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(), """
33
from bzrlib.transport.memory import MemoryTransport
36
from cStringIO import StringIO
38
from bzrlib.inter import InterObject
39
from bzrlib.textmerge import TextMerge
42
class VersionedFile(object):
43
"""Versioned text file storage.
45
A versioned file manages versions of line-based text files,
46
keeping track of the originating version for each line.
48
To clients the "lines" of the file are represented as a list of
49
strings. These strings will typically have terminal newline
50
characters, but this is not required. In particular files commonly
51
do not have a newline at the end of the file.
53
Texts are identified by a version-id string.
56
def __init__(self, access_mode):
58
self._access_mode = access_mode
61
def check_not_reserved_id(version_id):
62
revision.check_not_reserved_id(version_id)
64
def copy_to(self, name, transport):
65
"""Copy this versioned file to name on transport."""
66
raise NotImplementedError(self.copy_to)
69
"""Return a unsorted list of versions."""
70
raise NotImplementedError(self.versions)
72
def has_ghost(self, version_id):
73
"""Returns whether version is present as a ghost."""
74
raise NotImplementedError(self.has_ghost)
76
def has_version(self, version_id):
77
"""Returns whether version is present."""
78
raise NotImplementedError(self.has_version)
80
def add_lines(self, version_id, parents, lines, parent_texts=None,
81
left_matching_blocks=None, nostore_sha=None, random_id=False,
83
"""Add a single text on top of the versioned file.
85
Must raise RevisionAlreadyPresent if the new version is
86
already present in file history.
88
Must raise RevisionNotPresent if any of the given parents are
89
not present in file history.
91
:param lines: A list of lines. Each line must be a bytestring. And all
92
of them except the last must be terminated with \n and contain no
93
other \n's. The last line may either contain no \n's or a single
94
terminated \n. If the lines list does meet this constraint the add
95
routine may error or may succeed - but you will be unable to read
96
the data back accurately. (Checking the lines have been split
97
correctly is expensive and extremely unlikely to catch bugs so it
98
is not done at runtime unless check_content is True.)
99
:param parent_texts: An optional dictionary containing the opaque
100
representations of some or all of the parents of version_id to
101
allow delta optimisations. VERY IMPORTANT: the texts must be those
102
returned by add_lines or data corruption can be caused.
103
:param left_matching_blocks: a hint about which areas are common
104
between the text and its left-hand-parent. The format is
105
the SequenceMatcher.get_matching_blocks format.
106
:param nostore_sha: Raise ExistingContent and do not add the lines to
107
the versioned file if the digest of the lines matches this.
108
:param random_id: If True a random id has been selected rather than
109
an id determined by some deterministic process such as a converter
110
from a foreign VCS. When True the backend may choose not to check
111
for uniqueness of the resulting key within the versioned file, so
112
this should only be done when the result is expected to be unique
114
:param check_content: If True, the lines supplied are verified to be
115
bytestrings that are correctly formed lines.
116
:return: The text sha1, the number of bytes in the text, and an opaque
117
representation of the inserted version which can be provided
118
back to future add_lines calls in the parent_texts dictionary.
120
version_id = osutils.safe_revision_id(version_id)
121
parents = [osutils.safe_revision_id(v) for v in parents]
122
self._check_write_ok()
123
return self._add_lines(version_id, parents, lines, parent_texts,
124
left_matching_blocks, nostore_sha, random_id, check_content)
126
def _add_lines(self, version_id, parents, lines, parent_texts,
127
left_matching_blocks, nostore_sha, random_id, check_content):
128
"""Helper to do the class specific add_lines."""
129
raise NotImplementedError(self.add_lines)
131
def add_lines_with_ghosts(self, version_id, parents, lines,
132
parent_texts=None, nostore_sha=None, random_id=False,
134
"""Add lines to the versioned file, allowing ghosts to be present.
136
This takes the same parameters as add_lines and returns the same.
138
version_id = osutils.safe_revision_id(version_id)
139
parents = [osutils.safe_revision_id(v) for v in parents]
140
self._check_write_ok()
141
return self._add_lines_with_ghosts(version_id, parents, lines,
142
parent_texts, nostore_sha, random_id, check_content)
144
def _add_lines_with_ghosts(self, version_id, parents, lines, parent_texts,
145
nostore_sha, random_id, check_content):
146
"""Helper to do class specific add_lines_with_ghosts."""
147
raise NotImplementedError(self.add_lines_with_ghosts)
149
def check(self, progress_bar=None):
150
"""Check the versioned file for integrity."""
151
raise NotImplementedError(self.check)
153
def _check_lines_not_unicode(self, lines):
154
"""Check that lines being added to a versioned file are not unicode."""
156
if line.__class__ is not str:
157
raise errors.BzrBadParameterUnicode("lines")
159
def _check_lines_are_lines(self, lines):
160
"""Check that the lines really are full lines without inline EOL."""
162
if '\n' in line[:-1]:
163
raise errors.BzrBadParameterContainsNewline("lines")
165
def _check_write_ok(self):
166
"""Is the versioned file marked as 'finished' ? Raise if it is."""
168
raise errors.OutSideTransaction()
169
if self._access_mode != 'w':
170
raise errors.ReadOnlyObjectDirtiedError(self)
172
def enable_cache(self):
173
"""Tell this versioned file that it should cache any data it reads.
175
This is advisory, implementations do not have to support caching.
179
def clear_cache(self):
180
"""Remove any data cached in the versioned file object.
182
This only needs to be supported if caches are supported
186
def clone_text(self, new_version_id, old_version_id, parents):
187
"""Add an identical text to old_version_id as new_version_id.
189
Must raise RevisionNotPresent if the old version or any of the
190
parents are not present in file history.
192
Must raise RevisionAlreadyPresent if the new version is
193
already present in file history."""
194
new_version_id = osutils.safe_revision_id(new_version_id)
195
old_version_id = osutils.safe_revision_id(old_version_id)
196
parents = [osutils.safe_revision_id(v) for v in parents]
197
self._check_write_ok()
198
return self._clone_text(new_version_id, old_version_id, parents)
200
def _clone_text(self, new_version_id, old_version_id, parents):
201
"""Helper function to do the _clone_text work."""
202
raise NotImplementedError(self.clone_text)
204
def create_empty(self, name, transport, mode=None):
205
"""Create a new versioned file of this exact type.
207
:param name: the file name
208
:param transport: the transport
209
:param mode: optional file mode.
211
raise NotImplementedError(self.create_empty)
213
def get_format_signature(self):
214
"""Get a text description of the data encoding in this file.
218
raise NotImplementedError(self.get_format_signature)
220
def make_mpdiffs(self, version_ids):
221
"""Create multiparent diffs for specified versions."""
222
knit_versions = set()
223
for version_id in version_ids:
224
knit_versions.add(version_id)
225
knit_versions.update(self.get_parents(version_id))
226
lines = dict(zip(knit_versions,
227
self._get_lf_split_line_list(knit_versions)))
229
for version_id in version_ids:
230
target = lines[version_id]
231
parents = [lines[p] for p in self.get_parents(version_id)]
233
left_parent_blocks = self._extract_blocks(version_id,
236
left_parent_blocks = None
237
diffs.append(multiparent.MultiParent.from_lines(target, parents,
241
def _extract_blocks(self, version_id, source, target):
244
def add_mpdiffs(self, records):
245
"""Add mpdiffs to this VersionedFile.
247
Records should be iterables of version, parents, expected_sha1,
248
mpdiff. mpdiff should be a MultiParent instance.
250
# Does this need to call self._check_write_ok()? (IanC 20070919)
252
mpvf = multiparent.MultiMemoryVersionedFile()
254
for version, parent_ids, expected_sha1, mpdiff in records:
255
versions.append(version)
256
mpvf.add_diff(mpdiff, version, parent_ids)
257
needed_parents = set()
258
for version, parent_ids, expected_sha1, mpdiff in records:
259
needed_parents.update(p for p in parent_ids
260
if not mpvf.has_version(p))
261
for parent_id, lines in zip(needed_parents,
262
self._get_lf_split_line_list(needed_parents)):
263
mpvf.add_version(lines, parent_id, [])
264
for (version, parent_ids, expected_sha1, mpdiff), lines in\
265
zip(records, mpvf.get_line_list(versions)):
266
if len(parent_ids) == 1:
267
left_matching_blocks = list(mpdiff.get_matching_blocks(0,
268
mpvf.get_diff(parent_ids[0]).num_lines()))
270
left_matching_blocks = None
271
_, _, version_text = self.add_lines(version, parent_ids, lines,
272
vf_parents, left_matching_blocks=left_matching_blocks)
273
vf_parents[version] = version_text
274
for (version, parent_ids, expected_sha1, mpdiff), sha1 in\
275
zip(records, self.get_sha1s(versions)):
276
if expected_sha1 != sha1:
277
raise errors.VersionedFileInvalidChecksum(version)
279
def get_sha1(self, version_id):
280
"""Get the stored sha1 sum for the given revision.
282
:param version_id: The name of the version to lookup
284
raise NotImplementedError(self.get_sha1)
286
def get_sha1s(self, version_ids):
287
"""Get the stored sha1 sums for the given revisions.
289
:param version_ids: The names of the versions to lookup
290
:return: a list of sha1s in order according to the version_ids
292
raise NotImplementedError(self.get_sha1s)
294
def get_suffixes(self):
295
"""Return the file suffixes associated with this versioned file."""
296
raise NotImplementedError(self.get_suffixes)
298
def get_text(self, version_id):
299
"""Return version contents as a text string.
301
Raises RevisionNotPresent if version is not present in
304
return ''.join(self.get_lines(version_id))
305
get_string = get_text
307
def get_texts(self, version_ids):
308
"""Return the texts of listed versions as a list of strings.
310
Raises RevisionNotPresent if version is not present in
313
return [''.join(self.get_lines(v)) for v in version_ids]
315
def get_lines(self, version_id):
316
"""Return version contents as a sequence of lines.
318
Raises RevisionNotPresent if version is not present in
321
raise NotImplementedError(self.get_lines)
323
def _get_lf_split_line_list(self, version_ids):
324
return [StringIO(t).readlines() for t in self.get_texts(version_ids)]
326
def get_ancestry(self, version_ids, topo_sorted=True):
327
"""Return a list of all ancestors of given version(s). This
328
will not include the null revision.
330
This list will not be topologically sorted if topo_sorted=False is
333
Must raise RevisionNotPresent if any of the given versions are
334
not present in file history."""
335
if isinstance(version_ids, basestring):
336
version_ids = [version_ids]
337
raise NotImplementedError(self.get_ancestry)
339
def get_ancestry_with_ghosts(self, version_ids):
340
"""Return a list of all ancestors of given version(s). This
341
will not include the null revision.
343
Must raise RevisionNotPresent if any of the given versions are
344
not present in file history.
346
Ghosts that are known about will be included in ancestry list,
347
but are not explicitly marked.
349
raise NotImplementedError(self.get_ancestry_with_ghosts)
351
def get_graph(self, version_ids=None):
352
"""Return a graph from the versioned file.
354
Ghosts are not listed or referenced in the graph.
355
:param version_ids: Versions to select.
356
None means retrieve all versions.
358
if version_ids is None:
359
return dict(self.iter_parents(self.versions()))
361
pending = set(osutils.safe_revision_id(v) for v in version_ids)
363
this_iteration = pending
365
for version, parents in self.iter_parents(this_iteration):
366
result[version] = parents
367
for parent in parents:
373
def get_graph_with_ghosts(self):
374
"""Return a graph for the entire versioned file.
376
Ghosts are referenced in parents list but are not
379
raise NotImplementedError(self.get_graph_with_ghosts)
381
def get_parents(self, version_id):
382
"""Return version names for parents of a version.
384
Must raise RevisionNotPresent if version is not present in
387
raise NotImplementedError(self.get_parents)
389
def get_parents_with_ghosts(self, version_id):
390
"""Return version names for parents of version_id.
392
Will raise RevisionNotPresent if version_id is not present
395
Ghosts that are known about will be included in the parent list,
396
but are not explicitly marked.
398
raise NotImplementedError(self.get_parents_with_ghosts)
400
def annotate_iter(self, version_id):
401
"""Yield list of (version-id, line) pairs for the specified
404
Must raise RevisionNotPresent if the given version is
405
not present in file history.
407
raise NotImplementedError(self.annotate_iter)
409
def annotate(self, version_id):
410
return list(self.annotate_iter(version_id))
412
def join(self, other, pb=None, msg=None, version_ids=None,
413
ignore_missing=False):
414
"""Integrate versions from other into this versioned file.
416
If version_ids is None all versions from other should be
417
incorporated into this versioned file.
419
Must raise RevisionNotPresent if any of the specified versions
420
are not present in the other file's history unless ignore_missing
421
is supplied in which case they are silently skipped.
423
self._check_write_ok()
424
return InterVersionedFile.get(other, self).join(
430
def iter_lines_added_or_present_in_versions(self, version_ids=None,
432
"""Iterate over the lines in the versioned file from version_ids.
434
This may return lines from other versions, and does not return the
435
specific version marker at this point. The api may be changed
436
during development to include the version that the versioned file
437
thinks is relevant, but given that such hints are just guesses,
438
its better not to have it if we don't need it.
440
If a progress bar is supplied, it may be used to indicate progress.
441
The caller is responsible for cleaning up progress bars (because this
444
NOTES: Lines are normalised: they will all have \n terminators.
445
Lines are returned in arbitrary order.
447
raise NotImplementedError(self.iter_lines_added_or_present_in_versions)
449
def iter_parents(self, version_ids):
450
"""Iterate through the parents for many version ids.
452
:param version_ids: An iterable yielding version_ids.
453
:return: An iterator that yields (version_id, parents). Requested
454
version_ids not present in the versioned file are simply skipped.
455
The order is undefined, allowing for different optimisations in
456
the underlying implementation.
458
for version_id in version_ids:
460
yield version_id, tuple(self.get_parents(version_id))
461
except errors.RevisionNotPresent:
464
def transaction_finished(self):
465
"""The transaction that this file was opened in has finished.
467
This records self.finished = True and should cause all mutating
472
def plan_merge(self, ver_a, ver_b):
473
"""Return pseudo-annotation indicating how the two versions merge.
475
This is computed between versions a and b and their common
478
Weave lines present in none of them are skipped entirely.
481
killed-base Dead in base revision
482
killed-both Killed in each revision
485
unchanged Alive in both a and b (possibly created in both)
488
ghost-a Killed in a, unborn in b
489
ghost-b Killed in b, unborn in a
490
irrelevant Not in either revision
492
raise NotImplementedError(VersionedFile.plan_merge)
494
def weave_merge(self, plan, a_marker=TextMerge.A_MARKER,
495
b_marker=TextMerge.B_MARKER):
496
return PlanWeaveMerge(plan, a_marker, b_marker).merge_lines()[0]
499
class PlanWeaveMerge(TextMerge):
500
"""Weave merge that takes a plan as its input.
502
This exists so that VersionedFile.plan_merge is implementable.
503
Most callers will want to use WeaveMerge instead.
506
def __init__(self, plan, a_marker=TextMerge.A_MARKER,
507
b_marker=TextMerge.B_MARKER):
508
TextMerge.__init__(self, a_marker, b_marker)
511
def _merge_struct(self):
516
def outstanding_struct():
517
if not lines_a and not lines_b:
519
elif ch_a and not ch_b:
522
elif ch_b and not ch_a:
524
elif lines_a == lines_b:
527
yield (lines_a, lines_b)
529
# We previously considered either 'unchanged' or 'killed-both' lines
530
# to be possible places to resynchronize. However, assuming agreement
531
# on killed-both lines may be too aggressive. -- mbp 20060324
532
for state, line in self.plan:
533
if state == 'unchanged':
534
# resync and flush queued conflicts changes if any
535
for struct in outstanding_struct():
541
if state == 'unchanged':
544
elif state == 'killed-a':
547
elif state == 'killed-b':
550
elif state == 'new-a':
553
elif state == 'new-b':
557
assert state in ('irrelevant', 'ghost-a', 'ghost-b',
558
'killed-base', 'killed-both'), state
559
for struct in outstanding_struct():
563
class WeaveMerge(PlanWeaveMerge):
564
"""Weave merge that takes a VersionedFile and two versions as its input."""
566
def __init__(self, versionedfile, ver_a, ver_b,
567
a_marker=PlanWeaveMerge.A_MARKER, b_marker=PlanWeaveMerge.B_MARKER):
568
plan = versionedfile.plan_merge(ver_a, ver_b)
569
PlanWeaveMerge.__init__(self, plan, a_marker, b_marker)
572
class InterVersionedFile(InterObject):
573
"""This class represents operations taking place between two VersionedFiles.
575
Its instances have methods like join, and contain
576
references to the source and target versionedfiles these operations can be
579
Often we will provide convenience methods on 'versionedfile' which carry out
580
operations with another versionedfile - they will always forward to
581
InterVersionedFile.get(other).method_name(parameters).
585
"""The available optimised InterVersionedFile types."""
587
def join(self, pb=None, msg=None, version_ids=None, ignore_missing=False):
588
"""Integrate versions from self.source into self.target.
590
If version_ids is None all versions from source should be
591
incorporated into this versioned file.
593
Must raise RevisionNotPresent if any of the specified versions
594
are not present in the other file's history unless ignore_missing is
595
supplied in which case they are silently skipped.
598
# - if the target is empty, just add all the versions from
599
# source to target, otherwise:
600
# - make a temporary versioned file of type target
601
# - insert the source content into it one at a time
603
if not self.target.versions():
606
# Make a new target-format versioned file.
607
temp_source = self.target.create_empty("temp", MemoryTransport())
609
version_ids = self._get_source_version_ids(version_ids, ignore_missing)
610
graph = self.source.get_graph(version_ids)
611
order = tsort.topo_sort(graph.items())
612
pb = ui.ui_factory.nested_progress_bar()
615
# TODO for incremental cross-format work:
616
# make a versioned file with the following content:
617
# all revisions we have been asked to join
618
# all their ancestors that are *not* in target already.
619
# the immediate parents of the above two sets, with
620
# empty parent lists - these versions are in target already
621
# and the incorrect version data will be ignored.
622
# TODO: for all ancestors that are present in target already,
623
# check them for consistent data, this requires moving sha1 from
625
# TODO: remove parent texts when they are not relevant any more for
626
# memory pressure reduction. RBC 20060313
627
# pb.update('Converting versioned data', 0, len(order))
629
for index, version in enumerate(order):
630
pb.update('Converting versioned data', index, total)
631
_, _, parent_text = target.add_lines(version,
632
self.source.get_parents(version),
633
self.source.get_lines(version),
634
parent_texts=parent_texts)
635
parent_texts[version] = parent_text
637
# this should hit the native code path for target
638
if target is not self.target:
639
return self.target.join(temp_source,
649
def _get_source_version_ids(self, version_ids, ignore_missing):
650
"""Determine the version ids to be used from self.source.
652
:param version_ids: The caller-supplied version ids to check. (None
653
for all). If None is in version_ids, it is stripped.
654
:param ignore_missing: if True, remove missing ids from the version
655
list. If False, raise RevisionNotPresent on
656
a missing version id.
657
:return: A set of version ids.
659
if version_ids is None:
660
# None cannot be in source.versions
661
return set(self.source.versions())
663
version_ids = [osutils.safe_revision_id(v) for v in version_ids]
665
return set(self.source.versions()).intersection(set(version_ids))
667
new_version_ids = set()
668
for version in version_ids:
671
if not self.source.has_version(version):
672
raise errors.RevisionNotPresent(version, str(self.source))
674
new_version_ids.add(version)
675
return new_version_ids
added
removed
bzrlib/versionedfile.py
