1
# Copyright (C) 2007-2011 Canonical Ltd
3
# This program is free software; you can redistribute it and/or modify
4
# it under the terms of the GNU General Public License as published by
5
# the Free Software Foundation; either version 2 of the License, or
6
# (at your option) any later version.
8
# This program is distributed in the hope that it will be useful,
9
# but WITHOUT ANY WARRANTY; without even the implied warranty of
10
# MERCHANTABILITY or FITNESS FOR A PARTICULAR PURPOSE. See the
11
# GNU General Public License for more details.
13
# You should have received a copy of the GNU General Public License
14
# along with this program; if not, write to the Free Software
15
# Foundation, Inc., 51 Franklin Street, Fifth Floor, Boston, MA 02110-1301 USA
17
from __future__ import absolute_import
22
from .lazy_import import lazy_import
24
lazy_import(globals(), """
41
def topo_iter_keys(vf, keys=None):
44
parents = vf.get_parent_map(keys)
45
return _topo_iter(parents, keys)
47
def topo_iter(vf, versions=None):
49
versions = vf.versions()
50
parents = vf.get_parent_map(versions)
51
return _topo_iter(parents, versions)
53
def _topo_iter(parents, versions):
56
def pending_parents(version):
57
if parents[version] is None:
59
return [v for v in parents[version] if v in versions and
61
for version_id in versions:
62
if parents[version_id] is None:
65
for parent_id in parents[version_id]:
66
descendants.setdefault(parent_id, []).append(version_id)
67
cur = [v for v in versions if len(pending_parents(v)) == 0]
70
for version_id in cur:
71
if version_id in seen:
73
if len(pending_parents(version_id)) != 0:
75
next.extend(descendants.get(version_id, []))
81
class MultiParent(object):
82
"""A multi-parent diff"""
86
def __init__(self, hunks=None):
93
return "MultiParent(%r)" % self.hunks
95
def __eq__(self, other):
96
if self.__class__ is not other.__class__:
98
return (self.hunks == other.hunks)
101
def from_lines(text, parents=(), left_blocks=None):
102
"""Produce a MultiParent from a list of lines and parents"""
104
matcher = patiencediff.PatienceSequenceMatcher(None, parent,
106
return matcher.get_matching_blocks()
108
if left_blocks is None:
109
left_blocks = compare(parents[0])
110
parent_comparisons = [left_blocks] + [compare(p) for p in
113
parent_comparisons = []
115
new_text = NewText([])
117
block_iter = [iter(i) for i in parent_comparisons]
118
diff = MultiParent([])
121
return next(block_iter[p])
122
except StopIteration:
124
cur_block = [next_block(p) for p, i in enumerate(block_iter)]
125
while cur_line < len(text):
127
for p, block in enumerate(cur_block):
131
while j + n <= cur_line:
132
block = cur_block[p] = next_block(p)
140
offset = cur_line - j
146
if best_match is None or n > best_match.num_lines:
147
best_match = ParentText(p, i, j, n)
148
if best_match is None:
149
new_text.lines.append(text[cur_line])
152
if len(new_text.lines) > 0:
153
diff.hunks.append(new_text)
154
new_text = NewText([])
155
diff.hunks.append(best_match)
156
cur_line += best_match.num_lines
157
if len(new_text.lines) > 0:
158
diff.hunks.append(new_text)
161
def get_matching_blocks(self, parent, parent_len):
162
for hunk in self.hunks:
163
if not isinstance(hunk, ParentText) or hunk.parent != parent:
165
yield (hunk.parent_pos, hunk.child_pos, hunk.num_lines)
166
yield parent_len, self.num_lines(), 0
168
def to_lines(self, parents=()):
169
"""Contruct a fulltext from this diff and its parents"""
170
mpvf = MultiMemoryVersionedFile()
171
for num, parent in enumerate(parents):
172
mpvf.add_version(BytesIO(parent).readlines(), num, [])
173
mpvf.add_diff(self, 'a', list(range(len(parents))))
174
return mpvf.get_line_list(['a'])[0]
177
def from_texts(cls, text, parents=()):
178
"""Produce a MultiParent from a text and list of parent text"""
179
return cls.from_lines(BytesIO(text).readlines(),
180
[BytesIO(p).readlines() for p in parents])
183
"""Yield text lines for a patch"""
184
for hunk in self.hunks:
185
for line in hunk.to_patch():
189
return len(''.join(self.to_patch()))
191
def zipped_patch_len(self):
192
return len(gzip_string(self.to_patch()))
195
def from_patch(cls, text):
196
"""Create a MultiParent from its string form"""
197
return cls._from_patch(BytesIO(text))
200
def _from_patch(lines):
201
"""This is private because it is essential to split lines on \n only"""
202
line_iter = iter(lines)
207
cur_line = next(line_iter)
208
except StopIteration:
210
if cur_line[0] == 'i':
211
num_lines = int(cur_line.split(' ')[1])
212
hunk_lines = [next(line_iter) for _ in range(num_lines)]
213
hunk_lines[-1] = hunk_lines[-1][:-1]
214
hunks.append(NewText(hunk_lines))
215
elif cur_line[0] == '\n':
216
hunks[-1].lines[-1] += '\n'
218
if not (cur_line[0] == 'c'):
219
raise AssertionError(cur_line[0])
220
parent, parent_pos, child_pos, num_lines =\
221
[int(v) for v in cur_line.split(' ')[1:]]
222
hunks.append(ParentText(parent, parent_pos, child_pos,
224
return MultiParent(hunks)
226
def range_iterator(self):
227
"""Iterate through the hunks, with range indicated
229
kind is "new" or "parent".
230
for "new", data is a list of lines.
231
for "parent", data is (parent, parent_start, parent_end)
232
:return: a generator of (start, end, kind, data)
235
for hunk in self.hunks:
236
if isinstance(hunk, NewText):
238
end = start + len(hunk.lines)
242
start = hunk.child_pos
243
end = start + hunk.num_lines
244
data = (hunk.parent, hunk.parent_pos, hunk.parent_pos +
246
yield start, end, kind, data
250
"""The number of lines in the output text"""
252
for hunk in reversed(self.hunks):
253
if isinstance(hunk, ParentText):
254
return hunk.child_pos + hunk.num_lines + extra_n
255
extra_n += len(hunk.lines)
258
def is_snapshot(self):
259
"""Return true of this hunk is effectively a fulltext"""
260
if len(self.hunks) != 1:
262
return (isinstance(self.hunks[0], NewText))
265
class NewText(object):
266
"""The contents of text that is introduced by this text"""
268
__slots__ = ['lines']
270
def __init__(self, lines):
273
def __eq__(self, other):
274
if self.__class__ is not other.__class__:
276
return (other.lines == self.lines)
279
return 'NewText(%r)' % self.lines
282
yield 'i %d\n' % len(self.lines)
283
for line in self.lines:
288
class ParentText(object):
289
"""A reference to text present in a parent text"""
291
__slots__ = ['parent', 'parent_pos', 'child_pos', 'num_lines']
293
def __init__(self, parent, parent_pos, child_pos, num_lines):
295
self.parent_pos = parent_pos
296
self.child_pos = child_pos
297
self.num_lines = num_lines
300
return dict(parent=self.parent, parent_pos=self.parent_pos,
301
child_pos=self.child_pos, num_lines=self.num_lines)
304
return ('ParentText(%(parent)r, %(parent_pos)r, %(child_pos)r,'
305
' %(num_lines)r)' % self._as_dict())
307
def __eq__(self, other):
308
if self.__class__ is not other.__class__:
310
return self._as_dict() == other._as_dict()
313
yield ('c %(parent)d %(parent_pos)d %(child_pos)d %(num_lines)d\n'
317
class BaseVersionedFile(object):
318
"""Pseudo-VersionedFile skeleton for MultiParent"""
320
def __init__(self, snapshot_interval=25, max_snapshots=None):
323
self._snapshots = set()
324
self.snapshot_interval = snapshot_interval
325
self.max_snapshots = max_snapshots
328
return iter(self._parents)
330
def has_version(self, version):
331
return version in self._parents
333
def do_snapshot(self, version_id, parent_ids):
334
"""Determine whether to perform a snapshot for this version"""
335
if self.snapshot_interval is None:
337
if self.max_snapshots is not None and\
338
len(self._snapshots) == self.max_snapshots:
340
if len(parent_ids) == 0:
342
for ignored in range(self.snapshot_interval):
343
if len(parent_ids) == 0:
345
version_ids = parent_ids
347
for version_id in version_ids:
348
if version_id not in self._snapshots:
349
parent_ids.extend(self._parents[version_id])
353
def add_version(self, lines, version_id, parent_ids,
354
force_snapshot=None, single_parent=False):
355
"""Add a version to the versionedfile
357
:param lines: The list of lines to add. Must be split on '\n'.
358
:param version_id: The version_id of the version to add
359
:param force_snapshot: If true, force this version to be added as a
360
snapshot version. If false, force this version to be added as a
361
diff. If none, determine this automatically.
362
:param single_parent: If true, use a single parent, rather than
365
if force_snapshot is None:
366
do_snapshot = self.do_snapshot(version_id, parent_ids)
368
do_snapshot = force_snapshot
370
self._snapshots.add(version_id)
371
diff = MultiParent([NewText(lines)])
374
parent_lines = self.get_line_list(parent_ids[:1])
376
parent_lines = self.get_line_list(parent_ids)
377
diff = MultiParent.from_lines(lines, parent_lines)
378
if diff.is_snapshot():
379
self._snapshots.add(version_id)
380
self.add_diff(diff, version_id, parent_ids)
381
self._lines[version_id] = lines
383
def get_parents(self, version_id):
384
return self._parents[version_id]
386
def make_snapshot(self, version_id):
387
snapdiff = MultiParent([NewText(self.cache_version(version_id))])
388
self.add_diff(snapdiff, version_id, self._parents[version_id])
389
self._snapshots.add(version_id)
391
def import_versionedfile(self, vf, snapshots, no_cache=True,
392
single_parent=False, verify=False):
393
"""Import all revisions of a versionedfile
395
:param vf: The versionedfile to import
396
:param snapshots: If provided, the revisions to make snapshots of.
397
Otherwise, this will be auto-determined
398
:param no_cache: If true, clear the cache after every add.
399
:param single_parent: If true, omit all but one parent text, (but
400
retain parent metadata).
402
if not (no_cache or not verify):
404
revisions = set(vf.versions())
405
total = len(revisions)
406
with ui.ui_factory.nested_progress_bar() as pb:
407
while len(revisions) > 0:
409
for revision in revisions:
410
parents = vf.get_parents(revision)
411
if [p for p in parents if p not in self._parents] != []:
413
lines = [a + ' ' + l for a, l in
414
vf.annotate(revision)]
415
if snapshots is None:
416
force_snapshot = None
418
force_snapshot = (revision in snapshots)
419
self.add_version(lines, revision, parents, force_snapshot,
426
if not (lines == self.get_line_list([revision])[0]):
427
raise AssertionError()
429
pb.update(gettext('Importing revisions'),
430
(total - len(revisions)) + len(added), total)
431
revisions = [r for r in revisions if r not in added]
433
def select_snapshots(self, vf):
434
"""Determine which versions to add as snapshots"""
438
for version_id in topo_iter(vf):
439
potential_build_ancestors = set(vf.get_parents(version_id))
440
parents = vf.get_parents(version_id)
441
if len(parents) == 0:
442
snapshots.add(version_id)
443
build_ancestors[version_id] = set()
445
for parent in vf.get_parents(version_id):
446
potential_build_ancestors.update(build_ancestors[parent])
447
if len(potential_build_ancestors) > self.snapshot_interval:
448
snapshots.add(version_id)
449
build_ancestors[version_id] = set()
451
build_ancestors[version_id] = potential_build_ancestors
454
def select_by_size(self, num):
455
"""Select snapshots for minimum output size"""
456
num -= len(self._snapshots)
457
new_snapshots = self.get_size_ranking()[-num:]
458
return [v for n, v in new_snapshots]
460
def get_size_ranking(self):
461
"""Get versions ranked by size"""
463
new_snapshots = set()
464
for version_id in self.versions():
465
if version_id in self._snapshots:
467
diff_len = self.get_diff(version_id).patch_len()
468
snapshot_len = MultiParent([NewText(
469
self.cache_version(version_id))]).patch_len()
470
versions.append((snapshot_len - diff_len, version_id))
474
def import_diffs(self, vf):
475
"""Import the diffs from another pseudo-versionedfile"""
476
for version_id in vf.versions():
477
self.add_diff(vf.get_diff(version_id), version_id,
478
vf._parents[version_id])
480
def get_build_ranking(self):
481
"""Return revisions sorted by how much they reduce build complexity"""
484
for version_id in topo_iter(self):
485
could_avoid[version_id] = set()
486
if version_id not in self._snapshots:
487
for parent_id in self._parents[version_id]:
488
could_avoid[version_id].update(could_avoid[parent_id])
489
could_avoid[version_id].update(self._parents)
490
could_avoid[version_id].discard(version_id)
491
for avoid_id in could_avoid[version_id]:
492
referenced_by.setdefault(avoid_id, set()).add(version_id)
493
available_versions = list(self.versions())
495
while len(available_versions) > 0:
496
available_versions.sort(key=lambda x:
497
len(could_avoid[x]) *
498
len(referenced_by.get(x, [])))
499
selected = available_versions.pop()
500
ranking.append(selected)
501
for version_id in referenced_by[selected]:
502
could_avoid[version_id].difference_update(
503
could_avoid[selected])
504
for version_id in could_avoid[selected]:
505
referenced_by[version_id].difference_update(
506
referenced_by[selected]
510
def clear_cache(self):
513
def get_line_list(self, version_ids):
514
return [self.cache_version(v) for v in version_ids]
516
def cache_version(self, version_id):
518
return self._lines[version_id]
521
diff = self.get_diff(version_id)
523
reconstructor = _Reconstructor(self, self._lines, self._parents)
524
reconstructor.reconstruct_version(lines, version_id)
525
self._lines[version_id] = lines
529
class MultiMemoryVersionedFile(BaseVersionedFile):
530
"""Memory-backed pseudo-versionedfile"""
532
def __init__(self, snapshot_interval=25, max_snapshots=None):
533
BaseVersionedFile.__init__(self, snapshot_interval, max_snapshots)
536
def add_diff(self, diff, version_id, parent_ids):
537
self._diffs[version_id] = diff
538
self._parents[version_id] = parent_ids
540
def get_diff(self, version_id):
542
return self._diffs[version_id]
544
raise errors.RevisionNotPresent(version_id, self)
550
class MultiVersionedFile(BaseVersionedFile):
551
"""Disk-backed pseudo-versionedfile"""
553
def __init__(self, filename, snapshot_interval=25, max_snapshots=None):
554
BaseVersionedFile.__init__(self, snapshot_interval, max_snapshots)
555
self._filename = filename
556
self._diff_offset = {}
558
def get_diff(self, version_id):
559
start, count = self._diff_offset[version_id]
560
infile = open(self._filename + '.mpknit', 'rb')
563
sio = BytesIO(infile.read(count))
566
zip_file = gzip.GzipFile(None, mode='rb', fileobj=sio)
568
file_version_id = zip_file.readline()
569
content = zip_file.read()
570
return MultiParent.from_patch(content)
574
def add_diff(self, diff, version_id, parent_ids):
575
outfile = open(self._filename + '.mpknit', 'ab')
577
outfile.seek(0, 2) # workaround for windows bug:
578
# .tell() for files opened in 'ab' mode
579
# before any write returns 0
580
start = outfile.tell()
582
zipfile = gzip.GzipFile(None, mode='ab', fileobj=outfile)
583
zipfile.writelines(itertools.chain(
584
['version %s\n' % version_id], diff.to_patch()))
590
self._diff_offset[version_id] = (start, end-start)
591
self._parents[version_id] = parent_ids
595
os.unlink(self._filename + '.mpknit')
597
if e.errno != errno.ENOENT:
600
os.unlink(self._filename + '.mpidx')
602
if e.errno != errno.ENOENT:
606
open(self._filename + '.mpidx', 'wb').write(bencode.bencode(
607
(self._parents, list(self._snapshots), self._diff_offset)))
610
self._parents, snapshots, self._diff_offset = bencode.bdecode(
611
open(self._filename + '.mpidx', 'rb').read())
612
self._snapshots = set(snapshots)
615
class _Reconstructor(object):
616
"""Build a text from the diffs, ancestry graph and cached lines"""
618
def __init__(self, diffs, lines, parents):
621
self.parents = parents
624
def reconstruct(self, lines, parent_text, version_id):
625
"""Append the lines referred to by a ParentText to lines"""
626
parent_id = self.parents[version_id][parent_text.parent]
627
end = parent_text.parent_pos + parent_text.num_lines
628
return self._reconstruct(lines, parent_id, parent_text.parent_pos,
631
def _reconstruct(self, lines, req_version_id, req_start, req_end):
632
"""Append lines for the requested version_id range"""
633
# stack of pending range requests
634
if req_start == req_end:
636
pending_reqs = [(req_version_id, req_start, req_end)]
637
while len(pending_reqs) > 0:
638
req_version_id, req_start, req_end = pending_reqs.pop()
639
# lazily allocate cursors for versions
640
if req_version_id in self.lines:
641
lines.extend(self.lines[req_version_id][req_start:req_end])
644
start, end, kind, data, iterator = self.cursor[req_version_id]
646
iterator = self.diffs.get_diff(req_version_id).range_iterator()
647
start, end, kind, data = next(iterator)
648
if start > req_start:
649
iterator = self.diffs.get_diff(req_version_id).range_iterator()
650
start, end, kind, data = next(iterator)
652
# find the first hunk relevant to the request
653
while end <= req_start:
654
start, end, kind, data = next(iterator)
655
self.cursor[req_version_id] = start, end, kind, data, iterator
656
# if the hunk can't satisfy the whole request, split it in two,
657
# and leave the second half for later.
659
pending_reqs.append((req_version_id, end, req_end))
662
lines.extend(data[req_start - start: (req_end - start)])
664
# If the hunk is a ParentText, rewrite it as a range request
665
# for the parent, and make it the next pending request.
666
parent, parent_start, parent_end = data
667
new_version_id = self.parents[req_version_id][parent]
668
new_start = parent_start + req_start - start
669
new_end = parent_end + req_end - end
670
pending_reqs.append((new_version_id, new_start, new_end))
672
def reconstruct_version(self, lines, version_id):
673
length = self.diffs.get_diff(version_id).num_lines()
674
return self._reconstruct(lines, version_id, 0, length)
677
def gzip_string(lines):
679
data_file = gzip.GzipFile(None, mode='wb', fileobj=sio)
680
data_file.writelines(lines)
682
return sio.getvalue()
added
removed
breezy/multiparent.py
