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(b''.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
first_char = cur_line[0:1]
211
if first_char == b'i':
212
num_lines = int(cur_line.split(b' ')[1])
213
hunk_lines = [next(line_iter) for _ in range(num_lines)]
214
hunk_lines[-1] = hunk_lines[-1][:-1]
215
hunks.append(NewText(hunk_lines))
216
elif first_char == b'\n':
217
hunks[-1].lines[-1] += b'\n'
219
if not (first_char == b'c'):
220
raise AssertionError(first_char)
221
parent, parent_pos, child_pos, num_lines =\
222
[int(v) for v in cur_line.split(b' ')[1:]]
223
hunks.append(ParentText(parent, parent_pos, child_pos,
225
return MultiParent(hunks)
227
def range_iterator(self):
228
"""Iterate through the hunks, with range indicated
230
kind is "new" or "parent".
231
for "new", data is a list of lines.
232
for "parent", data is (parent, parent_start, parent_end)
233
:return: a generator of (start, end, kind, data)
236
for hunk in self.hunks:
237
if isinstance(hunk, NewText):
239
end = start + len(hunk.lines)
243
start = hunk.child_pos
244
end = start + hunk.num_lines
245
data = (hunk.parent, hunk.parent_pos, hunk.parent_pos +
247
yield start, end, kind, data
251
"""The number of lines in the output text"""
253
for hunk in reversed(self.hunks):
254
if isinstance(hunk, ParentText):
255
return hunk.child_pos + hunk.num_lines + extra_n
256
extra_n += len(hunk.lines)
259
def is_snapshot(self):
260
"""Return true of this hunk is effectively a fulltext"""
261
if len(self.hunks) != 1:
263
return (isinstance(self.hunks[0], NewText))
266
class NewText(object):
267
"""The contents of text that is introduced by this text"""
269
__slots__ = ['lines']
271
def __init__(self, lines):
274
def __eq__(self, other):
275
if self.__class__ is not other.__class__:
277
return (other.lines == self.lines)
280
return 'NewText(%r)' % self.lines
283
yield b'i %d\n' % len(self.lines)
284
for line in self.lines:
289
class ParentText(object):
290
"""A reference to text present in a parent text"""
292
__slots__ = ['parent', 'parent_pos', 'child_pos', 'num_lines']
294
def __init__(self, parent, parent_pos, child_pos, num_lines):
296
self.parent_pos = parent_pos
297
self.child_pos = child_pos
298
self.num_lines = num_lines
301
return {b'parent': self.parent,
302
b'parent_pos': self.parent_pos,
303
b'child_pos': self.child_pos,
304
b'num_lines': self.num_lines}
307
return ('ParentText(%(parent)r, %(parent_pos)r, %(child_pos)r,'
308
' %(num_lines)r)' % self._as_dict())
310
def __eq__(self, other):
311
if self.__class__ is not other.__class__:
313
return self._as_dict() == other._as_dict()
316
yield (b'c %(parent)d %(parent_pos)d %(child_pos)d %(num_lines)d\n'
320
class BaseVersionedFile(object):
321
"""Pseudo-VersionedFile skeleton for MultiParent"""
323
def __init__(self, snapshot_interval=25, max_snapshots=None):
326
self._snapshots = set()
327
self.snapshot_interval = snapshot_interval
328
self.max_snapshots = max_snapshots
331
return iter(self._parents)
333
def has_version(self, version):
334
return version in self._parents
336
def do_snapshot(self, version_id, parent_ids):
337
"""Determine whether to perform a snapshot for this version"""
338
if self.snapshot_interval is None:
340
if self.max_snapshots is not None and\
341
len(self._snapshots) == self.max_snapshots:
343
if len(parent_ids) == 0:
345
for ignored in range(self.snapshot_interval):
346
if len(parent_ids) == 0:
348
version_ids = parent_ids
350
for version_id in version_ids:
351
if version_id not in self._snapshots:
352
parent_ids.extend(self._parents[version_id])
356
def add_version(self, lines, version_id, parent_ids,
357
force_snapshot=None, single_parent=False):
358
"""Add a version to the versionedfile
360
:param lines: The list of lines to add. Must be split on '\n'.
361
:param version_id: The version_id of the version to add
362
:param force_snapshot: If true, force this version to be added as a
363
snapshot version. If false, force this version to be added as a
364
diff. If none, determine this automatically.
365
:param single_parent: If true, use a single parent, rather than
368
if force_snapshot is None:
369
do_snapshot = self.do_snapshot(version_id, parent_ids)
371
do_snapshot = force_snapshot
373
self._snapshots.add(version_id)
374
diff = MultiParent([NewText(lines)])
377
parent_lines = self.get_line_list(parent_ids[:1])
379
parent_lines = self.get_line_list(parent_ids)
380
diff = MultiParent.from_lines(lines, parent_lines)
381
if diff.is_snapshot():
382
self._snapshots.add(version_id)
383
self.add_diff(diff, version_id, parent_ids)
384
self._lines[version_id] = lines
386
def get_parents(self, version_id):
387
return self._parents[version_id]
389
def make_snapshot(self, version_id):
390
snapdiff = MultiParent([NewText(self.cache_version(version_id))])
391
self.add_diff(snapdiff, version_id, self._parents[version_id])
392
self._snapshots.add(version_id)
394
def import_versionedfile(self, vf, snapshots, no_cache=True,
395
single_parent=False, verify=False):
396
"""Import all revisions of a versionedfile
398
:param vf: The versionedfile to import
399
:param snapshots: If provided, the revisions to make snapshots of.
400
Otherwise, this will be auto-determined
401
:param no_cache: If true, clear the cache after every add.
402
:param single_parent: If true, omit all but one parent text, (but
403
retain parent metadata).
405
if not (no_cache or not verify):
407
revisions = set(vf.versions())
408
total = len(revisions)
409
with ui.ui_factory.nested_progress_bar() as pb:
410
while len(revisions) > 0:
412
for revision in revisions:
413
parents = vf.get_parents(revision)
414
if [p for p in parents if p not in self._parents] != []:
416
lines = [a + b' ' + l for a, l in
417
vf.annotate(revision)]
418
if snapshots is None:
419
force_snapshot = None
421
force_snapshot = (revision in snapshots)
422
self.add_version(lines, revision, parents, force_snapshot,
429
if not (lines == self.get_line_list([revision])[0]):
430
raise AssertionError()
432
pb.update(gettext('Importing revisions'),
433
(total - len(revisions)) + len(added), total)
434
revisions = [r for r in revisions if r not in added]
436
def select_snapshots(self, vf):
437
"""Determine which versions to add as snapshots"""
441
for version_id in topo_iter(vf):
442
potential_build_ancestors = set(vf.get_parents(version_id))
443
parents = vf.get_parents(version_id)
444
if len(parents) == 0:
445
snapshots.add(version_id)
446
build_ancestors[version_id] = set()
448
for parent in vf.get_parents(version_id):
449
potential_build_ancestors.update(build_ancestors[parent])
450
if len(potential_build_ancestors) > self.snapshot_interval:
451
snapshots.add(version_id)
452
build_ancestors[version_id] = set()
454
build_ancestors[version_id] = potential_build_ancestors
457
def select_by_size(self, num):
458
"""Select snapshots for minimum output size"""
459
num -= len(self._snapshots)
460
new_snapshots = self.get_size_ranking()[-num:]
461
return [v for n, v in new_snapshots]
463
def get_size_ranking(self):
464
"""Get versions ranked by size"""
466
new_snapshots = set()
467
for version_id in self.versions():
468
if version_id in self._snapshots:
470
diff_len = self.get_diff(version_id).patch_len()
471
snapshot_len = MultiParent([NewText(
472
self.cache_version(version_id))]).patch_len()
473
versions.append((snapshot_len - diff_len, version_id))
477
def import_diffs(self, vf):
478
"""Import the diffs from another pseudo-versionedfile"""
479
for version_id in vf.versions():
480
self.add_diff(vf.get_diff(version_id), version_id,
481
vf._parents[version_id])
483
def get_build_ranking(self):
484
"""Return revisions sorted by how much they reduce build complexity"""
487
for version_id in topo_iter(self):
488
could_avoid[version_id] = set()
489
if version_id not in self._snapshots:
490
for parent_id in self._parents[version_id]:
491
could_avoid[version_id].update(could_avoid[parent_id])
492
could_avoid[version_id].update(self._parents)
493
could_avoid[version_id].discard(version_id)
494
for avoid_id in could_avoid[version_id]:
495
referenced_by.setdefault(avoid_id, set()).add(version_id)
496
available_versions = list(self.versions())
498
while len(available_versions) > 0:
499
available_versions.sort(key=lambda x:
500
len(could_avoid[x]) *
501
len(referenced_by.get(x, [])))
502
selected = available_versions.pop()
503
ranking.append(selected)
504
for version_id in referenced_by[selected]:
505
could_avoid[version_id].difference_update(
506
could_avoid[selected])
507
for version_id in could_avoid[selected]:
508
referenced_by[version_id].difference_update(
509
referenced_by[selected]
513
def clear_cache(self):
516
def get_line_list(self, version_ids):
517
return [self.cache_version(v) for v in version_ids]
519
def cache_version(self, version_id):
521
return self._lines[version_id]
524
diff = self.get_diff(version_id)
526
reconstructor = _Reconstructor(self, self._lines, self._parents)
527
reconstructor.reconstruct_version(lines, version_id)
528
self._lines[version_id] = lines
532
class MultiMemoryVersionedFile(BaseVersionedFile):
533
"""Memory-backed pseudo-versionedfile"""
535
def __init__(self, snapshot_interval=25, max_snapshots=None):
536
BaseVersionedFile.__init__(self, snapshot_interval, max_snapshots)
539
def add_diff(self, diff, version_id, parent_ids):
540
self._diffs[version_id] = diff
541
self._parents[version_id] = parent_ids
543
def get_diff(self, version_id):
545
return self._diffs[version_id]
547
raise errors.RevisionNotPresent(version_id, self)
553
class MultiVersionedFile(BaseVersionedFile):
554
"""Disk-backed pseudo-versionedfile"""
556
def __init__(self, filename, snapshot_interval=25, max_snapshots=None):
557
BaseVersionedFile.__init__(self, snapshot_interval, max_snapshots)
558
self._filename = filename
559
self._diff_offset = {}
561
def get_diff(self, version_id):
562
start, count = self._diff_offset[version_id]
563
with open(self._filename + '.mpknit', 'rb') as infile:
565
sio = BytesIO(infile.read(count))
566
with gzip.GzipFile(None, mode='rb', fileobj=sio) as zip_file:
567
file_version_id = zip_file.readline()
568
content = zip_file.read()
569
return MultiParent.from_patch(content)
571
def add_diff(self, diff, version_id, parent_ids):
572
with open(self._filename + '.mpknit', 'ab') as outfile:
573
outfile.seek(0, 2) # workaround for windows bug:
574
# .tell() for files opened in 'ab' mode
575
# before any write returns 0
576
start = outfile.tell()
577
with gzip.GzipFile(None, mode='ab', fileobj=outfile) as zipfile:
578
zipfile.writelines(itertools.chain(
579
[b'version %s\n' % version_id], diff.to_patch()))
581
self._diff_offset[version_id] = (start, end-start)
582
self._parents[version_id] = parent_ids
586
os.unlink(self._filename + '.mpknit')
588
if e.errno != errno.ENOENT:
591
os.unlink(self._filename + '.mpidx')
593
if e.errno != errno.ENOENT:
597
open(self._filename + '.mpidx', 'wb').write(bencode.bencode(
598
(self._parents, list(self._snapshots), self._diff_offset)))
601
self._parents, snapshots, self._diff_offset = bencode.bdecode(
602
open(self._filename + '.mpidx', 'rb').read())
603
self._snapshots = set(snapshots)
606
class _Reconstructor(object):
607
"""Build a text from the diffs, ancestry graph and cached lines"""
609
def __init__(self, diffs, lines, parents):
612
self.parents = parents
615
def reconstruct(self, lines, parent_text, version_id):
616
"""Append the lines referred to by a ParentText to lines"""
617
parent_id = self.parents[version_id][parent_text.parent]
618
end = parent_text.parent_pos + parent_text.num_lines
619
return self._reconstruct(lines, parent_id, parent_text.parent_pos,
622
def _reconstruct(self, lines, req_version_id, req_start, req_end):
623
"""Append lines for the requested version_id range"""
624
# stack of pending range requests
625
if req_start == req_end:
627
pending_reqs = [(req_version_id, req_start, req_end)]
628
while len(pending_reqs) > 0:
629
req_version_id, req_start, req_end = pending_reqs.pop()
630
# lazily allocate cursors for versions
631
if req_version_id in self.lines:
632
lines.extend(self.lines[req_version_id][req_start:req_end])
635
start, end, kind, data, iterator = self.cursor[req_version_id]
637
iterator = self.diffs.get_diff(req_version_id).range_iterator()
638
start, end, kind, data = next(iterator)
639
if start > req_start:
640
iterator = self.diffs.get_diff(req_version_id).range_iterator()
641
start, end, kind, data = next(iterator)
643
# find the first hunk relevant to the request
644
while end <= req_start:
645
start, end, kind, data = next(iterator)
646
self.cursor[req_version_id] = start, end, kind, data, iterator
647
# if the hunk can't satisfy the whole request, split it in two,
648
# and leave the second half for later.
650
pending_reqs.append((req_version_id, end, req_end))
653
lines.extend(data[req_start - start: (req_end - start)])
655
# If the hunk is a ParentText, rewrite it as a range request
656
# for the parent, and make it the next pending request.
657
parent, parent_start, parent_end = data
658
new_version_id = self.parents[req_version_id][parent]
659
new_start = parent_start + req_start - start
660
new_end = parent_end + req_end - end
661
pending_reqs.append((new_version_id, new_start, new_end))
663
def reconstruct_version(self, lines, version_id):
664
length = self.diffs.get_diff(version_id).num_lines()
665
return self._reconstruct(lines, version_id, 0, length)
668
def gzip_string(lines):
670
with gzip.GzipFile(None, mode='wb', fileobj=sio) as data_file:
671
data_file.writelines(lines)
672
return sio.getvalue()
added
removed
breezy/multiparent.py
