1
# Copyright (C) 2007 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., 59 Temple Place, Suite 330, Boston, MA 02111-1307 USA
17
from bzrlib.lazy_import import lazy_import
19
lazy_import(globals(), """
23
from StringIO import StringIO
30
from bzrlib.util import bencode
32
from bzrlib.tuned_gzip import GzipFile
35
def topo_iter(vf, versions=None):
39
versions = vf.versions()
40
def pending_parents(version):
41
return [v for v in vf.get_parents(version) if v in versions and
43
for version_id in versions:
44
for parent_id in vf.get_parents(version_id):
45
descendants.setdefault(parent_id, []).append(version_id)
46
cur = [v for v in versions if len(pending_parents(v)) == 0]
49
for version_id in cur:
50
if version_id in seen:
52
if len(pending_parents(version_id)) != 0:
54
next.extend(descendants.get(version_id, []))
58
assert len(seen) == len(versions)
61
class MultiParent(object):
62
"""A multi-parent diff"""
64
def __init__(self, hunks=None):
71
return "MultiParent(%r)" % self.hunks
73
def __eq__(self, other):
74
if self.__class__ is not other.__class__:
76
return (self.hunks == other.hunks)
79
def from_lines(text, parents=(), left_blocks=None):
80
"""Produce a MultiParent from a list of lines and parents"""
82
matcher = patiencediff.PatienceSequenceMatcher(None, parent,
84
return matcher.get_matching_blocks()
86
if left_blocks is None:
87
left_blocks = compare(parents[0])
88
parent_comparisons = [left_blocks] + [compare(p) for p in
91
parent_comparisons = []
93
new_text = NewText([])
95
block_iter = [iter(i) for i in parent_comparisons]
96
diff = MultiParent([])
99
return block_iter[p].next()
100
except StopIteration:
102
cur_block = [next_block(p) for p, i in enumerate(block_iter)]
103
while cur_line < len(text):
105
for p, block in enumerate(cur_block):
109
while j + n < cur_line:
110
block = cur_block[p] = next_block(p)
118
offset = cur_line - j
124
if best_match is None or n > best_match.num_lines:
125
best_match = ParentText(p, i, j, n)
126
if best_match is None:
127
new_text.lines.append(text[cur_line])
130
if len(new_text.lines) > 0:
131
diff.hunks.append(new_text)
132
new_text = NewText([])
133
diff.hunks.append(best_match)
134
cur_line += best_match.num_lines
135
if len(new_text.lines) > 0:
136
diff.hunks.append(new_text)
139
def to_lines(self, parents=()):
140
"""Contruct a fulltext from this diff and its parents"""
141
mpvf = MultiMemoryVersionedFile()
142
for num, parent in enumerate(parents):
143
mpvf.add_version(StringIO(parent).readlines(), num, [])
144
mpvf.add_diff(self, 'a', range(len(parents)))
145
return mpvf.get_line_list(['a'])[0]
148
def from_texts(cls, text, parents=()):
149
"""Produce a MultiParent from a text and list of parent text"""
150
return cls.from_lines(StringIO(text).readlines(),
151
[StringIO(p).readlines() for p in parents])
154
"""Yield text lines for a patch"""
155
for hunk in self.hunks:
156
for line in hunk.to_patch():
160
return len(''.join(self.to_patch()))
162
def zipped_patch_len(self):
163
return len(gzip_string(self.to_patch()))
166
def from_patch(cls, text):
167
"""Create a MultiParent from its string form"""
168
return cls._from_patch(StringIO(text))
171
def _from_patch(lines):
172
"""This is private because it is essential to split lines on \n only"""
173
line_iter = iter(lines)
178
cur_line = line_iter.next()
179
except StopIteration:
181
if cur_line[0] == 'i':
182
num_lines = int(cur_line.split(' ')[1])
183
hunk_lines = [line_iter.next() for x in xrange(num_lines)]
184
hunk_lines[-1] = hunk_lines[-1][:-1]
185
hunks.append(NewText(hunk_lines))
186
elif cur_line[0] == '\n':
187
hunks[-1].lines[-1] += '\n'
189
assert cur_line[0] == 'c', cur_line[0]
190
parent, parent_pos, child_pos, num_lines =\
191
[int(v) for v in cur_line.split(' ')[1:]]
192
hunks.append(ParentText(parent, parent_pos, child_pos,
194
return MultiParent(hunks)
196
def range_iterator(self):
197
"""Iterate through the hunks, with range indicated
199
kind is "new" or "parent".
200
for "new", data is a list of lines.
201
for "parent", data is (parent, parent_start, parent_end)
202
:return: a generator of (start, end, kind, data)
205
for hunk in self.hunks:
206
if isinstance(hunk, NewText):
208
end = start + len(hunk.lines)
212
start = hunk.child_pos
213
end = start + hunk.num_lines
214
data = (hunk.parent, hunk.parent_pos, hunk.parent_pos +
216
yield start, end, kind, data
220
"""The number of lines in the output text"""
222
for hunk in reversed(self.hunks):
223
if isinstance(hunk, ParentText):
224
return hunk.child_pos + hunk.num_lines + extra_n
225
extra_n += len(hunk.lines)
228
def is_snapshot(self):
229
"""Return true of this hunk is effectively a fulltext"""
230
if len(self.hunks) != 1:
232
return (isinstance(self.hunks[0], NewText))
235
class NewText(object):
236
"""The contents of text that is introduced by this text"""
238
def __init__(self, lines):
241
def __eq__(self, other):
242
if self.__class__ is not other.__class__:
244
return (other.lines == self.lines)
247
return 'NewText(%r)' % self.lines
250
yield 'i %d\n' % len(self.lines)
251
for line in self.lines:
256
class ParentText(object):
257
"""A reference to text present in a parent text"""
259
def __init__(self, parent, parent_pos, child_pos, num_lines):
261
self.parent_pos = parent_pos
262
self.child_pos = child_pos
263
self.num_lines = num_lines
266
return 'ParentText(%(parent)r, %(parent_pos)r, %(child_pos)r,'\
267
' %(num_lines)r)' % self.__dict__
269
def __eq__(self, other):
270
if self.__class__ != other.__class__:
272
return (self.__dict__ == other.__dict__)
275
yield 'c %(parent)d %(parent_pos)d %(child_pos)d %(num_lines)d\n'\
279
class BaseVersionedFile(object):
280
"""Pseudo-VersionedFile skeleton for MultiParent"""
282
def __init__(self, snapshot_interval=25, max_snapshots=None):
285
self._snapshots = set()
286
self.snapshot_interval = snapshot_interval
287
self.max_snapshots = max_snapshots
290
return iter(self._parents)
292
def has_version(self, version):
293
return version in self._parents
295
def do_snapshot(self, version_id, parent_ids):
296
"""Determine whether to perform a a snapshot for this version"""
297
if self.snapshot_interval is None:
299
if self.max_snapshots is not None and\
300
len(self._snapshots) == self.max_snapshots:
302
if len(parent_ids) == 0:
304
for ignored in xrange(self.snapshot_interval):
305
if len(parent_ids) == 0:
307
version_ids = parent_ids
309
for version_id in version_ids:
310
if version_id not in self._snapshots:
311
parent_ids.extend(self._parents[version_id])
315
def add_version(self, lines, version_id, parent_ids,
316
force_snapshot=None, single_parent=False):
317
"""Add a version to the versionedfile
319
:param lines: The list of lines to add. Must be split on '\n'.
320
:param version_id: The version_id of the version to add
321
:param force_snapshot: If true, force this version to be added as a
322
snapshot version. If false, force this version to be added as a
323
diff. If none, determine this automatically.
324
:param single_parent: If true, use a single parent, rather than
327
if force_snapshot is None:
328
do_snapshot = self.do_snapshot(version_id, parent_ids)
330
do_snapshot = force_snapshot
332
self._snapshots.add(version_id)
333
diff = MultiParent([NewText(lines)])
336
parent_lines = self.get_line_list(parent_ids[:1])
338
parent_lines = self.get_line_list(parent_ids)
339
diff = MultiParent.from_lines(lines, parent_lines)
340
if diff.is_snapshot():
341
self._snapshots.add(version_id)
342
self.add_diff(diff, version_id, parent_ids)
343
self._lines[version_id] = lines
345
def get_parents(self, version_id):
346
return self._parents[version_id]
348
def make_snapshot(self, version_id):
349
snapdiff = MultiParent([NewText(self.cache_version(version_id))])
350
self.add_diff(snapdiff, version_id, self._parents[version_id])
351
self._snapshots.add(version_id)
353
def import_versionedfile(self, vf, snapshots, no_cache=True,
354
single_parent=False, verify=False):
355
"""Import all revisions of a versionedfile
357
:param vf: The versionedfile to import
358
:param snapshots: If provided, the revisions to make snapshots of.
359
Otherwise, this will be auto-determined
360
:param no_cache: If true, clear the cache after every add.
361
:param single_parent: If true, omit all but one parent text, (but
362
retain parent metadata).
364
assert no_cache or not verify
365
revisions = set(vf.versions())
366
total = len(revisions)
367
pb = ui.ui_factory.nested_progress_bar()
369
while len(revisions) > 0:
371
for revision in revisions:
372
parents = vf.get_parents(revision)
373
if [p for p in parents if p not in self._parents] != []:
375
lines = [a + ' ' + l for a, l in
376
vf.annotate_iter(revision)]
377
if snapshots is None:
378
force_snapshot = None
380
force_snapshot = (revision in snapshots)
381
self.add_version(lines, revision, parents, force_snapshot,
388
assert lines == self.get_line_list([revision])[0]
390
pb.update('Importing revisions',
391
(total - len(revisions)) + len(added), total)
392
revisions = [r for r in revisions if r not in added]
396
def select_snapshots(self, vf):
397
"""Determine which versions to add as snapshots"""
401
for version_id in topo_iter(vf):
402
potential_build_ancestors = set(vf.get_parents(version_id))
403
parents = vf.get_parents(version_id)
404
if len(parents) == 0:
405
snapshots.add(version_id)
406
build_ancestors[version_id] = set()
408
for parent in vf.get_parents(version_id):
409
potential_build_ancestors.update(build_ancestors[parent])
410
if len(potential_build_ancestors) > self.snapshot_interval:
411
snapshots.add(version_id)
412
build_ancestors[version_id] = set()
414
build_ancestors[version_id] = potential_build_ancestors
417
def select_by_size(self, num):
418
"""Select snapshots for minimum output size"""
419
num -= len(self._snapshots)
420
new_snapshots = self.get_size_ranking()[-num:]
421
return [v for n, v in new_snapshots]
423
def get_size_ranking(self):
424
"""Get versions ranked by size"""
426
new_snapshots = set()
427
for version_id in self.versions():
428
if version_id in self._snapshots:
430
diff_len = self.get_diff(version_id).patch_len()
431
snapshot_len = MultiParent([NewText(
432
self.cache_version(version_id))]).patch_len()
433
versions.append((snapshot_len - diff_len, version_id))
437
def import_diffs(self, vf):
438
"""Import the diffs from another pseudo-versionedfile"""
439
for version_id in vf.versions():
440
self.add_diff(vf.get_diff(version_id), version_id,
441
vf._parents[version_id])
443
def get_build_ranking(self):
444
"""Return revisions sorted by how much they reduce build complexity"""
447
for version_id in topo_iter(self):
448
could_avoid[version_id] = set()
449
if version_id not in self._snapshots:
450
for parent_id in self._parents[version_id]:
451
could_avoid[version_id].update(could_avoid[parent_id])
452
could_avoid[version_id].update(self._parents)
453
could_avoid[version_id].discard(version_id)
454
for avoid_id in could_avoid[version_id]:
455
referenced_by.setdefault(avoid_id, set()).add(version_id)
456
available_versions = list(self.versions())
458
while len(available_versions) > 0:
459
available_versions.sort(key=lambda x:
460
len(could_avoid[x]) *
461
len(referenced_by.get(x, [])))
462
selected = available_versions.pop()
463
ranking.append(selected)
464
for version_id in referenced_by[selected]:
465
could_avoid[version_id].difference_update(
466
could_avoid[selected])
467
for version_id in could_avoid[selected]:
468
referenced_by[version_id].difference_update(
469
referenced_by[selected]
473
def clear_cache(self):
476
def get_line_list(self, version_ids):
477
return [self.cache_version(v) for v in version_ids]
479
def cache_version(self, version_id):
481
return self._lines[version_id]
484
diff = self.get_diff(version_id)
486
reconstructor = _Reconstructor(self, self._lines,
488
reconstructor.reconstruct_version(lines, version_id)
489
self._lines[version_id] = lines
493
class MultiMemoryVersionedFile(BaseVersionedFile):
494
"""Memory-backed pseudo-versionedfile"""
496
def __init__(self, snapshot_interval=25, max_snapshots=None):
497
BaseVersionedFile.__init__(self, snapshot_interval, max_snapshots)
500
def add_diff(self, diff, version_id, parent_ids):
501
self._diffs[version_id] = diff
502
self._parents[version_id] = parent_ids
504
def get_diff(self, version_id):
505
return self._diffs[version_id]
511
class MultiVersionedFile(BaseVersionedFile):
512
"""Disk-backed pseudo-versionedfile"""
514
def __init__(self, filename, snapshot_interval=25, max_snapshots=None):
515
BaseVersionedFile.__init__(self, snapshot_interval, max_snapshots)
516
self._filename = filename
517
self._diff_offset = {}
519
def get_diff(self, version_id):
520
start, count = self._diff_offset[version_id]
521
infile = open(self._filename + '.mpknit', 'rb')
524
sio = StringIO(infile.read(count))
527
zip_file = GzipFile(None, mode='rb', fileobj=sio)
529
file_version_id = zip_file.readline()
530
return MultiParent.from_patch(zip_file.read())
534
def add_diff(self, diff, version_id, parent_ids):
535
outfile = open(self._filename + '.mpknit', 'ab')
537
start = outfile.tell()
539
zipfile = GzipFile(None, mode='ab', fileobj=outfile)
540
zipfile.writelines(itertools.chain(
541
['version %s\n' % version_id], diff.to_patch()))
547
self._diff_offset[version_id] = (start, end-start)
548
self._parents[version_id] = parent_ids
552
os.unlink(self._filename + '.mpknit')
554
if e.errno != errno.ENOENT:
557
os.unlink(self._filename + '.mpidx')
559
if e.errno != errno.ENOENT:
563
open(self._filename + '.mpidx', 'wb').write(bencode.bencode(
564
(self._parents, list(self._snapshots), self._diff_offset)))
567
self._parents, snapshots, self._diff_offset = bencode.bdecode(
568
open(self._filename + '.mpidx', 'rb').read())
569
self._snapshots = set(snapshots)
572
class _Reconstructor(object):
573
"""Build a text from the diffs, ancestry graph and cached lines"""
575
def __init__(self, diffs, lines, parents):
578
self.parents = parents
581
def reconstruct(self, lines, parent_text, version_id):
582
"""Append the lines referred to by a ParentText to lines"""
583
parent_id = self.parents[version_id][parent_text.parent]
584
end = parent_text.parent_pos + parent_text.num_lines
585
return self._reconstruct(lines, parent_id, parent_text.parent_pos,
588
def _reconstruct(self, lines, req_version_id, req_start, req_end):
589
"""Append lines for the requested version_id range"""
590
# stack of pending range requests
591
if req_start == req_end:
593
pending_reqs = [(req_version_id, req_start, req_end)]
594
while len(pending_reqs) > 0:
595
req_version_id, req_start, req_end = pending_reqs.pop()
596
# lazily allocate cursors for versions
598
start, end, kind, data, iterator = self.cursor[req_version_id]
600
iterator = self.diffs.get_diff(req_version_id).range_iterator()
601
start, end, kind, data = iterator.next()
602
if start > req_start:
603
iterator = self.diffs.get_diff(req_version_id).range_iterator()
604
start, end, kind, data = iterator.next()
606
# find the first hunk relevant to the request
607
while end <= req_start:
608
start, end, kind, data = iterator.next()
609
self.cursor[req_version_id] = start, end, kind, data, iterator
610
# if the hunk can't satisfy the whole request, split it in two,
611
# and leave the second half for later.
613
pending_reqs.append((req_version_id, end, req_end))
616
lines.extend(data[req_start - start: (req_end - start)])
618
# If the hunk is a ParentText, rewrite it as a range request
619
# for the parent, and make it the next pending request.
620
parent, parent_start, parent_end = data
621
new_version_id = self.parents[req_version_id][parent]
622
new_start = parent_start + req_start - start
623
new_end = parent_end + req_end - end
624
pending_reqs.append((new_version_id, new_start, new_end))
626
def reconstruct_version(self, lines, version_id):
627
length = self.diffs.get_diff(version_id).num_lines()
628
return self._reconstruct(lines, version_id, 0, length)
631
def gzip_string(lines):
633
data_file = GzipFile(None, mode='wb', fileobj=sio)
634
data_file.writelines(lines)
636
return sio.getvalue()
added
removed
bzrlib/multiparent.py
