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
self._check_write_ok()
121
return self._add_lines(version_id, parents, lines, parent_texts,
122
left_matching_blocks, nostore_sha, random_id, check_content)
124
def _add_lines(self, version_id, parents, lines, parent_texts,
125
left_matching_blocks, nostore_sha, random_id, check_content):
126
"""Helper to do the class specific add_lines."""
127
raise NotImplementedError(self.add_lines)
129
def add_lines_with_ghosts(self, version_id, parents, lines,
130
parent_texts=None, nostore_sha=None, random_id=False,
132
"""Add lines to the versioned file, allowing ghosts to be present.
134
This takes the same parameters as add_lines and returns the same.
136
self._check_write_ok()
137
return self._add_lines_with_ghosts(version_id, parents, lines,
138
parent_texts, nostore_sha, random_id, check_content)
140
def _add_lines_with_ghosts(self, version_id, parents, lines, parent_texts,
141
nostore_sha, random_id, check_content):
142
"""Helper to do class specific add_lines_with_ghosts."""
143
raise NotImplementedError(self.add_lines_with_ghosts)
145
def check(self, progress_bar=None):
146
"""Check the versioned file for integrity."""
147
raise NotImplementedError(self.check)
149
def _check_lines_not_unicode(self, lines):
150
"""Check that lines being added to a versioned file are not unicode."""
152
if line.__class__ is not str:
153
raise errors.BzrBadParameterUnicode("lines")
155
def _check_lines_are_lines(self, lines):
156
"""Check that the lines really are full lines without inline EOL."""
158
if '\n' in line[:-1]:
159
raise errors.BzrBadParameterContainsNewline("lines")
161
def _check_write_ok(self):
162
"""Is the versioned file marked as 'finished' ? Raise if it is."""
164
raise errors.OutSideTransaction()
165
if self._access_mode != 'w':
166
raise errors.ReadOnlyObjectDirtiedError(self)
168
def enable_cache(self):
169
"""Tell this versioned file that it should cache any data it reads.
171
This is advisory, implementations do not have to support caching.
175
def clear_cache(self):
176
"""Remove any data cached in the versioned file object.
178
This only needs to be supported if caches are supported
182
def clone_text(self, new_version_id, old_version_id, parents):
183
"""Add an identical text to old_version_id as new_version_id.
185
Must raise RevisionNotPresent if the old version or any of the
186
parents are not present in file history.
188
Must raise RevisionAlreadyPresent if the new version is
189
already present in file history."""
190
self._check_write_ok()
191
return self._clone_text(new_version_id, old_version_id, parents)
193
def _clone_text(self, new_version_id, old_version_id, parents):
194
"""Helper function to do the _clone_text work."""
195
raise NotImplementedError(self.clone_text)
197
def create_empty(self, name, transport, mode=None):
198
"""Create a new versioned file of this exact type.
200
:param name: the file name
201
:param transport: the transport
202
:param mode: optional file mode.
204
raise NotImplementedError(self.create_empty)
206
def get_format_signature(self):
207
"""Get a text description of the data encoding in this file.
211
raise NotImplementedError(self.get_format_signature)
213
def make_mpdiffs(self, version_ids):
214
"""Create multiparent diffs for specified versions."""
215
knit_versions = set()
216
for version_id in version_ids:
217
knit_versions.add(version_id)
218
knit_versions.update(self.get_parents(version_id))
219
lines = dict(zip(knit_versions,
220
self._get_lf_split_line_list(knit_versions)))
222
for version_id in version_ids:
223
target = lines[version_id]
224
parents = [lines[p] for p in self.get_parents(version_id)]
226
left_parent_blocks = self._extract_blocks(version_id,
229
left_parent_blocks = None
230
diffs.append(multiparent.MultiParent.from_lines(target, parents,
234
def _extract_blocks(self, version_id, source, target):
237
def add_mpdiffs(self, records):
238
"""Add mpdiffs to this VersionedFile.
240
Records should be iterables of version, parents, expected_sha1,
241
mpdiff. mpdiff should be a MultiParent instance.
243
# Does this need to call self._check_write_ok()? (IanC 20070919)
245
mpvf = multiparent.MultiMemoryVersionedFile()
247
for version, parent_ids, expected_sha1, mpdiff in records:
248
versions.append(version)
249
mpvf.add_diff(mpdiff, version, parent_ids)
250
needed_parents = set()
251
for version, parent_ids, expected_sha1, mpdiff in records:
252
needed_parents.update(p for p in parent_ids
253
if not mpvf.has_version(p))
254
for parent_id, lines in zip(needed_parents,
255
self._get_lf_split_line_list(needed_parents)):
256
mpvf.add_version(lines, parent_id, [])
257
for (version, parent_ids, expected_sha1, mpdiff), lines in\
258
zip(records, mpvf.get_line_list(versions)):
259
if len(parent_ids) == 1:
260
left_matching_blocks = list(mpdiff.get_matching_blocks(0,
261
mpvf.get_diff(parent_ids[0]).num_lines()))
263
left_matching_blocks = None
264
_, _, version_text = self.add_lines(version, parent_ids, lines,
265
vf_parents, left_matching_blocks=left_matching_blocks)
266
vf_parents[version] = version_text
267
for (version, parent_ids, expected_sha1, mpdiff), sha1 in\
268
zip(records, self.get_sha1s(versions)):
269
if expected_sha1 != sha1:
270
raise errors.VersionedFileInvalidChecksum(version)
272
def get_sha1(self, version_id):
273
"""Get the stored sha1 sum for the given revision.
275
:param version_id: The name of the version to lookup
277
raise NotImplementedError(self.get_sha1)
279
def get_sha1s(self, version_ids):
280
"""Get the stored sha1 sums for the given revisions.
282
:param version_ids: The names of the versions to lookup
283
:return: a list of sha1s in order according to the version_ids
285
raise NotImplementedError(self.get_sha1s)
287
def get_suffixes(self):
288
"""Return the file suffixes associated with this versioned file."""
289
raise NotImplementedError(self.get_suffixes)
291
def get_text(self, version_id):
292
"""Return version contents as a text string.
294
Raises RevisionNotPresent if version is not present in
297
return ''.join(self.get_lines(version_id))
298
get_string = get_text
300
def get_texts(self, version_ids):
301
"""Return the texts of listed versions as a list of strings.
303
Raises RevisionNotPresent if version is not present in
306
return [''.join(self.get_lines(v)) for v in version_ids]
308
def get_lines(self, version_id):
309
"""Return version contents as a sequence of lines.
311
Raises RevisionNotPresent if version is not present in
314
raise NotImplementedError(self.get_lines)
316
def _get_lf_split_line_list(self, version_ids):
317
return [StringIO(t).readlines() for t in self.get_texts(version_ids)]
319
def get_ancestry(self, version_ids, topo_sorted=True):
320
"""Return a list of all ancestors of given version(s). This
321
will not include the null revision.
323
This list will not be topologically sorted if topo_sorted=False is
326
Must raise RevisionNotPresent if any of the given versions are
327
not present in file history."""
328
if isinstance(version_ids, basestring):
329
version_ids = [version_ids]
330
raise NotImplementedError(self.get_ancestry)
332
def get_ancestry_with_ghosts(self, version_ids):
333
"""Return a list of all ancestors of given version(s). This
334
will not include the null revision.
336
Must raise RevisionNotPresent if any of the given versions are
337
not present in file history.
339
Ghosts that are known about will be included in ancestry list,
340
but are not explicitly marked.
342
raise NotImplementedError(self.get_ancestry_with_ghosts)
344
def get_graph(self, version_ids=None):
345
"""Return a graph from the versioned file.
347
Ghosts are not listed or referenced in the graph.
348
:param version_ids: Versions to select.
349
None means retrieve all versions.
351
if version_ids is None:
352
return dict(self.iter_parents(self.versions()))
354
pending = set(version_ids)
356
this_iteration = pending
358
for version, parents in self.iter_parents(this_iteration):
359
result[version] = parents
360
for parent in parents:
366
def get_graph_with_ghosts(self):
367
"""Return a graph for the entire versioned file.
369
Ghosts are referenced in parents list but are not
372
raise NotImplementedError(self.get_graph_with_ghosts)
374
def get_parents(self, version_id):
375
"""Return version names for parents of a version.
377
Must raise RevisionNotPresent if version is not present in
380
raise NotImplementedError(self.get_parents)
382
def get_parents_with_ghosts(self, version_id):
383
"""Return version names for parents of version_id.
385
Will raise RevisionNotPresent if version_id is not present
388
Ghosts that are known about will be included in the parent list,
389
but are not explicitly marked.
391
raise NotImplementedError(self.get_parents_with_ghosts)
393
def annotate_iter(self, version_id):
394
"""Yield list of (version-id, line) pairs for the specified
397
Must raise RevisionNotPresent if the given version is
398
not present in file history.
400
raise NotImplementedError(self.annotate_iter)
402
def annotate(self, version_id):
403
return list(self.annotate_iter(version_id))
405
def join(self, other, pb=None, msg=None, version_ids=None,
406
ignore_missing=False):
407
"""Integrate versions from other into this versioned file.
409
If version_ids is None all versions from other should be
410
incorporated into this versioned file.
412
Must raise RevisionNotPresent if any of the specified versions
413
are not present in the other file's history unless ignore_missing
414
is supplied in which case they are silently skipped.
416
self._check_write_ok()
417
return InterVersionedFile.get(other, self).join(
423
def iter_lines_added_or_present_in_versions(self, version_ids=None,
425
"""Iterate over the lines in the versioned file from version_ids.
427
This may return lines from other versions. Each item the returned
428
iterator yields is a tuple of a line and a text version that that line
429
is present in (not introduced in).
431
Ordering of results is in whatever order is most suitable for the
432
underlying storage format.
434
If a progress bar is supplied, it may be used to indicate progress.
435
The caller is responsible for cleaning up progress bars (because this
438
NOTES: Lines are normalised: they will all have \n terminators.
439
Lines are returned in arbitrary order.
441
:return: An iterator over (line, version_id).
443
raise NotImplementedError(self.iter_lines_added_or_present_in_versions)
445
def iter_parents(self, version_ids):
446
"""Iterate through the parents for many version ids.
448
:param version_ids: An iterable yielding version_ids.
449
:return: An iterator that yields (version_id, parents). Requested
450
version_ids not present in the versioned file are simply skipped.
451
The order is undefined, allowing for different optimisations in
452
the underlying implementation.
454
for version_id in version_ids:
456
yield version_id, tuple(self.get_parents(version_id))
457
except errors.RevisionNotPresent:
460
def transaction_finished(self):
461
"""The transaction that this file was opened in has finished.
463
This records self.finished = True and should cause all mutating
468
def plan_merge(self, ver_a, ver_b):
469
"""Return pseudo-annotation indicating how the two versions merge.
471
This is computed between versions a and b and their common
474
Weave lines present in none of them are skipped entirely.
477
killed-base Dead in base revision
478
killed-both Killed in each revision
481
unchanged Alive in both a and b (possibly created in both)
484
ghost-a Killed in a, unborn in b
485
ghost-b Killed in b, unborn in a
486
irrelevant Not in either revision
488
raise NotImplementedError(VersionedFile.plan_merge)
490
def weave_merge(self, plan, a_marker=TextMerge.A_MARKER,
491
b_marker=TextMerge.B_MARKER):
492
return PlanWeaveMerge(plan, a_marker, b_marker).merge_lines()[0]
495
class _PlanMergeVersionedFile(object):
496
"""A VersionedFile for uncommitted and committed texts.
498
It is intended to allow merges to be planned with working tree texts.
499
It implements only the small part of the VersionedFile interface used by
500
PlanMerge. It falls back to multiple versionedfiles for data not stored in
501
_PlanMergeVersionedFile itself.
504
def __init__(self, file_id, fallback_versionedfiles=None):
507
:param file_id: Used when raising exceptions.
508
:param fallback_versionedfiles: If supplied, the set of fallbacks to
509
use. Otherwise, _PlanMergeVersionedFile.fallback_versionedfiles
510
can be appended to later.
512
self._file_id = file_id
513
if fallback_versionedfiles is None:
514
self.fallback_versionedfiles = []
516
self.fallback_versionedfiles = fallback_versionedfiles
520
def plan_merge(self, ver_a, ver_b, base=None):
521
"""See VersionedFile.plan_merge"""
522
from merge import _PlanMerge
524
return _PlanMerge(ver_a, ver_b, self).plan_merge()
525
old_plan = list(_PlanMerge(ver_a, base, self).plan_merge())
526
new_plan = list(_PlanMerge(ver_a, ver_b, self).plan_merge())
527
return _PlanMerge._subtract_plans(old_plan, new_plan)
529
def plan_lca_merge(self, ver_a, ver_b, base=None):
530
from merge import _PlanLCAMerge
531
graph = self._get_graph()
532
new_plan = _PlanLCAMerge(ver_a, ver_b, self, graph).plan_merge()
535
old_plan = _PlanLCAMerge(ver_a, base, self, graph).plan_merge()
536
return _PlanLCAMerge._subtract_plans(list(old_plan), list(new_plan))
538
def add_lines(self, version_id, parents, lines):
539
"""See VersionedFile.add_lines
541
Lines are added locally, not fallback versionedfiles. Also, ghosts are
542
permitted. Only reserved ids are permitted.
544
if not revision.is_reserved_id(version_id):
545
raise ValueError('Only reserved ids may be used')
547
raise ValueError('Parents may not be None')
549
raise ValueError('Lines may not be None')
550
self._parents[version_id] = parents
551
self._lines[version_id] = lines
553
def get_lines(self, version_id):
554
"""See VersionedFile.get_ancestry"""
555
lines = self._lines.get(version_id)
556
if lines is not None:
558
for versionedfile in self.fallback_versionedfiles:
560
return versionedfile.get_lines(version_id)
561
except errors.RevisionNotPresent:
564
raise errors.RevisionNotPresent(version_id, self._file_id)
566
def get_ancestry(self, version_id, topo_sorted=False):
567
"""See VersionedFile.get_ancestry.
569
Note that this implementation assumes that if a VersionedFile can
570
answer get_ancestry at all, it can give an authoritative answer. In
571
fact, ghosts can invalidate this assumption. But it's good enough
572
99% of the time, and far cheaper/simpler.
574
Also note that the results of this version are never topologically
575
sorted, and are a set.
578
raise ValueError('This implementation does not provide sorting')
579
parents = self._parents.get(version_id)
581
for vf in self.fallback_versionedfiles:
583
return vf.get_ancestry(version_id, topo_sorted=False)
584
except errors.RevisionNotPresent:
587
raise errors.RevisionNotPresent(version_id, self._file_id)
588
ancestry = set([version_id])
589
for parent in parents:
590
ancestry.update(self.get_ancestry(parent, topo_sorted=False))
593
def get_parents(self, version_id):
594
"""See VersionedFile.get_parents"""
595
parents = self._parents.get(version_id)
596
if parents is not None:
598
for versionedfile in self.fallback_versionedfiles:
600
return versionedfile.get_parents(version_id)
601
except errors.RevisionNotPresent:
604
raise errors.RevisionNotPresent(version_id, self._file_id)
606
def _get_graph(self):
607
from graph import DictParentsProvider, Graph, _StackedParentsProvider
608
from repofmt.knitrepo import _KnitParentsProvider
609
parent_providers = [DictParentsProvider(self._parents)]
610
for vf in self.fallback_versionedfiles:
611
parent_providers.append(_KnitParentsProvider(vf))
612
return Graph(_StackedParentsProvider(parent_providers))
615
class PlanWeaveMerge(TextMerge):
616
"""Weave merge that takes a plan as its input.
618
This exists so that VersionedFile.plan_merge is implementable.
619
Most callers will want to use WeaveMerge instead.
622
def __init__(self, plan, a_marker=TextMerge.A_MARKER,
623
b_marker=TextMerge.B_MARKER):
624
TextMerge.__init__(self, a_marker, b_marker)
627
def _merge_struct(self):
632
def outstanding_struct():
633
if not lines_a and not lines_b:
635
elif ch_a and not ch_b:
638
elif ch_b and not ch_a:
640
elif lines_a == lines_b:
643
yield (lines_a, lines_b)
645
# We previously considered either 'unchanged' or 'killed-both' lines
646
# to be possible places to resynchronize. However, assuming agreement
647
# on killed-both lines may be too aggressive. -- mbp 20060324
648
for state, line in self.plan:
649
if state == 'unchanged':
650
# resync and flush queued conflicts changes if any
651
for struct in outstanding_struct():
657
if state == 'unchanged':
660
elif state == 'killed-a':
663
elif state == 'killed-b':
666
elif state == 'new-a':
669
elif state == 'new-b':
672
elif state == 'conflicted-a':
675
elif state == 'conflicted-b':
679
assert state in ('irrelevant', 'ghost-a', 'ghost-b',
680
'killed-base', 'killed-both'), state
681
for struct in outstanding_struct():
685
class WeaveMerge(PlanWeaveMerge):
686
"""Weave merge that takes a VersionedFile and two versions as its input."""
688
def __init__(self, versionedfile, ver_a, ver_b,
689
a_marker=PlanWeaveMerge.A_MARKER, b_marker=PlanWeaveMerge.B_MARKER):
690
plan = versionedfile.plan_merge(ver_a, ver_b)
691
PlanWeaveMerge.__init__(self, plan, a_marker, b_marker)
694
class InterVersionedFile(InterObject):
695
"""This class represents operations taking place between two VersionedFiles.
697
Its instances have methods like join, and contain
698
references to the source and target versionedfiles these operations can be
701
Often we will provide convenience methods on 'versionedfile' which carry out
702
operations with another versionedfile - they will always forward to
703
InterVersionedFile.get(other).method_name(parameters).
707
"""The available optimised InterVersionedFile types."""
709
def join(self, pb=None, msg=None, version_ids=None, ignore_missing=False):
710
"""Integrate versions from self.source into self.target.
712
If version_ids is None all versions from source should be
713
incorporated into this versioned file.
715
Must raise RevisionNotPresent if any of the specified versions
716
are not present in the other file's history unless ignore_missing is
717
supplied in which case they are silently skipped.
720
# - if the target is empty, just add all the versions from
721
# source to target, otherwise:
722
# - make a temporary versioned file of type target
723
# - insert the source content into it one at a time
725
if not self.target.versions():
728
# Make a new target-format versioned file.
729
temp_source = self.target.create_empty("temp", MemoryTransport())
731
version_ids = self._get_source_version_ids(version_ids, ignore_missing)
732
graph = self.source.get_graph(version_ids)
733
order = tsort.topo_sort(graph.items())
734
pb = ui.ui_factory.nested_progress_bar()
737
# TODO for incremental cross-format work:
738
# make a versioned file with the following content:
739
# all revisions we have been asked to join
740
# all their ancestors that are *not* in target already.
741
# the immediate parents of the above two sets, with
742
# empty parent lists - these versions are in target already
743
# and the incorrect version data will be ignored.
744
# TODO: for all ancestors that are present in target already,
745
# check them for consistent data, this requires moving sha1 from
747
# TODO: remove parent texts when they are not relevant any more for
748
# memory pressure reduction. RBC 20060313
749
# pb.update('Converting versioned data', 0, len(order))
751
for index, version in enumerate(order):
752
pb.update('Converting versioned data', index, total)
753
_, _, parent_text = target.add_lines(version,
754
self.source.get_parents(version),
755
self.source.get_lines(version),
756
parent_texts=parent_texts)
757
parent_texts[version] = parent_text
759
# this should hit the native code path for target
760
if target is not self.target:
761
return self.target.join(temp_source,
771
def _get_source_version_ids(self, version_ids, ignore_missing):
772
"""Determine the version ids to be used from self.source.
774
:param version_ids: The caller-supplied version ids to check. (None
775
for all). If None is in version_ids, it is stripped.
776
:param ignore_missing: if True, remove missing ids from the version
777
list. If False, raise RevisionNotPresent on
778
a missing version id.
779
:return: A set of version ids.
781
if version_ids is None:
782
# None cannot be in source.versions
783
return set(self.source.versions())
786
return set(self.source.versions()).intersection(set(version_ids))
788
new_version_ids = set()
789
for version in version_ids:
792
if not self.source.has_version(version):
793
raise errors.RevisionNotPresent(version, str(self.source))
795
new_version_ids.add(version)
796
return new_version_ids
added
removed
bzrlib/versionedfile.py
