« back to all changes in this revision
Viewing changes to bzrlib/weave.py
- browse files at revision 895
- compare with another revision
- download diff
- download tarball
- view history from revision 895
- Committer: Martin Pool
- Date: 2005-07-11 05:46:55 UTC
- Revision ID: mbp@sourcefrog.net-20050711054655-47ea25e1f75a8813
- ignore tmp dir
- files added:
- bzr-man.py
- bzrlib/selftest
- effbot
- effbot/org
- patches
- plugins
- plugins/changeset
- tools
- files removed:
- bzrlib/cache.py
- files renamed:
- contrib/bash/bzr => contrib/bash/bzr.simple
- files modified:
- .bzrignore
added
removed
bzrlib/weave.py
1
#! /usr/bin/python
2
3
# Copyright (C) 2005 Canonical Ltd
4
5
# This program is free software; you can redistribute it and/or modify
6
# it under the terms of the GNU General Public License as published by
7
# the Free Software Foundation; either version 2 of the License, or
8
# (at your option) any later version.
9
10
# This program is distributed in the hope that it will be useful,
11
# but WITHOUT ANY WARRANTY; without even the implied warranty of
12
# MERCHANTABILITY or FITNESS FOR A PARTICULAR PURPOSE. See the
13
# GNU General Public License for more details.
14
15
# You should have received a copy of the GNU General Public License
16
# along with this program; if not, write to the Free Software
17
# Foundation, Inc., 59 Temple Place, Suite 330, Boston, MA 02111-1307 USA
18
19
# Author: Martin Pool <mbp@canonical.com>
20
21
22
"""Weave - storage of related text file versions"""
23
24
# TODO: Perhaps have copy method for Weave instances?
25
26
# XXX: If we do weaves this way, will a merge still behave the same
27
# way if it's done in a different order? That's a pretty desirable
28
# property.
29
30
# TODO: How to write these to disk? One option is cPickle, which
31
# would be fast but less friendly to C, and perhaps not portable. Another is
32
33
# TODO: Nothing here so far assumes the lines are really \n newlines,
34
# rather than being split up in some other way. We could accomodate
35
# binaries, perhaps by naively splitting on \n or perhaps using
36
# something like a rolling checksum.
37
38
# TODO: Perhaps track SHA-1 in the header for protection? This would
39
# be redundant with it being stored in the inventory, but perhaps
40
# usefully so?
41
42
# TODO: Track version names as well as indexes.
43
44
# TODO: Probably do transitive expansion when specifying parents?
45
46
# TODO: Separate out some code to read and write weaves.
47
48
# TODO: End marker for each version so we can stop reading?
49
50
# TODO: Check that no insertion occurs inside a deletion that was
51
# active in the version of the insertion.
52
53
# TODO: Perhaps a special slower check() method that verifies more
54
# nesting constraints and the MD5 of each version?
55
56
57
58
try:
59
set
60
frozenset
61
except NameError:
62
from sets import Set, ImmutableSet
63
set = Set
64
frozenset = ImmutableSet
65
del Set, ImmutableSet
66
67
68
class WeaveError(Exception):
69
"""Exception in processing weave"""
70
71
72
class WeaveFormatError(WeaveError):
73
"""Weave invariant violated"""
74
75
76
class Weave(object):
77
"""weave - versioned text file storage.
78
79
A Weave manages versions of line-based text files, keeping track
80
of the originating version for each line.
81
82
To clients the "lines" of the file are represented as a list of strings.
83
These strings will typically have terminal newline characters, but
84
this is not required. In particular files commonly do not have a newline
85
at the end of the file.
86
87
Texts can be identified in either of two ways:
88
89
* a nonnegative index number.
90
91
* a version-id string.
92
93
Typically the index number will be valid only inside this weave and
94
the version-id is used to reference it in the larger world.
95
96
The weave is represented as a list mixing edit instructions and
97
literal text. Each entry in _l can be either a string (or
98
unicode), or a tuple. If a string, it means that the given line
99
should be output in the currently active revisions.
100
101
If a tuple, it gives a processing instruction saying in which
102
revisions the enclosed lines are active. The tuple has the form
103
(instruction, version).
104
105
The instruction can be '{' or '}' for an insertion block, and '['
106
and ']' for a deletion block respectively. The version is the
107
integer version index. There is no replace operator, only deletes
108
and inserts.
109
110
Constraints/notes:
111
112
* A later version can delete lines that were introduced by any
113
number of ancestor versions; this implies that deletion
114
instructions can span insertion blocks without regard to the
115
insertion block's nesting.
116
117
* Similarly, deletions need not be properly nested with regard to
118
each other, because they might have been generated by
119
independent revisions.
120
121
* Insertions are always made by inserting a new bracketed block
122
into a single point in the previous weave. This implies they
123
can nest but not overlap, and the nesting must always have later
124
insertions on the inside.
125
126
* It doesn't seem very useful to have an active insertion
127
inside an inactive insertion, but it might happen.
128
129
* Therefore, all instructions are always"considered"; that
130
is passed onto and off the stack. An outer inactive block
131
doesn't disable an inner block.
132
133
* Lines are enabled if the most recent enclosing insertion is
134
active and none of the enclosing deletions are active.
135
136
* There is no point having a deletion directly inside its own
137
insertion; you might as well just not write it. And there
138
should be no way to get an earlier version deleting a later
139
version.
140
141
_l
142
Text of the weave.
143
144
_v
145
List of parents, indexed by version number.
146
It is only necessary to store the minimal set of parents for
147
each version; the parent's parents are implied.
148
149
_sha1s
150
List of hex SHA-1 of each version, or None if not recorded.
151
"""
152
def __init__(self):
153
self._l = []
154
self._v = []
155
self._sha1s = []
156
157
158
def __eq__(self, other):
159
if not isinstance(other, Weave):
160
return False
161
return self._v == other._v \
162
and self._l == other._l
163
164
165
def __ne__(self, other):
166
return not self.__eq__(other)
167
168
169
def add(self, parents, text):
170
"""Add a single text on top of the weave.
171
172
Returns the index number of the newly added version.
173
174
parents
175
List or set of direct parent version numbers.
176
177
text
178
Sequence of lines to be added in the new version."""
179
## self._check_versions(parents)
180
## self._check_lines(text)
181
idx = len(self._v)
182
183
import sha
184
s = sha.new()
185
for l in text:
186
s.update(l)
187
sha1 = s.hexdigest()
188
del s
189
190
if parents:
191
ancestors = self.inclusions(parents)
192
delta = self._delta(ancestors, text)
193
194
# offset gives the number of lines that have been inserted
195
# into the weave up to the current point; if the original edit instruction
196
# says to change line A then we actually change (A+offset)
197
offset = 0
198
199
for i1, i2, newlines in delta:
200
assert 0 <= i1
201
assert i1 <= i2
202
assert i2 <= len(self._l)
203
204
# the deletion and insertion are handled separately.
205
# first delete the region.
206
if i1 != i2:
207
self._l.insert(i1+offset, ('[', idx))
208
self._l.insert(i2+offset+1, (']', idx))
209
offset += 2
210
# is this OK???
211
212
if newlines:
213
# there may have been a deletion spanning up to
214
# i2; we want to insert after this region to make sure
215
# we don't destroy ourselves
216
i = i2 + offset
217
self._l[i:i] = [('{', idx)] \
218
+ newlines \
219
+ [('}', idx)]
220
offset += 2 + len(newlines)
221
222
self._addversion(parents)
223
else:
224
# special case; adding with no parents revision; can do this
225
# more quickly by just appending unconditionally
226
self._l.append(('{', idx))
227
self._l += text
228
self._l.append(('}', idx))
229
230
self._addversion(None)
231
232
self._sha1s.append(sha1)
233
234
return idx
235
236
237
def inclusions(self, versions):
238
"""Return set of all ancestors of given version(s)."""
239
i = set(versions)
240
v = max(versions)
241
try:
242
while v >= 0:
243
if v in i:
244
# include all its parents
245
i.update(self._v[v])
246
v -= 1
247
return i
248
except IndexError:
249
raise ValueError("version %d not present in weave" % v)
250
251
252
def minimal_parents(self, version):
253
"""Find the minimal set of parents for the version."""
254
included = self._v[version]
255
if not included:
256
return []
257
258
li = list(included)
259
li.sort(reverse=True)
260
261
mininc = []
262
gotit = set()
263
264
for pv in li:
265
if pv not in gotit:
266
mininc.append(pv)
267
gotit.update(self.inclusions(pv))
268
269
assert mininc[0] >= 0
270
assert mininc[-1] < version
271
return mininc
272
273
274
def _addversion(self, parents):
275
if parents:
276
self._v.append(parents)
277
else:
278
self._v.append(frozenset())
279
280
281
def _check_lines(self, text):
282
if not isinstance(text, list):
283
raise ValueError("text should be a list, not %s" % type(text))
284
285
for l in text:
286
if not isinstance(l, basestring):
287
raise ValueError("text line should be a string or unicode, not %s"
288
% type(l))
289
290
291
292
def _check_versions(self, indexes):
293
"""Check everything in the sequence of indexes is valid"""
294
for i in indexes:
295
try:
296
self._v[i]
297
except IndexError:
298
raise IndexError("invalid version number %r" % i)
299
300
301
def annotate(self, index):
302
return list(self.annotate_iter(index))
303
304
305
def annotate_iter(self, version):
306
"""Yield list of (index-id, line) pairs for the specified version.
307
308
The index indicates when the line originated in the weave."""
309
for origin, lineno, text in self._extract([version]):
310
yield origin, text
311
312
313
def _extract(self, versions):
314
"""Yield annotation of lines in included set.
315
316
Yields a sequence of tuples (origin, lineno, text), where
317
origin is the origin version, lineno the index in the weave,
318
and text the text of the line.
319
320
The set typically but not necessarily corresponds to a version.
321
"""
322
included = self.inclusions(versions)
323
324
istack = []
325
dset = set()
326
327
lineno = 0 # line of weave, 0-based
328
329
isactive = None
330
331
WFE = WeaveFormatError
332
333
for l in self._l:
334
if isinstance(l, tuple):
335
c, v = l
336
isactive = None
337
if c == '{':
338
assert v not in istack
339
istack.append(v)
340
elif c == '}':
341
oldv = istack.pop()
342
assert oldv == v
343
elif c == '[':
344
if v in included:
345
assert v not in dset
346
dset.add(v)
347
else:
348
assert c == ']'
349
if v in included:
350
assert v in dset
351
dset.remove(v)
352
else:
353
assert isinstance(l, basestring)
354
if isactive is None:
355
isactive = (not dset) and istack and (istack[-1] in included)
356
if isactive:
357
yield istack[-1], lineno, l
358
lineno += 1
359
360
if istack:
361
raise WFE("unclosed insertion blocks at end of weave",
362
istack)
363
if dset:
364
raise WFE("unclosed deletion blocks at end of weave",
365
dset)
366
367
368
def get_iter(self, version):
369
"""Yield lines for the specified version."""
370
for origin, lineno, line in self._extract([version]):
371
yield line
372
373
374
def get(self, index):
375
return list(self.get_iter(index))
376
377
378
def mash_iter(self, included):
379
"""Return composed version of multiple included versions."""
380
included = frozenset(included)
381
for origin, lineno, text in self._extract(included):
382
yield text
383
384
385
def dump(self, to_file):
386
from pprint import pprint
387
print >>to_file, "Weave._l = ",
388
pprint(self._l, to_file)
389
print >>to_file, "Weave._v = ",
390
pprint(self._v, to_file)
391
392
393
394
def numversions(self):
395
l = len(self._v)
396
assert l == len(self._sha1s)
397
return l
398
399
400
def check(self, progress_bar=None):
401
# check no circular inclusions
402
for version in range(self.numversions()):
403
inclusions = list(self._v[version])
404
if inclusions:
405
inclusions.sort()
406
if inclusions[-1] >= version:
407
raise WeaveFormatError("invalid included version %d for index %d"
408
% (inclusions[-1], version))
409
410
# try extracting all versions; this is a bit slow and parallel
411
# extraction could be used
412
import sha
413
nv = self.numversions()
414
for version in range(nv):
415
if progress_bar:
416
progress_bar.update('checking text', version, nv)
417
s = sha.new()
418
for l in self.get_iter(version):
419
s.update(l)
420
hd = s.hexdigest()
421
expected = self._sha1s[version]
422
if hd != expected:
423
raise WeaveError("mismatched sha1 for version %d; "
424
"got %s, expected %s"
425
% (version, hd, expected))
426
427
# TODO: check insertions are properly nested, that there are
428
# no lines outside of insertion blocks, that deletions are
429
# properly paired, etc.
430
431
432
433
def merge(self, merge_versions):
434
"""Automerge and mark conflicts between versions.
435
436
This returns a sequence, each entry describing alternatives
437
for a chunk of the file. Each of the alternatives is given as
438
a list of lines.
439
440
If there is a chunk of the file where there's no diagreement,
441
only one alternative is given.
442
"""
443
444
# approach: find the included versions common to all the
445
# merged versions
446
raise NotImplementedError()
447
448
449
450
def _delta(self, included, lines):
451
"""Return changes from basis to new revision.
452
453
The old text for comparison is the union of included revisions.
454
455
This is used in inserting a new text.
456
457
Delta is returned as a sequence of
458
(weave1, weave2, newlines).
459
460
This indicates that weave1:weave2 of the old weave should be
461
replaced by the sequence of lines in newlines. Note that
462
these line numbers are positions in the total weave and don't
463
correspond to the lines in any extracted version, or even the
464
extracted union of included versions.
465
466
If line1=line2, this is a pure insert; if newlines=[] this is a
467
pure delete. (Similar to difflib.)
468
"""
469
# basis a list of (origin, lineno, line)
470
basis_lineno = []
471
basis_lines = []
472
for origin, lineno, line in self._extract(included):
473
basis_lineno.append(lineno)
474
basis_lines.append(line)
475
476
# add a sentinal, because we can also match against the final line
477
basis_lineno.append(len(self._l))
478
479
# XXX: which line of the weave should we really consider
480
# matches the end of the file? the current code says it's the
481
# last line of the weave?
482
483
from difflib import SequenceMatcher
484
s = SequenceMatcher(None, basis_lines, lines)
485
486
# TODO: Perhaps return line numbers from composed weave as well?
487
488
for tag, i1, i2, j1, j2 in s.get_opcodes():
489
##print tag, i1, i2, j1, j2
490
491
if tag == 'equal':
492
continue
493
494
# i1,i2 are given in offsets within basis_lines; we need to map them
495
# back to offsets within the entire weave
496
real_i1 = basis_lineno[i1]
497
real_i2 = basis_lineno[i2]
498
499
assert 0 <= j1
500
assert j1 <= j2
501
assert j2 <= len(lines)
502
503
yield real_i1, real_i2, lines[j1:j2]
504
505
506
507
def weave_info(filename, out):
508
"""Show some text information about the weave."""
509
from weavefile import read_weave
510
wf = file(filename, 'rb')
511
w = read_weave(wf)
512
# FIXME: doesn't work on pipes
513
weave_size = wf.tell()
514
print >>out, "weave file size %d bytes" % weave_size
515
print >>out, "weave contains %d versions" % len(w._v)
516
517
total = 0
518
print '%6s %6s %8s %40s %20s' % ('ver', 'lines', 'bytes', 'sha1', 'parents')
519
for i in (6, 6, 8, 40, 20):
520
print '-' * i,
521
print
522
for i in range(len(w._v)):
523
text = w.get(i)
524
lines = len(text)
525
bytes = sum((len(a) for a in text))
526
sha1 = w._sha1s[i]
527
print '%6d %6d %8d %40s' % (i, lines, bytes, sha1),
528
for pv in w._v[i]:
529
print pv,
530
print
531
total += bytes
532
533
print >>out, "versions total %d bytes" % total
534
print >>out, "compression ratio %.3f" % (float(total)/float(weave_size))
535
536
537
def usage():
538
print """bzr weave tool
539
540
Experimental tool for weave algorithm.
541
542
usage:
543
weave init WEAVEFILE
544
Create an empty weave file
545
weave get WEAVEFILE VERSION
546
Write out specified version.
547
weave check WEAVEFILE
548
Check consistency of all versions.
549
weave info WEAVEFILE
550
Display table of contents.
551
weave add WEAVEFILE [BASE...] < NEWTEXT
552
Add NEWTEXT, with specified parent versions.
553
weave annotate WEAVEFILE VERSION
554
Display origin of each line.
555
weave mash WEAVEFILE VERSION...
556
Display composite of all selected versions.
557
weave merge WEAVEFILE VERSION1 VERSION2 > OUT
558
Auto-merge two versions and display conflicts.
559
560
example:
561
562
% weave init foo.weave
563
% vi foo.txt
564
% weave add foo.weave < foo.txt
565
added version 0
566
567
(create updated version)
568
% vi foo.txt
569
% weave get foo.weave 0 | diff -u - foo.txt
570
% weave add foo.weave 0 < foo.txt
571
added version 1
572
573
% weave get foo.weave 0 > foo.txt (create forked version)
574
% vi foo.txt
575
% weave add foo.weave 0 < foo.txt
576
added version 2
577
578
% weave merge foo.weave 1 2 > foo.txt (merge them)
579
% vi foo.txt (resolve conflicts)
580
% weave add foo.weave 1 2 < foo.txt (commit merged version)
581
582
"""
583
584
585
586
def main(argv):
587
import sys
588
import os
589
from weavefile import write_weave, read_weave
590
from bzrlib.progress import ProgressBar
591
592
#import psyco
593
#psyco.full()
594
595
cmd = argv[1]
596
597
def readit():
598
return read_weave(file(argv[2], 'rb'))
599
600
if cmd == 'help':
601
usage()
602
elif cmd == 'add':
603
w = readit()
604
# at the moment, based on everything in the file
605
parents = map(int, argv[3:])
606
lines = sys.stdin.readlines()
607
ver = w.add(parents, lines)
608
write_weave(w, file(argv[2], 'wb'))
609
print 'added version %d' % ver
610
elif cmd == 'init':
611
fn = argv[2]
612
if os.path.exists(fn):
613
raise IOError("file exists")
614
w = Weave()
615
write_weave(w, file(fn, 'wb'))
616
elif cmd == 'get': # get one version
617
w = readit()
618
sys.stdout.writelines(w.get_iter(int(argv[3])))
619
620
elif cmd == 'mash': # get composite
621
w = readit()
622
sys.stdout.writelines(w.mash_iter(map(int, argv[3:])))
623
624
elif cmd == 'annotate':
625
w = readit()
626
# newline is added to all lines regardless; too hard to get
627
# reasonable formatting otherwise
628
lasto = None
629
for origin, text in w.annotate(int(argv[3])):
630
text = text.rstrip('\r\n')
631
if origin == lasto:
632
print ' | %s' % (text)
633
else:
634
print '%5d | %s' % (origin, text)
635
lasto = origin
636
637
elif cmd == 'info':
638
weave_info(argv[2], sys.stdout)
639
640
elif cmd == 'check':
641
w = readit()
642
pb = ProgressBar()
643
w.check(pb)
644
pb.clear()
645
646
elif cmd == 'inclusions':
647
w = readit()
648
print ' '.join(map(str, w.inclusions([int(argv[3])])))
649
650
elif cmd == 'parents':
651
w = readit()
652
print ' '.join(map(str, w._v[int(argv[3])]))
653
654
elif cmd == 'merge':
655
if len(argv) != 5:
656
usage()
657
return 1
658
659
w = readit()
660
v1, v2 = map(int, argv[3:5])
661
662
basis = w.inclusions([v1]).intersection(w.inclusions([v2]))
663
664
base_lines = list(w.mash_iter(basis))
665
a_lines = list(w.get(v1))
666
b_lines = list(w.get(v2))
667
668
from bzrlib.merge3 import Merge3
669
m3 = Merge3(base_lines, a_lines, b_lines)
670
671
name_a = 'version %d' % v1
672
name_b = 'version %d' % v2
673
sys.stdout.writelines(m3.merge_lines(name_a=name_a, name_b=name_b))
674
else:
675
raise ValueError('unknown command %r' % cmd)
676
677
678
if __name__ == '__main__':
679
import sys
680
sys.exit(main(sys.argv))
Loggerhead is a web-based interface for Breezy
Version: 2.0.1