's revision data, and configuration information
2284
        such as the make_working_trees setting.
1534.1.33 by Robert Collins
Move copy_content_into into InterRepository and InterWeaveRepo, and disable the default codepath test as we have optimised paths for all current combinations.
2285
        
2286
        This is a destructive operation! Do not use it on existing 
2287
        repositories.
2288
2289
        :param revision_id: Only copy the content needed to construct
2290
                            revision_id and its parents.
2291
        """
2292
        try:
2293
            self.target.set_make_working_trees(self.source.make_working_trees())
2294
        except NotImplementedError:
2295
            pass
1759.2.2 by Jelmer Vernooij
Revert some of my spelling fixes and fix some typos after review by Aaron.
2296
        # but don't bother fetching if we have the needed data now.
1996.3.20 by John Arbash Meinel
[merge] bzr.dev 2063
2297
        if (revision_id not in (None, _mod_revision.NULL_REVISION) and 
1534.1.33 by Robert Collins
Move copy_content_into into InterRepository and InterWeaveRepo, and disable the default codepath test as we have optimised paths for all current combinations.
2298
            self.target.has_revision(revision_id)):
2299
            return
2300
        self.target.fetch(self.source, revision_id=revision_id)
2301
2302
    @needs_write_lock
2949.1.2 by Robert Collins
* Fetch with pack repositories will no longer read the entire history graph.
2303
    def fetch(self, revision_id=None, pb=None, find_ghosts=False):
1910.7.20 by Andrew Bennetts
Merge from bzr.dev
2304
        """See InterRepository.fetch()."""
1563.2.31 by Robert Collins
Convert Knit repositories to use knits.
2305
        from bzrlib.fetch import GenericRepoFetcher
1534.1.31 by Robert Collins
Deprecated fetch.fetch and fetch.greedy_fetch for branch.fetch, and move the Repository.fetch internals to InterRepo and InterWeaveRepo.
2306
        mutter("Using fetch logic to copy between %s(%s) and %s(%s)",
2858.2.1 by Martin Pool
Remove most calls to safe_file_id and safe_revision_id.
2307
               self.source, self.source._format, self.target,
1910.2.15 by Aaron Bentley
Back out inter.get changes, make optimizers an ordered list
2308
               self.target._format)
1563.2.31 by Robert Collins
Convert Knit repositories to use knits.
2309
        f = GenericRepoFetcher(to_repository=self.target,
2310
                               from_repository=self.source,
2311
                               last_revision=revision_id,
2312
                               pb=pb)
1534.1.33 by Robert Collins
Move copy_content_into into InterRepository and InterWeaveRepo, and disable the default codepath test as we have optimised paths for all current combinations.
2313
        return f.count_copied, f.failed_revisions
1534.1.31 by Robert Collins
Deprecated fetch.fetch and fetch.greedy_fetch for branch.fetch, and move the Repository.fetch internals to InterRepo and InterWeaveRepo.
2314
1910.2.15 by Aaron Bentley
Back out inter.get changes, make optimizers an ordered list
2315
2241.1.12 by Martin Pool
Restore InterWeaveRepo
2316
class InterWeaveRepo(InterSameDataRepository):
2850.3.1 by Robert Collins
Move various weave specific code out of the base Repository class to weaverepo.py.
2317
    """Optimised code paths between Weave based repositories.
2318
    
2319
    This should be in bzrlib/repofmt/weaverepo.py but we have not yet
2320
    implemented lazy inter-object optimisation.
2321
    """
2241.1.12 by Martin Pool
Restore InterWeaveRepo
2322
2241.1.13 by Martin Pool
Re-register InterWeaveRepo, fix test integration, add test for it
2323
    @classmethod
2241.1.12 by Martin Pool
Restore InterWeaveRepo
2324
    def _get_repo_format_to_test(self):
2325
        from bzrlib.repofmt import weaverepo
2326
        return weaverepo.RepositoryFormat7()
2327
2328
    @staticmethod
2329
    def is_compatible(source, target):
2330
        """Be compatible with known Weave formats.
2331
        
2332
        We don't test for the stores being of specific types because that
2333
        could lead to confusing results, and there is no need to be 
2334
        overly general.
2335
        """
2336
        from bzrlib.repofmt.weaverepo import (
2337
                RepositoryFormat5,
2338
                RepositoryFormat6,
2339
                RepositoryFormat7,
2340
                )
2341
        try:
2342
            return (isinstance(source._format, (RepositoryFormat5,
2343
                                                RepositoryFormat6,
2344
                                                RepositoryFormat7)) and
2345
                    isinstance(target._format, (RepositoryFormat5,
2346
                                                RepositoryFormat6,
2347
                                                RepositoryFormat7)))
2348
        except AttributeError:
2349
            return False
2350
    
2351
    @needs_write_lock
2387.1.1 by Robert Collins
Remove the --basis parameter to clone etc. (Robert Collins)
2352
    def copy_content(self, revision_id=None):
2241.1.12 by Martin Pool
Restore InterWeaveRepo
2353
        """See InterRepository.copy_content()."""
2354
        # weave specific optimised path:
2387.1.1 by Robert Collins
Remove the --basis parameter to clone etc. (Robert Collins)
2355
        try:
2356
            self.target.set_make_working_trees(self.source.make_working_trees())
2357
        except NotImplementedError:
2358
            pass
2359
        # FIXME do not peek!
2360
        if self.source.control_files._transport.listable():
2361
            pb = ui.ui_factory.nested_progress_bar()
2241.1.12 by Martin Pool
Restore InterWeaveRepo
2362
            try:
2387.1.1 by Robert Collins
Remove the --basis parameter to clone etc. (Robert Collins)
2363
                self.target.weave_store.copy_all_ids(
2364
                    self.source.weave_store,
2365
                    pb=pb,
2366
                    from_transaction=self.source.get_transaction(),
2367
                    to_transaction=self.target.get_transaction())
2368
                pb.update('copying inventory', 0, 1)
2369
                self.target.control_weaves.copy_multi(
2370
                    self.source.control_weaves, ['inventory'],
2371
                    from_transaction=self.source.get_transaction(),
2372
                    to_transaction=self.target.get_transaction())
2373
                self.target._revision_store.text_store.copy_all_ids(
2374
                    self.source._revision_store.text_store,
2375
                    pb=pb)
2376
            finally:
2377
                pb.finished()
2378
        else:
2241.1.12 by Martin Pool
Restore InterWeaveRepo
2379
            self.target.fetch(self.source, revision_id=revision_id)
2380
2381
    @needs_write_lock
2949.1.2 by Robert Collins
* Fetch with pack repositories will no longer read the entire history graph.
2382
    def fetch(self, revision_id=None, pb=None, find_ghosts=False):
2241.1.12 by Martin Pool
Restore InterWeaveRepo
2383
        """See InterRepository.fetch()."""
2384
        from bzrlib.fetch import GenericRepoFetcher
2385
        mutter("Using fetch logic to copy between %s(%s) and %s(%s)",
2386
               self.source, self.source._format, self.target, self.target._format)
2387
        f = GenericRepoFetcher(to_repository=self.target,
2388
                               from_repository=self.source,
2389
                               last_revision=revision_id,
2390
                               pb=pb)
2391
        return f.count_copied, f.failed_revisions
2392
2393
    @needs_read_lock
3010.1.5 by Robert Collins
Test that missing_revision_ids handles the case of the source not having the requested revision correctly with and without find_ghosts.
2394
    def missing_revision_ids(self, revision_id=None, find_ghosts=True):
2241.1.12 by Martin Pool
Restore InterWeaveRepo
2395
        """See InterRepository.missing_revision_ids()."""
2396
        # we want all revisions to satisfy revision_id in source.
2397
        # but we don't want to stat every file here and there.
2398
        # we want then, all revisions other needs to satisfy revision_id 
2399
        # checked, but not those that we have locally.
2400
        # so the first thing is to get a subset of the revisions to 
2401
        # satisfy revision_id in source, and then eliminate those that
2402
        # we do already have. 
2403
        # this is slow on high latency connection to self, but as as this
2404
        # disk format scales terribly for push anyway due to rewriting 
2405
        # inventory.weave, this is considered acceptable.
2406
        # - RBC 20060209
2407
        if revision_id is not None:
2408
            source_ids = self.source.get_ancestry(revision_id)
2409
            assert source_ids[0] is None
2410
            source_ids.pop(0)
2411
        else:
2412
            source_ids = self.source._all_possible_ids()
2413
        source_ids_set = set(source_ids)
2414
        # source_ids is the worst possible case we may need to pull.
2415
        # now we want to filter source_ids against what we actually
2416
        # have in target, but don't try to check for existence where we know
2417
        # we do not have a revision as that would be pointless.
2418
        target_ids = set(self.target._all_possible_ids())
2419
        possibly_present_revisions = target_ids.intersection(source_ids_set)
2420
        actually_present_revisions = set(self.target._eliminate_revisions_not_present(possibly_present_revisions))
2421
        required_revisions = source_ids_set.difference(actually_present_revisions)
2422
        required_topo_revisions = [rev_id for rev_id in source_ids if rev_id in required_revisions]
2423
        if revision_id is not None:
2424
            # we used get_ancestry to determine source_ids then we are assured all
2425
            # revisions referenced are present as they are installed in topological order.
2426
            # and the tip revision was validated by get_ancestry.
2427
            return required_topo_revisions
2428
        else:
2429
            # if we just grabbed the possibly available ids, then 
2430
            # we only have an estimate of whats available and need to validate
2431
            # that against the revision records.
2432
            return self.source._eliminate_revisions_not_present(required_topo_revisions)
2433
2434
1910.2.15 by Aaron Bentley
Back out inter.get changes, make optimizers an ordered list
2435
class InterKnitRepo(InterSameDataRepository):
1563.2.31 by Robert Collins
Convert Knit repositories to use knits.
2436
    """Optimised code paths between Knit based repositories."""
2437
2241.1.6 by Martin Pool
Move Knit repositories into the submodule bzrlib.repofmt.knitrepo and
2438
    @classmethod
2241.1.7 by Martin Pool
rename method
2439
    def _get_repo_format_to_test(self):
2241.1.6 by Martin Pool
Move Knit repositories into the submodule bzrlib.repofmt.knitrepo and
2440
        from bzrlib.repofmt import knitrepo
2441
        return knitrepo.RepositoryFormatKnit1()
1563.2.31 by Robert Collins
Convert Knit repositories to use knits.
2442
2443
    @staticmethod
2444
    def is_compatible(source, target):
2445
        """Be compatible with known Knit formats.
2446
        
1759.2.2 by Jelmer Vernooij
Revert some of my spelling fixes and fix some typos after review by Aaron.
2447
        We don't test for the stores being of specific types because that
1563.2.31 by Robert Collins
Convert Knit repositories to use knits.
2448
        could lead to confusing results, and there is no need to be 
2449
        overly general.
2450
        """
2592.3.28 by Robert Collins
Make InterKnitOptimiser be used between any same-model knit repository.
2451
        from bzrlib.repofmt.knitrepo import RepositoryFormatKnit
1563.2.31 by Robert Collins
Convert Knit repositories to use knits.
2452
        try:
2592.3.28 by Robert Collins
Make InterKnitOptimiser be used between any same-model knit repository.
2453
            are_knits = (isinstance(source._format, RepositoryFormatKnit) and
2454
                isinstance(target._format, RepositoryFormatKnit))
1563.2.31 by Robert Collins
Convert Knit repositories to use knits.
2455
        except AttributeError:
2456
            return False
2592.3.28 by Robert Collins
Make InterKnitOptimiser be used between any same-model knit repository.
2457
        return are_knits and InterRepository._same_model(source, target)
1563.2.31 by Robert Collins
Convert Knit repositories to use knits.
2458
2459
    @needs_write_lock
2949.1.2 by Robert Collins
* Fetch with pack repositories will no longer read the entire history graph.
2460
    def fetch(self, revision_id=None, pb=None, find_ghosts=False):
1563.2.31 by Robert Collins
Convert Knit repositories to use knits.
2461
        """See InterRepository.fetch()."""
2462
        from bzrlib.fetch import KnitRepoFetcher
2463
        mutter("Using fetch logic to copy between %s(%s) and %s(%s)",
2464
               self.source, self.source._format, self.target, self.target._format)
2465
        f = KnitRepoFetcher(to_repository=self.target,
2466
                            from_repository=self.source,
2467
                            last_revision=revision_id,
2468
                            pb=pb)
2469
        return f.count_copied, f.failed_revisions
2470
2471
    @needs_read_lock
3010.1.5 by Robert Collins
Test that missing_revision_ids handles the case of the source not having the requested revision correctly with and without find_ghosts.
2472
    def missing_revision_ids(self, revision_id=None, find_ghosts=True):
1563.2.31 by Robert Collins
Convert Knit repositories to use knits.
2473
        """See InterRepository.missing_revision_ids()."""
2474
        if revision_id is not None:
2475
            source_ids = self.source.get_ancestry(revision_id)
1963.2.6 by Robey Pointer
pychecker is on crack; go back to using 'is None'.
2476
            assert source_ids[0] is None
1668.1.14 by Martin Pool
merge olaf - InvalidRevisionId fixes
2477
            source_ids.pop(0)
1563.2.31 by Robert Collins
Convert Knit repositories to use knits.
2478
        else:
2850.3.1 by Robert Collins
Move various weave specific code out of the base Repository class to weaverepo.py.
2479
            source_ids = self.source.all_revision_ids()
1563.2.31 by Robert Collins
Convert Knit repositories to use knits.
2480
        source_ids_set = set(source_ids)
2481
        # source_ids is the worst possible case we may need to pull.
2482
        # now we want to filter source_ids against what we actually
1759.2.2 by Jelmer Vernooij
Revert some of my spelling fixes and fix some typos after review by Aaron.
2483
        # have in target, but don't try to check for existence where we know
1563.2.31 by Robert Collins
Convert Knit repositories to use knits.
2484
        # we do not have a revision as that would be pointless.
2850.3.1 by Robert Collins
Move various weave specific code out of the base Repository class to weaverepo.py.
2485
        target_ids = set(self.target.all_revision_ids())
1563.2.31 by Robert Collins
Convert Knit repositories to use knits.
2486
        possibly_present_revisions = target_ids.intersection(source_ids_set)
2487
        actually_present_revisions = set(self.target._eliminate_revisions_not_present(possibly_present_revisions))
2488
        required_revisions = source_ids_set.difference(actually_present_revisions)
2489
        required_topo_revisions = [rev_id for rev_id in source_ids if rev_id in required_revisions]
2490
        if revision_id is not None:
2491
            # we used get_ancestry to determine source_ids then we are assured all
2492
            # revisions referenced are present as they are installed in topological order.
2493
            # and the tip revision was validated by get_ancestry.
2494
            return required_topo_revisions
2495
        else:
2496
            # if we just grabbed the possibly available ids, then 
2497
            # we only have an estimate of whats available and need to validate
2498
            # that against the revision records.
2499
            return self.source._eliminate_revisions_not_present(required_topo_revisions)
2500
1910.2.17 by Aaron Bentley
Get fetching from 1 to 2 under test
2501
2592.3.88 by Robert Collins
Move Pack repository logic to bzrlib.repofmt.pack_repo.
2502
class InterPackRepo(InterSameDataRepository):
2503
    """Optimised code paths between Pack based repositories."""
2504
2505
    @classmethod
2506
    def _get_repo_format_to_test(self):
2507
        from bzrlib.repofmt import pack_repo
2592.3.224 by Martin Pool
Rename GraphKnitRepository etc to KnitPackRepository
2508
        return pack_repo.RepositoryFormatKnitPack1()
2592.3.88 by Robert Collins
Move Pack repository logic to bzrlib.repofmt.pack_repo.
2509
2510
    @staticmethod
2511
    def is_compatible(source, target):
2512
        """Be compatible with known Pack formats.
2513
        
2514
        We don't test for the stores being of specific types because that
2515
        could lead to confusing results, and there is no need to be 
2516
        overly general.
2517
        """
2518
        from bzrlib.repofmt.pack_repo import RepositoryFormatPack
2519
        try:
2592.3.91 by Robert Collins
Incrementally closing in on a correct fetch for packs.
2520
            are_packs = (isinstance(source._format, RepositoryFormatPack) and
2592.3.88 by Robert Collins
Move Pack repository logic to bzrlib.repofmt.pack_repo.
2521
                isinstance(target._format, RepositoryFormatPack))
2522
        except AttributeError:
2523
            return False
2592.3.91 by Robert Collins
Incrementally closing in on a correct fetch for packs.
2524
        return are_packs and InterRepository._same_model(source, target)
2592.3.88 by Robert Collins
Move Pack repository logic to bzrlib.repofmt.pack_repo.
2525
2526
    @needs_write_lock
2949.1.2 by Robert Collins
* Fetch with pack repositories will no longer read the entire history graph.
2527
    def fetch(self, revision_id=None, pb=None, find_ghosts=False):
2592.3.88 by Robert Collins
Move Pack repository logic to bzrlib.repofmt.pack_repo.
2528
        """See InterRepository.fetch()."""
2951.1.2 by Robert Collins
Partial refactoring of pack_repo to create a Packer object for packing.
2529
        from bzrlib.repofmt.pack_repo import Packer
2592.3.88 by Robert Collins
Move Pack repository logic to bzrlib.repofmt.pack_repo.
2530
        mutter("Using fetch logic to copy between %s(%s) and %s(%s)",
2531
               self.source, self.source._format, self.target, self.target._format)
2592.3.90 by Robert Collins
Slightly broken, but branch and fetch performance is now roughly on par (for bzr.dev) with knits - should be much faster for large repos.
2532
        self.count_copied = 0
2533
        if revision_id is None:
2534
            # TODO:
2535
            # everything to do - use pack logic
2536
            # to fetch from all packs to one without
2592.3.93 by Robert Collins
Steps toward filtering revisions/inventories/texts during fetch.
2537
            # inventory parsing etc, IFF nothing to be copied is in the target.
2592.3.90 by Robert Collins
Slightly broken, but branch and fetch performance is now roughly on par (for bzr.dev) with knits - should be much faster for large repos.
2538
            # till then:
2539
            revision_ids = self.source.all_revision_ids()
2592.3.93 by Robert Collins
Steps toward filtering revisions/inventories/texts during fetch.
2540
            # implementing the TODO will involve:
2541
            # - detecting when all of a pack is selected
2542
            # - avoiding as much as possible pre-selection, so the
2543
            # more-core routines such as create_pack_from_packs can filter in
2544
            # a just-in-time fashion. (though having a HEADS list on a
2545
            # repository might make this a lot easier, because we could
2546
            # sensibly detect 'new revisions' without doing a full index scan.
2592.3.91 by Robert Collins
Incrementally closing in on a correct fetch for packs.
2547
        elif _mod_revision.is_null(revision_id):
2548
            # nothing to do:
3010.1.5 by Robert Collins
Test that missing_revision_ids handles the case of the source not having the requested revision correctly with and without find_ghosts.
2549
            return (0, [])
2592.3.90 by Robert Collins
Slightly broken, but branch and fetch performance is now roughly on par (for bzr.dev) with knits - should be much faster for large repos.
2550
        else:
2551
            try:
2949.1.2 by Robert Collins
* Fetch with pack repositories will no longer read the entire history graph.
2552
                revision_ids = self.missing_revision_ids(revision_id,
2553
                    find_ghosts=find_ghosts)
2592.3.90 by Robert Collins
Slightly broken, but branch and fetch performance is now roughly on par (for bzr.dev) with knits - should be much faster for large repos.
2554
            except errors.NoSuchRevision:
2555
                raise errors.InstallFailed([revision_id])
2592.3.232 by Martin Pool
Disambiguate two member variables called _packs into _packs_by_name and _pack_collection
2556
        packs = self.source._pack_collection.all_packs()
2951.1.2 by Robert Collins
Partial refactoring of pack_repo to create a Packer object for packing.
2557
        pack = Packer(self.target._pack_collection, packs, '.fetch',
2558
            revision_ids).pack()
2592.3.91 by Robert Collins
Incrementally closing in on a correct fetch for packs.
2559
        if pack is not None:
2592.3.232 by Martin Pool
Disambiguate two member variables called _packs into _packs_by_name and _pack_collection
2560
            self.target._pack_collection._save_pack_names()
2592.3.108 by Robert Collins
Autopack after pack to pack fetching too.
2561
            # Trigger an autopack. This may duplicate effort as we've just done
2562
            # a pack creation, but for now it is simpler to think about as
2563
            # 'upload data, then repack if needed'.
2592.3.232 by Martin Pool
Disambiguate two member variables called _packs into _packs_by_name and _pack_collection
2564
            self.target._pack_collection.autopack()
3010.1.5 by Robert Collins
Test that missing_revision_ids handles the case of the source not having the requested revision correctly with and without find_ghosts.
2565
            return (pack.get_revision_count(), [])
2592.3.91 by Robert Collins
Incrementally closing in on a correct fetch for packs.
2566
        else:
3010.1.5 by Robert Collins
Test that missing_revision_ids handles the case of the source not having the requested revision correctly with and without find_ghosts.
2567
            return (0, [])
2592.3.88 by Robert Collins
Move Pack repository logic to bzrlib.repofmt.pack_repo.
2568
2569
    @needs_read_lock
2949.1.2 by Robert Collins
* Fetch with pack repositories will no longer read the entire history graph.
2570
    def missing_revision_ids(self, revision_id=None, find_ghosts=True):
2571
        """See InterRepository.missing_revision_ids().
2572
        
2573
        :param find_ghosts: Find ghosts throughough the ancestry of
2574
            revision_id.
2575
        """
2576
        if not find_ghosts and revision_id is not None:
2577
            graph = self.source.get_graph()
2578
            missing_revs = set()
2579
            searcher = graph._make_breadth_first_searcher([revision_id])
2580
            target_index = \
2581
                self.target._pack_collection.revision_index.combined_index
2582
            null_set = frozenset([_mod_revision.NULL_REVISION])
2583
            while True:
2584
                try:
2585
                    next_revs = set(searcher.next())
2586
                except StopIteration:
2587
                    break
2588
                next_revs.difference_update(null_set)
2949.1.4 by Robert Collins
Use a list comprehension not a generator.
2589
                target_keys = [(key,) for key in next_revs]
2949.1.3 by Robert Collins
Review feedback.
2590
                have_revs = frozenset(node[1][0] for node in
2591
                    target_index.iter_entries(target_keys))
2949.1.2 by Robert Collins
* Fetch with pack repositories will no longer read the entire history graph.
2592
                missing_revs.update(next_revs - have_revs)
2593
                searcher.stop_searching_any(have_revs)
3010.1.5 by Robert Collins
Test that missing_revision_ids handles the case of the source not having the requested revision correctly with and without find_ghosts.
2594
            if next_revs - have_revs == set([revision_id]):
2595
                # we saw the start rev itself, but no parents from it (or
2596
                # next_revs would have been updated to e.g. set(). We remove
2597
                # have_revs because if we found revision_id locally we
2598
                # stop_searching at the first time around.
2599
                raise errors.NoSuchRevision(self.source, revision_id)
2949.1.2 by Robert Collins
* Fetch with pack repositories will no longer read the entire history graph.
2600
            return missing_revs
2601
        elif revision_id is not None:
2592.3.88 by Robert Collins
Move Pack repository logic to bzrlib.repofmt.pack_repo.
2602
            source_ids = self.source.get_ancestry(revision_id)
2603
            assert source_ids[0] is None
2604
            source_ids.pop(0)
2605
        else:
2592.3.151 by Robert Collins
Use the revision index, not the inventory index, for missing and fetch operations.
2606
            source_ids = self.source.all_revision_ids()
2592.3.88 by Robert Collins
Move Pack repository logic to bzrlib.repofmt.pack_repo.
2607
        # source_ids is the worst possible case we may need to pull.
2608
        # now we want to filter source_ids against what we actually
2609
        # have in target, but don't try to check for existence where we know
2610
        # we do not have a revision as that would be pointless.
2592.3.151 by Robert Collins
Use the revision index, not the inventory index, for missing and fetch operations.
2611
        target_ids = set(self.target.all_revision_ids())
2592.3.241 by Martin Pool
Remove redundant set construction from InterPackRepo.missing_revisions
2612
        return [r for r in source_ids if (r not in target_ids)]
2592.3.88 by Robert Collins
Move Pack repository logic to bzrlib.repofmt.pack_repo.
2613
2614
1910.2.24 by Aaron Bentley
Got intra-repository fetch working between model1 and 2 for all types
2615
class InterModel1and2(InterRepository):
2616
2241.1.6 by Martin Pool
Move Knit repositories into the submodule bzrlib.repofmt.knitrepo and
2617
    @classmethod
2241.1.7 by Martin Pool
rename method
2618
    def _get_repo_format_to_test(self):
2241.1.6 by Martin Pool
Move Knit repositories into the submodule bzrlib.repofmt.knitrepo and
2619
        return None
1910.2.24 by Aaron Bentley
Got intra-repository fetch working between model1 and 2 for all types
2620
2621
    @staticmethod
2622
    def is_compatible(source, target):
2305.2.1 by Andrew Bennetts
Use repo.supports_rich_root() everywhere rather than
2623
        if not source.supports_rich_root() and target.supports_rich_root():
1910.2.24 by Aaron Bentley
Got intra-repository fetch working between model1 and 2 for all types
2624
            return True
2625
        else:
2626
            return False
2627
2628
    @needs_write_lock
2949.1.2 by Robert Collins
* Fetch with pack repositories will no longer read the entire history graph.
2629
    def fetch(self, revision_id=None, pb=None, find_ghosts=False):
1910.2.24 by Aaron Bentley
Got intra-repository fetch working between model1 and 2 for all types
2630
        """See InterRepository.fetch()."""
2631
        from bzrlib.fetch import Model1toKnit2Fetcher
2632
        f = Model1toKnit2Fetcher(to_repository=self.target,
2633
                                 from_repository=self.source,
2634
                                 last_revision=revision_id,
2635
                                 pb=pb)
2636
        return f.count_copied, f.failed_revisions
2637
1910.2.26 by Aaron Bentley
Fix up some test cases
2638
    @needs_write_lock
2387.1.1 by Robert Collins
Remove the --basis parameter to clone etc. (Robert Collins)
2639
    def copy_content(self, revision_id=None):
1910.2.26 by Aaron Bentley
Fix up some test cases
2640
        """Make a complete copy of the content in self into destination.
2641
        
2642
        This is a destructive operation! Do not use it on existing 
2643
        repositories.
2644
2645
        :param revision_id: Only copy the content needed to construct
2646
                            revision_id and its parents.
2647
        """
2648
        try:
2649
            self.target.set_make_working_trees(self.source.make_working_trees())
2650
        except NotImplementedError:
2651
            pass
2652
        # but don't bother fetching if we have the needed data now.
1996.3.20 by John Arbash Meinel
[merge] bzr.dev 2063
2653
        if (revision_id not in (None, _mod_revision.NULL_REVISION) and 
1910.2.26 by Aaron Bentley
Fix up some test cases
2654
            self.target.has_revision(revision_id)):
2655
            return
2656
        self.target.fetch(self.source, revision_id=revision_id)
2657
1910.2.24 by Aaron Bentley
Got intra-repository fetch working between model1 and 2 for all types
2658
1910.2.17 by Aaron Bentley
Get fetching from 1 to 2 under test
2659
class InterKnit1and2(InterKnitRepo):
2660
2241.1.6 by Martin Pool
Move Knit repositories into the submodule bzrlib.repofmt.knitrepo and
2661
    @classmethod
2241.1.7 by Martin Pool
rename method
2662
    def _get_repo_format_to_test(self):
2241.1.6 by Martin Pool
Move Knit repositories into the submodule bzrlib.repofmt.knitrepo and
2663
        return None
1910.2.24 by Aaron Bentley
Got intra-repository fetch working between model1 and 2 for all types
2664
1910.2.17 by Aaron Bentley
Get fetching from 1 to 2 under test
2665
    @staticmethod
2666
    def is_compatible(source, target):
2255.2.211 by Robert Collins
Remove knit2 repository format- it has never been supported.
2667
        """Be compatible with Knit1 source and Knit3 target"""
2668
        from bzrlib.repofmt.knitrepo import RepositoryFormatKnit3
1910.2.17 by Aaron Bentley
Get fetching from 1 to 2 under test
2669
        try:
2592.3.88 by Robert Collins
Move Pack repository logic to bzrlib.repofmt.pack_repo.
2670
            from bzrlib.repofmt.knitrepo import (RepositoryFormatKnit1,
2671
                RepositoryFormatKnit3)
2592.3.224 by Martin Pool
Rename GraphKnitRepository etc to KnitPackRepository
2672
            from bzrlib.repofmt.pack_repo import (RepositoryFormatKnitPack1,
2673
                RepositoryFormatKnitPack3)
2592.3.220 by Robert Collins
Relax Knit1 to Knit3 conversion restrictions to allow Knit1 to Pack3.
2674
            return (isinstance(source._format,
2592.3.224 by Martin Pool
Rename GraphKnitRepository etc to KnitPackRepository
2675
                    (RepositoryFormatKnit1, RepositoryFormatKnitPack1)) and
2592.3.220 by Robert Collins
Relax Knit1 to Knit3 conversion restrictions to allow Knit1 to Pack3.
2676
                isinstance(target._format,
2592.3.224 by Martin Pool
Rename GraphKnitRepository etc to KnitPackRepository
2677
                    (RepositoryFormatKnit3, RepositoryFormatKnitPack3))
2592.3.220 by Robert Collins
Relax Knit1 to Knit3 conversion restrictions to allow Knit1 to Pack3.
2678
                )
1910.2.17 by Aaron Bentley
Get fetching from 1 to 2 under test
2679
        except AttributeError:
2680
            return False
2681
2682
    @needs_write_lock
2949.1.2 by Robert Collins
* Fetch with pack repositories will no longer read the entire history graph.
2683
    def fetch(self, revision_id=None, pb=None, find_ghosts=False):
1910.2.17 by Aaron Bentley
Get fetching from 1 to 2 under test
2684
        """See InterRepository.fetch()."""
2685
        from bzrlib.fetch import Knit1to2Fetcher
2686
        mutter("Using fetch logic to copy between %s(%s) and %s(%s)",
2687
               self.source, self.source._format, self.target, 
2688
               self.target._format)
2689
        f = Knit1to2Fetcher(to_repository=self.target,
2690
                            from_repository=self.source,
2691
                            last_revision=revision_id,
2692
                            pb=pb)
2693
        return f.count_copied, f.failed_revisions
2694
2695
2996.2.1 by Aaron Bentley
Add KnitRepositoryFormat4
2696
class InterDifferingSerializer(InterKnitRepo):
2697
2698
    @classmethod
2699
    def _get_repo_format_to_test(self):
2700
        return None
2701
2702
    @staticmethod
2703
    def is_compatible(source, target):
2704
        """Be compatible with Knit2 source and Knit3 target"""
2705
        if source.supports_rich_root() != target.supports_rich_root():
2706
            return False
2707
        # Ideally, we'd support fetching if the source had no tree references
2708
        # even if it supported them...
2709
        if (getattr(source, '_format.supports_tree_reference', False) and
2710
            not getattr(target, '_format.supports_tree_reference', False)):
2711
            return False
2712
        return True
2713
2714
    @needs_write_lock
2715
    def fetch(self, revision_id=None, pb=None, find_ghosts=False):
2716
        """See InterRepository.fetch()."""
2717
        revision_ids = self.target.missing_revision_ids(self.source,
2718
                                                        revision_id)
2996.2.2 by Aaron Bentley
Create install_revisions function
2719
        def revisions_iterator():
2720
            for current_revision_id in revision_ids:
2721
                revision = self.source.get_revision(current_revision_id)
2722
                tree = self.source.revision_tree(current_revision_id)
2723
                try:
2724
                    signature = self.source.get_signature_text(
2725
                        current_revision_id)
2726
                except errors.NoSuchRevision:
2727
                    signature = None
2728
                yield revision, tree, signature
2729
        install_revisions(self.target, revisions_iterator())
2996.2.1 by Aaron Bentley
Add KnitRepositoryFormat4
2730
        return len(revision_ids), 0
2731
2732
2535.3.12 by Andrew Bennetts
Add a first cut of a get_data_stream method to Repository.
2733
class InterRemoteToOther(InterRepository):
2734
2735
    def __init__(self, source, target):
2736
        InterRepository.__init__(self, source, target)
2737
        self._real_inter = None
2738
2739
    @staticmethod
2740
    def is_compatible(source, target):
2535.3.41 by Andrew Bennetts
Add tests for InterRemoteToOther.is_compatible.
2741
        if not isinstance(source, remote.RemoteRepository):
2742
            return False
2743
        source._ensure_real()
2744
        real_source = source._real_repository
2535.3.21 by Andrew Bennetts
All tests passing.
2745
        # Is source's model compatible with target's model, and are they the
2746
        # same format?  Currently we can only optimise fetching from an
2747
        # identical model & format repo.
2535.3.41 by Andrew Bennetts
Add tests for InterRemoteToOther.is_compatible.
2748
        assert not isinstance(real_source, remote.RemoteRepository), (
2749
            "We don't support remote repos backed by remote repos yet.")
2750
        return real_source._format == target._format
2751
2752
    @needs_write_lock
2949.1.2 by Robert Collins
* Fetch with pack repositories will no longer read the entire history graph.
2753
    def fetch(self, revision_id=None, pb=None, find_ghosts=False):
2535.3.41 by Andrew Bennetts
Add tests for InterRemoteToOther.is_compatible.
2754
        """See InterRepository.fetch()."""
2755
        from bzrlib.fetch import RemoteToOtherFetcher
2756
        mutter("Using fetch logic to copy between %s(remote) and %s(%s)",
2757
               self.source, self.target, self.target._format)
2758
        # TODO: jam 20070210 This should be an assert, not a translate
2759
        revision_id = osutils.safe_revision_id(revision_id)
2760
        f = RemoteToOtherFetcher(to_repository=self.target,
2761
                                 from_repository=self.source,
2762
                                 last_revision=revision_id,
2763
                                 pb=pb)
2764
        return f.count_copied, f.failed_revisions
2535.3.12 by Andrew Bennetts
Add a first cut of a get_data_stream method to Repository.
2765
2766
    @classmethod
2767
    def _get_repo_format_to_test(self):
2768
        return None
2769
2770
2771
class InterOtherToRemote(InterRepository):
2772
2773
    def __init__(self, source, target):
2774
        InterRepository.__init__(self, source, target)
2775
        self._real_inter = None
2776
2777
    @staticmethod
2778
    def is_compatible(source, target):
2779
        if isinstance(target, remote.RemoteRepository):
2780
            return True
2781
        return False
2782
2783
    def _ensure_real_inter(self):
2784
        if self._real_inter is None:
2785
            self.target._ensure_real()
2786
            real_target = self.target._real_repository
2787
            self._real_inter = InterRepository.get(self.source, real_target)
2788
    
2789
    def copy_content(self, revision_id=None):
2790
        self._ensure_real_inter()
2791
        self._real_inter.copy_content(revision_id=revision_id)
2792
2949.1.2 by Robert Collins
* Fetch with pack repositories will no longer read the entire history graph.
2793
    def fetch(self, revision_id=None, pb=None, find_ghosts=False):
2535.3.12 by Andrew Bennetts
Add a first cut of a get_data_stream method to Repository.
2794
        self._ensure_real_inter()
2795
        self._real_inter.fetch(revision_id=revision_id, pb=pb)
2796
2797
    @classmethod
2798
    def _get_repo_format_to_test(self):
2799
        return None
2800
2801
2996.2.1 by Aaron Bentley
Add KnitRepositoryFormat4
2802
InterRepository.register_optimiser(InterDifferingSerializer)
1910.2.15 by Aaron Bentley
Back out inter.get changes, make optimizers an ordered list
2803
InterRepository.register_optimiser(InterSameDataRepository)
2241.1.13 by Martin Pool
Re-register InterWeaveRepo, fix test integration, add test for it
2804
InterRepository.register_optimiser(InterWeaveRepo)
1563.2.31 by Robert Collins
Convert Knit repositories to use knits.
2805
InterRepository.register_optimiser(InterKnitRepo)
1910.2.24 by Aaron Bentley
Got intra-repository fetch working between model1 and 2 for all types
2806
InterRepository.register_optimiser(InterModel1and2)
1910.2.17 by Aaron Bentley
Get fetching from 1 to 2 under test
2807
InterRepository.register_optimiser(InterKnit1and2)
2592.3.90 by Robert Collins
Slightly broken, but branch and fetch performance is now roughly on par (for bzr.dev) with knits - should be much faster for large repos.
2808
InterRepository.register_optimiser(InterPackRepo)
2535.3.12 by Andrew Bennetts
Add a first cut of a get_data_stream method to Repository.
2809
InterRepository.register_optimiser(InterRemoteToOther)
2810
InterRepository.register_optimiser(InterOtherToRemote)
1534.1.31 by Robert Collins
Deprecated fetch.fetch and fetch.greedy_fetch for branch.fetch, and move the Repository.fetch internals to InterRepo and InterWeaveRepo.
2811
2812
1556.1.4 by Robert Collins
Add a new format for what will become knit, and the surrounding logic to upgrade repositories within metadirs, and tests for the same.
2813
class CopyConverter(object):
2814
    """A repository conversion tool which just performs a copy of the content.
2815
    
2816
    This is slow but quite reliable.
2817
    """
2818
2819
    def __init__(self, target_format):
2820
        """Create a CopyConverter.
2821
2822
        :param target_format: The format the resulting repository should be.
2823
        """
2824
        self.target_format = target_format
2825
        
2826
    def convert(self, repo, pb):
2827
        """Perform the conversion of to_convert, giving feedback via pb.
2828
2829
        :param to_convert: The disk object to convert.
2830
        :param pb: a progress bar to use for progress information.
2831
        """
2832
        self.pb = pb
2833
        self.count = 0
1596.2.22 by Robert Collins
Fetch changes to use new pb.
2834
        self.total = 4
1556.1.4 by Robert Collins
Add a new format for what will become knit, and the surrounding logic to upgrade repositories within metadirs, and tests for the same.
2835
        # this is only useful with metadir layouts - separated repo content.
2836
        # trigger an assertion if not such
2837
        repo._format.get_format_string()
2838
        self.repo_dir = repo.bzrdir
2839
        self.step('Moving repository to repository.backup')
2840
        self.repo_dir.transport.move('repository', 'repository.backup')
2841
        backup_transport =  self.repo_dir.transport.clone('repository.backup')
1910.2.12 by Aaron Bentley
Implement knit repo format 2
2842
        repo._format.check_conversion_target(self.target_format)
1556.1.4 by Robert Collins
Add a new format for what will become knit, and the surrounding logic to upgrade repositories within metadirs, and tests for the same.
2843
        self.source_repo = repo._format.open(self.repo_dir,
2844
            _found=True,
2845
            _override_transport=backup_transport)
2846
        self.step('Creating new repository')
2847
        converted = self.target_format.initialize(self.repo_dir,
2848
                                                  self.source_repo.is_shared())
2849
        converted.lock_write()
2850
        try:
2851
            self.step('Copying content into repository.')
2852
            self.source_repo.copy_content_into(converted)
2853
        finally:
2854
            converted.unlock()
2855
        self.step('Deleting old repository content.')
2856
        self.repo_dir.transport.delete_tree('repository.backup')
2857
        self.pb.note('repository converted')
2858
2859
    def step(self, message):
2860
        """Update the pb by a step."""
2861
        self.count +=1
2862
        self.pb.update(message, self.count, self.total)
1596.1.1 by Martin Pool
Use simple xml unescaping rather than importing xml.sax
2863
2864
1843.2.4 by Aaron Bentley
Switch to John Meinel's _unescape_xml implementation
2865
_unescape_map = {
2866
    'apos':"'",
2867
    'quot':'"',
2868
    'amp':'&',
2869
    'lt':'<',
2870
    'gt':'>'
2871
}
2872
2873
2874
def _unescaper(match, _map=_unescape_map):
2294.1.2 by John Arbash Meinel
Track down and add tests that all tree.commit() can handle
2875
    code = match.group(1)
2876
    try:
2877
        return _map[code]
2878
    except KeyError:
2879
        if not code.startswith('#'):
2880
            raise
2294.1.10 by John Arbash Meinel
Switch all apis over to utf8 file ids. All tests pass
2881
        return unichr(int(code[1:])).encode('utf8')
1843.2.4 by Aaron Bentley
Switch to John Meinel's _unescape_xml implementation
2882
2883
2884
_unescape_re = None
2885
2886
1596.1.1 by Martin Pool
Use simple xml unescaping rather than importing xml.sax
2887
def _unescape_xml(data):
1843.2.4 by Aaron Bentley
Switch to John Meinel's _unescape_xml implementation
2888
    """Unescape predefined XML entities in a string of data."""
2889
    global _unescape_re
2890
    if _unescape_re is None:
2120.2.1 by John Arbash Meinel
Remove tabs from source files, and add a test to keep it that way.
2891
        _unescape_re = re.compile('\&([^;]*);')
1843.2.4 by Aaron Bentley
Switch to John Meinel's _unescape_xml implementation
2892
    return _unescape_re.sub(_unescaper, data)
2745.6.3 by Aaron Bentley
Implement versionedfile checking for bzr check
2893
2894
3036.1.3 by Robert Collins
Privatise VersionedFileChecker.
2895
class _VersionedFileChecker(object):
2745.6.47 by Andrew Bennetts
Move check_parents out of VersionedFile.
2896
2988.1.6 by Robert Collins
Change the contract for VersionedFileChecker to consolidate related parameters rather than splitting them across two api calls. This allows better reuse of a single checker object.
2897
    def __init__(self, repository):
2745.6.47 by Andrew Bennetts
Move check_parents out of VersionedFile.
2898
        self.repository = repository
2988.1.8 by Robert Collins
Change check and reconcile to use the new _generate_text_key_index rather
2899
        self.text_index = self.repository._generate_text_key_index()
2745.6.49 by Andrew Bennetts
Get rid of bzrlib.repository._RevisionParentsProvider.
2900
    
2901
    def calculate_file_version_parents(self, revision_id, file_id):
2927.2.10 by Andrew Bennetts
More docstrings, elaborate a comment with an XXX, and remove a little bit of cruft.
2902
        """Calculate the correct parents for a file version according to
2903
        the inventories.
2904
        """
2988.1.8 by Robert Collins
Change check and reconcile to use the new _generate_text_key_index rather
2905
        parent_keys = self.text_index[(file_id, revision_id)]
2906
        if parent_keys == [_mod_revision.NULL_REVISION]:
2907
            return ()
2908
        # strip the file_id, for the weave api
2909
        return tuple([revision_id for file_id, revision_id in parent_keys])
2745.6.47 by Andrew Bennetts
Move check_parents out of VersionedFile.
2910
3036.1.2 by Robert Collins
Simplify the check_file_version_parents API some more. This has already changed in this release cycle.
2911
    def check_file_version_parents(self, weave, file_id):
2927.2.10 by Andrew Bennetts
More docstrings, elaborate a comment with an XXX, and remove a little bit of cruft.
2912
        """Check the parents stored in a versioned file are correct.
2913
2914
        It also detects file versions that are not referenced by their
2915
        corresponding revision's inventory.
2916
2927.2.14 by Andrew Bennetts
Tweaks suggested by review.
2917
        :returns: A tuple of (wrong_parents, dangling_file_versions).
2927.2.10 by Andrew Bennetts
More docstrings, elaborate a comment with an XXX, and remove a little bit of cruft.
2918
            wrong_parents is a dict mapping {revision_id: (stored_parents,
2919
            correct_parents)} for each revision_id where the stored parents
2927.2.14 by Andrew Bennetts
Tweaks suggested by review.
2920
            are not correct.  dangling_file_versions is a set of (file_id,
2921
            revision_id) tuples for versions that are present in this versioned
2922
            file, but not used by the corresponding inventory.
2927.2.10 by Andrew Bennetts
More docstrings, elaborate a comment with an XXX, and remove a little bit of cruft.
2923
        """
2927.2.3 by Andrew Bennetts
Add fulltexts to avoid bug 155730.
2924
        wrong_parents = {}
2988.1.8 by Robert Collins
Change check and reconcile to use the new _generate_text_key_index rather
2925
        unused_versions = set()
3036.1.2 by Robert Collins
Simplify the check_file_version_parents API some more. This has already changed in this release cycle.
2926
        for num, revision_id in enumerate(weave.versions()):
2927.2.6 by Andrew Bennetts
Make some more check tests pass.
2927
            try:
2988.1.8 by Robert Collins
Change check and reconcile to use the new _generate_text_key_index rather
2928
                correct_parents = self.calculate_file_version_parents(
2929
                    revision_id, file_id)
2930
            except KeyError:
3036.1.2 by Robert Collins
Simplify the check_file_version_parents API some more. This has already changed in this release cycle.
2931
                # The version is not part of the used keys.
2988.1.8 by Robert Collins
Change check and reconcile to use the new _generate_text_key_index rather
2932
                unused_versions.add(revision_id)
2933
            else:
2934
                try:
2935
                    knit_parents = tuple(weave.get_parents(revision_id))
2936
                except errors.RevisionNotPresent:
2937
                    knit_parents = None
2938
                if correct_parents != knit_parents:
2939
                    wrong_parents[revision_id] = (knit_parents, correct_parents)
2940
        return wrong_parents, unused_versions