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
56
self._cache_data = False
58
def copy_to(self, name, transport):
59
"""Copy this versioned file to name on transport."""
60
raise NotImplementedError(self.copy_to)
62
@deprecated_method(zero_eight)
64
"""Return a list of all the versions in this versioned file.
66
Please use versionedfile.versions() now.
68
return self.versions()
71
"""Return a unsorted list of versions."""
72
raise NotImplementedError(self.versions)
74
def has_ghost(self, version_id):
75
"""Returns whether version is present as a ghost."""
76
raise NotImplementedError(self.has_ghost)
78
def has_version(self, version_id):
79
"""Returns whether version is present."""
80
raise NotImplementedError(self.has_version)
82
def add_delta(self, version_id, parents, delta_parent, sha1, noeol, delta):
83
"""Add a text to the versioned file via a pregenerated delta.
85
:param version_id: The version id being added.
86
:param parents: The parents of the version_id.
87
:param delta_parent: The parent this delta was created against.
88
:param sha1: The sha1 of the full text.
89
:param delta: The delta instructions. See get_delta for details.
91
self._check_write_ok()
92
if self.has_version(version_id):
93
raise errors.RevisionAlreadyPresent(version_id, self)
94
return self._add_delta(version_id, parents, delta_parent, sha1, noeol, delta)
96
def _add_delta(self, version_id, parents, delta_parent, sha1, noeol, delta):
97
"""Class specific routine to add a delta.
99
This generic version simply applies the delta to the delta_parent and
102
# strip annotation from delta
104
for start, stop, delta_len, delta_lines in delta:
105
new_delta.append((start, stop, delta_len, [text for origin, text in delta_lines]))
106
if delta_parent is not None:
107
parent_full = self.get_lines(delta_parent)
110
new_full = self._apply_delta(parent_full, new_delta)
111
# its impossible to have noeol on an empty file
112
if noeol and new_full[-1][-1] == '\n':
113
new_full[-1] = new_full[-1][:-1]
114
self.add_lines(version_id, parents, new_full)
116
def add_lines(self, version_id, parents, lines, parent_texts=None):
117
"""Add a single text on top of the versioned file.
119
Must raise RevisionAlreadyPresent if the new version is
120
already present in file history.
122
Must raise RevisionNotPresent if any of the given parents are
123
not present in file history.
124
:param parent_texts: An optional dictionary containing the opaque
125
representations of some or all of the parents of
126
version_id to allow delta optimisations.
127
VERY IMPORTANT: the texts must be those returned
128
by add_lines or data corruption can be caused.
129
:return: An opaque representation of the inserted version which can be
130
provided back to future add_lines calls in the parent_texts
133
self._check_write_ok()
134
return self._add_lines(version_id, parents, lines, parent_texts)
136
def _add_lines(self, version_id, parents, lines, parent_texts):
137
"""Helper to do the class specific add_lines."""
138
raise NotImplementedError(self.add_lines)
140
def add_lines_with_ghosts(self, version_id, parents, lines,
142
"""Add lines to the versioned file, allowing ghosts to be present.
144
This takes the same parameters as add_lines.
146
self._check_write_ok()
147
return self._add_lines_with_ghosts(version_id, parents, lines,
150
def _add_lines_with_ghosts(self, version_id, parents, lines, parent_texts):
151
"""Helper to do class specific add_lines_with_ghosts."""
152
raise NotImplementedError(self.add_lines_with_ghosts)
154
def check(self, progress_bar=None):
155
"""Check the versioned file for integrity."""
156
raise NotImplementedError(self.check)
158
def _check_lines_not_unicode(self, lines):
159
"""Check that lines being added to a versioned file are not unicode."""
161
if line.__class__ is not str:
162
raise errors.BzrBadParameterUnicode("lines")
164
def _check_lines_are_lines(self, lines):
165
"""Check that the lines really are full lines without inline EOL."""
167
if '\n' in line[:-1]:
168
raise errors.BzrBadParameterContainsNewline("lines")
170
def _check_write_ok(self):
171
"""Is the versioned file marked as 'finished' ? Raise if it is."""
173
raise errors.OutSideTransaction()
174
if self._access_mode != 'w':
175
raise errors.ReadOnlyObjectDirtiedError(self)
177
def enable_cache(self):
178
"""Tell this versioned file that it should cache any data it reads.
180
This is advisory, implementations do not have to support caching.
184
def clear_cache(self):
185
"""Remove any data cached in the versioned file object.
187
This only needs to be supported if caches are supported
191
def clone_text(self, new_version_id, old_version_id, parents):
192
"""Add an identical text to old_version_id as new_version_id.
194
Must raise RevisionNotPresent if the old version or any of the
195
parents are not present in file history.
197
Must raise RevisionAlreadyPresent if the new version is
198
already present in file history."""
199
self._check_write_ok()
200
return self._clone_text(new_version_id, old_version_id, parents)
202
def _clone_text(self, new_version_id, old_version_id, parents):
203
"""Helper function to do the _clone_text work."""
204
raise NotImplementedError(self.clone_text)
206
def create_empty(self, name, transport, mode=None):
207
"""Create a new versioned file of this exact type.
209
:param name: the file name
210
:param transport: the transport
211
:param mode: optional file mode.
213
raise NotImplementedError(self.create_empty)
215
def fix_parents(self, version, new_parents):
216
"""Fix the parents list for version.
218
This is done by appending a new version to the index
219
with identical data except for the parents list.
220
the parents list must be a superset of the current
223
self._check_write_ok()
224
return self._fix_parents(version, new_parents)
226
def _fix_parents(self, version, new_parents):
227
"""Helper for fix_parents."""
228
raise NotImplementedError(self.fix_parents)
230
def get_delta(self, version):
231
"""Get a delta for constructing version from some other version.
233
:return: (delta_parent, sha1, noeol, delta)
234
Where delta_parent is a version id or None to indicate no parent.
236
raise NotImplementedError(self.get_delta)
238
def get_deltas(self, versions):
239
"""Get multiple deltas at once for constructing versions.
241
:return: dict(version_id:(delta_parent, sha1, noeol, delta))
242
Where delta_parent is a version id or None to indicate no parent, and
243
version_id is the version_id created by that delta.
246
for version in versions:
247
result[version] = self.get_delta(version)
250
def get_sha1(self, version_id):
251
"""Get the stored sha1 sum for the given revision.
253
:param name: The name of the version to lookup
255
raise NotImplementedError(self.get_sha1)
257
def get_suffixes(self):
258
"""Return the file suffixes associated with this versioned file."""
259
raise NotImplementedError(self.get_suffixes)
261
def get_text(self, version_id):
262
"""Return version contents as a text string.
264
Raises RevisionNotPresent if version is not present in
267
return ''.join(self.get_lines(version_id))
268
get_string = get_text
270
def get_texts(self, version_ids):
271
"""Return the texts of listed versions as a list of strings.
273
Raises RevisionNotPresent if version is not present in
276
return [''.join(self.get_lines(v)) for v in version_ids]
278
def get_lines(self, version_id):
279
"""Return version contents as a sequence of lines.
281
Raises RevisionNotPresent if version is not present in
284
raise NotImplementedError(self.get_lines)
286
def get_ancestry(self, version_ids):
287
"""Return a list of all ancestors of given version(s). This
288
will not include the null revision.
290
Must raise RevisionNotPresent if any of the given versions are
291
not present in file history."""
292
if isinstance(version_ids, basestring):
293
version_ids = [version_ids]
294
raise NotImplementedError(self.get_ancestry)
296
def get_ancestry_with_ghosts(self, version_ids):
297
"""Return a list of all ancestors of given version(s). This
298
will not include the null revision.
300
Must raise RevisionNotPresent if any of the given versions are
301
not present in file history.
303
Ghosts that are known about will be included in ancestry list,
304
but are not explicitly marked.
306
raise NotImplementedError(self.get_ancestry_with_ghosts)
308
def get_graph(self, version_ids=None):
309
"""Return a graph from the versioned file.
311
Ghosts are not listed or referenced in the graph.
312
:param version_ids: Versions to select.
313
None means retrieve all versions.
316
if version_ids is None:
317
for version in self.versions():
318
result[version] = self.get_parents(version)
320
pending = set(version_ids)
322
version = pending.pop()
323
if version in result:
325
parents = self.get_parents(version)
326
for parent in parents:
330
result[version] = parents
333
def get_graph_with_ghosts(self):
334
"""Return a graph for the entire versioned file.
336
Ghosts are referenced in parents list but are not
339
raise NotImplementedError(self.get_graph_with_ghosts)
341
@deprecated_method(zero_eight)
342
def parent_names(self, version):
343
"""Return version names for parents of a version.
345
See get_parents for the current api.
347
return self.get_parents(version)
349
def get_parents(self, version_id):
350
"""Return version names for parents of a version.
352
Must raise RevisionNotPresent if version is not present in
355
raise NotImplementedError(self.get_parents)
357
def get_parents_with_ghosts(self, version_id):
358
"""Return version names for parents of version_id.
360
Will raise RevisionNotPresent if version_id is not present
363
Ghosts that are known about will be included in the parent list,
364
but are not explicitly marked.
366
raise NotImplementedError(self.get_parents_with_ghosts)
368
def annotate_iter(self, version_id):
369
"""Yield list of (version-id, line) pairs for the specified
372
Must raise RevisionNotPresent if any of the given versions are
373
not present in file history.
375
raise NotImplementedError(self.annotate_iter)
377
def annotate(self, version_id):
378
return list(self.annotate_iter(version_id))
380
def _apply_delta(self, lines, delta):
381
"""Apply delta to lines."""
384
for start, end, count, delta_lines in delta:
385
lines[offset+start:offset+end] = delta_lines
386
offset = offset + (start - end) + count
389
def join(self, other, pb=None, msg=None, version_ids=None,
390
ignore_missing=False):
391
"""Integrate versions from other into this versioned file.
393
If version_ids is None all versions from other should be
394
incorporated into this versioned file.
396
Must raise RevisionNotPresent if any of the specified versions
397
are not present in the other files history unless ignore_missing
398
is supplied when they are silently skipped.
400
self._check_write_ok()
401
return InterVersionedFile.get(other, self).join(
407
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
NOTES: Lines are normalised: they will all have \n terminators.
417
Lines are returned in arbitrary order.
419
raise NotImplementedError(self.iter_lines_added_or_present_in_versions)
421
def transaction_finished(self):
422
"""The transaction that this file was opened in has finished.
424
This records self.finished = True and should cause all mutating
429
@deprecated_method(zero_eight)
430
def walk(self, version_ids=None):
431
"""Walk the versioned file as a weave-like structure, for
432
versions relative to version_ids. Yields sequence of (lineno,
433
insert, deletes, text) for each relevant line.
435
Must raise RevisionNotPresent if any of the specified versions
436
are not present in the file history.
438
:param version_ids: the version_ids to walk with respect to. If not
439
supplied the entire weave-like structure is walked.
441
walk is deprecated in favour of iter_lines_added_or_present_in_versions
443
raise NotImplementedError(self.walk)
445
@deprecated_method(zero_eight)
446
def iter_names(self):
447
"""Walk the names list."""
448
return iter(self.versions())
450
def plan_merge(self, ver_a, ver_b):
451
"""Return pseudo-annotation indicating how the two versions merge.
453
This is computed between versions a and b and their common
456
Weave lines present in none of them are skipped entirely.
459
killed-base Dead in base revision
460
killed-both Killed in each revision
463
unchanged Alive in both a and b (possibly created in both)
466
ghost-a Killed in a, unborn in b
467
ghost-b Killed in b, unborn in a
468
irrelevant Not in either revision
470
raise NotImplementedError(VersionedFile.plan_merge)
472
def weave_merge(self, plan, a_marker=TextMerge.A_MARKER,
473
b_marker=TextMerge.B_MARKER):
474
return PlanWeaveMerge(plan, a_marker, b_marker).merge_lines()[0]
477
class PlanWeaveMerge(TextMerge):
478
"""Weave merge that takes a plan as its input.
480
This exists so that VersionedFile.plan_merge is implementable.
481
Most callers will want to use WeaveMerge instead.
484
def __init__(self, plan, a_marker=TextMerge.A_MARKER,
485
b_marker=TextMerge.B_MARKER):
486
TextMerge.__init__(self, a_marker, b_marker)
489
def _merge_struct(self):
494
def outstanding_struct():
495
if not lines_a and not lines_b:
497
elif ch_a and not ch_b:
500
elif ch_b and not ch_a:
502
elif lines_a == lines_b:
505
yield (lines_a, lines_b)
507
# We previously considered either 'unchanged' or 'killed-both' lines
508
# to be possible places to resynchronize. However, assuming agreement
509
# on killed-both lines may be too aggressive. -- mbp 20060324
510
for state, line in self.plan:
511
if state == 'unchanged':
512
# resync and flush queued conflicts changes if any
513
for struct in outstanding_struct():
519
if state == 'unchanged':
522
elif state == 'killed-a':
525
elif state == 'killed-b':
528
elif state == 'new-a':
531
elif state == 'new-b':
535
assert state in ('irrelevant', 'ghost-a', 'ghost-b',
536
'killed-base', 'killed-both'), state
537
for struct in outstanding_struct():
541
class WeaveMerge(PlanWeaveMerge):
542
"""Weave merge that takes a VersionedFile and two versions as its input"""
544
def __init__(self, versionedfile, ver_a, ver_b,
545
a_marker=PlanWeaveMerge.A_MARKER, b_marker=PlanWeaveMerge.B_MARKER):
546
plan = versionedfile.plan_merge(ver_a, ver_b)
547
PlanWeaveMerge.__init__(self, plan, a_marker, b_marker)
550
class InterVersionedFile(InterObject):
551
"""This class represents operations taking place between two versionedfiles..
553
Its instances have methods like join, and contain
554
references to the source and target versionedfiles these operations can be
557
Often we will provide convenience methods on 'versionedfile' which carry out
558
operations with another versionedfile - they will always forward to
559
InterVersionedFile.get(other).method_name(parameters).
563
"""The available optimised InterVersionedFile types."""
565
def join(self, pb=None, msg=None, version_ids=None, ignore_missing=False):
566
"""Integrate versions from self.source into self.target.
568
If version_ids is None all versions from source should be
569
incorporated into this versioned file.
571
Must raise RevisionNotPresent if any of the specified versions
572
are not present in the other files history unless ignore_missing is
573
supplied when they are silently skipped.
576
# - if the target is empty, just add all the versions from
577
# source to target, otherwise:
578
# - make a temporary versioned file of type target
579
# - insert the source content into it one at a time
581
if not self.target.versions():
584
# Make a new target-format versioned file.
585
temp_source = self.target.create_empty("temp", MemoryTransport())
587
version_ids = self._get_source_version_ids(version_ids, ignore_missing)
588
graph = self.source.get_graph(version_ids)
589
order = topo_sort(graph.items())
590
pb = ui.ui_factory.nested_progress_bar()
593
# TODO for incremental cross-format work:
594
# make a versioned file with the following content:
595
# all revisions we have been asked to join
596
# all their ancestors that are *not* in target already.
597
# the immediate parents of the above two sets, with
598
# empty parent lists - these versions are in target already
599
# and the incorrect version data will be ignored.
600
# TODO: for all ancestors that are present in target already,
601
# check them for consistent data, this requires moving sha1 from
603
# TODO: remove parent texts when they are not relevant any more for
604
# memory pressure reduction. RBC 20060313
605
# pb.update('Converting versioned data', 0, len(order))
606
# deltas = self.source.get_deltas(order)
607
for index, version in enumerate(order):
608
pb.update('Converting versioned data', index, len(order))
609
parent_text = target.add_lines(version,
610
self.source.get_parents(version),
611
self.source.get_lines(version),
612
parent_texts=parent_texts)
613
parent_texts[version] = parent_text
614
#delta_parent, sha1, noeol, delta = deltas[version]
615
#target.add_delta(version,
616
# self.source.get_parents(version),
621
#target.get_lines(version)
623
# this should hit the native code path for target
624
if target is not self.target:
625
return self.target.join(temp_source,
633
def _get_source_version_ids(self, version_ids, ignore_missing):
634
"""Determine the version ids to be used from self.source.
636
:param version_ids: The caller-supplied version ids to check. (None
637
for all). If None is in version_ids, it is stripped.
638
:param ignore_missing: if True, remove missing ids from the version
639
list. If False, raise RevisionNotPresent on
640
a missing version id.
641
:return: A set of version ids.
643
if version_ids is None:
644
# None cannot be in source.versions
645
return set(self.source.versions())
648
return set(self.source.versions()).intersection(set(version_ids))
650
new_version_ids = set()
651
for version in version_ids:
654
if not self.source.has_version(version):
655
raise errors.RevisionNotPresent(version, str(self.source))
657
new_version_ids.add(version)
658
return new_version_ids
661
class InterVersionedFileTestProviderAdapter(object):
662
"""A tool to generate a suite testing multiple inter versioned-file classes.
664
This is done by copying the test once for each InterVersionedFile provider
665
and injecting the transport_server, transport_readonly_server,
666
versionedfile_factory and versionedfile_factory_to classes into each copy.
667
Each copy is also given a new id() to make it easy to identify.
670
def __init__(self, transport_server, transport_readonly_server, formats):
671
self._transport_server = transport_server
672
self._transport_readonly_server = transport_readonly_server
673
self._formats = formats
675
def adapt(self, test):
677
for (interversionedfile_class,
678
versionedfile_factory,
679
versionedfile_factory_to) in self._formats:
680
new_test = deepcopy(test)
681
new_test.transport_server = self._transport_server
682
new_test.transport_readonly_server = self._transport_readonly_server
683
new_test.interversionedfile_class = interversionedfile_class
684
new_test.versionedfile_factory = versionedfile_factory
685
new_test.versionedfile_factory_to = versionedfile_factory_to
686
def make_new_test_id():
687
new_id = "%s(%s)" % (new_test.id(), interversionedfile_class.__name__)
688
return lambda: new_id
689
new_test.id = make_new_test_id()
690
result.addTest(new_test)
694
def default_test_list():
695
"""Generate the default list of interversionedfile permutations to test."""
696
from bzrlib.weave import WeaveFile
697
from bzrlib.knit import KnitVersionedFile
699
# test the fallback InterVersionedFile from annotated knits to weave
700
result.append((InterVersionedFile,
703
for optimiser in InterVersionedFile._optimisers:
704
result.append((optimiser,
705
optimiser._matching_file_from_factory,
706
optimiser._matching_file_to_factory
708
# if there are specific combinations we want to use, we can add them
added
removed
bzrlib/versionedfile.py
