1
# Copyright (C) 2005, 2006 by Canonical Ltd
4
# Johan Rydberg <jrydberg@gnu.org>
6
# This program is free software; you can redistribute it and/or modify
7
# it under the terms of the GNU General Public License as published by
8
# the Free Software Foundation; either version 2 of the License, or
9
# (at your option) any later version.
11
# This program is distributed in the hope that it will be useful,
12
# but WITHOUT ANY WARRANTY; without even the implied warranty of
13
# MERCHANTABILITY or FITNESS FOR A PARTICULAR PURPOSE. See the
14
# GNU General Public License for more details.
16
# You should have received a copy of the GNU General Public License
17
# along with this program; if not, write to the Free Software
18
# Foundation, Inc., 59 Temple Place, Suite 330, Boston, MA 02111-1307 USA
20
"""Versioned text file storage api."""
23
from copy import deepcopy
24
from unittest import TestSuite
27
import bzrlib.errors as errors
28
from bzrlib.inter import InterObject
29
from bzrlib.textmerge import TextMerge
30
from bzrlib.transport.memory import MemoryTransport
31
from bzrlib.tsort import topo_sort
33
from bzrlib.symbol_versioning import (deprecated_function,
39
class VersionedFile(object):
40
"""Versioned text file storage.
42
A versioned file manages versions of line-based text files,
43
keeping track of the originating version for each line.
45
To clients the "lines" of the file are represented as a list of
46
strings. These strings will typically have terminal newline
47
characters, but this is not required. In particular files commonly
48
do not have a newline at the end of the file.
50
Texts are identified by a version-id string.
53
def __init__(self, access_mode):
55
self._access_mode = access_mode
57
def copy_to(self, name, transport):
58
"""Copy this versioned file to name on transport."""
59
raise NotImplementedError(self.copy_to)
61
@deprecated_method(zero_eight)
63
"""Return a list of all the versions in this versioned file.
65
Please use versionedfile.versions() now.
67
return self.versions()
70
"""Return a unsorted list of versions."""
71
raise NotImplementedError(self.versions)
73
def has_ghost(self, version_id):
74
"""Returns whether version is present as a ghost."""
75
raise NotImplementedError(self.has_ghost)
77
def has_version(self, version_id):
78
"""Returns whether version is present."""
79
raise NotImplementedError(self.has_version)
81
def add_delta(self, version_id, parents, delta_parent, sha1, noeol, delta):
82
"""Add a text to the versioned file via a pregenerated delta.
84
:param version_id: The version id being added.
85
:param parents: The parents of the version_id.
86
:param delta_parent: The parent this delta was created against.
87
:param sha1: The sha1 of the full text.
88
:param delta: The delta instructions. See get_delta for details.
90
self._check_write_ok()
91
if self.has_version(version_id):
92
raise errors.RevisionAlreadyPresent(version_id, self)
93
return self._add_delta(version_id, parents, delta_parent, sha1, noeol, delta)
95
def _add_delta(self, version_id, parents, delta_parent, sha1, noeol, delta):
96
"""Class specific routine to add a delta.
98
This generic version simply applies the delta to the delta_parent and
101
# strip annotation from delta
103
for start, stop, delta_len, delta_lines in delta:
104
new_delta.append((start, stop, delta_len, [text for origin, text in delta_lines]))
105
if delta_parent is not None:
106
parent_full = self.get_lines(delta_parent)
109
new_full = self._apply_delta(parent_full, new_delta)
110
# its impossible to have noeol on an empty file
111
if noeol and new_full[-1][-1] == '\n':
112
new_full[-1] = new_full[-1][:-1]
113
self.add_lines(version_id, parents, new_full)
115
def add_lines(self, version_id, parents, lines, parent_texts=None):
116
"""Add a single text on top of the versioned file.
118
Must raise RevisionAlreadyPresent if the new version is
119
already present in file history.
121
Must raise RevisionNotPresent if any of the given parents are
122
not present in file history.
123
:param parent_texts: An optional dictionary containing the opaque
124
representations of some or all of the parents of
125
version_id to allow delta optimisations.
126
VERY IMPORTANT: the texts must be those returned
127
by add_lines or data corruption can be caused.
128
:return: An opaque representation of the inserted version which can be
129
provided back to future add_lines calls in the parent_texts
132
self._check_write_ok()
133
return self._add_lines(version_id, parents, lines, parent_texts)
135
def _add_lines(self, version_id, parents, lines, parent_texts):
136
"""Helper to do the class specific add_lines."""
137
raise NotImplementedError(self.add_lines)
139
def add_lines_with_ghosts(self, version_id, parents, lines,
141
"""Add lines to the versioned file, allowing ghosts to be present.
143
This takes the same parameters as add_lines.
145
self._check_write_ok()
146
return self._add_lines_with_ghosts(version_id, parents, lines,
149
def _add_lines_with_ghosts(self, version_id, parents, lines, parent_texts):
150
"""Helper to do class specific add_lines_with_ghosts."""
151
raise NotImplementedError(self.add_lines_with_ghosts)
153
def check(self, progress_bar=None):
154
"""Check the versioned file for integrity."""
155
raise NotImplementedError(self.check)
157
def _check_lines_not_unicode(self, lines):
158
"""Check that lines being added to a versioned file are not unicode."""
160
if line.__class__ is not str:
161
raise errors.BzrBadParameterUnicode("lines")
163
def _check_lines_are_lines(self, lines):
164
"""Check that the lines really are full lines without inline EOL."""
166
if '\n' in line[:-1]:
167
raise errors.BzrBadParameterContainsNewline("lines")
169
def _check_write_ok(self):
170
"""Is the versioned file marked as 'finished' ? Raise if it is."""
172
raise errors.OutSideTransaction()
173
if self._access_mode != 'w':
174
raise errors.ReadOnlyObjectDirtiedError(self)
176
def enable_cache(self):
177
"""Tell this versioned file that it should cache any data it reads.
179
This is advisory, implementations do not have to support caching.
183
def clear_cache(self):
184
"""Remove any data cached in the versioned file object.
186
This only needs to be supported if caches are supported
190
def clone_text(self, new_version_id, old_version_id, parents):
191
"""Add an identical text to old_version_id as new_version_id.
193
Must raise RevisionNotPresent if the old version or any of the
194
parents are not present in file history.
196
Must raise RevisionAlreadyPresent if the new version is
197
already present in file history."""
198
self._check_write_ok()
199
return self._clone_text(new_version_id, old_version_id, parents)
201
def _clone_text(self, new_version_id, old_version_id, parents):
202
"""Helper function to do the _clone_text work."""
203
raise NotImplementedError(self.clone_text)
205
def create_empty(self, name, transport, mode=None):
206
"""Create a new versioned file of this exact type.
208
:param name: the file name
209
:param transport: the transport
210
:param mode: optional file mode.
212
raise NotImplementedError(self.create_empty)
214
def fix_parents(self, version, new_parents):
215
"""Fix the parents list for version.
217
This is done by appending a new version to the index
218
with identical data except for the parents list.
219
the parents list must be a superset of the current
222
self._check_write_ok()
223
return self._fix_parents(version, new_parents)
225
def _fix_parents(self, version, new_parents):
226
"""Helper for fix_parents."""
227
raise NotImplementedError(self.fix_parents)
229
def get_delta(self, version):
230
"""Get a delta for constructing version from some other version.
232
:return: (delta_parent, sha1, noeol, delta)
233
Where delta_parent is a version id or None to indicate no parent.
235
raise NotImplementedError(self.get_delta)
237
def get_deltas(self, versions):
238
"""Get multiple deltas at once for constructing versions.
240
:return: dict(version_id:(delta_parent, sha1, noeol, delta))
241
Where delta_parent is a version id or None to indicate no parent, and
242
version_id is the version_id created by that delta.
245
for version in versions:
246
result[version] = self.get_delta(version)
249
def get_sha1(self, version_id):
250
"""Get the stored sha1 sum for the given revision.
252
:param name: The name of the version to lookup
254
raise NotImplementedError(self.get_sha1)
256
def get_suffixes(self):
257
"""Return the file suffixes associated with this versioned file."""
258
raise NotImplementedError(self.get_suffixes)
260
def get_text(self, version_id):
261
"""Return version contents as a text string.
263
Raises RevisionNotPresent if version is not present in
266
return ''.join(self.get_lines(version_id))
267
get_string = get_text
269
def get_texts(self, version_ids):
270
"""Return the texts of listed versions as a list of strings.
272
Raises RevisionNotPresent if version is not present in
275
return [''.join(self.get_lines(v)) for v in version_ids]
277
def get_lines(self, version_id):
278
"""Return version contents as a sequence of lines.
280
Raises RevisionNotPresent if version is not present in
283
raise NotImplementedError(self.get_lines)
285
def get_ancestry(self, version_ids):
286
"""Return a list of all ancestors of given version(s). This
287
will not include the null revision.
289
Must raise RevisionNotPresent if any of the given versions are
290
not present in file history."""
291
if isinstance(version_ids, basestring):
292
version_ids = [version_ids]
293
raise NotImplementedError(self.get_ancestry)
295
def get_ancestry_with_ghosts(self, version_ids):
296
"""Return a list of all ancestors of given version(s). This
297
will not include the null revision.
299
Must raise RevisionNotPresent if any of the given versions are
300
not present in file history.
302
Ghosts that are known about will be included in ancestry list,
303
but are not explicitly marked.
305
raise NotImplementedError(self.get_ancestry_with_ghosts)
307
def get_graph(self, version_ids=None):
308
"""Return a graph from the versioned file.
310
Ghosts are not listed or referenced in the graph.
311
:param version_ids: Versions to select.
312
None means retrieve all versions.
315
if version_ids is None:
316
for version in self.versions():
317
result[version] = self.get_parents(version)
319
pending = set(version_ids)
321
version = pending.pop()
322
if version in result:
324
parents = self.get_parents(version)
325
for parent in parents:
329
result[version] = parents
332
def get_graph_with_ghosts(self):
333
"""Return a graph for the entire versioned file.
335
Ghosts are referenced in parents list but are not
338
raise NotImplementedError(self.get_graph_with_ghosts)
340
@deprecated_method(zero_eight)
341
def parent_names(self, version):
342
"""Return version names for parents of a version.
344
See get_parents for the current api.
346
return self.get_parents(version)
348
def get_parents(self, version_id):
349
"""Return version names for parents of a version.
351
Must raise RevisionNotPresent if version is not present in
354
raise NotImplementedError(self.get_parents)
356
def get_parents_with_ghosts(self, version_id):
357
"""Return version names for parents of version_id.
359
Will raise RevisionNotPresent if version_id is not present
362
Ghosts that are known about will be included in the parent list,
363
but are not explicitly marked.
365
raise NotImplementedError(self.get_parents_with_ghosts)
367
def annotate_iter(self, version_id):
368
"""Yield list of (version-id, line) pairs for the specified
371
Must raise RevisionNotPresent if any of the given versions are
372
not present in file history.
374
raise NotImplementedError(self.annotate_iter)
376
def annotate(self, version_id):
377
return list(self.annotate_iter(version_id))
379
def _apply_delta(self, lines, delta):
380
"""Apply delta to lines."""
383
for start, end, count, delta_lines in delta:
384
lines[offset+start:offset+end] = delta_lines
385
offset = offset + (start - end) + count
388
def join(self, other, pb=None, msg=None, version_ids=None,
389
ignore_missing=False):
390
"""Integrate versions from other into this versioned file.
392
If version_ids is None all versions from other should be
393
incorporated into this versioned file.
395
Must raise RevisionNotPresent if any of the specified versions
396
are not present in the other files history unless ignore_missing
397
is supplied when they are silently skipped.
399
self._check_write_ok()
400
return InterVersionedFile.get(other, self).join(
406
def iter_lines_added_or_present_in_versions(self, version_ids=None,
408
"""Iterate over the lines in the versioned file from version_ids.
410
This may return lines from other versions, and does not return the
411
specific version marker at this point. The api may be changed
412
during development to include the version that the versioned file
413
thinks is relevant, but given that such hints are just guesses,
414
its better not to have it if we don't need it.
416
If a progress bar is supplied, it may be used to indicate progress.
417
The caller is responsible for cleaning up progress bars (because this
420
NOTES: Lines are normalised: they will all have \n terminators.
421
Lines are returned in arbitrary order.
423
raise NotImplementedError(self.iter_lines_added_or_present_in_versions)
425
def transaction_finished(self):
426
"""The transaction that this file was opened in has finished.
428
This records self.finished = True and should cause all mutating
433
@deprecated_method(zero_eight)
434
def walk(self, version_ids=None):
435
"""Walk the versioned file as a weave-like structure, for
436
versions relative to version_ids. Yields sequence of (lineno,
437
insert, deletes, text) for each relevant line.
439
Must raise RevisionNotPresent if any of the specified versions
440
are not present in the file history.
442
:param version_ids: the version_ids to walk with respect to. If not
443
supplied the entire weave-like structure is walked.
445
walk is deprecated in favour of iter_lines_added_or_present_in_versions
447
raise NotImplementedError(self.walk)
449
@deprecated_method(zero_eight)
450
def iter_names(self):
451
"""Walk the names list."""
452
return iter(self.versions())
454
def plan_merge(self, ver_a, ver_b):
455
"""Return pseudo-annotation indicating how the two versions merge.
457
This is computed between versions a and b and their common
460
Weave lines present in none of them are skipped entirely.
463
killed-base Dead in base revision
464
killed-both Killed in each revision
467
unchanged Alive in both a and b (possibly created in both)
470
ghost-a Killed in a, unborn in b
471
ghost-b Killed in b, unborn in a
472
irrelevant Not in either revision
474
raise NotImplementedError(VersionedFile.plan_merge)
476
def weave_merge(self, plan, a_marker=TextMerge.A_MARKER,
477
b_marker=TextMerge.B_MARKER):
478
return PlanWeaveMerge(plan, a_marker, b_marker).merge_lines()[0]
481
class PlanWeaveMerge(TextMerge):
482
"""Weave merge that takes a plan as its input.
484
This exists so that VersionedFile.plan_merge is implementable.
485
Most callers will want to use WeaveMerge instead.
488
def __init__(self, plan, a_marker=TextMerge.A_MARKER,
489
b_marker=TextMerge.B_MARKER):
490
TextMerge.__init__(self, a_marker, b_marker)
493
def _merge_struct(self):
498
def outstanding_struct():
499
if not lines_a and not lines_b:
501
elif ch_a and not ch_b:
504
elif ch_b and not ch_a:
506
elif lines_a == lines_b:
509
yield (lines_a, lines_b)
511
# We previously considered either 'unchanged' or 'killed-both' lines
512
# to be possible places to resynchronize. However, assuming agreement
513
# on killed-both lines may be too aggressive. -- mbp 20060324
514
for state, line in self.plan:
515
if state == 'unchanged':
516
# resync and flush queued conflicts changes if any
517
for struct in outstanding_struct():
523
if state == 'unchanged':
526
elif state == 'killed-a':
529
elif state == 'killed-b':
532
elif state == 'new-a':
535
elif state == 'new-b':
539
assert state in ('irrelevant', 'ghost-a', 'ghost-b',
540
'killed-base', 'killed-both'), state
541
for struct in outstanding_struct():
545
class WeaveMerge(PlanWeaveMerge):
546
"""Weave merge that takes a VersionedFile and two versions as its input"""
548
def __init__(self, versionedfile, ver_a, ver_b,
549
a_marker=PlanWeaveMerge.A_MARKER, b_marker=PlanWeaveMerge.B_MARKER):
550
plan = versionedfile.plan_merge(ver_a, ver_b)
551
PlanWeaveMerge.__init__(self, plan, a_marker, b_marker)
554
class InterVersionedFile(InterObject):
555
"""This class represents operations taking place between two versionedfiles..
557
Its instances have methods like join, and contain
558
references to the source and target versionedfiles these operations can be
561
Often we will provide convenience methods on 'versionedfile' which carry out
562
operations with another versionedfile - they will always forward to
563
InterVersionedFile.get(other).method_name(parameters).
567
"""The available optimised InterVersionedFile types."""
569
def join(self, pb=None, msg=None, version_ids=None, ignore_missing=False):
570
"""Integrate versions from self.source into self.target.
572
If version_ids is None all versions from source should be
573
incorporated into this versioned file.
575
Must raise RevisionNotPresent if any of the specified versions
576
are not present in the other files history unless ignore_missing is
577
supplied when they are silently skipped.
580
# - if the target is empty, just add all the versions from
581
# source to target, otherwise:
582
# - make a temporary versioned file of type target
583
# - insert the source content into it one at a time
585
if not self.target.versions():
588
# Make a new target-format versioned file.
589
temp_source = self.target.create_empty("temp", MemoryTransport())
591
version_ids = self._get_source_version_ids(version_ids, ignore_missing)
592
graph = self.source.get_graph(version_ids)
593
order = topo_sort(graph.items())
594
pb = ui.ui_factory.nested_progress_bar()
597
# TODO for incremental cross-format work:
598
# make a versioned file with the following content:
599
# all revisions we have been asked to join
600
# all their ancestors that are *not* in target already.
601
# the immediate parents of the above two sets, with
602
# empty parent lists - these versions are in target already
603
# and the incorrect version data will be ignored.
604
# TODO: for all ancestors that are present in target already,
605
# check them for consistent data, this requires moving sha1 from
607
# TODO: remove parent texts when they are not relevant any more for
608
# memory pressure reduction. RBC 20060313
609
# pb.update('Converting versioned data', 0, len(order))
610
# deltas = self.source.get_deltas(order)
611
for index, version in enumerate(order):
612
pb.update('Converting versioned data', index, len(order))
613
parent_text = target.add_lines(version,
614
self.source.get_parents(version),
615
self.source.get_lines(version),
616
parent_texts=parent_texts)
617
parent_texts[version] = parent_text
618
#delta_parent, sha1, noeol, delta = deltas[version]
619
#target.add_delta(version,
620
# self.source.get_parents(version),
625
#target.get_lines(version)
627
# this should hit the native code path for target
628
if target is not self.target:
629
return self.target.join(temp_source,
637
def _get_source_version_ids(self, version_ids, ignore_missing):
638
"""Determine the version ids to be used from self.source.
640
:param version_ids: The caller-supplied version ids to check. (None
641
for all). If None is in version_ids, it is stripped.
642
:param ignore_missing: if True, remove missing ids from the version
643
list. If False, raise RevisionNotPresent on
644
a missing version id.
645
:return: A set of version ids.
647
if version_ids is None:
648
# None cannot be in source.versions
649
return set(self.source.versions())
652
return set(self.source.versions()).intersection(set(version_ids))
654
new_version_ids = set()
655
for version in version_ids:
658
if not self.source.has_version(version):
659
raise errors.RevisionNotPresent(version, str(self.source))
661
new_version_ids.add(version)
662
return new_version_ids
665
class InterVersionedFileTestProviderAdapter(object):
666
"""A tool to generate a suite testing multiple inter versioned-file classes.
668
This is done by copying the test once for each InterVersionedFile provider
669
and injecting the transport_server, transport_readonly_server,
670
versionedfile_factory and versionedfile_factory_to classes into each copy.
671
Each copy is also given a new id() to make it easy to identify.
674
def __init__(self, transport_server, transport_readonly_server, formats):
675
self._transport_server = transport_server
676
self._transport_readonly_server = transport_readonly_server
677
self._formats = formats
679
def adapt(self, test):
681
for (interversionedfile_class,
682
versionedfile_factory,
683
versionedfile_factory_to) in self._formats:
684
new_test = deepcopy(test)
685
new_test.transport_server = self._transport_server
686
new_test.transport_readonly_server = self._transport_readonly_server
687
new_test.interversionedfile_class = interversionedfile_class
688
new_test.versionedfile_factory = versionedfile_factory
689
new_test.versionedfile_factory_to = versionedfile_factory_to
690
def make_new_test_id():
691
new_id = "%s(%s)" % (new_test.id(), interversionedfile_class.__name__)
692
return lambda: new_id
693
new_test.id = make_new_test_id()
694
result.addTest(new_test)
698
def default_test_list():
699
"""Generate the default list of interversionedfile permutations to test."""
700
from bzrlib.weave import WeaveFile
701
from bzrlib.knit import KnitVersionedFile
703
# test the fallback InterVersionedFile from annotated knits to weave
704
result.append((InterVersionedFile,
707
for optimiser in InterVersionedFile._optimisers:
708
result.append((optimiser,
709
optimiser._matching_file_from_factory,
710
optimiser._matching_file_to_factory
712
# if there are specific combinations we want to use, we can add them
added
removed
bzrlib/versionedfile.py
