1
# Copyright (C) 2005, 2006 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."""
22
from bzrlib.lazy_import import lazy_import
23
lazy_import(globals(), """
24
from copy import deepcopy
33
from bzrlib.transport.memory import MemoryTransport
36
from bzrlib.inter import InterObject
37
from bzrlib.textmerge import TextMerge
38
from bzrlib.symbol_versioning import (deprecated_function,
44
class VersionedFile(object):
45
"""Versioned text file storage.
47
A versioned file manages versions of line-based text files,
48
keeping track of the originating version for each line.
50
To clients the "lines" of the file are represented as a list of
51
strings. These strings will typically have terminal newline
52
characters, but this is not required. In particular files commonly
53
do not have a newline at the end of the file.
55
Texts are identified by a version-id string.
58
def __init__(self, access_mode):
60
self._access_mode = access_mode
63
def check_not_reserved_id(version_id):
64
revision.check_not_reserved_id(version_id)
66
def copy_to(self, name, transport):
67
"""Copy this versioned file to name on transport."""
68
raise NotImplementedError(self.copy_to)
70
@deprecated_method(zero_eight)
72
"""Return a list of all the versions in this versioned file.
74
Please use versionedfile.versions() now.
76
return self.versions()
79
"""Return a unsorted list of versions."""
80
raise NotImplementedError(self.versions)
82
def has_ghost(self, version_id):
83
"""Returns whether version is present as a ghost."""
84
raise NotImplementedError(self.has_ghost)
86
def has_version(self, version_id):
87
"""Returns whether version is present."""
88
raise NotImplementedError(self.has_version)
90
def add_delta(self, version_id, parents, delta_parent, sha1, noeol, delta):
91
"""Add a text to the versioned file via a pregenerated delta.
93
:param version_id: The version id being added.
94
:param parents: The parents of the version_id.
95
:param delta_parent: The parent this delta was created against.
96
:param sha1: The sha1 of the full text.
97
:param delta: The delta instructions. See get_delta for details.
99
self._check_write_ok()
100
if self.has_version(version_id):
101
raise errors.RevisionAlreadyPresent(version_id, self)
102
return self._add_delta(version_id, parents, delta_parent, sha1, noeol, delta)
104
def _add_delta(self, version_id, parents, delta_parent, sha1, noeol, delta):
105
"""Class specific routine to add a delta.
107
This generic version simply applies the delta to the delta_parent and
110
# strip annotation from delta
112
for start, stop, delta_len, delta_lines in delta:
113
new_delta.append((start, stop, delta_len, [text for origin, text in delta_lines]))
114
if delta_parent is not None:
115
parent_full = self.get_lines(delta_parent)
118
new_full = self._apply_delta(parent_full, new_delta)
119
# its impossible to have noeol on an empty file
120
if noeol and new_full[-1][-1] == '\n':
121
new_full[-1] = new_full[-1][:-1]
122
self.add_lines(version_id, parents, new_full)
124
def add_lines(self, version_id, parents, lines, parent_texts=None):
125
"""Add a single text on top of the versioned file.
127
Must raise RevisionAlreadyPresent if the new version is
128
already present in file history.
130
Must raise RevisionNotPresent if any of the given parents are
131
not present in file history.
132
:param parent_texts: An optional dictionary containing the opaque
133
representations of some or all of the parents of
134
version_id to allow delta optimisations.
135
VERY IMPORTANT: the texts must be those returned
136
by add_lines or data corruption can be caused.
137
:return: An opaque representation of the inserted version which can be
138
provided back to future add_lines calls in the parent_texts
141
self._check_write_ok()
142
return self._add_lines(version_id, parents, lines, parent_texts)
144
def _add_lines(self, version_id, parents, lines, parent_texts):
145
"""Helper to do the class specific add_lines."""
146
raise NotImplementedError(self.add_lines)
148
def add_lines_with_ghosts(self, version_id, parents, lines,
150
"""Add lines to the versioned file, allowing ghosts to be present.
152
This takes the same parameters as add_lines.
154
self._check_write_ok()
155
return self._add_lines_with_ghosts(version_id, parents, lines,
158
def _add_lines_with_ghosts(self, version_id, parents, lines, parent_texts):
159
"""Helper to do class specific add_lines_with_ghosts."""
160
raise NotImplementedError(self.add_lines_with_ghosts)
162
def check(self, progress_bar=None):
163
"""Check the versioned file for integrity."""
164
raise NotImplementedError(self.check)
166
def _check_lines_not_unicode(self, lines):
167
"""Check that lines being added to a versioned file are not unicode."""
169
if line.__class__ is not str:
170
raise errors.BzrBadParameterUnicode("lines")
172
def _check_lines_are_lines(self, lines):
173
"""Check that the lines really are full lines without inline EOL."""
175
if '\n' in line[:-1]:
176
raise errors.BzrBadParameterContainsNewline("lines")
178
def _check_write_ok(self):
179
"""Is the versioned file marked as 'finished' ? Raise if it is."""
181
raise errors.OutSideTransaction()
182
if self._access_mode != 'w':
183
raise errors.ReadOnlyObjectDirtiedError(self)
185
def enable_cache(self):
186
"""Tell this versioned file that it should cache any data it reads.
188
This is advisory, implementations do not have to support caching.
192
def clear_cache(self):
193
"""Remove any data cached in the versioned file object.
195
This only needs to be supported if caches are supported
199
def clone_text(self, new_version_id, old_version_id, parents):
200
"""Add an identical text to old_version_id as new_version_id.
202
Must raise RevisionNotPresent if the old version or any of the
203
parents are not present in file history.
205
Must raise RevisionAlreadyPresent if the new version is
206
already present in file history."""
207
self._check_write_ok()
208
return self._clone_text(new_version_id, old_version_id, parents)
210
def _clone_text(self, new_version_id, old_version_id, parents):
211
"""Helper function to do the _clone_text work."""
212
raise NotImplementedError(self.clone_text)
214
def create_empty(self, name, transport, mode=None):
215
"""Create a new versioned file of this exact type.
217
:param name: the file name
218
:param transport: the transport
219
:param mode: optional file mode.
221
raise NotImplementedError(self.create_empty)
223
def fix_parents(self, version, new_parents):
224
"""Fix the parents list for version.
226
This is done by appending a new version to the index
227
with identical data except for the parents list.
228
the parents list must be a superset of the current
231
self._check_write_ok()
232
return self._fix_parents(version, new_parents)
234
def _fix_parents(self, version, new_parents):
235
"""Helper for fix_parents."""
236
raise NotImplementedError(self.fix_parents)
238
def get_delta(self, version):
239
"""Get a delta for constructing version from some other version.
241
:return: (delta_parent, sha1, noeol, delta)
242
Where delta_parent is a version id or None to indicate no parent.
244
raise NotImplementedError(self.get_delta)
246
def get_deltas(self, versions):
247
"""Get multiple deltas at once for constructing versions.
249
:return: dict(version_id:(delta_parent, sha1, noeol, delta))
250
Where delta_parent is a version id or None to indicate no parent, and
251
version_id is the version_id created by that delta.
254
for version in versions:
255
result[version] = self.get_delta(version)
258
def get_sha1(self, version_id):
259
"""Get the stored sha1 sum for the given revision.
261
:param name: The name of the version to lookup
263
raise NotImplementedError(self.get_sha1)
265
def get_suffixes(self):
266
"""Return the file suffixes associated with this versioned file."""
267
raise NotImplementedError(self.get_suffixes)
269
def get_text(self, version_id):
270
"""Return version contents as a text string.
272
Raises RevisionNotPresent if version is not present in
275
return ''.join(self.get_lines(version_id))
276
get_string = get_text
278
def get_texts(self, version_ids):
279
"""Return the texts of listed versions as a list of strings.
281
Raises RevisionNotPresent if version is not present in
284
return [''.join(self.get_lines(v)) for v in version_ids]
286
def get_lines(self, version_id):
287
"""Return version contents as a sequence of lines.
289
Raises RevisionNotPresent if version is not present in
292
raise NotImplementedError(self.get_lines)
294
def get_ancestry(self, version_ids):
295
"""Return a list of all ancestors of given version(s). This
296
will not include the null revision.
298
Must raise RevisionNotPresent if any of the given versions are
299
not present in file history."""
300
if isinstance(version_ids, basestring):
301
version_ids = [version_ids]
302
raise NotImplementedError(self.get_ancestry)
304
def get_ancestry_with_ghosts(self, version_ids):
305
"""Return a list of all ancestors of given version(s). This
306
will not include the null revision.
308
Must raise RevisionNotPresent if any of the given versions are
309
not present in file history.
311
Ghosts that are known about will be included in ancestry list,
312
but are not explicitly marked.
314
raise NotImplementedError(self.get_ancestry_with_ghosts)
316
def get_graph(self, version_ids=None):
317
"""Return a graph from the versioned file.
319
Ghosts are not listed or referenced in the graph.
320
:param version_ids: Versions to select.
321
None means retrieve all versions.
324
if version_ids is None:
325
for version in self.versions():
326
result[version] = self.get_parents(version)
328
pending = set(version_ids)
330
version = pending.pop()
331
if version in result:
333
parents = self.get_parents(version)
334
for parent in parents:
338
result[version] = parents
341
def get_graph_with_ghosts(self):
342
"""Return a graph for the entire versioned file.
344
Ghosts are referenced in parents list but are not
347
raise NotImplementedError(self.get_graph_with_ghosts)
349
@deprecated_method(zero_eight)
350
def parent_names(self, version):
351
"""Return version names for parents of a version.
353
See get_parents for the current api.
355
return self.get_parents(version)
357
def get_parents(self, version_id):
358
"""Return version names for parents of a version.
360
Must raise RevisionNotPresent if version is not present in
363
raise NotImplementedError(self.get_parents)
365
def get_parents_with_ghosts(self, version_id):
366
"""Return version names for parents of version_id.
368
Will raise RevisionNotPresent if version_id is not present
371
Ghosts that are known about will be included in the parent list,
372
but are not explicitly marked.
374
raise NotImplementedError(self.get_parents_with_ghosts)
376
def annotate_iter(self, version_id):
377
"""Yield list of (version-id, line) pairs for the specified
380
Must raise RevisionNotPresent if any of the given versions are
381
not present in file history.
383
raise NotImplementedError(self.annotate_iter)
385
def annotate(self, version_id):
386
return list(self.annotate_iter(version_id))
388
def _apply_delta(self, lines, delta):
389
"""Apply delta to lines."""
392
for start, end, count, delta_lines in delta:
393
lines[offset+start:offset+end] = delta_lines
394
offset = offset + (start - end) + count
397
def join(self, other, pb=None, msg=None, version_ids=None,
398
ignore_missing=False):
399
"""Integrate versions from other into this versioned file.
401
If version_ids is None all versions from other should be
402
incorporated into this versioned file.
404
Must raise RevisionNotPresent if any of the specified versions
405
are not present in the other files history unless ignore_missing
406
is supplied when they are silently skipped.
408
self._check_write_ok()
409
return InterVersionedFile.get(other, self).join(
415
def iter_lines_added_or_present_in_versions(self, version_ids=None,
417
"""Iterate over the lines in the versioned file from version_ids.
419
This may return lines from other versions, and does not return the
420
specific version marker at this point. The api may be changed
421
during development to include the version that the versioned file
422
thinks is relevant, but given that such hints are just guesses,
423
its better not to have it if we don't need it.
425
If a progress bar is supplied, it may be used to indicate progress.
426
The caller is responsible for cleaning up progress bars (because this
429
NOTES: Lines are normalised: they will all have \n terminators.
430
Lines are returned in arbitrary order.
432
raise NotImplementedError(self.iter_lines_added_or_present_in_versions)
434
def transaction_finished(self):
435
"""The transaction that this file was opened in has finished.
437
This records self.finished = True and should cause all mutating
442
@deprecated_method(zero_eight)
443
def walk(self, version_ids=None):
444
"""Walk the versioned file as a weave-like structure, for
445
versions relative to version_ids. Yields sequence of (lineno,
446
insert, deletes, text) for each relevant line.
448
Must raise RevisionNotPresent if any of the specified versions
449
are not present in the file history.
451
:param version_ids: the version_ids to walk with respect to. If not
452
supplied the entire weave-like structure is walked.
454
walk is deprecated in favour of iter_lines_added_or_present_in_versions
456
raise NotImplementedError(self.walk)
458
@deprecated_method(zero_eight)
459
def iter_names(self):
460
"""Walk the names list."""
461
return iter(self.versions())
463
def plan_merge(self, ver_a, ver_b):
464
"""Return pseudo-annotation indicating how the two versions merge.
466
This is computed between versions a and b and their common
469
Weave lines present in none of them are skipped entirely.
472
killed-base Dead in base revision
473
killed-both Killed in each revision
476
unchanged Alive in both a and b (possibly created in both)
479
ghost-a Killed in a, unborn in b
480
ghost-b Killed in b, unborn in a
481
irrelevant Not in either revision
483
raise NotImplementedError(VersionedFile.plan_merge)
485
def weave_merge(self, plan, a_marker=TextMerge.A_MARKER,
486
b_marker=TextMerge.B_MARKER):
487
return PlanWeaveMerge(plan, a_marker, b_marker).merge_lines()[0]
490
class PlanWeaveMerge(TextMerge):
491
"""Weave merge that takes a plan as its input.
493
This exists so that VersionedFile.plan_merge is implementable.
494
Most callers will want to use WeaveMerge instead.
497
def __init__(self, plan, a_marker=TextMerge.A_MARKER,
498
b_marker=TextMerge.B_MARKER):
499
TextMerge.__init__(self, a_marker, b_marker)
502
def _merge_struct(self):
507
def outstanding_struct():
508
if not lines_a and not lines_b:
510
elif ch_a and not ch_b:
513
elif ch_b and not ch_a:
515
elif lines_a == lines_b:
518
yield (lines_a, lines_b)
520
# We previously considered either 'unchanged' or 'killed-both' lines
521
# to be possible places to resynchronize. However, assuming agreement
522
# on killed-both lines may be too aggressive. -- mbp 20060324
523
for state, line in self.plan:
524
if state == 'unchanged':
525
# resync and flush queued conflicts changes if any
526
for struct in outstanding_struct():
532
if state == 'unchanged':
535
elif state == 'killed-a':
538
elif state == 'killed-b':
541
elif state == 'new-a':
544
elif state == 'new-b':
548
assert state in ('irrelevant', 'ghost-a', 'ghost-b',
549
'killed-base', 'killed-both'), state
550
for struct in outstanding_struct():
554
class WeaveMerge(PlanWeaveMerge):
555
"""Weave merge that takes a VersionedFile and two versions as its input"""
557
def __init__(self, versionedfile, ver_a, ver_b,
558
a_marker=PlanWeaveMerge.A_MARKER, b_marker=PlanWeaveMerge.B_MARKER):
559
plan = versionedfile.plan_merge(ver_a, ver_b)
560
PlanWeaveMerge.__init__(self, plan, a_marker, b_marker)
563
class InterVersionedFile(InterObject):
564
"""This class represents operations taking place between two versionedfiles..
566
Its instances have methods like join, and contain
567
references to the source and target versionedfiles these operations can be
570
Often we will provide convenience methods on 'versionedfile' which carry out
571
operations with another versionedfile - they will always forward to
572
InterVersionedFile.get(other).method_name(parameters).
576
"""The available optimised InterVersionedFile types."""
578
def join(self, pb=None, msg=None, version_ids=None, ignore_missing=False):
579
"""Integrate versions from self.source into self.target.
581
If version_ids is None all versions from source should be
582
incorporated into this versioned file.
584
Must raise RevisionNotPresent if any of the specified versions
585
are not present in the other files history unless ignore_missing is
586
supplied when they are silently skipped.
589
# - if the target is empty, just add all the versions from
590
# source to target, otherwise:
591
# - make a temporary versioned file of type target
592
# - insert the source content into it one at a time
594
if not self.target.versions():
597
# Make a new target-format versioned file.
598
temp_source = self.target.create_empty("temp", MemoryTransport())
600
version_ids = self._get_source_version_ids(version_ids, ignore_missing)
601
graph = self.source.get_graph(version_ids)
602
order = tsort.topo_sort(graph.items())
603
pb = ui.ui_factory.nested_progress_bar()
606
# TODO for incremental cross-format work:
607
# make a versioned file with the following content:
608
# all revisions we have been asked to join
609
# all their ancestors that are *not* in target already.
610
# the immediate parents of the above two sets, with
611
# empty parent lists - these versions are in target already
612
# and the incorrect version data will be ignored.
613
# TODO: for all ancestors that are present in target already,
614
# check them for consistent data, this requires moving sha1 from
616
# TODO: remove parent texts when they are not relevant any more for
617
# memory pressure reduction. RBC 20060313
618
# pb.update('Converting versioned data', 0, len(order))
619
# deltas = self.source.get_deltas(order)
620
for index, version in enumerate(order):
621
pb.update('Converting versioned data', index, len(order))
622
parent_text = target.add_lines(version,
623
self.source.get_parents(version),
624
self.source.get_lines(version),
625
parent_texts=parent_texts)
626
parent_texts[version] = parent_text
627
#delta_parent, sha1, noeol, delta = deltas[version]
628
#target.add_delta(version,
629
# self.source.get_parents(version),
634
#target.get_lines(version)
636
# this should hit the native code path for target
637
if target is not self.target:
638
return self.target.join(temp_source,
646
def _get_source_version_ids(self, version_ids, ignore_missing):
647
"""Determine the version ids to be used from self.source.
649
:param version_ids: The caller-supplied version ids to check. (None
650
for all). If None is in version_ids, it is stripped.
651
:param ignore_missing: if True, remove missing ids from the version
652
list. If False, raise RevisionNotPresent on
653
a missing version id.
654
:return: A set of version ids.
656
if version_ids is None:
657
# None cannot be in source.versions
658
return set(self.source.versions())
661
return set(self.source.versions()).intersection(set(version_ids))
663
new_version_ids = set()
664
for version in version_ids:
667
if not self.source.has_version(version):
668
raise errors.RevisionNotPresent(version, str(self.source))
670
new_version_ids.add(version)
671
return new_version_ids
674
class InterVersionedFileTestProviderAdapter(object):
675
"""A tool to generate a suite testing multiple inter versioned-file classes.
677
This is done by copying the test once for each InterVersionedFile provider
678
and injecting the transport_server, transport_readonly_server,
679
versionedfile_factory and versionedfile_factory_to classes into each copy.
680
Each copy is also given a new id() to make it easy to identify.
683
def __init__(self, transport_server, transport_readonly_server, formats):
684
self._transport_server = transport_server
685
self._transport_readonly_server = transport_readonly_server
686
self._formats = formats
688
def adapt(self, test):
689
result = unittest.TestSuite()
690
for (interversionedfile_class,
691
versionedfile_factory,
692
versionedfile_factory_to) in self._formats:
693
new_test = deepcopy(test)
694
new_test.transport_server = self._transport_server
695
new_test.transport_readonly_server = self._transport_readonly_server
696
new_test.interversionedfile_class = interversionedfile_class
697
new_test.versionedfile_factory = versionedfile_factory
698
new_test.versionedfile_factory_to = versionedfile_factory_to
699
def make_new_test_id():
700
new_id = "%s(%s)" % (new_test.id(), interversionedfile_class.__name__)
701
return lambda: new_id
702
new_test.id = make_new_test_id()
703
result.addTest(new_test)
707
def default_test_list():
708
"""Generate the default list of interversionedfile permutations to test."""
709
from bzrlib.weave import WeaveFile
710
from bzrlib.knit import KnitVersionedFile
712
# test the fallback InterVersionedFile from annotated knits to weave
713
result.append((InterVersionedFile,
716
for optimiser in InterVersionedFile._optimisers:
717
result.append((optimiser,
718
optimiser._matching_file_from_factory,
719
optimiser._matching_file_to_factory
721
# if there are specific combinations we want to use, we can add them
added
removed
bzrlib/versionedfile.py
